A Comprehensive Epidemiological Study of Opisthorchiasis in a Rural Community of Thailand

Picha Suwannahitatorn

Imperial College London

Department of Infectious Disease Epidemiology

Thesis submitted for PhD examination

2015

Abstract

Opisthorchiasis is a zoonotic parasitic infection caused by human liver flukes. Opisthorchis viverrini (OV) is endemic in Southeast Asia along Mekong basin which estimated 9 million people are infected. Transmission to humans occurs through the consumption of uncooked fish; cyprinoid or white-scale freshwater fish containing infective stage metacercariae. The infection is mainly asymptomatic. Adult flukes could live up in the bile duct in the absence of treatment. Chronic infection is strongly related with bile duct cancer or cholangiocarcinoma. The International Agency for Research on Cancer has declared that Opisthorchis viverrini is a member of group 1 agent, carcinogenic to humans. Thailand has the highest incidence of cholangiocarcinoma in the world. However, opisthorchiasis is acknowledged as a neglected and underestimated disease on the global scale.

In the present day, the epidemiological data on opisthorchiasis from OV infection in Thailand are considered outdated. An extensive cross-sectional study was undertaken from 1984 to 2001 under the National Control Program, but data on incidence is very limited. Moreover, infection dynamics are still poorly understood.

This project aimed to comprehensively study the infection dynamics of OV infection using multiple tools in order to explore the infection in various aspects. Statistical models were developed to explore epidemiological data; prevalence, incidence and infection intensity, with risk factors for acquiring the infection. Infection dynamics will be described using mathematical modelling. The qualitative technique, by interviewing and group discussions, will be used to explain the cause of uncontrolled infection rate in bio-psycho-social aspect. Integrated study results will be used to develop community intervention strategies under the framework of the public health planning model. The overall outcome will be valuable for Thailand National Health Policy and epidemiological data will provide the basis for further rigorous academic research.

Declaration of originality

I hereby declare that this project is entirely my own work and produced under the supervision of Prof Christl Donnelly, Prof Joanne Webster and Prof Steven Riley.

Co-researchers and collaborators who participated in this work are acknowledged as stated below.

1. Prof Saovaanee Leelayoova, PhD Research investigator 2. Prof Mathirut Mungthin, MD PhD Research investigator 3. Assist Prof Ram Rangsin, MD DrPH Research investigator 4. Assoc Prof Paanjit Taamasri, MSc Research investigator 5. Dr Phunlerd Piyaraj, MD PhD Research investigator 6. Assoc Prof Wirote Areekul, MD Research collaborator 7. Assist Prof Pote Aimpun, MD PhD Research collaborator 8. Assist Prof Suthee Panichkupl, MD Research collaborator 9. Dr Tanongsan Tienthavorn, MD PhD Research collaborator 10. Mr Tawee Naaglor, MSc Laboratory manager 11. Mrs Worarachanee Imjaijitt, MSc Data unit manager

Any forms of additional source of information including published works, figures or quotes have been fully referenced in the thesis.

Picha Suwannahitatorn, MD MSc

October 2015

The copyright of this thesis rests with the author and is made available under a Creative Commons Attribution Non-Commercial No Derivatives licence. Researchers are free to copy, distribute or transmit the thesis on the condition that they attribute it, that they do not use it for commercial purposes and that they do not alter, transform or build upon it. For any reuse or redistribution, researchers must make clear to others the licence terms of this work.

Contents

Abstract ...... 2

Declaration of originality ...... 3

List of figures ...... 11

List of tables ...... 13

List of abbreviations and acronyms ...... 15

Acknowledgments ...... 17

Chapter 1 Introduction ...... 19

1.1 Overview ...... 19 1.2 Biology of Opisthorchis viverrini ...... 20 1.2.1 Morphology and life cycle ...... 20 1.2.2 Opisthorchiasis and cholangiocarcinoma ...... 23 1.3 Epidemiology of OV infection ...... 24 1.3.1 Current situation in the Mekong region ...... 24 1.3.2 Epidemiology of OV infection in Thailand ...... 27 1.3.2.1 Prevalence of OV infection ...... 27 1.3.2.2 Infection intensity of OV infection ...... 35 1.3.2.3 Incidence of OV infection ...... 36 1.4 Control of OV infection in Thailand ...... 39 1.4.1 Timeline for OV infection control in Thailand ...... 39 1.4.1.1 Survey phase ...... 39 1.4.1.2 National Control Program phase ...... 40 1.4.1.3 Passive surveillance strategy phase ...... 44 1.5 Main conceptual framework...... 45 1.6 Study aims and thesis outline ...... 47 1.7 Project outline ...... 48 1.7.1 Background of study community ...... 48 1.7.2 Background of community healthcare system ...... 51 1.7.3 Overview of study methods ...... 52

Chapter 2 Epidemiology of Opisthorchis viverrini infection: distribution and risk factors ...... 53

2.1 Introduction ...... 55 2.2 Methodology ...... 57 2.2.1 Study areas and population ...... 57 2.2.3 Questionnaires ...... 59 2.2.4 Collection of stool specimen ...... 59 2.2.5 Data analysis ...... 60 2.2.6 Model for prevalence and incidence data ...... 60 2.2.7 Model for infection intensity ...... 61 2.2.8 Model analysis and fitting ...... 66 2.2.9 Ethical consideration ...... 66 2.3 Results ...... 67 2.3.1 Population characteristics ...... 67 2.3.1.1 Response from the study ...... 67 2.3.1.2 Demographic data ...... 71 2.3.2 Uncooked fish consumption behaviors...... 73 2.3.3 Distribution and risk factors of OV infection ...... 76 2.3.3.1 Distribution of prevalence ...... 76 2.3.3.2 Distribution of incidence ...... 77 2.3.4 Evaluation of risk factors for acquiring OV infection ...... 79 2.3.4.1 Univariable analysis of risk factors for acquiring OV infection ...... 79 2.3.4.2 Model fitting ...... 80 2.3.4.3 Multivariable analysis of risk factors for acquiring OV infection ...... 81 2.3.5 Study of infection intensity ...... 85 2.3.5.1 Prevalence and intensity profiles ...... 86 2.3.5.2 The prevalence-age structure ...... 92 2.3.5.3 The age-prevalence-intensity structure ...... 93 2.3.5.4 Estimation of parasite aggregation ...... 95 2.3.5.5 Infection dynamics of infection intensity...... 96 2.3.5.6 Relationship of infection intensity and risk factors using count model ...... 98 2.4 Discussion ...... 103 2.4.1 Study design and response ...... 103 2.4.2 Distribution of OV infection ...... 106 2.4.2.1 Prevalence of OV infection ...... 106 7

2.4.2.2 Incidence of OV infection ...... 107 2.4.2.3 Intensity of the infection ...... 107 2.4.3 Risk factors for acquiring the infection ...... 109 2.5 Conclusion ...... 112 Chapter 3 Infection dynamics of OV infection...... 113

3.1 Introduction ...... 115 3.1.1 The relationship of prevalence and incidence ...... 117 3.2 Methodology ...... 119 3.2.1 Model definition ...... 119 3.2.2 Parameters of the model ...... 122 3.2.3 Fitting the model ...... 125 3.2.4 Basic reproductive number ...... 127 3.2.5 Estimation of treatment effectiveness using individual-level data ...... 128 3.3 Results ...... 130 3.3.1 Infection dynamics of OV infection ...... 130 3.3.2 Estimation of treatment effectiveness using population data and mathematical model ……………………………………………………………………………………………………………………………….132 3.3.3 Estimation of treatment effectiveness using individual-level data ...... 133

3.3.4 Basic reproductive number; R0 ...... 134 3.3.5 Estimating the effectiveness of the National Control Program using the model ...... 135 3.4 Discussion ...... 136 3.4.1 Infection dynamics and study limitation ...... 136 3.4.2 Basic reproductive number ...... 137 3.4.3 Treatment effectiveness ...... 138 3.4.4 Treatment failure ...... 138 3.5 Conclusion ...... 140 Chapter 4 Evaluation of risk behaviours for OV infection as determined by a qualitative approach ...... 141

4.1 Introduction ...... 143 4.2 Methodology ...... 146 4.2.1 Group interview ...... 146 4.2.2 In-depth interview ...... 146 4.2.3 Study population ...... 147 4.2.4 Study process ...... 148 8

4.2.4.1 Focus group discussion ...... 148 4.2.4.2 In-depth interview ...... 149 4.2.5 Data processing and analysis ...... 149 4.2.5.1 Thematic content analysis ...... 150 4.2.5.2 Framework analysis...... 150 4.2.5.3 Grounded theory ...... 150 4.3 Results ...... 153 4.3.1 Population characteristics ...... 153 4.3.2 Thematic contents ...... 156 4.3.2.1 Situation of the OV infection in the community ...... 156 4.3.2.2 Current situation of consumption behaviours ...... 157 4.3.2.3 Knowledge of the infection ...... 159 4.3.2.4 Perception of OV infection and its consequences ...... 161 4.3.2.5 Health concern of risk factors and the infection ...... 162 4.3.2.6 Diagnosis and treatment ...... 164 4.3.2.7 Role of National Control Program ...... 165 4.4 Discussion ...... 167 4.4.1 Social component integrated in the main conceptual framework ...... 168 4.4.2 Knowledge, attitude and perception of OV infection contributed by age structure.. 169 4.4.3 Attitude and perception towards risk behaviours ...... 171 4.4.4 Impact of OV infection on health, treatment and prevention ...... 174 4.5 Conclusion ...... 176 Chapter 5 Community-based intervention: PRECEDE-PROCEED model framework for controlling OV infection ...... 178

5.1 Introduction ...... 180 5.1.1 Overview of PRECEDE-PROCEED model ...... 181 5.2 Methodology ...... 185 5.2.1 Overview of study methods ...... 185 5.2.2 PRECEDE – PROCEED model establishment ...... 186 5.2.2.1 Step 1 Social diagnosis ...... 186 5.2.2.2 Step 2 Epidemiological diagnosis ...... 188 5.2.2.3 Step 3 Behavioural and environmental diagnosis ...... 188 5.2.2.4 Step 4 Educational and ecological diagnosis...... 188 5.2.2.5 Step 5 Administrative and policy assessment ...... 193 9

5.2.2.6 Step 6 Implementation ...... 194 5.2.2.7 Fieldwork methodology ...... 195 5.2.2.8 Step 7 Process evaluation ...... 203 5.2.2.9 Step 8 Impact evaluation ...... 203 5.2.2.10 Step 9 Outcome evaluation ...... 204 5.3 Results ...... 205 5.3.1 Step 1 Social diagnosis ...... 205 5.3.1.1 Demographic data ...... 205 5.3.1.2 Characteristics of the study populations from qualitative approach ...... 206 5.3.1.3 Precipitating factors for promoting uncooked fish consumption ...... 207 5.3.1.4 Precipitating factors to discontinue Koi pla consumption ...... 209 5.3.2 Step 2 Epidemiological diagnosis ...... 209 5.3.3 Step 3 Behavioural and environmental diagnosis ...... 209 5.3.4 Step 4 Educational and ecological diagnosis...... 210 5.3.5 Step 5 Administrative and policy assessment ...... 211 5.3.6 Step 6 Implementation ...... 213 5.3.7 Step 7 Process evaluation ...... 214 5.3.7.1 Summary of study flow and response rate ...... 214 5.3.7.2 Uncooked fish consumption from distribution and risk factors study ...... 219 5.3.7.3 Incidence of OV infection ...... 221 5.3.8 Step 8 Impact evaluation ...... 222 5.3.8.1 Evaluation of risk factors for acquiring OV infection ...... 222 5.3.8.2 Evaluation of impact of community intervention on Koi pla consumption ...... 223 5.3.9 Step 9 Outcome evaluation ...... 223 5.3.9.1 Evaluation of impact of community intervention on incidence of OV infection .... 223 5.3.9.2 Qualitative approach for evaluation of community intervention ...... 224 5.4 Discussion ...... 225 5.4.1 Study design and response ...... 225 5.4.2 Situation of OV infection ...... 228 5.4.2.1 Situation of Koi pla consumption ...... 229 5.4.2.2 Incidence of OV infection ...... 229 5.4.3 Community intervention ...... 230 5.4.3.1 Impact of intervention on Koi pla consumption ...... 230 5.4.3.2 Impact of intervention on incidence of OV infection ...... 232 10

5.4.4 Implication of PRECEDE – PROCEED model framework ...... 233 5.5 Conclusion ...... 236 Chapter 6 Discussion ...... 237

6.1 Overview ...... 237 6.1.1 Host and agent relationship ...... 239 6.1.2 Host and environment relationship ...... 239 6.1.3 Agent and environment relationship ...... 240 6.1.4 Development of planning model ...... 240 6.2 Summary of study findings ...... 242 6.2.1 Epidemiology of Opisthorchis viverrini infection: distribution and risk factors ...... 242 6.2.2 Infection dynamics of OV infection ...... 243 6.2.3 Evaluation of OV infection and risk behaviours using qualitative approach ...... 244 6.2.4 Community-based intervention: PRECEDE-PROCEED model framework for controlling OV infection ...... 246 6.3 Public health implication ...... 247 6.3.1 Primary prevention ...... 247 6.3.2 Secondary prevention ...... 248 6.4 Future work ...... 249 6.5 Conclusion ...... 252 References ...... 253

Appendices ...... 261

List of figures

Figure 1.1 Morphology of Opisthorchis viverrini egg and adult ...... 20 Figure 1.2 Life cycle of Opisthorchis viverrini ...... 21 Figure 1.3 Bithynia snail and cyprinoid fish ...... 22 Figure 1.4 Prevalence of opisthorchiasis in Southeast Asian and Southern China ...... 24 Figure 1.5 Strategic approach for liver fluke control ...... 40 Figure 1.6 Prevalence of opisthorchiasis under National Control Program Implementation 1987 - 2001 ...... 43 Figure 1.7 Main conceptual framework of study project ...... 45 Figure 1.8 Overview of study area ...... 48 Figure 2.1 Conceptual framework for distribution and risk factors of opisthorchiasis ...... 56 Figure 2.2 Flowchart of fieldwork ...... 58 Figure 2.4 Distribution of count data in hurdle model ...... 63 Figure 2.3 Distribution of count data in zero-inflated model ...... 63 Figure 2.5 Summarized flowchart for cross-sectional study ...... 67 Figure 2.6 Summarized flowchart for follow-up study ...... 69 Figure 2.7 Population pyramids ...... 71 Figure 2.8 Prevalence of OV infection distributed by age-group ...... 76 Figure 2.9 Incidence of OV infection distributed by age-group ...... 77 Figure 2.10 Prevalence and intensity of Na-yao 2002 cross-sectional study ...... 86 Figure 2.11 Frequency distribution of EPG from Na-yao 2002 cross-sectional study ...... 86 Figure 2.12 Prevalence and intensity of Na-yao 2004 follow-up study ...... 88 Figure 2.13 Frequency distribution of EPG from Na-yao 2004 follow-up study...... 88 Figure 2.14 Prevalence and intensity of Na-ngam and Na-isarn cross-sectional study ...... 90 Figure 2.15 Frequency distribution of EPG from Na-ngam and Na-isarn cross-sectional study ...... 90 Figure 2.16 Fitting –ln(s) with age ...... 92 Figure 2.17 k estimates from study areas ...... 95 Figure 2.18 Predicted prevalence from varying k and M for Na-yao cohort data ...... 96 Figure 2.19 Predicted incidence from varying k and M for Na-yao prevalence data ...... 96 Figure 2.20 Predicted prevalence from varying k and M for Na-ngam and Na-isarn prevalence data 97 Figure 3.1 Conceptual framework for infection dynamics...... 116 Figure 3.2 Fitting prevalence with age-structure incidence and fieldwork incidence ...... 118 Figure 3.3 Model diagram for opisthorchiasis infection dynamics ...... 119

Figure 3.4 Differential equations of S1IS2R model...... 122

Figure 3.5 Model diagram considered the R0 ...... 127 Figure 3.6 Diagram for estimation of treatment effectiveness using individual-level data ...... 128 Figure 3.7 Infection dynamics of prevalence in study area ...... 131 Figure 3.8 Estimation of treatment effectiveness from the model ...... 132 Figure 3.9 Basic reproductive number with various duration of infectiousness ...... 134 Figure 3.10 Fitting model with prevalence from National Control Program...... 135 Figure 4.1 Conceptual framework for qualitative approach ...... 144 Figure 4.2 Population pyramid of Na-ngam 2012 prevalence data ...... 153 Figure 4.3 Conceptual framework of opisthorchiasis ...... 167 12

Figure 4.4 Social component integrated to conceptual framework ...... 167 Figure 5.1 Diagram of PRECEDE – PROCEED model ...... 181 Figure 5.2 Conceptual framework for community intervention study ...... 183 Figure 5.3 Summary of study flowchart ...... 185 Figure 5.4 Summary of fieldwork flowchart ...... 194 Figure 5.5 Community intervention flowchart ...... 197 Figure 5.6 Distribution and risk factors study flowchart ...... 198 Figure 5.7 Population pyramid of Tung-heang 2012 data ...... 205 Figure 5.8 Implementation of fieldwork ...... 213 Figure 5.9 Summarized flowchart for community trial study ...... 214 Figure 5.10 Summarized flowchart for distribution and risk factors study ...... 217 Figure 5.11 Incidence of opisthorchiasis distributed by age-group ...... 221 Figure 5.12 PRECEDE – PROCEED model integrated to conceptual framework ...... 233 Figure 6.1 Integrated conceptual framework from comprehensive study ...... 237

List of tables

Table 1.1 Summary of prevalence and intensity of OV infection ...... 29 Table 1.2 Prevalence of OV infection in overseas Thai population ...... 32 Table 1.3 Summary of incidence of OV infection ...... 33 Table 1.4 Summary of annual prevalence during National Control Program implementation...... 42 Table 1.5 Summary of fieldwork data collection ...... 52 Table 2.1 Study areas for risk and distribution analysis ...... 57 Table 2.2 Response for cross-sectional study ...... 68 Table 2.3 Response for follow-up study ...... 70 Table 2.4 Population characteristics of cross-sectional study ...... 72 Table 2.5 Population characteristics of follow-up study ...... 72 Table 2.6 Multivariable analysis of population characteristics and uncooked fish consumption behaviours in cross-sectional study ...... 74 Table 2.7 Multivariable analysis of population characteristics and uncooked fish consumption behaviours in follow-up study ...... 75 Table 2.8 Proportional distribution of OV infection diagnosis categorised by stool examination methods ...... 78 Table 2.9 Model fitting for prevalence and follow-up study ...... 80 Table 2.10 Multivariable analysis of risk factors for acquiring OV infection of prevalence data from Na-isarn 2013 study ...... 81 Table 2.11 Multivariable analysis of risk factors for acquiring OV infection of prevalence data from Na-ngam 2012 study ...... 82 Table 2.12 Multivariable analysis of risk factors for acquiring OV infection of follow-up data from Na- yao 2012 study ...... 83 Table 2.13 Multivariable analysis of risk factors for acquiring OV infection of follow-up data from Na- ngam 2014 study ...... 84 Table 2.14 Summary of positive egg count diagnosed by Kato-Katz and FECT methods ...... 85 Table 2.15 Summary of Na-yao 2002 intensity data ...... 86 Table 2.16 Summary of Na-yao 2004 intensity data ...... 88 Table 2.17 Summary of Na-isarn and Na-ngam 2012 - 13 intensity data ...... 90 Table 2.18 Summary of prevalence-age structure ...... 92 Table 2.19 k estimates from study areas ...... 95 Table 2.20 Summary of count model for infection intensity analysis ...... 98 Table 2.21 Multivariable analysis of risk factors and infection intensity of Na-yao 2004 follow-up data ...... 99 Table 2.22 Multivariable analysis of risk factors and infection intensity of Na-isarn and Na-ngam 2012-13 prevalence data ...... 101 Table 2.23 Model fitting for Na-yao follow-up study ...... 102 Table 2.24 Model fitting for Na-isarn and Na-ngam cross-sectional study ...... 102 Table 3.1 Obtained incidence from fieldwork and estimated incidence from prevalence-age structure ...... 117 Table 3.2 Summary of fieldwork data for Na-yao area ...... 120 Table 3.3 Summary of fieldwork data for Na-ngam area ...... 121 14

Table 3.4 Model parameters ...... 123 Table 3.5 Estimation of duration of infectiousness and recovery rate ...... 130 Table 3.6 Estimation of treatment effectiveness from Na-yao 2002-04 data ...... 133 Table 3.7 Estimation of treatment effectiveness from Na-ngam 2012-14 data ...... 133 Table 3.8 Basic reproductive number ...... 134 Table 4.1 Selection criteria for focus group discussion ...... 147 Table 4.2 Analysis approach to qualitative data ...... 149 Table 4.3 Age structure classified by generations of Na-ngam 2012 prevalence data ...... 153 Table 4.4 Characteristics of focus group discussion participants ...... 154 Table 4.5 Characteristic of in-depth interview participants ...... 155 Table 4.6 Situation of the OV infection in the community ...... 156 Table 4.7 Current situation of consumption behaviours ...... 157 Table 4.8 Knowledge of the infection ...... 159 Table 4.9 Perception of OV infection and its consequences ...... 161 Table 4.10 Health concern of risk factors and the infection ...... 162 Table 4.11 Diagnosis and treatment ...... 164 Table 4.12 Role of National Control Program ...... 165 Table 4.13 Knowledge, attitude and perception of OV infection contributed by age structure ...... 169 Table 4.14 Common characteristics regarding knowledge, attitude and perception across generations ...... 172 Table 4.15 Impact of OV infection on health, treatment and prevention ...... 174 Table 5.1 Community mutual agreement...... 192 Table 5.2 Selection criteria for focus group discussion ...... 201 Table 5.3 Review of study process and method of assessment ...... 203 Table 5.4 Population characteristics of Tung-heang area ...... 206 Table 5.5 Response for community trial study ...... 215 Table 5.6 Population characteristics of community trial study ...... 216 Table 5.7 Response for distribution and risk factors study ...... 218 Table 5.8 Population characteristics of distribution and risk factors study ...... 219 Table 5.9 Distribution of uncooked fish consumption ...... 219 Table 5.10 Univariable and multivariable analysis of population characteristics and uncooked fish consumption behaviours ...... 220 Table 5.11 Univariable and multivariable analysis of risk factors for acquiring OV infection ...... 222

List of abbreviations and acronyms

AIC Alkaline information criterion

BPP Border Patrol Police

CAB Community advisory board

CCA Cholangiocarcinoma

CI Confidence interval

CIP Community intervention package

DALY Disability-adjusted life year

EPG Egg per gram

FECT Formalin-ether concentration technique

FGD Focus group discussion

HBM Health belief model

IARC International Agency for Research on Cancer

IRR Incidence rate ratio

IQR Interquartile range k parasite aggregation parameter

LC Latent class

M Mean infection intensity

MIF Minute intestinal fluke

MCMC Markov chain Monte Carlo

NBLH Negative binomial logit hurdle model

NBRM Negative binomial regression model

NTDs Neglected tropical diseases 16

OR Odds ratio

OV Opisthorchis viverrini

PCR Polymerase chain reaction

PDR People's Democratic Republic

PLH Poisson logit hurdle model

PPM PRECEDE-PROCEED model

PRM Poisson regression model

Q1 1st quartile

Q2 2nd quartile

Q3 3rd quartile

Q4 4th quartile

R0 Basic reproductive number

S.D. Standard deviation

SCT Social cognitive theory

ZINB Zero-inflated negative binomial model

ZIP Zero-inflated Poisson model

Acknowledgments

I would like to gratefully thank my supervisors; Prof Christl Donnelly and co-supervisors Prof Joanne Webster and Prof Steven Riley for all their support. Prof Christl Donnelly has used her expertise in statistics to guide me through this project for technical support in rigorous data analysis. Prof Joanne Webster provided her knowledge supporting me in the aspect of parasitology and Prof Steven Riley has guided me to use fieldwork data to formulate mathematical models. Moreover, the entire project is carefully supervised through my PhD. They have enabled me to use all my potential to achieve this challenging task.

Before I came to Imperial College London, I believed this University to be one of the top universities, and from my earliest experiences during my MSc course in 2011, this was quickly confirmed by the wealth of inspiration, the perfect atmosphere conducive to learning and abundance of friendly, warm and resourceful colleagues supporting my growth and development. I knew right away, that studying at ICL will stand to be one of my best decisions throughout life. I consider every moment here as precious and I wish to express my gratitude to the entire college.

I acknowledge and gratefully thank the research team and members of staff from Thailand for their collaboration. I would also like to thank staff members from the Department of Parasitology and Department of Military and Community Medicine, Phramongkutklao College of Medicine for cooperation and Phramongkutklao Research Fund for financial support to my research.

I would like to thank former head of Department of Parasitology; Prof Saovaanee Leelayoova and my academic mentor Prof Mathirut Mungthin for motivation and inspiration to commence work on the neglected opisthorchiasis disease. I also thank my current head of department; Assoc Prof Paanjit Taamasri and my colleague Dr Phunlerd Piyaraj for all the support since the beginning.

The fieldwork could not have been done without the initiation of a community project started in 2002 under the visionary researchers from Department of Military and Community Medicine; Assoc Prof Wirote Areekul (former head of department) and Assist Prof Ram Rangsin (current head of department). Research colleague; Assist Pote Aimpun and Assist Suthee Panichkupl provided full support for my research.

I would like to address Mr Tawee Naaglor and Mrs Worarachanee Imjaijitt as potential collaborators. Mr Tawee’s skill in laboratory work and Mrs Worarachanee’s expertise in biostatistics are outstanding. I am very grateful to have them in the research team. 18

On behalf of the research team, I acknowledge the collaboration of the authorities of The Chachoengsao Provincial Health Office, The Sanamchaikhet District Health Office, Local health volunteers and all participants from Na-yao, Na-ngam, Na-isarn and Tung-heang village for their great contributions.

I would like to thank my parents and my younger sister. It is a tough decision for choosing academic pathway from my medical doctor background. From the moment I decided to focus on research, rather than medical practice, the way was not smooth and there were many challenging tasks. Without them, I would not have succeeded to accomplish my academic life. Their love and care are unconditional and infinite. The overwhelming gratitude also extends to the support from all of my family members.

I sincerely want to thank those who give me true love and care. Every element of your support helps me to develop this work along the way. For those who sometimes make my life difficult, I forgive them.

Lastly, I wish to grant this contribution to the Royal Thai Government, Ministry of Science and Royal Thai Army Medical Department for their sponsorship and making this opportunity come true.

Picha Suwannahitatorn, MD MSc

October 2015

Chapter 1 Introduction

1.1 Overview Opisthorchiasis is a zoonotic parasitic infection caused by human liver flukes. Three major species are responsible for the infection (1-4): Opisthorchis viverrini, Opisthorchis felineus and Clonorchis sinensis (disease is also called clonorchiasis). An estimated 700 million people are at risk worldwide (5). Opisthorchis viverrini is endemic in Southeast Asian along the Mekong Basin (6-9) including Thailand, Lao PDR and Cambodia where an estimated 9 million people are infected (5, 9-11).

OV is a pathologically important food-borne trematode (12, 13), leading to infection within the hepato-biliary system. Transmission to humans occurs through the consumption of uncooked fish; cyprinoid or white-scale freshwater fish containing infective stage metacercariae (2, 14). After infection, OV can live in the bile duct in the absence of treatment (15-17). Many studies have indicated that chronic infection is strongly related to a bile duct cancer, cholangiocarcinoma; CCA (17-21). The International Agency for Research on Cancer (IARC) has declared that OV is a member of group 1 agents, carcinogenic to humans (5, 18). Thailand is the highest ranked in the world for CCA incidence, with ranges from 93.8 to 317.6 /100000 person-years (15-17, 20, 22-24). However, opisthorchiasis is acknowledged as a neglected and underestimated disease on the global scale (10, 11, 25).

In Thailand, the prevalence of OV infection was extensively studied in the past especially during 1984 – 2001 under the National Control Program implementation (26-28). The latest national survey was performed in 2001 (28). The record shows the geographical variation in infection prevalence is not homogenously distributed through the country. The endemic areas are confined to locations where intermediate hosts and uncooked fish consumption behaviours are present (1, 5, 15, 29).

In the present day, the data on disease epidemiology are considered outdated and the infection dynamics are still poorly understood (10, 11, 25, 30).

1.2 Biology of Opisthorchis viverrini 1.2.1 Morphology and life cycle Human liver flukes are digenetic trematodes belonging to the family Opisthorchiidae (2, 3, 13, 31). The 3 majors species found to be infectious to humans are Opisthorchis viverrini, Opisthorchis felineus and Clonorchis sinensis. Others species are rarely reported (13).

Figure 1.1 Morphology of Opisthorchis viverrini egg and adult

Image of egg and adult OV as shown in Figure 1.1 were obtained by myself at Department of Parasitology, Phramongkutklao College of Medicine (October, 2015).

The adult worm (Figure 1.1, right) of OV is 7.0 (5.4-10.2) X 1.5(0.8-1.9) mm on average. The fluke morphology is dorso-ventrally flattened, thin and transparent with reddish-blue colour. Eggs are oval (Figure 1.1, left), 17 X 15 microns, with Ill-developed operculum and an aboperculum knob at the bottom. The life cycle is complex (12, 32-36) and comprises multiple intermediate hosts involving several stages of parasitic growth development as shown in Figure 1.2.

Figure 1.2 Life cycle of Opisthorchis viverrini

Figure 1.2 is obtained from free-access database from Laboratory Identification of Parasites of Public Health Concern, US CDC, http://www.dpd.cdc.gov/dpdx/html/opisthorchiasis.htm.

The life cycle of Opisthorchis viverrini was fully described in 1965 (2). Embryonated eggs are released with the faeces (1) in natural water bodies such as ponds, rivers or flooded rice fields. In the water, eggs are ingested by Bithynia snails (2) serving as the first intermediate host. Miracidia (2a) hatch from the eggs and grow in many stages as seen in the Figure, miracidia (2a)  sporocysts  radiae (2c)  cercariae (2d). The growth and development from miracidia to cercariae takes approximately 2 months after snail infection. Free-swimming cercariae are released from the snails (3) and become encysted under freshwater fish muscle layer or scales becoming infective metacercariae (4) in 21 days. Cyprinoid freshwater fish serve as the second intermediate host.

Uncooked freshwater fish containing encysted metacercariae are consumed by humans. Cysts are digested and become excysted in the duodenal part of the small bowl (5). Metacercariae will travel to the biliary duct through the ampulla of Vater (the opening of common bile duct to the intestine) where they grow to fully developed adults. The adult flukes are ready to reproduce by laying eggs in 3 – 4 weeks. In total, it would take 4 months to complete the full life cycle. 22

There are two medically important stages involving human explained below.

The infective stage (i): encysted metacercariae in fish muscles or scales serve as the infective stage when humans consume raw fish.

Diagnostic stage (d): opisthorchiasis can be diagnosed by identifying fluke eggs in faecal materials or bile content. More importantly, eggs release parasite offspring into the environment.

Figure 1.3 Bithynia snail and cyprinoid fish

Image of Bithynia snail (left) and cyprinoid fish (right) as shown in Figure 1.3 were obtained by myself at Department of Parasitology, Phramongkutklao College of Medicine (October, 2015). Figure 1.3 shows the general appearance of the Bithynia snail and cyprinoid fish. The prevalence of cercariae in snails is typically low; 0.05 - 0.07% (32). The diversity and abundance of snails are strongly related to seasonal variation mainly depending on rainfall. The population of snails are abundant during the rainy season, which is also the time of rice planting and dramatically decrease during the dry season. Therefore, controlling snails is impractical due to the seasonal variation in the snail population (32, 37).

Unlike snails, the prevalence of infection in cyprinoid fish is high (38). Up to 95% of cyprinoid fish species harbour the metacercariae. In Thailand, more than 15 species are found in natural water bodies and serve as a good source of food.

The reservoir hosts are also reported in various species of animals including dogs, cats, rats and pigs. These reservoir hosts are responsible for the transmission when they excrete contaminated stool in the environment. Studies have shown that reservoir hosts might have some important role as an area with minimal transmission from humans that still maintains re-infection rates (9, 39). 23

1.2.2 Opisthorchiasis and cholangiocarcinoma

Once the flukes become fully grown, they reside in the biliary duct. The host’s immune system will extensively response to the pathogen producing inflammatory mediators also damaging the host’s own tissues (12, 21). The parasites can evade immune mechanisms. However, the recitative host response from chronic inflammation causes abnormal tissue growth: biliary epithelial cell metaplasia. Long-term infection also induces accumulation of tissue fibrosis; a scar to the epithelial wall. Many cytokines (IL-6, TNF-α) are responsible for altering the regular cell cycle to abnormal cell proliferation including the mucus glands and fibrous tissue proliferation ending in advanced periductal fibrosis (12, 15, 16). Carcinogenesis of OV is explained by secretion of growth factor from the parasite. Many studies and experiments have demonstrated an ES product (excretory/secretory) called OV-GRN-1 can stimulate abnormal cell proliferation and create a suitable environment for tumour development (17, 22).

Opisthorchiasis is typically asymptomatic (15, 20). Abdominal discomfort is also seen but the symptom is nonspecific. Even though bile duct thickening is always found, it does not cause bile passage blockage so jaundice rarely occurs from chronic infection. The simple way to diagnose infection is through stool examination. Although adult flukes can be detected by bile duct aspiration, the procedure is invasive and is impractical in clinical use. Opisthorchiasis can be treated with a single dose of praziquantel. Drug side effects are typically minimal including nausea or drowsiness. A field study in Thailand shows a high drug efficacy of 90 – 95% (28, 40).

Cholangiocarcinoma is a primary malignancy of bile duct (16, 17, 19, 23). It is an extensively invasive tumour and often causes metastasis. No definitive treatment is currently available result in death within a short period of time. Early stage is always asymptomatic or causes nonspecific manifestations, which are difficult to detect in general health practice. Late stage tumour grow and invade the liver and major blood vessels causing extensive metastasis, which is always non- resectable (15, 17, 20, 22). Bile duct obstruction results in jaundice and liver failure. Prognosis is very poor (41), median 5-year rate is approximately 30 % for non-resectable tumour. The IARC has declared that Opisthorchis viverrini is carcinogenic to humans (2, 11, 42). The burden from OV infection and subsequent CCA costs 120 million USD annually for medical care and lost productivity (11) and results in 74,070 DALYs (43).

1.3 Epidemiology of OV infection 1.3.1 Current situation in the Mekong region

Liver flukes are geographically distributed with regard to intermediate hosts including snails and cyprinoid fish (33, 34, 37, 44, 45). Opisthorchis viverrini and Clonorchis sinensis are prevalent along the Mekong region including Southeast Asian and Southern China (1, 5)

Figure 1.4 Prevalence of opisthorchiasis in Southeast Asian and Southern China

Figure 1.4 is adapted from Sithithaworn et al. (5) depicting the prevalence of Opisthorchis viverrini and Clonorchis sinensis, which is geographically-related to the Mekong River. The Mekong River origin is located in southern China with an estimated length of 4000 km ranking it the 12th – longest river in the world. From the Tibetan Plateau, the river runs through China's Yunnan Province and flows through Southeast Asian forming a natural border along Thailand, Lao PDR and Myanmar, then running eastward along the Thai and Laos PDR border and through the country of Cambodia then continuing to the South Chinese Sea. The river serves as an important water reservoir along the Mekong Basin. 25

Opisthorchis viverrini is abundant in area A below the grey dotted line (A) including Thailand, Lao PDR, Cambodia and Vietnam while Clonorchis sinensis is prevalent in area B within Southern China. Clonorchiasis is not common in Southeast Asian except as reported in northern Vietnam bordering China. However, sporadic cases are reported from central Thailand from PCR technique and within the study area by Traub et al. (46).

Opisthorchis viverrini has been reported in Thailand for almost 100 years. In the past, multiple surveys have shown a nearly 100% prevalence in northern and northeastern provinces (27, 28). The prevalence is relatively low in central and southern Thailand. Biologically, natural intermediate hosts can be found across the country (11, 47). However, further studies indicate the relationship of uncooked fish consumption behaviors and high prevalence of OV infection (4, 29) related to geographical population characteristics.

With the aid of foreign support and governmental concern (26-28, 48), the National Control Program was implemented from 1984 – 2001 mainly focusing on case diagnosis and treatment. The national prevalence was dramatically decreased to 10%. However, no on-going surveillance was conducted due to the national policy and economic burden.

Accuracy regarding the field diagnostic methods, mainly performed with simple smear and Kato-Katz technique, is a concerned issue. Concentration and molecular techniques indicate that simple smear and Kato-Katz are less sensitive, which could lead to gross underestimation of the infection prevalence (49, 50).

For Lao PDR, the prevalence is particularly higher in lowland rather than highland areas. The Laotian cultural context is considered similar to Thai especially in the northeastern and also northern provinces along the Mekong Basin. Thai and Lao languages are partially similar; therefore, cultural- transfers across the border are common. The practice of uncooked fish consumption is similar and prevalence of OV infection along the basin is also high (29, 51, 52). However, a recent study has evaluated the efficacy of several drug choices against OV infection in Lao PDR showing promise for future implementation of the National Control Program (53).

In Cambodia, data on OV infection are considered limited. As shown in Figure 1.4, the data are available only along the Mekong Basin in limited part of Cambodia. Therefore, no accurate estimate exists on the situation of the infection (5). Additionally, numbers of Cambodian immigrants are currently settled in Thailand and also travel freely across the border. The given population dynamics might impact transmission dynamics especially in the future when Southeast Asia will become an economic community with free population movement. 26

Vietnam, has reported both Opisthorchis viverrini and Clonorchis sinensis transmission. Clonorchiasis is reported in the northern region (54) bordering southern China. Consumption of uncooked fish is also reported in Vietnamese populations (5). However, no official data is available on national prevalence.

Moreover, sporadic cases of OV infection have been occasionally reported in other native Southeast Asian populations such as Malaysia and Singapore (55). Moreover, OV infection has also been reported among overseas Thais who work abroad (discussed in Section 1.3.2.1 and Table 1.4).

1.3.2 Epidemiology of OV infection in Thailand 1.3.2.1 Prevalence of OV infection

As seen in Table 1.1, prevalence and intensity have been estimated across the country. With regard to Figure 1.1, the prevalence is high in the north and northeastern of Thailand. Most of the reports emphasized these areas considered endemic. Previous reports from 1980 – 1990 also focus on the assessment to support the National Control Program.

The first report of OV infection in Thailand was described by Leiper and Kerr (56) in 1911 where they found that 17% of 230 adult male prisoners in Chiang Mai, Northern Thailand were infected with Opisthorchis felineus. It was later confirmed by Sadun (39) in 1965 that reported cases in Thailand were from Opisthorchis viverrini not Opisthorchis felineus.

Regional surveys aimed to assess the burden of opisthorchiasis from 1950 – 1987. The surveys supported by the US government from 1950 – 1958 promoted the intestinal helminth and liver fluke control program in the National policy. After a field trial of treatment was conducted the National Control Program of OV infection was then finally initiated in 1984 (discussed in Section 1.4). The data collected from 1984 – 1987 in 4 provinces indicated that the prevalence of OV infection was 63.6% and the nationwide treatment was started in 1987 (26) under the 6th National Public Health Development Plan (1987 – 1991) which reported the prevalence before the treatment was 33.9%.

National surveys were reported twice by the 7th (1992 – 1996) and 8th National Public Health Development Plans (1997 – 2001) (26-28, 48).

For the 7th National Public Health Development Plan 3,007,125 subjects were examined with Kato- Katz method reporting a national prevalence of opisthorchiasis of 21.5%. The prevalence for each region was 32.5 % in the North, 15.3% in the Northeast, 16.7% in Central region and 0.1% in the South.

The National Control Program finished in 2001 at the end of the 8th National Public Health Development Plan, and the reported national prevalence was 9.7%. Regional prevalence was 19.3% in the North, 12.4% in the Northeast, 3.8% in the Central region and 0% in the South.

Additionally, national surveys were based on Kato-Katz technique. A report based on the FECT method was noted from 1982 – 89 in a study conducted on human autopsy cases from fatal car accidents where stool specimens were directly collected from the large intestine (26).

From 1987 – 2001, prevalence was reported mostly in the northern and northeastern regions. In 1985, the reported prevalence in Khon Kaen was 38.1% in the study of self-pay scheme for diagnosis 28

and treatment of OV infection (57). A report from Ramathibodi Hospital, Bangkok investigated patients of northeast origin, living in rural and urban areas, who visited or were admitted. The result should that the prevalence of OV infection among rural dwellers was significantly higher than among those living in urban areas (79.4% and 54.8%, respectively) (58). A study in Loei and Nong Khai Provinces in Northeastern Thailand reported that the people living far from the river had a higher prevalence of the infection than those living on the bank despite the result that bank dwellers consumed more uncooked fish (59). 29

Table 1.1 Summary of prevalence and intensity of OV infection

Study Study Diagnostic Prevalence Intensity Details Author Study area period population method (%) (EPG) Jongsuksuntikul (26-28, Northeast; 1984-87 629522 Kato-Katz 63.6 National survey initiated by Department of 48) Khon Kaen, Roi-Et, Communicable Disease Control Sakol Nakorn and Ubon Ratchathani province Northeast; 1994 1912 Kato-Katz 18.6 1371.0 Thai-German Community Health Development 19 provinces through Parasitic Control Project Consumption of uncooked fish 7.0-42.0 % Northeast 1987-91 5238062 Kato-Katz 33.9 6th National Public Health Development Plan (1987–1991) Nationwide 1992-96 3007125 Kato-Katz 21.5 876.3 7th National Public Health Development North 32.6 1226.4 Plan (1992–1996) Northeast 15.3 268.5 Central 16.7 1034.6 South 0.1 N/A Nationwide 1997- 1062725 Kato-Katz 9.7 8th National Public Health Development North 2001 19.3 Plan (1997–2001) Northeast 12.4 Consumption of uncooked fish 15.0-59.8 % Central 3.8 South 0.0 Sithithaworn (60) Northeast; 1982-89 181 Stoll’s dilution 66.9 53.3 Study conducted in human autopsy cases from Khon Kaen province technique fatal accidents FECT Feces collected from rectum or terminal colon 30

Study Study Diagnostic Prevalence Intensity Details Author Study area period population method (%) (EPG) Mongkolintra (57) Northeast 1985 1417 N/A 38.1 Kurathong (58) Northeast; 1987 126 Kato-Katz 54.8 Urban area Northeast; 433 79.4 Rural area Tesana (59) Northeast; 1991 N/A N/A 41.3 Prevalence of infection among the people Loei and Nong Khai residing far from the rivers was higher than those province residing on the banks Maleewong (61) Northeast; 1991 2412 N/A 66.4 uninfected: 33% Nakhon Phanom light: 59% moderate: 7% heavy: 1% Radomyos (62, 63) Northeastern; 1994 681 Stool 92.4 16 provinces sedimentation North; 1998 431 technique 11.6 16 provinces Sriamporn (64) Northeast; 1990- 18393 Stoll’s dilution 24.5 Study aimed to determine the association of Khon Kaen province 2001 technique prevalence of OV infection and incidence of FECT cholangiocarcinoma Waree (65) Northern; 2000 584 FECT 10.8 Phitsanulok province Wiwanitkit (66) Northeast; 2001 183 FECT 8.1 Udonthani province Waikagul (67) Northern; 2001 1010 N/A 1.7 31

Study Study Diagnostic Prevalence Intensity Details Author Study area period population method (%) (EPG) Nan province Saksirisampant (68) Various regions 2001 2213 FECT 28.9 Health screening program in Thai workers for oversea employment conducted in The King Chulalongkorn Memorial Hospital in Bangkok Rangsin (69) Central; 2002 668 Simple-smear 21.3 Study served as baseline population for cohort Chachoengsao Kato-Katz study province FECT Suwannahitatorn (70) Central; 2007 1024 Simple-smear 18.6 Study served as baseline population for cohort Chachoengsao Kato-Katz study province FECT 2011 1038 Simple-smear 6.2 Unpublished data Kato-Katz Study served as baseline population for PhD FECT research project Tungtrongchitr (71) Northeast; 2007 479 Kato-Katz 14.8 Study compared sensitivity of direct simple smear Khon Kaen province to Kato-Katz technique Northeast; 1124 Simple smear 32.0 Ubon Ratchathani Kato-Katz province Kaewpitoon (72, 73) Northeast; 2011 333 Kato-Katz 9.9 Study conducted in age ≥ 60 years population Surin province Northeast; 2011 1168 Kato-Katz 2.5 Nakhon Ratchasima province Saengsawang (8) Northeast; 2012 1569 Kato-Katz 38.7 Yasothon province 32

Table 1.2 Prevalence of OV infection in overseas Thai population

Study Study Diagnostic Prevalence (%) Intensity Details Author Study area period population method (EPG) Hira (74) Kuwait 1982-86 N/A FECT 10 cases were Case series from annual report of in-patient and reported out-patient screening program Estimated screening specimen 18000 per annum from 4 hospitals Peng (75) Taiwan 1991-92 1364 FECT 7.0 Thai laborers joining the Six-Year National Construction Project in Taiwan

Cheng (76) Taiwan 1992-93 302 FECT 43.0 Among infected cases, 77.1 % known to have infection and 67.4 % received treatment before arrival Lo (77) Taiwan 1993-94 N/A FECT 16.6 Greenberg (78) Israel 1994 98 N/A 56.1 Wang (79, 80) Taiwan 1992-96 7670 FECT 7.0 Taiwan 1998-99 702 FECT 2.1 Maneeboonyang (81) Thai-Myanmar 2004 286 FECT 19.2 Thai male army personnel operating along Thai- border Myanmar border

Table 1.3 Summary of incidence of OV infection

Study Study Diagnostic Incidence (%) Intensity Details Author Study area period population method (EPG) Upatham (82) Northeast; Khon 1980 - 1651 Stoll’s 46.4 % during Incidence is significantly higher in male than Brockelman (83) Kaen province 1982 dilution 1980 – 1981 female technique 19.4 % during Brockelman used catalytic infection model for fit 1981 - 1982 age-prevalence profile Sornmani (84) Northeast; Khon 1984 942 Trial Stoll’s 2.0 /month in A pilot trial to assess praziquantel efficacy against Kaen province group dilution trail group OV infection for community implementation 442 technique 5.0 /month in Control control group group Upatham (40) Northeast; Khon 1987 - 704 Stoll’s 87.7 % of re- Re-infection rate was assessed 1 year after Kaen province 1988 dilution infection rate treatment with praziquantel technique after 1 year follow-up Rangsin (69) Central; 2002 - 526 Simple-smear 21.6/100 Chachoengsao 2004 Kato-Katz person-years province FECT Suwannahitatorn (70) Central; 2007 - 980 Simple-smear 21.4/100 Chachoengsao 2009 Kato-Katz person-years province FECT 34

Concentration techniques were used from 1994 – 2001 studies. Radomyos el al. (62, 63) reported prevalence from stool sedimentation technique: 92.4% from 16 northeastern provinces and 11.6% from 16 northern provinces. A cohort study was conducted among 18,393 subjects from 1990 – 2001 to measure the incidence of CCA (64); the study aimed to determine the association of OV infection prevalence with incidence of CCA. The reported prevalence was 24.5% using FECT. More reports with FECT technique were in Phitsanuloke (65) and Nan (67) Provinces in Northern Thailand with prevalence of 10.8% and 1.7%, respectively. Another survey in Udon Thani, northeastern Thailand (66) reported a prevalence of 8.1%.

In 2002, Rangsin et al. (69) initiated a health study in a rural area of Chachoengsao, central Thailand. The multiple fieldwork projects reported on the prevalence of OV infection. Published data (69, 70) indicated that prevalence was 21.3% in 2002 and 18.6% in 2007. Fieldwork prevalence will further discussed.

A study in 2007 (71) compared the sensitivity of direct simple smear and Kato-Katz technique reporting a prevalence of 14.8% in Khon Kaen and 32.0% in Ubon Ratchathani Provinces. Prevalence surveys in northeastern provinces reported various geographic prevalence of 9.9 – 38.7 % from 2011 – 2012 (72, 73).

Additional prevalence was reported among overseas Thais (Table 1.2) who worked in Taiwan, Israel and Kuwait. All of them were examined in their working country during employment. The population movement caused impact on population dynamics and parasite transmission dynamics. In 2001, a health screening program was conducted for Thai workers before overseas employment (68) reporting that 28.9% of 2213 screened candidates were infected with OV infection.

The study of Thai workers in Taiwan was under the Six-Year National Construction Project from 1991 – 1997 (75, 77, 79) indicating the prevalence ranged from 7.0 to 43.0%. Additional health screening from 1998 – 1999 reported a prevalence of 2.1% (76). The study suggested the prominent characteristics of particular local Thai uncooked fish dishes were associated with OV infection.

The infection was also reported from Middle East countries, which were nonendemic area, and where uncooked fish consumption was not a regular habit. One case report was described in Kuwait (74) for 10 Thai migrants infected with Opisthorchis viverrini. The screening test for Thai workers in Israel (78) found a prevalence of 56.1%.

Maneeboonyang (81) reported the prevalence among Thai male army personnel operating along the Thai-Myanmar border in Ratchaburi Province was 19.2%. The area was endemic but almost all personnel were northeastern Thais. 35

1.3.2.2 Infection intensity of OV infection

As seen in Table 1.1, available data for infection intensity were considered limited during the implementation period of the National Control Program from 1992 – 1996 (26, 27). A 1994 survey supported by the Thai-German Community Health Development though Parasitic Control Project reported a mean EPG of 1371.0 among 1912 subjects. From the 7th National Public Health Development Plan from 1992 – 1996, the national mean EPG was 876.3 and the highest mean EPG was in the northern region.

A study conducted among human autopsies reported an EPG of 53.3 (60) from stool samples. A study conducted in 14 villages in Nakhon Phanom Province near the Mekong Basin on the Thai-Lao PDR border (61) in 1994 reported a prevalence of 66.4%. Considering the infected cases, 59% of cases were light intensity, only 1% had heavy infection compared with the 1996 National Survey reporting a heavy infection of 1.9% (27).

A notable study of infection intensity was conducted in 1988 in the evaluation of praziquantel treatment where the intensity of infection was measured between pre- and posttreatment. The result stated that high pre-treatment intensity subjects were more likely to have a heavier intensity of re-infection indicating the predisposing factors contributing to heavy infection (40).

Ramsey (85) and Haswell-Elkins (86) conducted stool expulsion with chemotherapy from 1989 - 1991 and found that heavy intensity was aggregated in a small proportion of the infected population and over-dispersion distribution was suggested. However, no recent data regarding infection intensity and its distribution characteristics of OV infection have been reported in Thailand. This study provides up-to-date data regarding the infection intensity.

1.3.2.3 Incidence of OV infection Although data on prevalence are available, estimates of the incidence of OV infection are extremely limited. From Table 1.3, notable studies of incidence were conducted from 1980 – 1982 in a rural area of Khon Kaen Province, northeastern Thailand. Upatham et al. (82) conducted a fieldwork measuring incidence within 2 periods: 1980 – 81 and 1981 – 82. The incidence was 46.4% and 19.4%, respectively. Brockelman et al. (83) later used a catalytic model for fitting age-prevalence data from the fieldwork to estimate incidence from mathematical modeling compared with directly-estimated incidence (discussed in Chapter 2).

Sornmani et al. (84) conducted a pilot clinical trial to assess the efficacy of praziquantel in 1984 to implement the treatment in the National Control Program. The result indicated that incidence in the control group was 5.0% monthly compared with 2.0% monthly in the trial group. Upatham (40) assessed the re-infection rate after praziquantel treatment. One-year follow up in a cohort population revealed an average re-infection rate of 87.7% yearly.

Rangsin (69) reported the incidence of OV infection in northeastern-originated community in Chachoengsao Province, central Thailand was 21.6/100 person-years from 2002 – 2004. Suwannahitatorn (70) later re-evaluated the incidence in the same study area and reported the incidence of 21.4/100 person-years from 2007 – 2009. Moreover, both Rangsin and Suwannahitatorn cohort studies were conducted with similar methodology. Only negative cases for OV infection from baseline cross-sectional study were re-examined in follow-up study which could resulted in bias when the incidence of OV might be underestimated.

1.3.3 Source of infection From Section 1.2.1, mode of infection is consumption of uncooked fish containing metacercariae. Therefore, the primary mode of infection will be intentionally consumed uncooked fish (2). Utensils and hands contaminated with fish scraps are also responsible for the infection.

OV infection is endemic in the north and northeastern parts of Thailand where eating habits closely follow local traditions. Uncooked fish is considered popular resulting in various dishes based on this method of preparation. In Thai including the local dialect, the word “Pla” means fish;

1) Instantly prepared uncooked fish: freshwater fish is chopped and mixed with spicy ingredients. Raw fresh meat is usually denatured by lime juice due to acidity which dramatically changes its colour to a cooked-like texture. One Ill-known popular dish is Koi pla (chopped raw fish salad). Koi pla is always consumed instantly. The variation of Koi pla is Larb pla, fish meat that is thoroughly chopped and mixed with particular ingredients. 37

2) Moderately fermented fish: fish meat is preserved in highly-concentrated salt for a few days to one week. The period of preservation varies and might affect the viability of metacercariae. Studies have reported that fish preserved less than 7 days were likely to be infective (47, 87) The viability of metacercariae depends on the concentration of preserving solution and the period of preservation (88), which can be extended to minute intestinal flukes as Ill. Popular dishes for moderately fermented fish are Pla som and Pla jom.

3) Extensively fermented fish: freshwater fish are preserved in highly concentrated salt solution for 3 - 6 months. The dish is widely known as Pla ra. Chunks of fish could be consumed uncooked and the remaining fish sauce is often used as a daily ingredient for local dishes (29, 87, 89).

Various choices of uncooked dishes result in a variety of eating practices. The pattern of consumption behavior differs regarding the particular uncooked fish dish. Koi pla is consumed as a main dish and limited to special occasions such as social drinking, religious ceremonies or festive celebrations. However, the dish is popular with respect to a high proportion of individuals who have experienced this dish. Pla ra itself has different patterns of consumption; the fish sauce is the main ingredient for local dishes, for example, local Thai papaya salad or well-known Som tum is dressed with Pla ra sauce for local consumption. The consumption behaviors are considered related to social factors. However, the underlying social influence dynamics have not yet been extensively studied (9). Despite the importance of epidemiological data, a study sociocultural dynamics relating to Opisthorchis viverrini in the Mekong Basin (29) by Grundy-Warr et al. suggested the importance of social dynamics, which should be incorporated with parasite infection dynamics, human behaviors and public health awareness.

Consumption of Koi pla and Pla ra were firstly described by Sadun in 1955 (39) for the method of preparation. Association of Koi pla consumption with risk of OV infection are mostly addressed by cross-sectional studies (25, 58, 90, 91). The prevalence of Koi pla consumption was high in the north and northeastern Thailand, up to 80%. The pattern was still observed even among overseas Thai workers, who were from endemic areas of OV infection (68, 79, 80). Based on the cross-sectional design, the causal relationship of uncooked fish consumption and OV infection could not be fully explained due to the exposure and outcome is measured at the same time. However, the National Control Program has implemented interventions on avoiding uncooked fish based on given biological properties consistent with the mode of infection. Without a clear underlying explanation regarding the epidemiology of OV infection and its risk factors, the intervention package could have issues with efficacy especially with primary prevention, which could prevent newly-infected individuals. 38

A cohort design can provide better understanding in terms of causal-relationship and data with which to estimate the incidence of infection, which is important considering the role of prevention and control campaigns. As described in Section 1.3.2.3, the data on incidence are scant and available data are considered outdated. The incidence from follow-up studies conducted in Na-yao area rom 2002 – 2009 (69, 70) provided up-to-date data and highlighted the relationship of Koi pla consumption and OV infection. My study will analyze more data on risk factors related to infection intensity, which could provide more understanding of infection dynamics.

1.4 Control of OV infection in Thailand 1.4.1 Timeline for OV infection control in Thailand The history of OV infection dates back to 1950 when the field survey was initiated by the government. The timeline to manage OV infection in national level can be divided in 3 main phases described below.

1.4.1.1 Survey phase This phase included the assessment of parasitic infection burden, field study of prevention and field trial of chemotherapy.

First, a wide-scale survey was conducted from 1950 – 1958 by the Ministry of Health with technical and financial support from the US government. The Helminthiasis Control Unit was organized to diagnose and treat intestinal parasitic infections including OV infection. The service was provided to 5 provinces, 4 provinces in the northeastern region (Nakhon Ratchasima, Udon Thani, Sakon Nakhon and Ubon Ratchathani) and Songkhla Province in the Southern region.

The support from the US government was terminated in 1958 resulting in the termination of The Helminthiasis Control Unit. However, the service scheme was still active and was later integrated with the rural health development project of the Health Development Division. The main strategy still targeted diagnosis and treatment.

The implementation of prevention campaigns was conducted from 1967 – 1974 under the Liver Fluke Control Unit. The trial targeted primary prevention in discontinuing uncooked fish consumption and preparation. The pilot study was conducted in Sakon Nakhon Province emphasizing health education including introducing methods for cooked fish preparation and also distributed cooking pots to support cooking equipment.

So far, the implementation of OV infection control from 1950 – 1974 included health education as primary prevention and diagnosis and treatment as secondary prevention.

From 1980 - 1983, praziquantel efficacy was firstly assessed as a widely used chemotherapy. The Faculty of Tropical Medicine, Mahidol University initially reported the efficacy of a single dose of 40 mg /1 kg body weight for praziquantel treatment against OV infection that provided a cure rate of 91%.

The collaboration was established between Mahidol University and the Department of Communicable Disease, Ministry of Health to extend the study to field trials which reported a cure rate with praziquantel of 95.5%. 40

Furthermore, the Helminthiasis Section of the Department of Communicable Disease set up liver fluke control units in 4 northeastern provinces; Khon Kaen, Roi Et, Sakon Nakhon and Ubon Ratchathani. The result should that treatment was provided to 400,452 infected cases from 629,522 infected cases (approximately 63.6%, Table 1.4) from 1984 – 1987.

1.4.1.2 National Control Program phase

This phase included the implementation of the National Control Strategy following 3 main approaches. This phase emphasized the control program and evaluated the reduction in prevalence. The scheme was extended to national scale where the activity was prioritized and incorporated in the National Public Health Development Plan from 1987 – 2001. The National Control Program was integrated from the 6th to 8th National Public Health Development Plans following strategic activity.

Figure 1.5 Strategic approach for liver fluke control

Figure 1.5 is adapted from Jongsuksuntigul et al. (28) summarizing 3 main strategic concepts for controlling liver flukes. The primary endpoint was to reduce prevalence of OV infection. The campaign included primary and secondary prevention.

Primary prevention: the prevention scheme focused on providing health education to promote cooked fish consumption. 41

Secondary prevention: the mobile stool examination unit was organized to provide service once a year including diagnosis and treatment for OV infection. A small fee was collected from the participants for the diagnosis (10 baht – 20p) and additional 10 baht for positive result which later treated with praziquantel. Hygienic defecation was promoted through health education to reduce the transmission of fluke egg in natural water bodies.

Community participation was engaged by the mobile stool examination in which the strategy prioritized case seeking and treatment. The activity continued for consecutive years. If prevalence decreased below 10% in any community, the program switched to passive surveillance strategy.

Therefore, the outcome from primary prevention was not directly determined as the incidence of infection was not obtained. Moreover, outcome from providing health education was not yet quantified. Re-infection rate and infection intensity were not measured as well; therefore, the infection dynamics regarding the program efficacy solely relied on the report of prevalence of infection.

Table 1.4 Summary of annual prevalence during National Control Program implementation

National Public Health Year No. positive/total N Prevalence (%) 95% CI* Development Plan Pre-implementation 1984-87 400452 / 629522 63.6 63.5 – 63.7 1988 160308 / 450677 35.6 35.4 – 35.7 1989 407309 / 1343110 30.3 30.2 – 30.4 6th 1990 533147 / 2193275 24.3 24.2 – 24.4 1991 437241 / 2747011 15.9 15.8 – 16.0 1992 363689 / 2876423 12.6 12.6 – 12.7 1993 311809 / 2834725 11.0 10.9 – 11.0 7th 1994 300720 / 2766837 10.9 10.8 – 10.9 1995 330780 / 3130563 10.6 10.5 – 10.6 1996 365010 / 3007125 12.1 12.1 – 12.2 1997 125805 / 1532678 8.2 8.1 – 8.3 1998 128037 / 1375915 9.3 9.2 – 9.4 8th 1999 110920 / 1234408 9.0 8.9 – 9.0 2000 97419 / 1448877 6.7 6.6 – 6.8 2001 100059 / 1062725 9.5 9.2 – 9.6 * revised from original paper

The prevalence from Table 1.4 was modified from Jongsuksuntigul P et al. (28). The prevalence in 1995 was corrected to 10.6% from direct calculation; 330780/3130563 as it reported 11.0% from the original Table. The 95% CI was calculated by binomial distribution regarding the reported results.

A large-scale field trial of praziquantel efficacy was conducted from 1984 – 87. The successful cure rate led to expanding the treatment coverage to the National Control Program implementation, initiated in the 6th National Public Health Development Plan from 1987 – 1991. The program was applied to all health facilities in the northeastern region of Thailand under the supervision of the Department of Communicable Disease Control. Stool examination was conducted among 5,238,062 participants of whom 1,774,929 were diagnosed with OV infection and treated with praziquantel; 33.89% (95% CI: 33.84 – 33.93).

Moreover, the Promotion of Community Health through Parasite Control Project was conducted by the support of Germany to provide stool examination and treatment in 7 northeastern provinces. A total of 1,839,813 cases were examined while 531,174 cases were positive and treated; 28.87% (95% CI: 28.81 – 28.94). 43

As shown in Table 1.4, the success in reducing the prevalence from 63.6 to 15.9% by the end of the 6th plan resulted in continuing the programs in the following 7th (1992 – 1996) and 8th (1997 – 2001) National Public Health Development Plans. The implementation covered all provinces in the northeast and some provinces in the central region resulting in 42 provinces receiving coverage (55.3% from 76 provinces). From1992 – 1994, 5,238,062 cases were examined and 632,869 infected cases were treated; 12.08% (95% CI: 12.05 – 12.11).

0.7 0.64

0.6

0.5

0.4 0.36

0.30 0.3 0.24 Prevalence (as proportion) proportion) Prevalence (as

0.2 0.16

0.13 0.12 0.11 0.11 0.11 0.09 0.09 0.08 0.09 0.1 0.07

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 Year Years

Figure 1.6 Prevalence of opisthorchiasis under National Control Program Implementation 1987 - 2001

Figure 1.6 shows the prevalence trend from the data in Table 1.4. From 1997, the prevalence fell under 10% and continued to remain below 10% for 5 consecutive years until the end of the 8th National Public Health Development Plan as shown in Table 1.6. Therefore, the program was considered to have achieved the goal and then switched to passive surveillance strategy. Figure 1.6 shows the prevalence trend from the data in Table 1.4. From 1997, the prevalence fell under 10 % and continued to remain below 10 % for 5 consecutive years until the end of 8th National Public Health Development Plan as shown in Table 1.6. Therefore, the program was considered archiving goal and then switched to passive surveillance strategy.

1.4.1.3 Passive surveillance strategy phase From the 9th National Public Health Development Plan to the current 11th plan (2001 up to the present), the control strategy was merged with rural health services. The scheme focuses on targeted areas where the infection is still a burden. Therefore, the priorities of the control activity depend on the situation and policy in each particular province.

From Table 1.1, the cross-sectional study conducted after 2001 focused on the northeastern and northern provinces which were identified as endemic areas. This study was conducted from 2002 – 2007 in Chachoengsao Province, central Thailand, which was considered as a non-endemic area showing that prevalence and incidence were high (69, 70). This issue raised a question regarding the infection dynamics of OV infection for the distribution and the factors contributing to the infection.

1.5 Main conceptual framework

Intervention Prevention Diagnosis and Treatment

Risk behaviors Opisthorchiasis

Social influence

Figure 1.7 Main conceptual framework of study project

In this project, opisthorchiasis will be defined as OV infection which do not examine the clinical disease, but rather emphasize on epidemiological parameter in infection dynamics.

From reviewed literature and available data, gaps were presented with regard to infection dynamics of OV infection. A comprehensive study was required to establish up-to-date knowledge in various aspects.

The study aim was to explore the infection dynamics of OV infection to fill the gap with regard to various aspects of the infection. Based on the main conceptual framework in Figure 1.7, the infection dynamics were considered a complex interaction shaped by multiple factors as described below.

 OV infection risk is affected by risk behaviors and also shaped by risk factors. The aim was to explore the epidemiology of the infection using fieldwork study with rigorous methods. The 46

results included the up-to-date prevalence and incidence of infection. The risk factors were identified with a better explanation of the causal relationship through the cohort study design.  The infection intensity was examined. Other studies addressed the over-dispersion and density- dependent property (85, 86), which will be further explored in this thesis. Statistical modeling was formulated to analyze the relationship between risk factors and OV infection with respect to its intensity distribution.  The relationship between prevalence and incidence was described with mathematical modeling. The model was fitted to fieldwork data to explore the infection dynamics.  Socio-culture dynamics have been important (29) in determining the prevalence of risk behaviors. The qualitative approach was introduced to obtain social-related data. The qualitative method allowed me to collect and analyze underlying bio-psycho-social aspects of the infection and explore responses and interactions from individual to community levels to and with the infection.  Quantitative and qualitative methods were combined as comprehensive data to design a planning model for public health implications to conduct a pilot community trial study to explore the effectiveness and feasibility of the intervention.  Effectiveness of primary prevention was assessed through community trials and secondary prevention was assessed by statistical and mathematical modeling fitted to fieldwork data to determine the cure rate from praziquantel treatment.  Comprehensive data provided more insights about infection dynamics. The introduction of qualitative methods, novel statistical and mathematical modeling completed the gap of research questions addressed in previous studies, which was crucial for the future implementation of OV infection studies in Thailand.

1.6 Study aims and thesis outline The project aims to comprehensively study the infection dynamics of Opisthorchiasis from OV infection by using multiple approaches with regard to statistical analysis, mathematical modelling, social science and public health planning.

The thesis comprises 6 chapters. Sub-objectives are outlined below.

Chapter 1: The first chapter aims to explain the current situation of OV infection in Thailand and the role of the National Control Policy.

Chapter 2: This chapter explores the distribution of OV infection included prevalence, incidence and infection intensity. The age-prevalence-intensity models were formulated allowing for over- dispersion (for example, using zero-inflated distributions), a common characteristic of helminthic density distributions.

Chapter 3: Chapter three constructs mathematical modelling to explore the infection dynamics and the effectiveness of praziquantel treatment using obtained fieldwork data.

Chapter 4: This chapter explores the complexity of bio-psycho-social aspects to explain risk- behaviour relationships using qualitative methods.

Chapter 5: This chapter introduces the principle and theory behind the community-based approach, the relationship between disease biology, human factors and environmental context.

Chapter 6: The final chapter summarizes the overall picture of disease burden, what the current distribution and risk factors are and how intervention would impact on disease and community. Research data would extend knowledge of disease epidemiology and reflect the current situation of OV infection to initiate a community intervention package (CIP) and expand this work on the national scale.

1.7 Project outline The study area was located in Tha-kradan Subdistrict, Sanamchaikhet District, Chacheongsao Province in central Thailand as shown in Figure 1.8. Chacheongsao Province is about 100 km east of Bangkok. However, the research site was isolated and remotely located from the centre of the province.

Figure 1.8 Overview of study area

The fieldwork was conducted in 4 areas; Na-yao, Na-ngam, Na-isarn and Tung-heang Villages. The total population consisted of approximately 5000 residents.

1.7.1 Background of study community Before the settlement of Tha-kradan and neighboring villages, the geographical area was covered with jungle comprising natural ponds and small rivers making it rich in natural resources. Therefore, a private sawmill was later opened and operated under governmental authorization for the wood industry. The “Sukapoom” was the first operational sawmill located in a nearby area along the riverside to produce raw wooden materials for the industrial sector.

Most of the workers were employed from the northeastern region of Thailand. Some of them were working in the wood factory as a part-time job during off-harvest season and returned to their hometown when the rice farming started in the early rainy season. 49

The northeastern region of Thailand was originally arid and droughty and sometimes made the locals find it difficult to make a living and the Tha-kradan area was also more resourceful and more suitable for agriculture. Therefore, some native villagers chose to emigrate and settle permanently in the Tha-kradan area.

The very first settlement comprised small groups of people from various provinces in the northeastern moving to the Tha-kradan area in 1975; they claimed and possessed large areas of empty lands. The following immigrants purchased lands from them at a very cheap price; 40 pence for 1 rai (local Thai unit, approximately 1600 m2 or 0.4 acres). Each household owned more than 100 rai because title deeds for land ownership were not yet available.

The first settled community was peaceful but not uniformly organized. Each household lived on their own creating a problem when the community grew bigger. In 1983, the Border Patrol Police (BPP) managed to take care of the community safety. Because the area was close to the country’s border, incidences of drug smugglers and conflicts were encountered from minority tribes, so the BPP setup a community center and relocated the outer households to be gathered together as a whole community.

In 1984, the community leader cooperated with the BPP to organize the landscape planning; areas were designated for schools and a temple later built by community members. The landscape was divided in 4 rai squares (800x800 m). Each block was surrounded by road and filled with 4 households. The school was under BPP supervision; their staff served as teachers and also maintained border security.

From 1989 - 91, the community grew rapidly; current households attracted their relatives to move in. The lands were purchased at much higher prices. Compared with 1975, the pricing was 20 times higher in 1983. Within 15 years, the land price was 100 times higher in 1990.

The rapidly growing community resulted in resource management. Even though the landscape was Ill-planned, the community setting was rural. They still lacked fundamental infrastructure.

In 1994, HRH Princess Maha Chakri Sirindhorn visited the BPP School and noticed that poverty still affected the villagers’ quality of life. Agriculture lacked adequate technology resulting in underproduction and irrigation was poorly managed. HRH the princess took the BPP School under her patronage and also setup an office for community development. The scheme was under her major royal projects covering all BPP schools located along the national border. The office for community development has since been continuously operated and she has makes regular visits to Na-yao BPP School until the present. 50

Agriculture has remained a major occupation since the community foundation. In the early period, most agricultural works involved the timber industry as they were primarily sawmill workers. After the forest was reclaimed in the very first period, the landscape was transformed into lowlands suitable for rice field. As they kept moving to higher ground, the upland became suitable for farm plants such as corn and sesame.

The forest area dramatically decreased raising environmental concerns. The authorities later declared the remaining forest was a conservation area, which ended the timber industry in the area. As a consequence, agricultural areas expanded. The villagers started to grow various kinds of plants including cassava, soybeans, jute, green beans and sunflowers. However, corn was still a major agricultural product. The lowland was exclusively suitable for rice fields because it needed large amounts of water. The authorities supported the villagers by promoting rice seed and proper irrigation.

As the community grew, the commercial sector also expanded. The central market was set up to exchange local products. Local groceries imported goods such as clothes and household appliances from external suppliers so the villagers did not need to travel downtown. The local cooperative was formed in 1995 to facilitate the distribution of agricultural products.

Up to the present, the villages are still considered a rural environment in context where the majority of residents still preserve Northeastern culture. The inflow of modernization has resulted in the development of infrastructure and thus started the transitional period to urbanization. However, the local traditions including the local dialect, local folk customs, and local calendar ceremonies and eating behaviours are still widely maintained.

1.7.2 Background of community healthcare system

In the beginning, no official health care facilities existed in the community either modern medicine physician or traditional doctor. Villagers relied on self-care or herbal medicine. In 1984, a major outbreak of malaria resulted in more than 200 cases with over 20 fatalities. The fatal outbreak triggered health concerns within the community.

In 1987, local clinics were opened by two medical doctors providing the first local healthcare. One of them was obstetrics, so modern delivery was also available. Morbidity and mortality from malaria were still reported annually but continued to decrease. In 1990, the malarial unit was setup to provide immediate diagnosis and treatment to the villagers and local clinics also grew.

The local healthcare center was established in 1995 located within the temple area in the middle of the community. Three staff were stationed at the facility providing primary health care and some secondary healthcare services. One of the staff was a medical doctor who also opened his own private clinic. However, the facility was not always equipped with medical doctors. The majority of staff was health officers and nurses.

The local healthcare center was under district hospital supervision. Medical doctors paid regular visits especially for chronic disease appointments. Any cases beyond the facility’s scope were transferred to higher level healthcare. The closest facility was the district hospital, a 1-hr drive away. The road was the only exclusive method of transportation for referrals.

1.7.3 Overview of study methods

Table 1.5 Summary of fieldwork data collection

Study outcome Availability of study outcome Area Year Source of data Study type Risk factor Outcome Egg count assessment 2002 Secondary Cross-sectional Prevalence Yes No 2004 Secondary Follow-up Incidence Yes Yes Na-yao 2007 Secondary Cross-sectional Prevalence No No 2011 Secondary Cross-sectional Prevalence No Yes 2013 Primary Follow-up Incidence No Yes Cross-sectional Prevalence Yes Yes 2012 Primary Na-ngam Qualitative Qualitative No Yes 2014 Primary Follow-up Incidence No Yes Na-isarn 2013 Primary Cross-sectional Prevalence Yes Yes 2010 Secondary Cross-sectional Prevalence No Yes Tung-heang Community trial Incidence No Yes 2012 Primary Qualitative Qualitative No Yes

As seen in Table 1.5, the method of data collection was based on fieldwork study since 2007. Data obtained from fieldwork came from 2 sources; primary data was directly obtained by the research team. Secondary data was obtained from earlier fieldwork. The method was a mixed-method design comprising quantitative and qualitative methods.

The quantitative method consisted of two approaches: cross-sectional and cohort design. Enrolled participants were required to provide stool examination and undergo questionnaire assessment.

A cross-sectional study was conducted in Na-ngam and Na-isarn Areas to assess disease burden. A follow-up study was conducted in Na-yao and Na-ngam Areas for which the corresponding cross- sectional study served as baseline. Each participant was required to complete a standardized questionnaire to assess basic demographic data and risk behaviours for acquiring OV infection.

The qualitative techniques used in this study were focus group discussions and in-depth interviews conducted in Na-yao and Tung-heang Area.

For community trials, a planning model was applied to the community to systematically engage OV infection. The community intervention was developed with an effort to build community participation concerning local and environmental context under health behavioural theory to archive goals to reduce OV infection through behavioural modification by reducing risk behaviours. 53

Chapter 2 Epidemiology of Opisthorchis viverrini infection: distribution and risk factors Abstract

Opisthorchis viverrini infection or OV infection is a pathologically important food-borne trematode, where transmission to humans occurs through the consumption of uncooked fish containing infective stage metacercariae leading to infection within the hepato-biliary system. Chronic infection is strongly associated with to a bile duct cancer or cholangiocarcinoma, where Thailand holds the highest ranked incidence in the world. Cholangiocarcinoma can be partially prevented through prevention and control measures against the parasite itself. However, OV infection is acknowledged as a neglected and underestimated disease on a global scale and the available epidemiological data are considered outdated.

This study aimed to provide the most up-to–date and comprehensive study of the epidemiology of OV infection in Thailand. Prevalence and incidence were directly measured through both cross- sectional and follow-up studies in three study areas: Na-yao, Na-isarn and Na-ngam Areas during 2012 – 2014 in population 1 – 87 years of age. Risk factors for acquiring the infection were assessed by age-group and uncooked fish consumption and identified through statistical modelling. Infection intensity was determined to reflect the infection burden through eggs per gram. Regression models for count outcome were used to examine patterns of intensity distribution including Poisson and negative binomial distributions. Compound model, zero-inflated and hurdle models were incorporated to incorporate over-dispersed and zero excess data.

For age structure, quartiles were used for equally-proportioned age groups. The prevalence of OV infection was 7.85% (95% CI: 4.96 – 10.75) in Na-isarn and 9.29% (95% CI: 6.64 – 11.93) in Na-ngam area. The incidence rate was 7.98/100 person-years (95% CI: 5.49 – 11.20) in Na-yao and 6.80/100 person-years (95% CI: 4.68 – 9.54). The infection intensity data were over-dispersed, suggesting that the use of negative binomial distribution was more appropriate than Poisson distribution. A negative binomial hurdle model was the model of choice for zero excess and over-dispersed count data in this study. Final models included logistic regression model for prevalence data, Poisson model for cohort data and negative binomial logit hurdle model (NBLH) for intensity data with two covariates: age group and consumption of Koi pla (local instantly-prepared uncooked fish). The results demonstrated that consumption of Koi pla was a potential risk factor for acquiring the infection as it showed a significant association from 5 of 6 models. Koi pla consumption was also shown to increases the risk of OV infection and infection intensity when incorporated with the NBLH model. 54

Increasing age-groups linearly increased infection risk when compared with the 1st quartile age- group in the NBLH model but not for infection intensity. Prevalence linearly increased with age. Incidence was calculated from prevalence through the regression model. The age-prevalence- intensity relationship indicated an age-structure component within the intensity and level of parasite aggregation. Assuming incidence is uniform in all ages, the intensity tended to increase and the parasite spread more equally.

Up-to-date information on disease dynamics has a potential impact on public health and further ongoing extensive research in this field is essential to provide effective public health management.

2.1 Introduction

OV infection is a parasitic infection caused by human liver flukes. Three major species are responsible for the infection (1-4): Opisthorchis viverrini (OV), Opisthorchis felineus and Clonorchis sinensis (infection is also called clonorchiasis). As described in Section 1.3, OV infection is a pathologically important food-borne trematode, where transmission to humans occurs through the consumption of uncooked fish containing infective stage metacercariae (12, 13), leading to infection within the hepato-biliary system (15-17). Many studies have indicated that chronic infection is strongly related to a bile duct cancer, cholangiocarcinoma CCA (17, 19-21, 92). In Thailand, knowledge of the current incidence of infection is extremely limited and the available prevalence data are considered outdated in many areas (5, 9, 30, 40, 42, 69, 83). The record showed geographical variation in infection prevalence. The prevalence was higher in the North (19.3%) and the Northeast (15.7%), where people regularly consume dishes containing uncooked freshwater fish (27, 28). Consumption patterns are strongly influenced by culture and traditional beliefs and a similar pattern was also observed throughout the Mekong basin region. Endemic areas were thus revealed to be confined to where intermediate hosts and uncooked fish consumption behaviours are present.

In late 2002, Rangsin et al. (69) conducted a health survey in Na-yao Area, a rural community located in central Thailand, and found an infection prevalence of 21.3% (95% CI: 18.2 – 24.6). The negative cases were enrolled in a follow-up study in February 2004. The incidence of OV infection and risk factors were evaluated by a community-based cohort design showing the incidence was 21.6/100 person-years. In 2007 - 2009, Suwannahitatorn et al. (70) re-evaluated the situation in the same area found prevalence was 18.6% and incidence was at 21.4/100 person-years. This indicated a high prevalence area outside endemic areas where the national control strategy is no longer targeted.

The study of the prevalence and incidence can be extended to infection intensity. Studies of helminths such as Schistosoma haematobium, S. mansoni, Ascaris lumbricoidis, Trichuris trichiura and hookworm, for example (93-95), all tend to be characterized by a pattern of infection intensity with over-dispersed distributions, where few individuals are infected with a large burden of worm load while the majority are infected with light or no infection. Such a typical over-dispersion characteristic of helminth infection raise technical issues to analyses, such as in relation to excess zero count outcomes and parasite aggregation. Conventional statistical models such as the Poisson model may be technically problematic to properly explain such a distribution. Extended models can therefore be introduced to count model families to explore the issues such as negative binomial distribution or compound model. 56

My hypothesis is that Opisthorchis viverrini infection will be characterized by an over-dispersed distribution pattern. Therefore, compound models including zero-inflated and hurdle models were introduced in this study to characterise infection intensity. The model framework was useful for excess zero data (96, 97) and are applicable in many areas, such as economics, ecology and agriculture. Zero-inflated models are also becoming more popular in medicine (97). For helminth epidemiology, however, the application of the compound model remains limited (95).

Intervention

Prevention Diagnosis and Treatment

Risk factors Opisthorchiasis

Social influence

Figure 2.1 Conceptual framework for distribution and risk factors of opisthorchiasis

As illustrated in Figure 2.1, I will explore the distribution of OV and evaluate risk factors contributing to the infection though statistical modelling. The study aimed to:

 Explore the prevalence and incidence of OV and evaluate risk factors for acquiring the infection, and  Explore the infection intensity; distribution pattern, population dynamics of age-prevalence- intensity data and apply models for over-dispersion and excess zeros in count outcomes. 57

2.2 Methodology

2.2.1 Study areas and population

The study area was located in Na-yao, Na-isarn and Na-isarn area located in Tha-kradan Subdistrict, Sanamchaikhet District, Chacheongsao Province in central Thailand. Population characteristic can be referred to Section 1.7.1 for details.

Table 2.1 Study areas for risk and distribution analysis

Data collection Infection Area Year Source of data Study type Intensity outcome Risk behaviours (egg count) (yes/no) 2002 Secondary Cross-sectional Yes Yes No 2004 Secondary Follow-up Yes Yes Yes Na-yao 2011 Secondary Cross-sectional Yes No Yes 2013 Primary Follow-up Yes No Yes 2012 Primary Cross-sectional Yes Yes Yes Na-ngam 2014 Primary Follow-up Yes No Yes Na-isarn 2013 Primary Cross-sectional Yes Yes Yes

This chapter comprises two sources of data. Primary data were obtained from the fieldwork from 2012 - 2014. Secondary data were obtained from the Na-yao area, which was surveyed before and conducted as a continuing study as summarized in Table 2.1.

The Na-yao 2002 - 04 published study (69) by Rangsin and colleagues described incidence and risk factors associated with OV.

A cross-sectional study was conducted in Na-ngam and Na-isarn areas to assess disease burden by comprehensive survey.

Follow-up population was enrolled from corresponding baseline cross-sectional study. A follow-up study was conducted using a community-based retrospective cohort design in Na-yao and Na-ngam areas. From baseline cross-sectional study, negative cases for OV infection were eligible for a current follow-up to assess the incidence of OV. Enrolled participants were also required to report their risk behaviours retrospectively from the time they joined the survey

2.2.2 Fieldwork procedure

Eligible subject

Excluded from exclusion criteria

Informed consent

Data management Enrolled subject  Questionnaire data Receive survey package entry  Stool container  Stool examination data  Questionnaire entry Data analysis  Matching result  Subject excluded for Stool examination incomplete data

Negative for parasitic Receive stool examination infection result  Health education

Positive for parasitic infection  Health education  Receive medical treatment

Figure 2.2 Flowchart of fieldwork

The fieldwork protocol for cross-sectional and follow-up studies was similar, as illustrated in Figure 2.2. Fieldwork was organized from the community health centre in the middle of the research area where a field laboratory and data processing unit were set up. The team worked together with village heath volunteers to co-operate with the research participants. Eligible participants were invited in the study. The enrolment process included providing written informed consent forms and some participants were excluded from the study by exclusion criteria.

After the enrolment process, each participant received a survey package which included a questionnaire and a sealable container for collecting a stool specimen. Stool examination was performed after specimen collection. The stool examination result and questionnaire data were recorded and matched for subsequent data analysis. Participants received their personal stool examination result; reporting intestinal pathogenic parasite infection such as helminth and protozoa infection, and health education based on National Health Guidelines. Participants with OV received proper medical treatment depending on their identified pathogen. 59

2.2.3 Questionnaires

Each participant was required to complete a standardized questionnaire to assess basic demographic data and risk behaviours for acquiring the infection. The questionnaire had been validated and used earlier in the same study area. Basic demographic data included sex, age and occupation. Risk behaviours were determined by assessing uncooked-fish eating behaviour patterns (69, 70, 98). Uncooked fish preparations are classified in two major groups as described below.

1) Instantly prepared uncooked fish

A well-known popular dish is Koi pla (chopped raw fish salad). Freshwater fish is chopped and mixed with spicy ingredients. Raw fresh meat is usually denatured by lime juice due to acidity, which dramatically changes its colour to a cooked-like texture. Koi pla is always consumed instantly.

2) Extensively fermented fish

Pla ra is a freshwater fish preserved in a highly concentrated salt solution for 3-6 months. Chunks of fish could be consumed uncooked and remaining fish sauce is often used as daily ingredient for local dishes (29, 87, 89).

2.2.4 Collection of stool specimen After receiving a sealable container, participants collected stool samples with a scoop provided. Then the container was sealed and transferred to a village health volunteer and delivered to the field lab. Each participant provided 1 stool sample, processed by 3 examination methods. Two methods were performed in the field: 1) Wet preparation and 2) Kato-Katz technique. Both processes were conducted within six hours after collection. The specimen was examined under a light microscope to identify intestinal parasite eggs, focusing on the presence of OV eggs by experienced examiners. Then specimens were kept in a refrigerator and transported to the Department of Parasitology, Phramongkutklao College of Medicine for processing for the third examination method with formalin-ethyl acetate concentration technique and also examined under light microscopy. At least one positive result from any method was interpreted as positive for the infection (69, 70). Additionally, OV eggs were counted under Kato-Katz technique; the specimen were examined throughout the slide and reported as total egg per slide.

As described in Section 1.2.1, the morphology of OV eggs is distinct from other intestinal helminths, but could be physically similar to MIF under the microscope. Evidence indicated that no MIF was presented in the study area (46), but no further confirmation was made from the molecular study in the research project. All OV-like eggs were assumed to be Opisthorchis viverrini eggs. 60

2.2.5 Data analysis

Each participant was matched for questionnaire and specimen using a code-embedded identification system. Stool examiners were blinded to the individual who provided the specimen. The data management unit matched all corresponding data (questionnaire and stool examination results) for data analysis. Incomplete questionnaires with missing data on relevant exposures, i.e., sex, age, occupation and consumption behaviours, were excluded to perform risk factor analysis.

Descriptive statistics were used to describe data; outcome was reported as frequency and percentage. Statistical analysis was reported with significance level of 0.05 and 95% confidence intervals.

Outcomes of infection were considered in two ways: binary outcome as infected or not and egg count as number of eggs per slide for infection intensity. Distribution of OV was reported as prevalence (%) in the cross-sectional study and incidence (/100 person-years) for the follow-up study. Egg intensity was reported as eggs per gram (EPG). The number of eggs per slide was multiplied by 24 to obtain EPG. The distribution was explored by histogram to visualize the pattern of intensity.

2.2.6 Model for prevalence and incidence data

Univariable analysis was initially performed to examine the association of OV of each exposure of interest and providing crude association. Exposures included sex, age, occupation and uncooked fish consumption behaviours. The association between univariable risk factors and the infection was quantified using Pearson’s chi-square.

Multivariable analysis was performed using regression modelling for adjusting risk factors. Risk factors were adjusted for age based on previous works (69, 70). Factors with p value <0.2 from univariable analysis were also considered for inclusion in the multivariable model.

Instant fish preparation (Koi pla) was recognised as a risk factor from previous works in this project (69, 70). However, extensively fermented fish (Pla ra) was considered to be included into the model to control for confounder based on the data that they were also consumed on a regular basis.

For the cross-sectional study, outcomes were reported as categorical data (infected/uninfected). A logistic regression model was used for model fitting. Risk factor associations were described using odds ratios. 61

For the follow-up study, each infected case was counted as a singular event. All participants were followed for the specified follow-up time (14-15 months). Poisson regression was developed to study incidence outcomes with risk factors and reported as incidence rate ratio.

2.2.7 Model for infection intensity

Biologically, adult OV worms reside in the human host bile duct (2). Therefore, direct worm counts for OV infection are not available for practical fieldwork (58, 86). Therefore, egg per gram measurements of intensity are performed, where it is assumed, for logistical reasons that the amount of eggs excreted in the stool is associated with the intensity of worm load presented in the host bile duct (95).

Data were analyzed using a count model approach. Studies across many species of helminthes have addressed the problem of over-dispersion data (93-95, 99-101),which results in large amounts of zero count eggs for which classical Poisson distribution may not be appropriate. Therefore, extensive models for count data were used here to explore the characteristics of over-dispersion and evaluate the association of risk factors and outcomes, and also the age-prevalence-intensity relationship.

The regression models for count data and related equations were based on model framework by Zeileis et al. (102) published for count data analysis in R.

The count data model was based on the concept of the generalized linear model (GLM) framework

(102, 103) which explains the dependence of a scalar variable yi (when i = 1, 2,…, n) on a vector of regressors xi with the probability density function as shown in Equation 2.1 below.

푦∙휆−푏(휆) 푓(푦; 휆, 휙) = exp ( ) + 푐(푦, 휙), where (2.1) 휙

 The conditional distribution of yi|xi follows a linear exponential family.  λ is a canonical parameter.  휙 is a dispersion parameter.  The function b() and c() refer to a particular family which will be used; normal, binomial or Poisson distribution. ′ Mean was defined as 퐸[푦𝑖|푥𝑖] = 푏 (휆𝑖) or μi and variance could be defined as 푉퐴푅[푦𝑖|푥𝑖] = 휙 ⋅

푏′′(휆𝑖); therefore, the distribution of yi is determined by μi.

⊺ The dependence of the conditional mean 퐸[푦𝑖|푥𝑖] = 휇𝑖 on regressors xi was given by 푔(휇𝑖) = 푥𝑖 훽 where g() was a link function and 훽 was a vector of regression coefficient. 62

The simple model for count data within the GLM framework was the Poisson distribution with probability density function as described in Equation 2.2 below.

exp(−휇)∙휇푦 푓(푦; 휇) = (2.2) 푦!

Poisson distribution refers to the number of eggs excreted in the stool specimen during the fieldwork period. As the mean egg count was low, most data consisted of low egg counts and few data held highly skewed values. When the mean egg count increased, the distribution approached a normal distribution pattern which meant equal to variance; V(μ) = μ, so the dispersion (φ) was fixed at 1. The canonical link between mean and linear predictor was defined as g(μ) = log(μ) resulting in a log-linear relationship.

In particular, the data included a larger amount of extreme values resulting in variance spreading greater than mean, called over-dispersion and is explained below.

1. Some individuals carried extremely high egg count. Many helminth infections exhibit a particular pattern when only few cases carry a heavy burden of infection while the remaining carry a few or zero egg counts, resulting in a visibly longer tail in distribution.

2. A large number of zero egg counts resulted in a heavily skewed distribution.

One way to deal with this problem was to assume a negative binomial distribution as shown in Equation 2.3 below.

Γ(푦+ 휃) 휇푦⋅휃휃 푓(푦; 휇, 휃) = ⋅ , where (2.3) Γ(휃)⋅푦! (휇+휃)푦+휃

 휃 is a shape parameter.  Γ() is a gamma function.

휇2 The 휙 is 1 but the variance could refer to 푉(휇) = 휇 + 휃

A negative binomial distribution has been proposed for a better choice for over-dispersion data regarding some parasite intensity data (94, 95) . Thus, the preliminary analysis will deal with the property of egg count outcomes from fieldwork data with classical Poisson and negative binomial distributions.

Compound model

Zero egg count raised an interesting issue as the zero result tended to be a major proportion of stool outcomes. Zero result could be derived from true or false origin whether the examination tools could truly capture the zero count with high sensitivity, or the light infection could be missed due to false negatives (95).

The compound model has been addressed to treat the issue of zero egg excretor from various assumptions (95).

Zero-inflated model Hurdle model Structural Structural 120 120 Sampling Sampling 100 100

80 80

60 60 Frequency Frequency 40 40

20 20

0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Number of egg count Number of egg count

Figure 2.3 Distribution of count data in zero- Figure 2.4 Distribution of count data in hurdle inflated model model

A s illustrated in Figures 2.3 and 2.4, the zero component was based on two assumptions: whether the zero components were sampling or structural. Structural zero is derived from true negative cases referring to uninfected individuals. The difference of the zero-inflated and a hurdle model depends on how the zero will be sampled. Basically, the hurdle model assumes that all zero egg outputs are true negative; therefore, a non-zero output should belong to infected cases with sampling distribution (Figure 2.4). However, the zero-inflated model assumes that a zero egg count could come from either true negative or false negative due to diagnostic sensitivity. A true negative result will refer to uninfected individuals. For infected cases, when the test fails to detect fluke eggs in stool, then the outcome will be negative as a result of a false negative. Therefore, the sampling of egg counts will include zero from the assumption.

Zero-inflated model

As shown in Figure 2.3, the zero-inflated model assumes that zero outcomes stem from different origins. The projection is a joint model consisting of two parts: one for the predicting zero and the other for the count model. 64

Some infected cases produced zero egg counts, i.e., zero excretor. This portion of zero excretor will be a sampling zero since it is assumed to happen by chance.

As a result, the zero-inflated model will comprise two components when one part produces only zero, a point mass at zero 퐼{0}(푦), and other part generates both zero and non-zero outcomes, a count component. A binary model is used to model the unobserved probability with 휋 = 푔−1(푧⊺훾). Therefore, the density of the inflated-model will be incorporated with point mass at zero and a count distribution, as shown below in Equation 2.4.

푓(푦; 푥, 푧, 훽, 훾) = 푓푧푒푟표(0; 푧, 훾) ⋅ 퐼{0}(푦) + (1 − 푓푧푒푟표(0; 푧, 훾)) ⋅ 푓푐표푢푛푡(푦; 푥, 훽), (2.4)

Where the probability of observing zero count is inflated with probability 휋 = 푓푧푒푟표(0; 푧, 훾) and a count model assumed to follow Poisson or negative binomial distributions depends on the property of count outcome, as referred to dispersion parameter. Regarding the mean, the regression function will be;

⊺ 휇𝑖 = 휋𝑖 ⋅ 0 + (1 − 휋𝑖) ⋅ exp(푥𝑖 훽),

Where the default link function; 푔(휋), based on the binomial, is logit link.

Hurdle model

The hurdle model is a compound model to deal with excess zero data with a different assumption from the zero-inflated model. The model comprises two components:

The first hurdle will determine the zero part of the result. The model assumes that all zero outcomes are structural. Therefore, zero egg excretor are truly uninfected cases. Thus, this component will assume the probability whether the individual will be infected or not following binary outcome (0/1) under the logit regression model, defined as zero hurdle model 푓푧푒푟표(푦; 푧, 훾), which generally follows binomial distribution.

The second hurdle will deal with count outcome as infection intensity. From the first hurdle, I assume that zero egg excretor is uninfected so the nonzero egg excretor will be truly infected. Therefore, the infected cases must produce a nonzero result since they cannot have zero egg count by assumption. The count outcome from this component will follow the truncated count model either Poisson or negative binomial distribution,; 푓푐표푢푛푡(푦; 푥, 훽) in Equation 2.5.

푓푧푒푟표(0; 푧, 훾) 푖푓 푦 = 0 푓ℎ푢푟푑푙푒(푦; 푥,푧,훽,훾) = { 푓푐표푢푛푡(푦;푥,훽) (2.5) (1 − 푓푧푒푟표(0; 푧, 훾)) ⋅ 푖푓 푦 > 0 1− 푓푐표푢푛푡(0;푥,훽) 65

The likelihood for zero hurdle and count model estimation can be maximized separately. The mean and regressors can be described as shown below in Equation 2.6.

⊺ log(휇) = 푥𝑖 훽 + log (1 − 푓푧푒푟표(0; 푧, 훾))) − log (1 − 푓푐표푢푛푡(0; 푥, 훽)). (2.6)

Based on the assumption regarding the zero component, the zero-inflated model will predict the probability for obtaining sampling zero when the hurdle model will predict the probability for obtaining nonzero outcome.

The results from model fitting are interpreted in light of the characteristics of the infection and method of stool examination.

2.2.8 Model analysis and fitting

Statistical modeling was performed using STATA/IC 14 and R 3.1.2. The zero-inflated and hurdle models were fitted using the functions countfit, hplogit, hnblogit in STATA and PSCL package in R (Political Science Computational Laboratory).

For prevalence and incidence data, covariates comprised five variables: sex, age, and occupation, consumption of Koi pla and consumption of Pla ra. All model building began with a full model and compared with a reduced model. The method of covariate selection was based on the following.

 All covariates were preliminarily explored with univariable analysis to obtain statistical value.  Full models were performed with corresponding analysis with five covariates.  Reduced model excluded non-significant covariates from full models.  The full model was compared with the reduced model by log-likelihood and AIC to obtain the best fitted model.

For intensity data, the compound models incorporated covariates in both components and covariates were selected from the best fitted model.

Over-dispersion was explored and tested. To test over-dispersion, the likelihood ratio test between Poisson and negative binomial model was performed. The over-dispersion parameter (α) was used. When α was zero, the negative binomial distribution was equivalent to standard Poisson distribution.

The Poisson model was assigned as a baseline model. The ∆AIC; difference of AIC to baseline model ≥10 indicated that the smallest AIC model was best fitted. The observed zero counts were compared to expected zero count to determine the performance of handling excess zero outcomes.

2.2.9 Ethical consideration

The research protocol of this study was approved by the Ethics Committee of the Royal Thai Army Medical Department (approval code S045h/54). Study participants or parents of participants under 18 years of age agreed to join the study after reading the data sheet provided in Thai and provided their written informed consent. Appropriate treatment was provided for those who were found to be positive for any intestinal parasitic infection.

2.3 Results

2.3.1 Population characteristics

2.3.1.1 Response from the study

Baseline population  Na-isarn and Na-ngam village  2012 - 2013  1514 Participants

1514 eligible subjects Informed for the consent

A Dispatched of survey package Consent form document Stool specimen container Questionnaire Did not return survey package B  Did not consented Collection of survey package  Unable to contact  Stool container (1303)  Questionnaire (1000)

C Processing of survey package  Stool examination  Questionnaire data entry Excluded from the analysis  Stool examination data entry Unmatched or incomplete  Matching for individual questionnaires (509) record

D Completed data records for analysis 794 records

Figure 2.5 Summarized flowchart for cross-sectional study

As illustrated in Figure 2.5, Consenting participants returned the survey package including stool specimens and questionnaires for 86.1% and 66.1%, respectively. Test on the equality of proportions between two study areas were reported with significant p-value.

Table 2.2 Response for cross-sectional study

Na-isarn 2013 Na-gnam 2012 p Total N % N % N % A. Survey package dispatched 583 931 1514 B. Consented and returned survey package Returned stool specimen 431 73.9 872 93.7 <0.001 1303 86.1 Returned questionnaire 354 60.7 646 69.4 <0.001 1000 66.1 C. Matching and excluded for 100 17.1 409 44.0 <0.001 509 33.7 incomplete questionnaire D. Completed questionnaire 331 56.8 463 49.7 0.01 794 52.4 and specimen

As illustrated in Table 2.2, the response for returned stool specimens was significantly higher in Na- ngam Area (93.7% and 73.9 %, respectively, p <0.001). After pairing stool examination results with matched questionnaires and excluding incomplete questionnaires, eligible records from Na-ngam were more numerous than Na-isarn Area (44.0 % and 17.1 %, respectively, p <0.001).

For the whole study area, completed questionnaires matched with specimens eligible for data analysis totalled 52.4 %. Only 13.9% of participants did not return the packages during the survey period and were unable to contact. Some of them could not provide the stool specimen at the time of collection. Returned questionnaire rate was also lower than stool specimens from both areas.

Baseline cross-sectional study  Na-yao and Na-ngam village Positive for OV infection  2011 - 2012 107 subjects  1910 Participants

Negative for OV infection 1803 subjects

A 1803 eligible subjects for a follow-up study

Invited to participate in B Excluded from the study retrospective cohort study in  Did not consented (8) 2013-2014  Deceased (15) 14-15 months follow-up  Permanent moved-out period (82) Informed consent

C Enrolled to the study 1698 subjects received survey package  Stool container  Questionnaire D E Loss to follow-up Collection of survey package  Temporary moved-out (361) 884 packages  Unable to contact/return specimen (453)

Processing of survey package  Stool examination F  Questionnaire data entry Excluded from the analysis  Stool examination data entry Unmatched or incomplete  Matching for individual questionnaires (138) record

G Completed data records for analysis 746 records

Figure 2.6 Summarized flowchart for follow-up study

Eligible participants comprised negative cases from the baseline survey conducted from 2013-14 with a total of 1803 subjects: 974 cases from Na-yao Area and 829 cases from Na-ngam Area.

As illustrated in Figure 2.6, the enrolment process was conducted at the time of fieldwork. Cases were enrolled retrospectively; therefore, deceased subjects or those with confirmed permanently moved-out status were excluded from the study. Test results on the equality of proportions between the two study areas were reported with significant p-value. 70

Table 2.3 Response for follow-up study

Na-yao 2013 Na-gnam 2014 Total N % N % p N % A. Eligible participants 974 829 1803 Consented 916 94.0 782 94.3 0.80 1698 94.2 B. Excluded participants Did not consent 5 0.5 3 0.4 0.63 8 0.4 Died during follow-up period 3 0.3 12 1.4 0.01 15 0.8 Permanently moved-out 50 5.1 32 3.9 0.20 82 4.5 C. Enrolled participants 916 782 1698 D. Loss to follow-up participants Temporarily moved-out during specimen collection 187 20.4 174 22.3 0.36 361 21.2 Unable to contact/return specimen 352 38.4 101 12.9 <0.001 453 26.7 E. Survey package collected 377 41.2 507 64.3 <0.001 884 52.1 F. Excluded for unmatched or incomplete 46 5.1 92 11.2 0.02 138 9.9 questionnaire G. Completed questionnaire and 331 36.1 415 53.1 <0.001 746 42.2 specimen

As shown Table 2.3, 94.2% of eligible cases consented to join the follow-up study. No significance was observed concerning consented participants for both areas (p = 0.80). The percentage of those who did not consent or confirmed as permanently moved-out was insignificant. Subjects who died during the follow-up period were more numerous in Na-gnam Area (p = 0.01). After the enrolment process, 916 participants for Na-yao and 782 participants for Na-ngam joined the follow-up study. The follow-up time for Na-yao area was 15 months and Na-ngam area, 14 months.

The survey package was dispatched to participants. Completed packages were returned with stool specimens and completed questionnaires to perform data analysis. The overall response for returned survey packages was 52.1 % and loss to follow-up participants comprised 47.9 %. The difference in response rate was significant (41.2 % and 64.3 %, respectively, p <0.001). The difference was mainly derived from the subjects whom I were unable to contact or did not return specimens. Overall excluded incomplete questionnaires totalled 9.9%, resulting in a lower response rate of 42.2%. The final response rates in Na-yao and Na-gnam were 36.1% and 53.1%, respectively.

2.3.1.2 Demographic data

Na-isarn 2013 cross-sectional study Na-ngam 2012 cross-sectional study

Na-yao 2013 follow-up study Na-ngam 2014 follow-up study

Figure 2.7 Population pyramids

As shown in Figure 2.7, the population pyramids showed an irregular pattern of age structure where the near-bottom class was relatively limited in size in all four study areas.

The pattern indicated that the majority of subjects were aged between 30 - 59 years. Considering the working age, which usually starts at early adolescence for Thai rural population, the age period 20 - 39 was relatively small in proportion compared with age ≥40 years.

For the Na-isarn 2013 cross-sectional study, no females were included in the 0 - 9 years age class. More females were in the middle to old age in all areas.

However, the age structure of the follow-up population needed to be compared with the baseline population because the individuals identified as defined at baseline were excluded.

Table 2.4 Population characteristics of cross-sectional study

Na-isarn Na-ngam Total Characteristic p N (%) N (%) N (%) Sex

Female 179 (54.1) 255 (55.1) 434 (55.1) 0.78 Male 152 (45.9) 208 (44.9) 360 (45.3) Age profile

Mean ± S.D. 44.6 ± 15.9 44.7 ± 18.6 44.7 ± 15.9 0.46 Median ± IQR 44.0 ± 20.0 46.0 ± 23.0 45.0 ± 21.0

Min – max 7 – 84 2 - 86 2 - 86 Age group (in years, as quartiles) Q1 0 – 36 88 (26.6) 0 – 35 117 (25.3) 0 – 35 200 (26.2) Q2 37 – 44 78 (23.6) 36 – 46 119 (25.7) 36 – 45 186 (24.4) Q3 45 – 56 88 (26.6) 47 – 58 116 (25.1) 46 – 57 202 (26.5) Q4 ≥ 57 77 (23.3) ≥ 59 111 (24.0) ≥ 58 175 (22.9) Occupation

Unemployed 30 (9.1) 45 (9.7) 75 (9.4)

Non-agriculture 81 (24.5) 147 (31.7) 228 (28.7) 0.06 Agriculture 220 (66.5) 271 (58.5) 491 (61.8)

Table 2.5 Population characteristics of follow-up study

Na-yao Na-ngam Total Characteristic p N (%) N (%) N (%) Sex Female 176 (53.2) 228 (54.9) 404 (54.2) 0.63 Male 155 (46.8) 187 (45.1) 342 (45.8) Age profile Mean ± S.D. 47.2 ± 19.1 43.3 ± 19.9 45.0 ± 19.6 0.88 Median ± IQR 50.0 ± 22 47.0 ± 23 48.0 ± 23 Min – max 1 – 85 2 – 87 1 - 87

Age group (in years, as quartiles) Q1 0 – 38 83 (25.1) 0 – 34 104 (25.1) 0 – 36 189 (25.3) Q2 39 – 50 86 (26.0) 35 – 47 117 (28.2) 37 – 48 190 (25.5) Q3 51 – 60 87 (23.6) 48 – 57 91 (21.9) 49 – 59 191 (25.6) Q4 ≥ 60 75 (22.7) ≥ 57 103 (24.8) ≥ 60 176 (23.6)

Occupation Unemployed 17 (5.1) 36 (8.7) 53 (7.1) Non-agriculture 100 (30.2) 146 (35.2) 246 (33.0) 0.33 Agriculture 214 (64.7) 233 (56.1) 447 (59.9)

The majority of population worked in the agricultural-related sector. As shown in Table 2.4, the overall mean age in the cross-sectional study was 44.7 ± 15.9 years with a range of 2 - 86 years and median age of 45.0 ± 21.0 years. As shown in Table 2.5, the follow-up study showed the overall mean age was 45.0 ± 19.6 years with a range of 1 - 87 years and median age was 48.0 ± 23.0 years. Due to the irregularity of age structure, age was categorized in four age groups by quartile intervals.

2.3.2 Uncooked fish consumption behaviors

From the uncooked fish consumption behaviors from the 2 prevalence and 2 follow-up studies; all four areas showed similar patterns of uncooked fish consumption behaviors. The range of consumption was 39.8 - 44.1 %. (Supplement Table S-1 and S-2)

However, Pla ra consumption was more popular: 82.1% from the cross-sectional study and 77.5% from the follow-up study. Moreover, consumption of Koi pla did not significantly differ for both areas within the same study method. The proportion of Pla ra consumption was significantly higher in Na-isarn in the cross-sectional study and in Na-ngam in the follow-up study.

Table 2.6 Multivariable analysis of population characteristics and uncooked fish consumption behaviours in cross-sectional study

Na-isarn 2013 Na-ngam 2012 Extensively Chopped raw fish Extensively Chopped raw fish salad fermented fish salad fermented fish (Koi pla) Characteristic (Pla ra) (Koi pla) (Pla ra) Adjusted Adjusted Adjusted Adjusted OR OR P OR P OR P P (95% CI) (95% CI) (95% CI) (95% CI) Sex Female 1 1 1 1 Male 3.46 1.91 4.20 1.30 (2.13-5.63) <0.01 (0.95-3.85) 0.07 (2.81-6.27) <0.01 (0.81-2.08) 0.28 Age group (as quartiles) Q1 1 1 1 1 Q2 2.18 0.91 2.18 2.47 (1.05-4.54) 0.04 (0.36-2.28) 0.84 (1.21-3.93) 0.01 (1.26-4.86) 0.01 Q3 3.18 3.40 1.86 2.45 (1.53-6.61) <0.01 (0.89-12.98) 0.07 (1.02-3.81) 0.04 (1.22-4.91) 0.01 Q4 2.64 0.47 2.54 1.12 (1.23-5.68) 0.01 (0.19-1.14) 0.10 (1.36-4.73) <0.01 (0.59-2.12) 0.74 Occupation Unemployed 1 1 1 1 Non - 1.33 1.27 1.02 1.00 agriculture (0.45-3.93) 0.60 (0.43-3.76) 0.67 (0.47-2.23) 0.96 (0.45-2.20) 0.99 Agriculture 2.13 2.72 1.26 2.13 (0.83-5.49) 0.12 (1.03-7.17) 0.04 (0.61-2.59) 0.53 (1.01-4.50) 0.04

Table 2.6 shows the multivariable analysis of population characteristics and uncooked fish consumption behaviours. Sex, age group and occupation were considered potential demographic factors based on previous studies contributing risk behaviours (69, 70). Age-group is referred to in quartiles as shown in Table 2.4. Both study areas show similar Koi pla consumption patterns, male sex and age-group were statistically significant factors. For age-group, the adjusted OR was highest in Q3 for Na-isarn and Q4 for Na-ngam area indicating that the OR could be higher in older age. No statistical significance for occupation and Koi pla consumption was found from both study areas.

For Pla ra, age was only significant for Q2 and Q3 in Na-ngam area. Agricultural-related work significantly increased risk of eating Pla ra in both areas compared with unemployment. 75

Table 2.7 Multivariable analysis of population characteristics and uncooked fish consumption behaviours in follow-up study

Na-yao 2013 Na-ngam 2014 Extensively Chopped raw fish Extensively Chopped raw fish salad fermented fish salad fermented fish Characteristic (Koi pla) (Pla ra) (Koi pla) (Pla ra) Adjusted RR Adjusted RR Adjusted RR Adjusted RR P P P P (95% CI) (95% CI) (95% CI) (95% CI) Sex Female 1 1 1 1 Male 1.20 0.27 0.93 0.57 1.56 <0.01 0.97 0.81 (0.86-1.67) (0.71-1.21) (1.14-2.13) (0.79-1.20) Age group (as quartiles) Q1 1 1 1 1 Q2 2.25 1.24 1.90 1.15 (1.22-4.17) 0.01 (0.83-1.86) 0.29 (1.11-3.26) 0.02 (0.84-1.59) 0.39 Q3 2.53 1.17 1.84 1.08 (1.38-4.64) <0.01 (0.77-1.76) 0.46 (1.04-3.26) 0.04 (0.76-1.53) 0.68 Q4 2.94 1.11 1.64 1.17 (1.58-5.45) <0.01 (0.72-1.71) 0.64 (0.92-2.93) 0.09 (0.84-1.65) 0.36 Occupation Unemployed 1 1 1 1 Non- 1.32 1.07 1.37 1.00 agriculture (0.53-3.30) 0.55 (0.52-2.18) 0.86 (0.59-3.21) 0.47 (0.95-1.54) 0.99 Agriculture 1.52 1.21 2.25 1.10 (0.65-3.55) 0.34 (0.62-2.36) 0.57 (1.02-4.94) 0.04 (0.73-0.64) 0.66

As illustrated in Table 2.7, age-groups showed that increasing age was associated with Koi pla consumption in both study areas when compared with referenced age-group except the Q4 age- group in Na-ngam study. Male gender and agricultural-related work were also statistically significant in Na-ngam area for Koi pla consumption. Age-group was referred to in quartiles as shown in Table 2.5. No significant association was observed for population characteristics and Pla ra consumption. 76

2.3.3 Distribution and risk factors of OV infection

2.3.3.1 Distribution of prevalence

Na-isarn 2013 cross-sectional study Na-ngam 2014 cross-sectional study

25 25

20 20

15 15

10 10 Prevalence (% wirh 95% CI)wirh Prevalence (% 5 95% CI)wirh Prevalence (% 5

0 0 1 2 3 4 1 2 3 4 0-36 37-44 45-56 ≥57 0-34 37-48 49-59 ≥60 AgeAge group-group AgeAge group-group

Figure 2.8 Prevalence of OV infection distributed by age-group

Prevalence of OV in Na-isarn was 7.85 % (95% CI: 4.96 - 10.75) and Na-ngam is 9.29 % (95% CI: 6.64 - 11.93), respectively. The overall prevalence was 8.69 % (95% CI: 6.82 - 10.87) and both prevalence did not vary significantly between the two study areas. The prevalence increased by age where the prevalence was higher in older age groups as shown in Figure 2.8. (Supplement Table S-3)

From the study, 1303 survey packages were returned for which stool samples were processed. However, only 794 specimens were matched to completed questionnaires to perform data analysis. Table 2.11 compared the prevalence of OV between two groups indicating that no significant difference for the completed and incomplete questionnaire groups for both areas. However, the prevalence was statistically higher among those completing questionnaires in the Na-isarn area. (Supplement Table S-4)

2.3.3.2 Distribution of incidence

Na-yao 2013 follow-up study Na-ngam2014 follow-up study

20 20

15 15 year wirh year 95% CI)wirh year wirh year 95% CI)wirh - - 10 10

5 5 Incidence (/100 (/100 person Incidence Incidence (/100 (/100 person Incidence 0 0 1 2 3 4 1 2 3 4 0-38 39-50 51-60 ≥61 0-36 37-48 49-59 ≥60 Age -groupgroup Age- group

Figure 2.9 Incidence of OV infection distributed by age-group

The incidence of OV infection was 7.98/100 person-years (95% CI: 5.49 – 11.20) in the Na-yao study and 6.80/100 person-years (95% CI: 4.68 – 9.54) in the Na-ngam study. The incidences from both areas did not significantly differ. (Supplement Table S-5)

Figure 2.9 shows the distribution of incidence by age-group. For Na-yao area, incidence increased in the younger age and remained relatively stable in the older age groups. Na-ngam study showed that the incidence was highest in the 35 - 47 age group.

The incidence rate of OV infection between the two groups indicating no significance for the completed and incomplete questionnaires groups. (Supplement Table S-6)

Table 2.8 Proportional distribution of OV infection diagnosis categorised by stool examination methods

Cross-sectional study Follow-up study Total Test N % N % N % N of cases 69 66 135 Simple smear 0 0.0 0 0.0 0 0.0 1 positive test Kato-Katz 0 0.0 0 0.0 0 0.0 FECT 47 68.1 43 65.2 90 66.7 Simple smear 0 0.0 0 0.0 0 0.0 + Kato-Katz Simple smear 2 positive tests 1 1.4 2 3.0 3 2.2 + FECT Kato-Katz 17 24.6 16 24.2 33 24.4 + FECT Simple smear 3 positive tests + Kato-Katz 4 5.8 5 7.6 9 6.7 + FECT

Table 2.8 reports positive stool examination results diagnosed by each method. From both study designs conducted in 4 areas; Na-yao, Na-isarn and Na-ngam from 2011 - 2014, 135 cases were diagnosed with OV; 69 cases from the cross-sectional study and 66 cases from the follow-up study. In all, 66.7% of diagnoses were from FECT only while the other 33.3% of cases were positive with a combination of two of three tests.

FECT was responsible for all positive cases either single or combination results. Regarding the combination of results, simple smear test results were positive in the diagnosis of 12 cases (8.9%) and Kato-Katz for 42 cases (31.1%). A total of 9 (6.7%) cases all tested positive. No case was diagnosed with simple smear or Kato-Katz alone as well as simple smear and Kato-Katz method.

2.3.4 Evaluation of risk factors for acquiring OV infection

2.3.4.1 Univariable analysis of risk factors for acquiring OV infection

Univariable analysis assesses the association between each factor to OV infection. The analysis indicates that sex, occupation and Pla ra consumption were not significantly associated with OV in all 4 areas. (Supplement Table S-7 and S-8)

From the cross-sectional study, the oldest age group, (Q4) was statistically significant for having the infection compared with the Q1 age group in both areas. However, increasing age was associated with increasing risk in Na-isarn area. Koi pla was significantly associated with the infection in both areas. (Supplement Table S-7)

The follow-up study showed a different pattern of risk factors from the cross-sectional study. The 35 - 47 age group was only observed for significant factors for OV in the Na-ngam 2014 study. Koi pla consumption was statistically significant in the Na-yao 2013 study. (Supplement Table S-8)

2.3.4.2 Model fitting

Table 2.9 Model fitting for prevalence and follow-up study

Cross-sectional study

Na-isarn 2013 Na-ngam 2012

Covariates 5 4 2 5 4 2 Log L -79.89 -80.03 -80.31 -133.87 -135.58 -136.38 AIC 177.78 174.06 170.61 285.75 285.16 282.76 Follow-up study

Na-yao 2013 Na-ngam 2014

Covariates 5 4 2 5 4 2 Log L -105.07 -105.13 -105.46 -112.00 -112.07 -112.38 AIC 228.14 224.25 220.92 242.01 238.15 234.77

Table 2.9 shows the model fitting with regard to covariate selection, age group and consumption of Koi pla was significant from univariable models. Therefore, they were selected in the final model. Sex and occupation revealed a potential risk obtained from the qualitative approach detailed in the Chapter 5.

Moreover, Pla ra is a very popular uncooked fish dish for which the consumption is approximately 80%. The qualitative approach also revealed that Pla ra was a main ingredient for local dishes. Additional analysis revealed that Pla ra held a significantly positive correlation with Koi pla in all study areas (correlation coefficient = 0.5 - 0.6, p <0.01).

The full multivariable model comprised five covariates as a baseline model; sex, age, occupation, Koi pla and Pla ra consumption.

The four-covariate model excluded the occupation covariate from the model. Sex remained in the model since previous studies showed the potential risk (70) besides the qualitative result and Pla ra was still included in the model as the role of main uncooked fish selection.

The final model selected only those two statistically significant covariates obtained from current fieldwork study: age and Koi pla consumption.

The results indicated no significant difference for log-likelihood ratio between models while the AIC was smallest in the two-covariate model suggesting that besides age and Koi pla consumption, the other three covariates may not contribute much influence in the model. 81

2.3.4.3 Multivariable analysis of risk factors for acquiring OV infection

Table 2.10 Multivariable analysis of risk factors for acquiring OV infection of prevalence data from Na-isarn 2013 study

5 covariates 4 covariates 2 covariates Characteristic Adjusted OR Adjusted OR Adjusted OR p p p (95% CI) (95% CI) (95% CI) Sex (female as reference) Male 0.74 (0.31 – 0.18) 0.51 0.75 (0.31 – 1.80) 0.52 Age group (as quartiles, age 0 – 36 as reference) 37 – 44 1.37 1.45 1.44 (0.21 – 8.69) 0.74 (0.23 – 9.04) 0.69 (0.23 – 8.97) 0.70 45 – 56 3.86 4.21 4.31 (7.60 – 19.63) 0.10 (0.87 – 20.35) 0.07 (0.89 – 20.80) 0.07 ≥ 57 5.33 5.58 5.43 (1.03 – 27.50) 0.045 (1.16 – 26.95) 0.032 (1.13 – 26.13) 0.034 Occupation (unemployed as reference) Non-agriculture 1.09 (0.15 – 7.89) 0.94 Agriculture 1.39 (0.28 – 6.94) 0.67 Fish menus Koi pla Yes 2.54 (1.01 – 6.39) 0.048 2.57 (1.02 – 6.47) 0.045 2.43 (1.01 – 5.80) 0.046 Pla ra Yes 1.31 (0.27 – 6.36) 0.74 1.36 (0.28 – 6.53) 0.70

Table 2.11 Multivariable analysis of risk factors for acquiring OV infection of prevalence data from Na-ngam 2012 study

5 covariates 4 covariates 2 covariates Characteristic Adjusted OR Adjusted OR Adjusted OR p p p (95% CI) (95% CI) (95% CI) Sex (female as reference) Male 1.31 (0.65 – 2.61) 0.45 1.38 (0.70 – 2.73) 0.36 Age group (as quartiles, age 0 - 35 as reference) 36 - 46 1.24 1.53 1.45 (0.42 – 3.69) 0.70 (0.53 – 4.45) 0.44 (0.50 – 4.18) 0.49 47 - 58 1.39 1.76 1.68 (0.47 – 4.08) 0.55 (0.62 – 5.04) 0.29 (0.59 – 4.76) 0.33 ≥ 59 2.10 2.66 2.61 (0.74 – 6.00) 0.17 (0.98 – 7.19) 0.054 (0.97 – 7.04) 0.057 Occupation (unemployed as reference) Non-agriculture 0.73 (0.18 – 2.85) 0.65 Agriculture 1.57 (1.16 – 5.15) 0.45 Fish menus Koi pla Yes 2.44 (1.16 – 5.15) 0.019 2.43 (1.16 – 5.12) 0.019 2.48 (1.28 – 4.82) 0.007 Pla ra Yes 0.63 (0.27 – 1.49) 0.29 0.69 (0.29 – 1.60) 0.39

Tables 2.10 and 2.11 show the multivariable analysis for prevalence data. From the Na-isarn study in Table 2.10, age ≥57 years old and consumption of Koi pla remained significant risk factors in the three models indicating their consistency when compared with the reference age-group.

For the Na-ngam study presented in Table 2.11, the consistency of Koi pla association was still observed in all three models. When adjusted for other covariates, age was not statistically significant.

Table 2.12 Multivariable analysis of risk factors for acquiring OV infection of follow-up data from Na- yao 2012 study

5 covariates 4 covariates 2 covariates Characteristic Adjusted IRR Adjusted IRR Adjusted IRR p p p (95% CI) (95% CI) (95% CI) Sex (female as reference) Male 1.23 (0.61 -2.48) 0.56 1.21 (0.61 – 2.42) 0.58 Age group (as quartiles, age 0 – 38 as reference) 39 – 50 11.36 1.38 1.35 (0.41 – 4.47) 0.61 (0.45 – 4.27) 0.58 (0.40 – 4.12) 0.60 51 – 60 1.37 1.40 1.40 (0.42 – 4.47) 0.60 (0.46 – 4.27) 0.55 (0.46 – 4.25) 0.55 ≥ 61 1.33 1.40 1.40 (0.38 – 4.65) 0.66 (0.45 – 4.39) 0.56 (0.45 – 4.37) 0.56 Occupation (unemployed as reference)

Non-agriculture 0.75 (0.14 – 3.98) 0.73 Agriculture 0.77 (0.17 – 3.51) 0.74 Fish menus Koi pla Yes 2.37 (1.10 – 5.12) 0.028 2.35 (1.09 – 5.06) 0.029 2.34 (1.10 – 4.99) 0.027 Pla ra Yes 0.81 (0.39 – 1.70) 0.58 0.81 (0.39 – 1.69) 0.57

Table 2.13 Multivariable analysis of risk factors for acquiring OV infection of follow-up data from Na- ngam 2014 study

5 covariates 4 covariates 2 covariates Characteristic Adjusted IRR Adjusted IRR Adjusted IRR p p p (95% CI) (95% CI) (95% CI) Sex (female as reference) Male 1.23 (0.61 – 2.47) 0.56 1.22 (0.61 – 2.45) 0.58 Age group (as quartiles age 0 – 34 as reference) 35 – 47 3.08 1.38 3.14 (0.95 - 10.03) 0.06 (0.45 – 4.27) 0.58 (1.02 – 9.65) 0.046 48 – 57 1.26 1.40 1.29 (0.31 – 5.16) 0.75 (0.46 – 4.27) 0.55 (0.34 – 4.91) 0.71 ≥ 58 1.70 1.40 1.83 (0.46 – 6.32) 0.43 (0.45 – 4.39) 0.56 (0.54 – 6.16) 0.33 Occupation (unemployed as reference)

Non-agriculture 0.76 (0.18 – 3.11) 0.70 Agriculture 0.80 (0.22 – 2.93) 0.70 Fish menus Koi pla Yes 1.31 (0.63 – 2.72) 0.47 1.30 (0.63 – 2.66) 0.48 1.39 (0.69 – 2.80) 0.35 Pla ra Yes 1.40 (0.42 – 4.67) 0.59 1.40 (0.42 – 4.67) 0.59

Table 2.12 shows the Na-yao follow-up data, and Koi-pla consumption was significantly associated with OV. Age-group was not associated with the infection in any models.

The Na-ngam follow-up data from Table 2.13 revealed different patterns from other areas. Consumption of Koi pla and Pla ra were not statistically significant. The five-covariate model showed no association with OV. When reduced to 4 and 2 covariates, the age group of 35 -47 was significantly associated with the infection when compared the reference age-group. 85

2.3.5 Study of infection intensity

From Table 2.1, four study areas with EPG results were eligible for infection intensity analysis.

Table 2.14 Summary of positive egg count diagnosed by Kato-Katz and FECT methods

Na-yao Na-isarn and Na-ngam Year 2002 2004 2013-14 Study design Prevalence Follow-up Prevalence Positive by Kato-Katz 17.14 20.77 4.19 (%, 95% CI)) (14.02-20.64) (16.25-25.29) (2.77-5.62) Positive by FECT 22.86 26.52 7.99 (%, 95% CI) (19.33-27.31) (21.60-31.43) (6.07-9.92) Sensitivity of Kato-Katz 74.98 78.32 52.44 compared to FECT as gold (72.53-75.58) (75.2-80.48) (45.63-56.65) standard (%, 95% CI) % of excess positive from FECT 5.72 5.75 3.80 (% of excess proportion) (25.02) (21.68) (47.56)

Table 2.14 reports the percentage of positive cases diagnosed by Kato-Katz or FECT. Both methods are microscopic-based where fluke eggs need to be identified. However, FECT is a more sensitive technique than Kato-Katz (p <0.05).

Considering FECT as a gold standard and responsible for all positive results (Table 2.8), sensitivity for Kato-Katz in Na-yao from 2002 - 04 study was 74.98 - 78.32%. Sensitivity for Na-isarn and Na-gnam was 52.44% or approximately half the proportion of FECT result.

To perform the infection intensity analysis, EPGs were calculated through the egg output per slide from Kato-Katz technique. Table 2.14 shows the prevalence and intensity from Na-ngam and Na- isarn data were also less than Na-yao 2002-24 data. As a result, I combined Na-isarn and Na-ngam prevalence data to gain sufficient positive results for intensity analysis.

2.3.5.1 Prevalence and intensity profiles

Table 2.15 Summary of Na-yao 2002 intensity data

Age group Prevalence (%) EPG, S.D. N Variance Range (years) (95% CI) (95% CI) 6.06 7.52, 61.87 0 – 11 99 3828.01 0 - 600 (2.26 – 12.73) (6.98 – 8.08) 20.38 12.09, 42.59 12 – 32 157 1813.84 0 - 312 (14.38 – 27.54 (11.54 – 12.63) 25.00 53.45, 263.00 33 – 46 132 69169.17 0 - 2376 (17.88 – 33.28) (52.21 – 54.72) 35.77 153.46, 460.01 ≥ 47 137 211606.80 0 - 3048 (27.77 – 44.40) (151.39 – 155.55) 22.86 58.51, 277.43 Total 525 76969.91 0 - 3048 (19.33 – 27.31) (57.86 – 59.17)

Na-yao 2002 cross-secitonal study Na-yao cross-sectional study

450 50 45 40 400 35 30 25 350 20 Prevalence Prevalence (%) 15 10 300 5 0 0 5 10 15 20 25 30 35 40 45 50 55 60 Age group 250

Frequency Na-yao 2002 cross-sectional study 200

180 160 150 140 120 100 100

EPG 80 60 50 40 20 0 0 0 5 10 15 20 25 30 35 40 45 50 55 60 Age group Number of egg count

Figure 2.10 Prevalence and intensity of Na-yao Figure 2.11 Frequency distribution of EPG from 2002 cross-sectional study Na-yao 2002 cross-sectional study

Table 2.15 shows that EPG ranged between 0 and 3048, but the majority of individuals harboured few eggs or zero eggs as shown in Figure 2.10. Only few cases carried large amounts of eggs; therefore, the distribution resulted in a high variance exceeding the mean egg count, i.e., EPG of 58.51 with S.D. of 277.43 and variance of 76969.91.

Figure 2.11 shows the prevalence of the infection rapidly increased in the very first years from 0 - 15 years of age and then slowly increased thereafter. The mean intensity showed a different pattern in that the egg count gradually increased by age at younger ages and raise quickly in older years. The egg count ≥100 was observed in the age group ≥ 47years.

Table 2.16 Summary of Na-yao 2004 intensity data

Incidence (%) EPG, S.D. Age group N Variance Range (95% CI) (95% CI) 8.23 5.54, 21.78 0-14 91 474.58 0 - 120 (3.64 – 15.58) (5.07 – 6.04) 25.11 52.8, 212.84 15-40 75 45301.62 0 - 1800 (16.58 – 35.46) (51.17 – 54.47) 23.18 41.33, 149.51 41-53 72 22353.01 0 - 1080 (14.68 – 33.75) (42.81 – 45.90) 36.42 188.48, 532.44 ≥54 75 283496.20 0 - 3048 (26.93 – 46.90) (185.39 – 191.61) 23.08 69.62, 296.68 Total 313 88017.10 0 - 3048 (18.87 – 27.77) (68.70 – 70.55)

Na-yao 2004 follow-up study Na-yao 2004 follow-up study

250 50 45 40 35 30 25 200 20 Incidence (%) Incidence 15 10 5 0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 150 Age group

Na-yao 2004 follow-up study Frequency

100 250

200

150

50 EPG 100

0 0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 Age group Number of egg count

Figure 2.12 Prevalence and intensity of Na-yao Figure 2.13 Frequency distribution of EPG from 2004 follow-up study Na-yao 2004 follow-up study

Figure 2.12 reveals the pattern of egg count distribution was similar to the cross-sectional study. Table 2.16 indicates the range of egg count was 0 - 3048 and mean intensity was 69.62 with S.D. of 296.68 and variance of 88017.10. The oldest age group still exhibited the highest egg intensity.

Figure 2.13 demonstrates the incidence increased from age 0 to age 35 and then decreased before continuing to reach the highest incidence at age group ≥54 years. The mean intensity showed the same pattern where intensity rose abruptly from the 41 - 53 age group to the ≥54 age group.

Table 2.17 Summary of Na-isarn and Na-ngam 2012 - 13 intensity data

Prevalence EPG, S.D. Age group N Variance Range (95% CI) (95% CI) 3.00 1.08, 6.93 0 – 35 200 48.03 0 - 72 (0.62-5.37) (0.94 – 1.23) 9.14 5.16, 28.75 36 – 45 186 826.32 0 - 312 (4.98-13.30) (4.84 – 5.50) 9.41 4.87, 30.22 46 – 57 202 913.23 0 - 408 (5.36-13.45) (4.57 – 5.19) 10.86 6.17, 29.86 ≥ 58 175 891.90 0 - 336 (6.23-15.49) 5.81 – 6.55) 7.99 4.25, 25.72 Total 763 661.50 0 - 408 (6.07-9.92) (4.10 – 4.40)

Na-ngam and Na-isarn 2012-13 cross-sectional study Na-ngam and Na-isarn 2012-13 cross-sectional study

750 18 16 700 14

650 12 10

600 8

Prevalence Prevalence (%) 6 550 4

500 2

0 450 0 5 10 15 20 25 30 35 40 45 50 55 60 65 Age group 400

Frequency 350 Na-ngam and Na-isarn 2012-13 cross-sectional study

300 7

6 250 5 200 4 EPG 150 3

100 2

1 50 0 0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 Age group Number of egg count

Figure 2.14 Prevalence and intensity of Na- Figure 2.15 Frequency distribution of EPG from ngam and Na-isarn cross-sectional study Na-ngam and Na-isarn cross-sectional study

The infection intensity in Na-isarn and Na-ngam from 2012 – 13 study was significantly lighter (p = 0.03) than the Na-yao 2002 study. Table 2.17 indicates EPG was 4.25 with S.D. of 25.72 and variance of 661.50. However, the intensity distribution was similar to Na-yao data where the majority of cases had zero egg count.

Prevalence and intensity showed similar patterns where they increased from zero to 40 years and slightly increased afterward as shown in Figure 2.15. 92

2.3.5.2 The prevalence-age structure

Assuming that incidence (λ) is constant over time and for all age groups; p(a) proportion of infected individuals at age a s(a) proportion of uninfected individuals at age a

Therefore;

푠(푎) = 1 − 푝(푎) (2.7)

From Equation 2.7, s(a), can be described using following expression;

푠(푎) = 푒−휆푎 (2.8)

Taking the natural logarithm on both sides of Equation 2.8;

− 푙푛(푠) = 휆푎 (2.9)

Refers to general linear function: Y = mX + c, the incidence (λ) from Equation 2.9 can be estimated as the slope of –ln (s) against age;

Na-yao 2002: fitting -ln(s) with age Na-isarn and Na-ngam: fitting -ln(s) with age

0.6 0.14

0.5 0.12 0.1 0.4 0.08 0.3 ln(s) ln(s) - - 0.06 0.2 0.04

0.1 0.02

0 0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 Age Age

Figure 2.16 Fitting –ln(s) with age

Table 2.18 Summary of prevalence-age structure

흀 Area R2 95% CI p - value (/person-years) Na-yao 2002 0.902 0.0070 0.0028 – 0.0113 0.013 Na-isarn and Na-ngam 2012 - 13 0.985 0.0019 0.0014 – 0.0023 0.001 93

Fitting a regression model for –ln(s) as a function of age indicates that a linear relationship exists in both cross-sectional study areas as shown in Figure 2.16. The regression coefficient could be used as the incidence rate. Table 2.18 reveals the estimated incidence for Na-yao was 0.0070/person-years (95% CI: 0.0028 - 0.0113) and Na-isarn and Na-ngam was 0.0019/person-years (95% CI: 0.0014 - 0.0023). By assuming that incidence is constant for all ages, the prevalence provides a linear relationship for age-structure from both study areas.

2.3.5.3 The age-prevalence-intensity structure

Figures 2.10, 2.12 and 2.14, which refer to distribution of EPG, indicates that a large proportion of egg count was zero and only few individuals carried large amounts of eggs. The standard Poisson model assumes equal mean and variance for the distribution. From three datasets from Tables 2.15 - 2.17, the variance of egg counts was much greater than the mean, indicating the over-dispersion distribution.

From Section 2.2.8, the likelihood ratio test of α = 0 showed a χ2 = 2385.77 with p-value <0.001 in Na-yao 2002 prevalence data. For Na-isarn and Na-ngam prevalence data, the likelihood ratio test gave χ2 = 325.46 with p-value <0.001. The α was significantly different from zero result in over- dispersion; therefore, Poisson distribution was inappropriate. The results confirmed the evidence of over-dispersion in the study for both prevalence and incidence data.

Over-dispersion has been observed in many helminth infections. The studies on hookworm, ascariasis and schistosomiasis (93-95, 101) indicate the practical use of negative binomial distribution to explore density-dependent patterns of infection dynamics.

The relationship between prevalence and infection intensity can be described by the following Equation 2.10 below.

푀(푎) 푝(푎) = 1 − (1 + )−푘 (2.10) 푘

Where the parameters are defined as p(a) proportion of infected individuals at age a

M(a) infection intensity at age a which referred to EPG k parasite aggregation parameter 94

As seen in Equation 2.10, I assume that the parasite distribution in the population is in negative binomial manner where k estimates refers to level of aggregation. Worm burden, defined as infection intensity (M), is estimated through the mean EPG for each age-group.

2.3.5.4 Estimation of parasite aggregation

0.09

0.08

0.07

0.06

0.05

K esimates K 0.04

0.03

0.02

0.01

0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 Age (year)

Na-yao prevalence data Na-yao cohort data Na-isarn/Na-ngam prevalence data

K for median age K for median age K for median age

Figure 2.17 k estimates from study areas

Table 2.19 k estimates from study areas

Na-yao 2002 cross-sectional Na-isarn/Na-ngam Na-yao 2004 follow-up study study cross-sectional study Age group Age group Age group k k k Quartile (as years, (as years, (as years, (95% CI) (95% CI) (95% CI) median) median) median) 0 – 11 0.009 0 – 14 0.014 0 – 35 0.006 Q1 (7) (0.002 – 0.023) (11) (0.005 – 0.032) (17) (0.002 – 0.015) 12 – 32 0.040 15 – 40 0.036 36 – 45 0.017 Q2 (13) (0.025 – 0.060) (32) (0.020 – 0.061) (41) (0.009 – 0.030) 33 – 46 0.040 41 – 53 0.041 46 – 57 0.018 Q3 (39.5) (0.026 – 0.059) (45) (0.024 – 0.067) (51) (0.010 – 0.030) ≥ 47 0.056 ≥ 54 0.056 ≥ 58 0.020 Q4 (57) (0.039 – 0.077) (61) (0.037 – 0.082) (64) (0.011 – 0.034) Median age 0.035 0.034 0.015 33 38 45 (as years) (0.028 – 0.043) (0.026 – 0.044) (0.011 – 0.020) 96

Table 2.19 shows the k estimates from 3 datasets. The pattern from Figure 2.17 shows a similar trend in that infection is highly aggregated in the younger age-group. As k increased, the infection became progressively less clustered for older age groups. For Na-yao prevalence data, k estimates became relatively stable from age group 12 - 32 and 33 - 46 years. When k is assumed to be constant at median age, k estimates ranged between 0.015 - 0.035.

2.3.5.5 Infection dynamics of infection intensity

Na-yao 2002 prevalence data Na-yao 2004 incidence data

0.4 0.4

0.35 0.35

0.3 0.3

0.25 0.25

0.2 0.2

Prevalence 0.15 0.15 Infected proportion Infected 0.1 0.1

0.05 0.05

0 0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 0 5 10 15 20 25 30 35 40 45 50 55 60 65 Age (year) Age (year)

variable M and constant k constant M and variable k variable M and constant k constant M and variable k

observed observed

Figure 2.18 Predicted prevalence from varying Figure 2.19 Predicted incidence from varying k

k and M for Na-yao cohort data and M for Na-yao prevalence data

Table 2.18, confirms that uniform incidence for all ages was fitted for linear prevalence-age structure. Figures 2.18 and 2.19 explore the infection dynamics whether infection intensity or parasite aggregation, which denoted by M and k, are show to be age-dependent or not. Constant k is derived using mean-age k and constant M is defined by mean-age EPG.

To explore the age-dependent structure with respect to infection dynamics of infection intensity, the assumptions are explained below.

1. When the degree of parasite aggregation; k, was assumed to be constant with age-varying intensity; predicted incidence with variable M was linearly fitted for observed incidence in Na- yao follow-up data with p-value = 0.032 (Figure 2.19). For prevalence data in Figure 2.18, the p- value = 0.058, which is not statistically significant. 2. When infection intensity; M, was assumed to be constant with age-varying parasite aggregation; predicted prevalence with variable k was linearly fitted for both datasets in from Figures 2.18 97

and 2.19 with p = 0.004 for cross-sectional study and p = 0.002 for follow-up study. Parasite aggregation was age-dependent when infection intensity was assumed to be constant.

Constant k overestimated the prevalence in younger ages, and predicted prevalence became higher than observed prevalence for the Q2 age-group. Predicted prevalence from constant M was overestimated from young age until the Q3 age group and became lower than observed prevalence in Q4.

Na-isarn and Na-ngam prevalence data

0.12

0.1

0.08

0.06

0.04 Infected proportion Infected

0.02

0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 Age (year) variable M and constant k constant M and variable k observed

Figure 2.20 Predicted prevalence from varying k and M for Na-ngam and Na-isarn prevalence data

Figure 2.20, Na-isarn and Na-ngam plots, show similar patterns to Na-yao follow-up data. The data was linearly fitted for both constant k; p = 0.033 and constant M; p = <0.001.

Parasite aggregation was age-dependent from all three datasets and infection intensity was age- dependent from 2 of 3 datasets under the assumption that incidence was constant for all age- groups.

Therefore, the prevalence and intensity tended to increase with age and the distribution of infection became less aggregated.

2.3.5.6 Relationship of infection intensity and risk factors using count model

Table 2.20 Summary of count model for infection intensity analysis

PRM NBRM ZIP ZINB PLH NBLH Poisson Negative Zero-inflated Zero-inflated Poisson logit Negative Model regression binomial Poisson negative hurdle binomial logit regression binomial hurdle Poisson Negative Poisson Negative Truncated Truncated Count model binomial binomial Poisson negative binomial Zero model Logit Logit Logit Logit

Table 2.20 summarizes 6 count models for infection intensity analysis; Poisson and Negative binomial model (NB), and compound model; zero-inflated Poisson (ZIP), zero-inflated negative binomial (ZINB), Poisson logit hurdle (PLH) and negative binomial logit hurdle (NBLH) model. The covariates were comprised of age-group categorized by quartiles and consumption of Koi pla, while the compound model incorporated the same set of covariates in both components of the model.

All models are based on the count model, reflecting the EPG as infection intensity. The compound model incorporated with zero modeling will reflect the infection probability.

Based on the concept of modeling zero component; the zero-inflated model describes the probability of observing zero value whereas the hurdle model predicts the probability of obtaining positive count value 1 in the logit part. For better interpretation and comparison, the results were transformed to the probability of obtaining 1 to provide more understanding with respect to risk factors.

Table 2.21 Multivariable analysis of risk factors and infection intensity of Na-yao 2004 follow-up data

Model PRM NBRM ZIP ZINB PLH NBLH Count model Age (years) 0-14 as reference 6.47 5.21 2.30 1.97 2.30 1.97 15-40 (5.89-7.10) (1.32-20.59) (2.09-2.52) (0.79-4.92) (2.09-2.52) (0.79-4.92) 4.39 4.49 2.13 1.97 2.13 1.97 41-53 (3.99-4.82) (1.13-17.90) (1.94-2.34) (0.74-4.74) (1.94-2.34) (0.74-4.73) 19.20 16.65 5.42 4.26 5.41 4.26 ≥54 (17.56-21.00) (3.99-66.94) (4.96-5.92) (1.76-10.33) (4.96-5.92) (1.75-10.33) Koi pla consumption 5.56 4.77 3.32 2.76 3.32 2.76 Yes (5.34-5.80) (1.70-13.40) (3.18-3.46) (1.62-4.70) (3.18-3.46) (1.62-4.70) Zero model Age (years) 0-14 as reference 3.54 3.51 3.63 3.63 15-40 (1.44-8.68) (1.43-8.63) (1.48-8.90) (1.48-8.90) 2.66 2.63 2.68 2.68 41-53 (1.06-6.71) (1.04-6.66) (1.06-6.77) (1.06-6.77) 6.45 6.34 6.01 6.01 ≥54 (2.68-15.53) (2.63-15.33) (2.74-15.92) (2.74-15.92) Koi pla consumption 2.54 2.51 2.54 2.54 Yes (1.44-4.46) (1.42-4.42) (1.45-4.47) (1.45-4.47)

As shown in Table 2.21, the probability of infection in zero modeling was significantly associated with Koi pla consumption in all 4 models; the OR was 2.54 (95% CI: 1.45 - 4.46) in the PLH and NBLH models. The OR in the ZIP and ZINB models was 2.54 (95% CI: 1.44 - 4.46) and 2.51 (95% CI: 1.42 - 4.42), respectively. All models reported similar results for age-groups. Older age-groups; Q2, Q3 and Q4, showed significantly increased risk of acquiring OV when compared with the Q1 age-group. The age group ≥54 years had highest risk for acquiring the infection; OR = 3.48 (95% CI: 1.34 - 9.01) from all 4 count models.

For the count model component, Koi pla was statistically significant in all models; infection intensity was higher among individuals who consumed Koi pla. Age groups showed similar patterns of effect 100

where older age increased intensity. Zero-inflated and hurdle models showed similar results with respect to the same distribution.

Older age groups; Q2, Q3 and Q4, were significantly related with intensity when compared with the Q1 age group in Poisson, NB and Poisson-based compound models; ZIP and PLH models. Age ≥54 years significantly increased infection intensity in all 6 models.

101

Table 2.22 Multivariable analysis of risk factors and infection intensity of Na-isarn and Na-ngam 2012-13 prevalence data

Model PRM NBRM ZIP ZINB PLH NBLH Count model Age (years) 0-35 as reference 3.31 2.32 1.19 0.99 1.19 0.99 36-45 (2.85-3.84) (0.41-13.28) (1.02-1.39) (0.44-2.25) (1.02-1.39) (0.44-2.25) 3.15 2.26 1.07 0.93 1.07 0.93 46-57 (2.71-3.66) (0.41-12.31) (0.92-1.24) (0.42-2.07) (0.92-1.24) (0.42-2.07) 4.35 3.61 1.21 1.17 1.21 1.17 ≥58 (3.75-5.04) (0.69-18.90) (1.04-1.40) (0.54-2.52) (1.04-1.40) (0.54-2.52) Koi pla consumption 3.09 3.01 1.86 1.95 1.86 1.95 Yes (2.87-3.34) (3.85-4.40) (1.72-2.01) (1.25-3.06) (1.72-2.01) (1.25-3.06) Zero model Age (years) 0-35 as reference 2.73 2.73 2.73 2.73 36-45 (1.03-7.20) (1.03-7.20) (1.03-7.20) (1.03-7.20) 2.83 2.83 2.83 2.83 46-57 (1.09-7.37) (1.09-7.37) (1.09-7.37) (1.09-7.37) 3.48 3.48 3.48 3.48 ≥58 (1.34-9.01) (1.34-9.01) (1.34-9.01) (1.34-9.01) Koi pla consumption 1.76 1.75 1.76 1.76 Yes (1.02-3.02) (1.02-3.00) (1.02-3.02) (1.02-3.02)

As shown in Table 2.22, the infection probability model within the compound model indicated that Koi pla consumption increased risk of infection by OR = 1.76 (95% CI: 1.02 - 3.02) in all 4 compound models. Moreover, increasing age group significantly increased risk of infection when compared with the Q1 age group. Risk was highest in age ≥58 years with OR = 3.48 (95% CI: 1.34 - 9.01). All models reported similar results in age group for infection probability.

For the count model component, Koi pla significantly increased infection intensity in all models except the NB model. Older age groups and Q2, Q3 and Q4, were observed to increase infection intensity in the Poisson-based model. All NB-based models did not provide any significance for age group association. 102

Table 2.23 Model fitting for Na-yao follow-up study

PRM NBRM ZIP ZINB PLH NBLH log L -35531.44 -730.24 -17191.51 -685.23 -16830.46 -685.88 AIC 71073 1472 34395 1392 33681 1394 ∆AIC 69600 36678 69680 37392 69680 Zero count 2.2 97.1 100 100 100 100 capturing (%)

Table 2.24 Model fitting for Na-isarn and Na-ngam cross-sectional study

PRM NBRM ZIP ZINB PLH NBLH log L -9193.62 -543.15 -1955.30 -501.79 -1955.30 -501.79 AIC 18397 1098 3931 1026 3931 1026 ∆AIC 17299 14467 17372 14467 17372 Zero count 14.1 99.8 100 100 100 100 capturing (%)

Model fitting for 2 datasets are described in Tables 2.23 and 2.24 revealing similar patterns. Evidence of over-dispersion from Section 2.3.5.3 indicated that the negative binomial distribution should be more appropriate than Poisson distribution. AIC of NB-based models was lower than Poisson-based models: 1472.48 for NB, 1392.46 for ZINB, 1393.76 for NBLH in the Na-yao study and 1098.3 for NB and 1025.59 for ZINB and HLNB in the Na-isarn and Na-ngam study. The ∆AIC shows the difference of AIC from each model to PRM used as the reference model. The largest ∆AIC means the particular model provides minimal value of AIC. From both datasets, ZINB and NBLH models provided the largest ∆AIC. Modeling the zero counts was conducted to compare the expected number of zero counts with the observed zeros by % of zero count capturing. The results indicated that Poisson model was not suitable when comparing with other models. Poisson model only achieved 2.2% of zero capturing in Table 2.23 and 14.1% in Table 2.24. NB models were much better in modeling zero counts; 97.1% and 99.8% from both datasets. However, all compound models performed better with 100% matching of expected zero to observed zero count.

The zero-inflated and hurdle models provided a better result than Poisson and negative binomial models. Based on fit statistics, ZINB and NBLH provided the best results from both datasets. However, NBLH might be slightly more preferable when considering the process of how the stool egg count is produced when the first process determining individuals are infected with OV or not, and the second process determines how many eggs are produced from infected individuals. 103

2.4 Discussion

From this project, extensive studies were conducted in three study areas with seven fieldwork projects highlighting various epidemiological data including prevalence, incidence, infection intensity and risk factors.

2.4.1 Study design and response

The cross-sectional study provided prevalence data from the Mekong Basin area where OV is endemic. However, many studies showed that prevalence of OV varies for different study areas and tended to be localized in subpopulations depending on consumption patterns and cultural influences (5, 10, 11, 29, 30, 69, 70). The results may not fully represent the whole picture of the Thai rural population. Many studies indicate a strong association between uncooked freshwater fish consumption and OV from cross-sectional designs. Causal relationship should be cautiously interpreted with respect to study method.

Cohort designs provide better evidence of a causal relationship which strengthens the evidence for potential risk factors. However, only negative cases from baseline survey were enrolled in the follow-up study while positive cases were excluded. Thus, the true population-level incidence rate might be underestimated because the high risk group was disregarded.

The cross-sectional study was conducted in two areas; Na-isarn and Na-ngam Areas. Basic demographic profile including sex, age and occupation showed no significant differences between the two areas. From the population pyramid, the constrictive pattern appears in late teenage to early adult stage, considered as the working age. Informal conversations with community leaders revealed that some villagers in this generation regularly moved out temporarily for job seeking in urban areas especially during off-season for rice planting. Additionally, the efficiency of the local database system was still limited for providing accurate population data. The population dynamics are later explored and discussed with the qualitative approach in Chapter 4.

Regarding the study response in Table 2.1, 1514 participants received the survey package including the research questionnaire and a container for stool collection. The return of the survey package was 86.1%. Although the response rate was significantly lower than in Na-isarn area, the rate was still over 70%. However, excluded questionnaires were approximately twice as high as in Na-ngam Area (44.0% and 17.1%, p <0.001) resulted in lower final response rate for Na-ngam Area. After exclusions for incomplete questionnaires, the effective response was 52.4%. The impact from incomplete questionnaires resulted in a loss response of 33.7%. The completeness of questionnaires 104

significantly affected the prevalence in the Na-isarn study (Supplement Table S-4), where the prevalence of OV of completed questionnaires group was significantly higher (p = 0.03) than the incomplete questionnaire group. However, study participants provided impressive cooperation for the high return rate of stool specimens while some limitations may have affected the completeness of the survey packages, for example, illiteracy.

The follow-up study in Na-yao and Na-ngam Areas required a more complicated enrolment process (Figure 2.6). Eligible participants comprised the population from baseline survey for each study area. The follow-up study was conducted retrospectively where the eligible participants were enrolled during the fieldwork period. The follow-up time was 15-17 months. From Table 2.3, eligible participants totalled 1803 subjects from baseline survey.

After enrolment, only 5.8% of baseline participants were excluded due to lack of consent, death and permanently moved-out accounting for 4.5 %. The returned survey packages were received at a 52.1% response rate. Loss to follow-up was 47.9%. Permanently moved-out participants did not significantly differ in the two areas. However, lost participants due to unable to contact/return specimen was statistically higher in Na-yao Area which might have resulted from less cooperation from the villagers.

Overall the percentage of questionnaires excluded was 33.7% for the cross-sectional and 9.9% for the follow-up study. Major obstacles were due to literacy and translation of the standard-Thai questionnaire to the Northeastern-Thai dialect spoken by locals. The Northeastern dialect posed some communication challenges. Local health volunteers and educated participants contributed greatly by helping to translate local words to the local Thai dialect. In addition, any information provided to the villagers needed to be translated back as well. From the loss of incomplete questionnaires statistically affected the results in the Na-isarn cross-sectional study when the three other study areas showed no significant difference from the prevalence or incidence outcome. (Supplement Table S-4 and S-6)

As I collected uncooked fish consumption data retrospectively, potential recall bias was possible. I collected exposure data and processed stool examination at the same time, so the outcome should be less affected by bias (69, 70).

Considering the stool examination, the diagnostic methods depend on detecting Opisthorchis viverrini eggs in the specimen. However, the study also aimed to provide diagnosis for other intestinal parasites including helminthic and protozoa infection for public health benefits for the study participants. 105

Among three diagnostic methods: simple smear, Kato-Katz and FECT, the Kato-Katz technique is considered a standard technique used in ambulatory settings such as healthcare facility and research fieldwork (49, 104-106). However, single Kato-Katz smear has low sensitivity for Opisthorchis viverrini diagnosis especially from light infections. Therefore, an increased number of smears or specimens would gain higher sensitivity. A study on praziquantel efficacy against Schistosoma mekongi and Opisthorchis viverrini revealed that conducting one specimen with triplicate Kato-Katz smears (1X3 scheme) or single Kato-Katz smear with 3 stool specimens (3X1 scheme) provided the same estimate of OV prevalence. However, collecting 3 stool specimens consecutively from one individual is likely to demand more time and workload, which seems unfeasible for such fieldwork (49).

I used similar diagnostic methods in both prevalence and follow-up studies. A single stool specimen was tested with three diagnostic methods instead of collecting multiple specimens assuming that I would achieve a higher response rate using the FECT technique that could improve sensitivity for detecting Opisthorchis viverrini eggs (107). Multiple studies have reported various effectiveness levels of FECT for diagnosing different species of trematode including schistosomiasis and clonorchiasis (50, 108) but information on Opisthorchis viverrini is still lacking. However, my results revealed that FECT was exclusively responsible for all diagnoses in this study while Kato-Katz was positive in 31.1% of OV diagnoses from combined results (Tables 2.14 and 2.22).

Many issues involve difficulty in microscopically distinguishing Opisthorchis viverrini eggs from other parasite eggs with similar structure such as a minute intestinal fluke (MIF) and co-infection of MIF and Opisthorchis viverrini is common in the Mekong Basin area (109, 110). However, MIF was not found in this area in the study of Traub et al. (46).

106

2.4.2 Distribution of OV infection

2.4.2.1 Prevalence of OV infection

The overall prevalence of OV was 8.69 % (95% CI: 6.82 - 10.87) which indicated that a disease burden still exists in the studied communities. The prevalence was slightly lower than the latest national survey in 2001 reporting a nationwide average of 9.4%. However, the prevalence was higher compared with regional average for the central region of Thailand (3.8%) (28). Regarding the well- preserved Northeastern environment, cultural background could play a major role in continuing risk behaviours, which could potentially maintain a higher prevalence in non-endemic area.

The youngest OV case was 2 years old. The prevalence for Q1 age group 0 - 36 years in Na-isarn 2013 was 3.41% (95% CI: 3.41 - 9.64) and 0 - 35 years in Na-ngam 2012 was 5.13% (95% CI: 1.90 - 10.83). A study from 1982 - 84 also revealed that the youngest age group had the lowest prevalence (90). Children were assumed less likely to acquire uncooked fish for consumption by themselves as their parents probably fed them intentionally or accidentally contaminated them with metacercariae from fish during food preparation (90, 111).

As seen in Figure 2.8, prevalence seems to increase with age. The prevalence remains relatively low at younger age and continues to increase by age. To maintain high prevalence in older age, the accumulation of parasites could play a major role for contributing such high prevalence. The infection period needs to be long from either long parasite life expectancy or sustained re-infection (99).

Prevalence-age structure was analysed in both study areas based on the assumption that the incidence was constant over time and for all age groups. Brockelman et al. (83) has explored the age- structure of OV prevalence in Northeastern Thailand and also estimated the incidence from the prevalence-age relationship. The Brockelman et al. (83) result stated that prevalence would increase accordingly with age from birth until the age of 15 years and remain relatively stable thereafter. My results showed that prevalence would continue to increase until older age.

I assume that the difference of infection pattern depended on the prevalence itself; the Brockelman study was conducted in an area with very high prevalence which became saturated at nearly 100 % before 15 years of age. My study revealed a relatively lower prevalence; therefore, age influence can be observed through older ages. Sustained prevalence has raised a question whether the infection maintains itself for a long period. Unlike other cross-sectional studies, this study revealed a relatively high prevalence of OV outside the traditional endemic area exhibiting some unique characteristics for infection. 107

2.4.2.2 Incidence of OV infection

Unlike cross-sectional studies, most published incidence data could be considered outdated. Surveys on the incidence of infection were not incorporated in the Thai national survey (27, 28). Moreover, the national survey scheme was suspended in 2001 and the program was exclusively active in high endemic area (28).

The incidence showed an increasing trend with age in the Na-yao 2013 study. However, a nonspecific pattern was observed in the Na-ngam 2014 study, where the incidence peaked at 35 - 47 years of age (Figure 2.9). The youngest case detected was 2 years old. Considering the cohort method, the 2-year-old case acquired the infection before age of 1 year. Infection in very young children provides an interesting issue since they are newly infected during a short follow-up period.

Na-ngam Area was the area conducted with both study designs during the project timeframe. The study began in 2012 where a cross-sectional study served as baseline for follow-up study in 2014. Na-ngam Area received a significantly higher response rate than Na-yao Area (64.3 % and 41.2 %, respectively, p <0.001).

As illustrated in Table 2.18, the prevalence-age structure allowed the incidence to be estimated from the prevalence trend (Equation 2.9). Interestingly, the estimated incidence differed from that estimated from the cohort data; 0.19/100 person-years (95% CI: 0.14 - 0.23) compared with 6.80/100 person-years (95% CI: 4.68 - 9.54) to be further explored in Chapter 3.

2.4.2.3 Intensity of the infection

The mean worm burden was indirectly assessed by the intensity of egg counts in stool specimens as EPG. From the stool examination methods, the Kato-Katz technique was less sensitive than FECT. Therefore, zero egg count was classified as negative even though showing a positive result from FECT result in less infected populations than all 3 methods combined (Table 2.8).

The standard method for stool egg count for OV infection is the Kato-Katz technique, also used to diagnose other helminth infections (71, 104, 108). However, diagnosis is made under microscope so detection of parasite egg relies on the expertise of the examiner. False negatives mainly come from two reasons: the examined portion of infected specimen does not contain eggs or the examiner missed the presence of eggs in stool. The examination is considered more difficult when the intensity is very light resulting in a low production of Opisthorchis viverrini eggs. From the three studies for which egg count results were available, i.e., Na-yao 2002 follow-up, Na-isarn and Na- ngam prevalence, the Kato-Katz test diagnostic sensitivity was 52.44 - 78.32 % (Table 2.14) 108

compared with FECT meaning about 21.68 - 47.56% of infected individuals were missed in the intensity analysis. The sensitivity for Na-isarn 2013 and Na-ngam 2012 cross-sectional study was also significantly lower than Na-yao 2002 follow-up due to the lighter intensity.

Many studies have indicated that parasitic infection is based on host heterogeneity regarding risk exposure and transmission dynamics (95, 112, 113). Few individuals harbour a large burden of worms, while remaining individuals carry minimal or no infections including zero excretor.

In general, standard Poisson distribution was used to analyse count data assuming equal mean and variance. Regarding egg count data, distribution was heavily skewed due to over-dispersion resulting in extremely large variances. The analysis indicated that negative binomial distribution was suitable to handle over-dispersed data. Besides the prevalence-age structure, intensity and parasite aggregation have played a major role in infection dynamics of helminth infection.

I explored the age-prevalence-intensity following the relationship as described in Section 2.3.5.3 with Equation 2.10. Assuming that incidence is uniform in all ages, infection intensity tended to increase with age while the parasite became less aggregated.

In this study, the level of parasite aggregation was evaluated as k estimates. The data indicated that k estimates increased with age as well, meaning that the flukes were highly aggregated at younger age (Figure 2.17). The parasite became less clustered as k increased. The estimates of k at median age from all three dataset ranged from 0.015 - 0.035 (Table 2.27). A study of S. mansoni reported an estimated k value of 0.03 - 0.5 (99).

Assuming that the infection was equally distributed in all ages (constant k), the infection intensity was statistically associated with linear age-structure from 2 of 3 datasets (Section 2.3.5.5). Furthermore, assuming constant M provides the result that the level of parasite aggregation was progressively less clustered with older age as well. The result highlighted that age-structure was involved in prevalence, intensity and parasite aggregation. At very first years of age, the parasite locally aggregated in particular hosts with relatively low intensity and the prevalence was still low as the infection occurred in few individuals. The infection became less aggregated and distributed more evenly with higher intensity among older hosts.

Bundy et al. (93, 94) reported in the study of intensity in three helminths (Ascariasis, hookworm and Trichuriasis) that despite the prevalence of all helminths becoming stable after age 5 years, the parasitic aggregation was also stable while the intensity seemed to fall at older ages. I assumed that the different pattern came from the reproductive biology of the parasite. As the individual host is a closed system, the flukes would need adequate resources to support their life and contribute to 109

reproduction. When the number of flukes reaches the critical level, the lack of resources and highly competitive environment will affect their ability to reproduce, resulting in a decline in egg output per fluke. The parasite becomes more equally-distributed because almost everyone is infected.

My fieldwork reflected a picture of lighter infection when the age-structure can be observed through the ages. Despite the mode of infection, children can acquire infection from a very young age. The k estimate ranged from 0.015 - 0.035. The infection seemed to progressively accumulate and be distributed across the population.

2.4.3 Risk factors for acquiring the infection

Infection status was assessed by logistic and Poisson regression models according to study methods. The model shows that using two covariates, age group and consumption of Koi pla, were appropriate for the analysis, which could extend to infection intensity outcome where the models were more complicated. The results summarize that consumption of Koi pla is a potential risk factor for acquiring the infection as it shows a significant association from 5 of 6 models (Section 2.3.4.3 and 2.3.5.6). Koi pla consumption also increases the risk of OV and infection intensity when incorporated with the NBLH model.

Overall picture reveals that Q4 age-group or oldest age is significantly related with increasing risk of infection from 3 of 6 studies; age ≥57 years in Na-isarn 2013 prevalence data, age ≥54 years in Na- yao 2004 cohort data and age ≥58 years in Na-isarn and Na-ngam prevalence data from the NBLH model. A trend was observed among age-groups in the NBLH model where risk of acquiring infection increased with age when compared with the reference age group. However, the same trend was not observed for infection intensity.

OV is biologically acquired by uncooked fish consumption where the infective stage; metacercariae, is found in raw fish meat. Consumption of these dishes might directly lead to increasing risk of the infection regarding method of preparation.

Two main local dishes from the study also contained raw fish meat for which various methods of preparation depend on how the fish is cooked and preserved. Koi pla is consumed immediately after mixing with ingredients. A study on survival of metacercariae by preserving fish meat under sodium chloride and acetic acid resembling the preparation method reveals that metacercariae were completely viable in the very first hours and completely inactive or presumed dead after two days of preservation. For Pla ra, it takes at least 3 - 6 months with extremely concentrated salt preservation, a conclusively unsuitable environment for the parasite to survive (87, 88). 110

Regarding the distribution of uncooked fish consumption (Supplement Table S-1 and S-2), only 12.5% from the cross-sectional study and 14.3% from the follow-up study did not report consumption of any uncooked fish dish within the study period, resulting in 85.7 - 87.5% of the population experiencing at least one dish. For Pla ra only, the consumption could reach as high as almost 90%. Therefore, a health campaign carrying a broad message to avoid all kinds of uncooked fish might result in confusion and be impractical for daily lifestyle of the local villagers.

Population characteristics play an important role for the uncooked fish consumption. Male sex and older age-group was significantly associated with consumption of Koi pla when compared with Q1, the youngest age-group. Agricultural-related work was related with consumption of Pla ra in some areas. In terms of consumption behaviours, the quantitative approach may not fully explain the pattern. Method of consumption was a matter of behaviour relating to attitude, social value and more importantly, traditional culture. While OV is more likely to be a global burden, consumption patterns seemed to be localized and environment-related (29, 30). A qualitative approach was able to explain and describe its epidemiology in terms of bio-psycho-social aspects (114). Concerning uncooked fish consumption in this community, a strong attachment was found to tradition and culture including main popular dishes as described in the results.

National Control Program suggests avoiding all uncooked fish to prevent the infection and provide treatment for infected cases (27, 28). Prevention campaigns might be impractical to villagers because they have to change most of their daily habit. Pla ra was the main ingredient in most all local Northeastern dishes while Koi pla was more likely a main course. In addition, the chemical properties of Pla ra, contain highly concentrated salt, and did not allow metacercariae to survive (47, 87, 89). As a result, implementation of controlling specific risks seemed to be more practical and convenient (69).

Compound model

Over-dispersion poses a challenge for analysis of egg intensity data. The negative binomial distribution provides a better fit than the classical Poisson model for the over-dispersed data observed in many helminth intensity studies (93, 94, 115).

The NBLH model separated the infection probability component of the model from the intensity component. In the logit model, where the probability is determined, risk for acquiring OV increases with older age-groups when compared with the reference age-group. For logistic and Poisson models, only particular age-group was associated while the trend was not clear. 111

If I assume that incidence is uniform at all ages, infection intensity tended to linearly increase with age when the parasite aggregation was assumed to be constant. The level of parasite aggregation also showed a linear trend with age when the infection became less clumped as age increased. Therefore, both intensity and parasite aggregation would interact accordingly to age and provide a different pattern from risk factor analysis. Moreover, it would be more realistic to think about incidence as a dynamic parameter rather than in constant terms (94).

Zero excretor may be the result of light infection in which small amounts of eggs are produced increasing the chance to miss the diagnosis under the microscope. The egg count method relies on Kato-Katz technique with single stool sample. Because only 42 mg of stool was used the chance was increased that the tested stool sample would not contain parasite egg as well.

The zero-inflated model included explicit consideration of the zero outcomes with sampling (from lightly infected individuals) and structural (from uninfected individuals) origin. Thus, the count part of the model will contain zero count intensity. Regarding the outcome, sampling zero could be the result of zero excretor where I assume that infected cases produce zero egg.

The zero excretor can be seen by comparing FECT to Kato-Katz results; the additional infected cases diagnosed by FECT and report zero count by Kato-Katz were 5.75% for Na-yao 2004 data and 3.80% for Na-isarn and Na-ngam 2012 – 13 data. However, based on the examination method, zero count was regarded as an uninfected case in the first place. Therefore, the infection intensity could be measured by Kato-Katz only.

The results point out that Kato-Katz underestimates the results when compared with FECT. This is considered a limitation to my results of intensity analysis.

The model only needs to handle zero excess as a cause of over-dispersion where few individuals carry a large burden of intensity while the remaining report zero or minimal intensity. The NBLH model with negative binomial in the count module showed a better fit to the data than PLH. Zero excretor may not interfere with Kato-Katz result. However, a false negative will definitely affect the outcome in underestimating prevalence and incidence in the compound model.

Consumption of Koi pla is statistically associated with infection intensity but age-groups do not provide any consistent pattern. Considering the method of data collection, the question on consumption behaviour was more likely a qualitative assessment whether the participants ever consumed uncooked fish dishes or not within the study period and frequency of consumption was not recorded. More data may be required on exposure in terms of consumption frequency for a better explanation for its association with infection intensity. 112

2.5 Conclusion

Extensive epidemiological study and statistical analyses were performed in a rural area of Thailand where OV infection is still a health burden. Prevalence and incidence are relatively high considering that the studied areas were considered non-endemic for the infection. Population characteristics played an important role in contributing to risk of infection, especially the consumption of particular traditional uncooked fish menus.

In summary, age-structure was observed in prevalence of OV with regard to intensity and level of parasite aggregation; population characteristics and risk factors were identified contributing to OV. Infection intensity showed the pattern of over-dispersion and compound models were introduced to handle the distribution issues and excess zeros outcomes.

Consumption of instant uncooked fish dishes such as Koi pla has been identified as linked with OV (1, 5, 15, 27-29). My results with two cohorts and additional supplement data from another 2 follow-up studies (69, 70) provided rational supporting evidence for the casual relationship of Koi pla, which is clearly identified as a potential risk factor for acquiring OV and should be prioritized as a major concern for prevention efforts.

Prevalence and consumption of uncooked fish are both age-related. Because the infection can apparently reside and remain for long periods, I could assume that flukes may have a long life expectancy or the host is constantly re-infected. Infected cases could accumulate in the long run and could effectively transmit the parasite in completing its life cycle. Moreover, newly-emerging cases are also capable of parasite transmission. Especially in the event of no treatment or intervention, the infection remained stable and increased the risk of cholangiocarcinoma, a bile duct cancer.

The study highlights the epidemiological view of OV in the rural community setting the majority area of Thailand. Data obtained from the field study could be useful for extensive epidemiological tools to gain more insights into the complexity of the biology and disease risk relationship providing greater understanding of the infection dynamics and parasite-host relationship. Disease surveillance might be useful for areas with a high burden of disease where the high risk population could be primarily identified with a history of uncooked fish consumption. Community-based intervention should be extensively considered with respect to National Control Programs, but the strategy needs to be adapted to the social/cultural settings, which change over time. Up-to-date information on disease dynamics has a potential impact on public health and further ongoing extensive research in this field is essential to provide effective public health management. 113

Chapter 3 Infection dynamics of OV infection

Abstract

Mathematical models were applied to explore the infection dynamics of OV infection in a rural area in Thailand. They were calibrated using epidemiological data obtained from fieldwork to gain more insights regarding the infection.

The model was based on the classic susceptible – infectious – recovered (SIR) model. The main extension was adding a second susceptible class to create primary susceptible individuals S1 who were infected for the first time. Infected individuals will naturally recover and become secondary susceptible (S2) again for re-infection with reduced infectivity. The model assumed that infected individuals recover because the Praziquantel treatment will gain longer immunity before potentially becoming re-infected. The force of infection was assumed to be uniform and the recovery rate depended on the parasite life expectancy.

The model was based on key data from the fieldwork. The fieldwork data included 2 study areas; Na- yao Area comprising 4 consecutive prevalence studies with an incidence rate of 22.1/100 person- years (95% CI: 17.6 – 27.3), Na-ngam Area comprised 2 prevalence studies with an incidence rate of 6.8/100 person-years (95% CI: 4.7 – 9.5). The model assumed that the infection was in the endemic phase where Na-yao prevalence was 22.9% (95% CI: 19.3 – 26.5) and Na-ngam was 9.3 % (95% CI: 6.9 – 11.6). The effectiveness of praziquantel treatment was estimated in Na-yao area where multiple surveys had been performed. Basic reproductive numbers (R0) were estimated from the model. Finally, the model was used to assess the effectiveness of the National Control Program.

Using model, the average infectious time was estimated in the range 1.51 – 1.81 years and that treatment effectiveness, serving as the cure rate of praziquantel was 80.1 – 92.5%. An alternative calculation from individual-level fieldwork data gave a similar estimate of 88.7 – 92.2 % of treatment effectiveness. R0 for Na-yao was 2.74 (95% CI: 2.61 – 2.89) and for Na-ngam was 2.30 (95% CI: 2.23 –

2.40). Compared with other trematodes; S. mansoni had an infectious period of 3 – 4 years and R0 1 – 2. The provided treatment with praziquantel appeared to yield the same effectiveness as that observed in previous studies at 80 – 95%. When the model was applied to the National Control Program data; uniform incidence was estimated to be 9.8 – 47.3/100 person-years. Re-infection played an important role for the chronic infection picture and shaped a sustained prevalence.

The model provided a good insight for infection dynamics of OV infection. However, parameter values using a likelihood approach were unable to be formally estimated and therefore the 114

numerical findings could have been more robust. The density-dependent pattern could have played an important role as it was a common characteristic of parasitic infection. Further studies with more data can put forward the understanding of model development with prospective up-to-date information. 115

3.1 Introduction

Mathematical modelling has been applied to almost all scientific branches. Medical science is one discipline where the application of mathematical modelling is constantly increasing including the field of infectious disease (116). Transmission is a natural process of inter-organism interaction when the infectious agent invades another organism; the host. Sometimes the process is pathogenic causing illness or disease (112, 117, 118). The challenging issues are that rates of transmission are influenced in a complicated way by the epidemiological triangle; host, pathogen and environment. None of them interact with transmission in a straightforward way such that it is easy to describe the system accurately with a mathematical formula.

Helminths are naturally complex organisms that mostly require multiple stages in their life cycle to grow and develop (116). Human infection with helminths usually occurs in tropical areas where intermediate hosts are abundant. Helminth infections are still considered to be neglected tropical diseases (NTDs). Infection is typically prevalent in underdeveloped and developing countries where resources are considerably limited (14, 43).

Some NTDs have contributed a huge impact on global scale. Schistosomiasis, or blood fluke, affects over 100 million people mostly in underdeveloped countries raising concerns from international authorities (119). Studies of its infection dynamics provide good insights on the disease (99, 112, 113, 117). Human and parasite factors have been incorporated in complex models capable of describing the infection in almost every aspect leading to massive intervention on prevention and control programs (120-123).

OV infection is regarded as an NTD because of the lack of resources and has been prioritized for control (10, 11, 30). Infection posts some serious complications and threatens the population’s health in endemic areas. OV infection is caused by Opisthorchis viverrini prevalent in Southeast Asia especially along the Mekong River Basin including Thailand, Lao PDR and Cambodia. The prevalence is very high in some areas and can reach up to 90% (29, 51).

Opisthorchis viverrini is a major pathological human liver fluke (along with other 2 species), endemic in other areas of the world. Almost 10 million people are affected by the infection. More importantly, the parasite is classified as a carcinogen causing cholangiocarcinoma; a lethal bile duct cancer. The incidence of cholangiocarcinoma in Thailand is the highest in the world (15-17, 20, 22- 24). In addition to the benefits from prevention and control, resulting in better hygiene and quality 116

of life, it is promising that cholangiocarcinoma can be prevented from Opisthorchis viverrini infection.

Intervention

Prevention Diagnosis and Treatment

A Uninfected Infected

(Risk behaviors) (Opisthorchiasis)

Social influence

Figure 3.1 Conceptual framework for infection dynamics. A and B are referred in the main text

This chapter aims to explore the infection dynamics of OV infection. Based on the conceptual framework shown in Figure 3.1, the relationship of risk behaviours and OV infection can be reflected as uninfected and infected individuals for the infection dynamics situation. Data obtained from fieldwork including prevalence and incidence of infection can be used to formulate a mathematical model to explore its interaction [A] as shown in Figure 3.1. Additionally, the effectiveness of treatment on OV infection can be evaluated [B] through follow-up study with regard to study design.

The relationship of prevalence and incidence has been discussed previously in Section 2.3.5.2 where a linear regression model was fitted with prevalence-age structure. The following section will discuss the interesting issues between directly-obtained incidence from the fieldwork and incidence from prevalence-age structure leading to the formulation of more robust mathematical modelling. 117

3.1.1 The relationship of prevalence and incidence

In general, disease burden can be measured by prevalence and incidence (124). In this study, prevalence is defined as the proportion of individuals being diagnosed for OV infection by stool examination over total studied population at a given point time. Incidence refers to the number of new cases during a follow-up period of time. Therefore, incidence can be thought of as the rate of emerging cases at a given time.

Prevalence is usually considered a less challenging way to measure disease burden. Incidence requires more effort because direct measurement needs a follow-up study to estimate newly- emerged cases from baseline data. Considering that and population measurement consumes considerable time and resources, and for field investigators to conduct at least two surveys to obtain a proper incidence is often difficult in some circumstances (83, 124).

Indirect methods for measuring the incidence of OV infection have been proposed by Brockelman et al. using age-prevalence data (40, 83). Their results demonstrated a good age-prevalence profile using a catalytic infection model following the regression method. Therefore, the single cross- sectional study could provide the incidence.

According to Brockelman’s work, the regression method should consider the following assumptions stated below;

● The infection should be in the endemic state where prevalence shows stable pattern. ● The incidence should remain stable as long as the oldest age population have been alive.

From Section 2.3.5.2 in Chapter 2, data from Na-yao 2002 – 04 and Na-ngam 2012 – 14 study shows that incidence can be fitted from prevalence-age structure.

Table 3.1 Obtained incidence from fieldwork and estimated incidence from prevalence-age structure

Incidence from prevalence-age structure Incidence from fieldwork Study area (/100 person-years) (/100 person-years) (95% CI) (95% CI) 0.70 22.05 Na-yao 2002 - 2004 (0.28 – 1.13) (17.57 – 27.34) 0.21 6.80 Na-ngam 2012 - 2014 (0.12 – 0.30) (4.68 – 9.54)

Table 3.1 shows that the incidence from prevalence-age structure is significantly lower than direct incidence obtained from fieldwork (p < 0.01). 118

Na-yao 2002: fitting prevalence Na-ngam 2012: fitting prevalence 100 100 90 90 80 80 70 70 60 60 50 50 40 40 Prevalence Prevalence (%) Prevalence (%) 30 30 20 20 10 10 0 0 0 5 10 15 20 25 30 35 40 45 50 55 0 5 10 15 20 25 30 35 40 45 50 55 60 65 Age Age

Observed Observed Predicted from age-structure incidence Predicted from age-structure incidence

Predicted from fieldwork incidence Predicted from fieldwork incidence

Figure 3.2 Fitting prevalence with age-structure incidence and fieldwork incidence

As shown in Figure 3.2, fitting prevalence with provided incidence with Equations 2.7 and 2.8 in Chapter 2 showed that incidence from prevalence-age structure provided a linear trend for underling prevalence; p – value = 0.012 and 0.005 for Na-yao and Na-ngam areas, respectively.

Therefore, the possibility of using a mathematical model to explore the relationship between prevalence and incidence from the fieldwork was considered to better understand infection dynamics.

The intentions for this section of the thesis are stated below.

● To explore the infection dynamics of OV infection with epidemiological data from the fieldwork emphasizing human population dynamics

●Figure To exploreSTYLEREF the 1 impact\s 3. SEQ of treatmentFigure \* ARABIC on infection \s 1 2 dynamics Fitting prevalence and apply with the findingsage-structure to the incidence National and fieldwork incidence Control Program data.

119

3.2 Methodology

3.2.1 Model definition

The infection dynamics are explored with a compartmental deterministic model with some parameters obtained from the literature and other obtained by comparing model output with observed field data. The individuals in the population were formed in subgroups as compartmental models and the infection dynamics were observed among these individuals uniformly. A differential equation was applied to describe the rate of change in the number of individuals over time.

The model structure was based on the SIR framework. That infectious individuals tended to be re- infected once they temporarily recovered was assumed (99).

Therefore the model is extended to S1IS2R framework; Primary susceptible – Infectious – Secondary susceptible – Recovered.

Recovered R(t) Deaths (m) Deaths (m) ε·r

Deaths (m) βI/N Births (b) Primary Infectious Susceptible I(t) γ (1-ε)·r S1(t)

μβI/N Secondary Susceptible S2(t)

Figure STYLEREF 1 \s 3. SEQ Figure \* ARABIC \s 1 3 Model diagram for Opisthorchiasis infectionDeaths (m) dynamics Figure 3.3 Model diagram for opisthorchiasis infection dynamics

Figure 3.2 show the model diagram with S1IS2R framework. The model assumes that the infection process starts with primary susceptible individuals (S1) where they have never been infected with OV infection before.

Then they become infectious (I); the infectious time depends on parasite life expectancy in the human host body. The incubation period of OV infection is relatively short; approximately 2 – 4 weeks (3, 9), compared with the entire infection process, so the pre-infectious compartment is not incorporated in the model. 120

Infectious individuals naturally recovered with rate r. To maintain its endemicity, recovered individuals are considered as re-infected, so the compartment was defined as secondary susceptible

(S2). That infectious individuals would have some immunological response to the infection was assumed; therefore, the secondary susceptible individuals would be re-infected with reduced infectivity (μ) as a result of partial immunity.

In case treatment was provided, a fraction (ε) of infectious individuals would recover and be immune for some period while the remaining fraction (1 - ε) of infectious individuals would proceed to the S2 compartment as a natural process of the infection. However, the immunity against infection was considered temporary as well; protective immunity will prevent the recovered individuals from being re-infected for some period. Waning of the immunity will move them to the S2 compartment again with the rate γ.

In addition, the model also aimed to describe the long-term transmission of the infection. Population demographic parameters such as birth rate and death rate were also incorporated in the model.

Table 3.2 Summary of fieldwork data for Na-yao area

Value Year Outcome Source (95% CI) 22.9 % 2002 Prevalence (19.3-26.5) 20.7 % Prevalence Revised data from (69) (17.6-23.9) 2004 22.1 /100 person-years Incidence (17.6-27.3) 18.6 % 2007 Prevalence Revised data from (70) (16.4-20.9) 10.7 % 2011 Prevalence Project fieldwork (8.7-12.7)

The data used in this chapter comprised 2 sets of fieldwork data: Na-yao and Na-ngam area. The Na- yao area has been extensively studied since 2002. The results from studies in 2002-04 and 2007-09 were published with incidence and risk factors of OV infection (69, 70) and the current 2011 - 13 study was included in this PhD project. The 4 point prevalence data available was used to observe infection dynamics. The incidence rate from the 2004 (69) data served as a force of infection of the study as shown in Table 3.4.

121

Table 3.3 Summary of fieldwork data for Na-ngam area

Value Year Outcome Source (95% CI) 9.3 % 2012 Prevalence (6.9 – 11.6) 8.1 % Prevalence Project fieldwork (6.2 - 10.0) 2014 6.8 /100 person-years Incidence (4.7 – 9.5)

As shown in Table 3.5, the summary cross-sectional study was conducted in Na-ngam Area in 2012 and a follow-up study in 2014 with 14 months follow-up period. Both studies used comprehensive surveys, so prevalence and incidence were applicable for the study. 122

3.2.2 Parameters of the model

Diagram of compartmental model in Figure 3.3 could refer to following differential equations in Figure 3.4 from Equation 3.1 to 3.4:

푑푆 (푡) 훽퐼(푡) 1 = 푏푁(푡) − 푆 (푡) − 푚푆 (푡) (3.1) 푑푡 푁(푡) 1 1

푑퐼(푡) 훽퐼(푡) 훽퐼(푡) = 푆 (푡) − 휀푟퐼(푡) − (1 − 휀)푟퐼(푡) + 휇 푆 (푡) − 푚퐼(푡) (3.2) 푑푡 푁(푡) 1 푁(푡) 2

푑푆 (푡) 훽퐼(푡) 2 = (1 − 휀)푟퐼(푡) − 휇 푆 (푡) + 훾푅(푡) − 푚푆 (푡) (3.3) 푑푡 푁(푡) 2 2

푑푅(푡) = 휀푟퐼(푡) − 훾푅(푡) − 푚푅(푡) (3.4) 푑푡

Figure 3.4 Differential equations of S1IS2R model

The definitions of population compartments are as follow:

푑푆1(푡) The rate of change in the number of primary susceptible individuals at time t 푑푡

푑퐼(푡) The rate of change in the number of infectious individuals at time t 푑푡

푑푆2(푡) The rate of change in the number of secondary susceptible individuals at time t 푑푡

푑푅(푡) The rate of change in the number of recovered individuals time t 푑푡

N(t) Total population size at time t which equals to S1(t) + I(t) + S2(t) + R(t).

123

Table 3.4 Model parameters

Parameters Definition Value ● Na-yao study: 0.221/person-years (95% CI: Force of infection; the rate at which primary 0.176 – 0.273) βI/N susceptible individuals becoming infected ● Na-ngam study: 0.068/person-years (95% per unit time at time t CI: 0.047 – 0.095) Recovery rate; the rate at which infectious r individuals recover from the infection per ● Fitted from the model unit time Fraction of infectious individuals being ● ε = 0 for fitting recovery rate ε recovered from treatment ● Fitting for the treatment effectiveness Fraction of infectious individuals being 1 - ε naturally recovered and become secondary susceptible Fraction of reduced infectivity due to μ 0.9 (40) immune modulation Force of re-infection; the rate at which μβI/N secondary susceptible individuals becoming Force of infection reduced by fraction μ re-infected per unit time at time t γ The rate of waning immunity 0.1/person-years (125-127) b Per capita birth rate 0.0129* m Per capita mortality rate 0.0072* * From World Bank database http://data.worldbank.org/country/thailand

The force of infection is determined from the rate at which susceptible individuals become infected per unit time at time t. The rate is obtained from the fieldwork where incidence rate (λ) was directly measured from the follow-up study from both Na-yao and Na-ngam studies.

The model framework assumed that force of infection was uniform. Individuals were mixed homogeneously regardless of demographic characteristics.

The incidence from both published follow-ups study (69, 70) in Na-yao area indicate that the incidence rate is fairly constant with value of 22.1/100 person-years (95% CI; 17.6 – 27.3) and 21.4/100 person-years (95% CI; 18.5-24.7), respectively. Therefore the model uses incidence rate from first follow-up from both 2004 Na-yao and 2014 Na-ngam study as a force of infection.

124

Force of infection is following frequency-dependent assumption where the risk of infection remains unchanged over increased population referring to the Equation 3.5;

퐼(푡) 휆(푡) = 훽 (3.5) 푁(푡)

훽 is the number of individuals effectively contacted by each person per unit time. Since I assume the incidence is constant, 훽 at time t can be indirectly calculated by Equation 3.6;

푁(푡) 훽(푡) = 휆 (3.6) 퐼(푡)

So, number of susceptible individuals getting infected is calculated by Equation 3.7;

퐼(푡) 푁(푡) 퐼(푡) 훽 푆(푡) = 휆 푆(푡) = 휆푆(푡) (3.7) 푁(푡) 퐼(푡) 푁(푡)

In this setting, 훽 will be vary in order to maintain constant incidence rate in growing population.

The force of infection was assumed to refer to infection of the primary susceptible individuals. Re- infection occurring in naturally-recovered individuals resulted in lower rates due to the partial influence from immune modulation.

The data on re-infection rate was very limited. The rate was estimated from the evaluation of re- infection rates after praziquantel treatment from previous studies (40) indicating that the force infection could be reduced to 90% from primary infection.

125

3.2.3 Fitting the model

From mathematical model, I fit the parameters by finding a point estimated value from the differential equation.

Little is known about OV life expectancy which reflects the duration of infectiousness. Therefore, the model was used to fit the duration of infectiousness considering the fieldwork data. The indirect evidence obtained by the immunological response from the infected host could be used to estimate a life expectancy at 1 – 10 years (9, 13, 45, 70, 128). Compared with other helminths, life expectancy of S. mansoni is 3 – 4 years and other soil-transmitted helminths are 1 – 4 years (99, 113).

Assuming that re-infection is a main feature of the infection, where no infected individuals could develop protective immunity (ε = 0), infected individuals should recover due to the parasite life expectancy solely and tend to be re-infected again.

The duration of infectiousness was fitted from the model and the rate of recovery; defined as r, duration of infectiousness was estimated from Equation 3.8;

1 퐷푢푟푎푡푖표푛 표푓 푖푛푓푒푐푡푖표푢푠푛푒푠푠 = (3.8) 푅푒푐표푣푒푟푦 푟푎푡푒

When treatment was provided with praziquantel, the fraction ε of infectious individuals recovered and remained immune for some time. Waning immunity will remove them with the rate γ to be re- infected again in compartment S2. The fraction ε estimated from the model reflected the effectiveness of the praziquantel treatment.

The period of immunity was assumed to be longer than the S2 period. However, the data on protective immunity was limited. Based on available data, the immunity was assumed to last 10 years (125-127).

Birth rate; b and mortality rate; m, are also incorporated to the model in order to explore the long- term infection dynamics.

The study emphasized the epidemiological view of the infection dynamics regarding human factors, so the result mainly accounted for disease distribution. The study outcome was reported as prevalence using Equation 3.9;

푁푢푚푏푒푟 표푓 푒푥𝑖푠푡𝑖푛푔 푐푎푠푒푠 표푛 푎 푠푝푒푐𝑖푓𝑖푐 푑푎푡푒 Prevalence = (3.9) 푁푢푚푏푒푟 표푓 푝푒표푝푙푒 𝑖푛 푡ℎ푒 푝표푝푢푙푎푡𝑖표푛 표푛 푡ℎ𝑖푠 푑푎푡푒

126

According to the study, the point prevalence constituted the proportion of infected individuals at time t determined from Equation 3.10;

퐼(푡) Prevalence(t) = (3.10) 푁(푡)

The model was used to calculate the impact of the National Control Program on annual prevalence from 1984 – 2001 (27, 28) based on estimates of treatment effectiveness from the mathematical model.

127

3.2.4 Basic reproductive number

Conventionally, the basic reproductive number or R0 was defined as the average number of secondary infectious individuals resulting from infectious persons and introduced in a totally susceptible population (112, 113, 116, 117).

In the model incorporated with population demographic data, assuming that the population will

increase over time, the R0 was defined as demonstrated by Equation 3.11;

푅0 = 훽퐷 (3.11)

Where β is contact rate and D is duration of infectiousness. Considering my model;

Recovered R(t) Deaths (m) Deaths (m) Figure STYLEREF 1 \s 3. SEQ Figure \* ARABIC \s 1ε·r 5 Model diagram considered the R0 Deaths (m) βI/N Births (b) Primary Infectious Susceptible I(t) γ (1-ε)·r S1(t)

μβI/N Secondary Susceptible S2(t)

Deaths (m)

Figure 3.5 Model diagram considered the R0

The contact rate is incorporated with 2 components as shown in Figure 3.5;

● Primary infection contact rate (β): obtained from force of infection of primary susceptible individuals. ● Re-infection contact rate (μβ): force of infection with reduced infectivity (μ) The duration of infectiousness depends on

● Recovery rate (r) ● Death rate (m)

The R0 for the infection would be as Equation 3.12;

훽 + 휇훽 푅 = (3.12) 0 푟 + 푚

128

3.2.5 Estimation of treatment effectiveness using individual-level data

Treatment Newly negative (A) Positive rfgf Still positive (B)

Newly positive (C)

Negative Still negative (D)

Baseline study Follow-up study

Figure 3.6 Diagram for estimation of treatment effectiveness using individual-level data

Treatment effectiveness could be conventionally estimated from the fieldwork to give an overview of the infection dynamics. Because post-treatment follow-up was not immediately assessed after the baseline study, the treatment effectiveness was calculated from the proportion of positive and negative cases reported in the follow-up study by assuming the conditions below.

● All positive cases received praziquantel treatment. ● Newly negative cases are cured as a result from treatment. ● Still positive cases come from treatment failure. ● Newly positive cases acquired the infection during follow-up period. ● Still negative cases did not acquire infection during follow-up period.

The eligible studies for assessing treatment effectiveness included the Na-yao study in 2002-04 and the Na-ngam study in 2012 – 14, where both baseline and follow-up studies were comprehensive surveys. The assumptions are based on all fieldworks being conducted with the same method and response rate being fairly equal (70 – 80 %).

As illustrated in Figure 3.6, the epidemiological outcome could be calculated as;

● Total survey was conducted in the follow-up study. However, the study design emphasized on the follow-up population (C + D). The portion of A + B are the additional population included in the total survey. I assume that the proportional population structure would remain the same. Therefore the proportion of A and B are indirectly estimated from prevalence and incidence measured from fieldworks. 129

푁푒푤푙푦 푝표푠𝑖푡𝑖푣푒 (퐶) ● Incidence of infection is , the negative cases from baseline study 퐵푎푠푒푙𝑖푛푒 푛푒푔푎푡𝑖푣푒 (퐶+퐷) will be followed-up. There would be a portion of baseline negative that becomes newly positive.

● Prevalence of follow-up study will be the sum of positive cases divided by total study 퐴푙푙 푝표푠𝑖푡𝑖푣푒 (퐵+퐶) population, 푇표푡푎푙 푠푡푢푑푦 푝표푝푢푙푎푡𝑖표푛 (퐴+퐵+퐶+퐷)

● After baseline survey, positive cases receive treatment and result in follow-up study. Newly

푁푒푤푙푦 푛푒푔푎푡푖푣푒 (퐴) negative cases are assumed to the cured from the treatment, , 퐵푎푠푒푙푖푛푒 푝표푠푖푡푖푣푒 (퐴+퐵) Treatment effectiveness will be the proportion of newly negative cases and baseline positive.

● I assume that still positive cases are the result of treatment failure, so it could be roughly 퐹표푙푙표푤−푢푝 푝표푠𝑖푡𝑖푣푒 (퐵) estimated with indirect calculation by, , However, the follow-up 퐵푎푠푒푙𝑖푛푒 푝표푠𝑖푡𝑖푣푒 (퐴+퐵) positive individuals were based on the total value; the original method was not intended to include the participants in this portion to be matched for treatment failure rate, and immediate posttreatment evaluation was not conducted. The calculation was a crude estimation.

130

3.3 Results 3.3.1 Infection dynamics of OV infection

Table 3.5 Estimation of duration of infectiousness and recovery rate

Fieldwork data Model estimation Incidence rate Duration of infectiousness Recovery rate Study area (/person-years) (years) (/person-years) (95% CI) (95% CI) (95% CI) 0.221 1.81 0.662 Na-yao (0.176 – 0.273) (1.21 - 1.84) (0.543 – 0.825) 0.068 1.51 0.595 Na-ngam (0.047 – 0.095) (1.15 - 2.22) (0.449 – 0.869)

As seen in Table 3.5, the model shows that the estimated recovery rate is 0.662 /100 person-years (95% CI: 0.543 – 0.825) for Na-yao data and 0.595/person-years (95% CI: 0.449 – 0.869) for Na-ngam data, respectively. Therefore, the duration of infectiousness will be 1.81 years (95% CI: 1.21 - 1.84) for Na-yao and 1.51 years (95% CI: 1.15 - 2.22) for Na-ngam area. 131

Figure 3.7 Infection dynamics of prevalence in study area

As shown in Table 3.5, the result is that duration of infectiousness allows the model to fit the data in the two different study populations as shown in Table 3.7. For the two study sites, the values of the duration of infectiousness were consistent with both the prevalence of infection and the transmission model. The parameters in the transmission model (other than the duration of infection) were well constrained by other evidence available in the literature (Table 3.2). Based on the assumption that the force of infection was uniform and the prevalence was in a stable phase, the duration of infectiousness from both studies was in the range 1.51 – 1.81 years (Table 3.3). The model solution found was consistent with these observations suggesting that prevalence increased within the first 10 years and then remained stable thereafter for both studies.

132

3.3.2 Estimation of treatment effectiveness using population data and mathematical model

1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 2020 2022 Year

Figure STYLEREF 1 \s 3. SEQ Figure \* ARABIC \s 1 8 Estimation of treatment efficacy from the model Figure 3.8 Estimation of treatment effectiveness from the model

The serial prevalence data from the Na-yao study (Table 3.4) is appropriate to estimate treatment effectiveness. The trend of prevalence can be observed through the 9-year study period with initial prevalence starting at 22.9% (95% CI: 19.3 – 26.5) and assumed to be in a stable phase. Subsequent prevalence with 95% CI was fitted in the model to estimate treatment effectiveness as a result of the ε fraction. Incidence was assumed to be uniform at 22.1/100 person-years (95% CI: 17.6 – 27.3).

Using the model solutions found above, the proportion of treated infectious individuals (ε) who entered a recovered stage was identified. Values between 80.1 – 92.5% for the effectiveness of treatment gave prevalence patterns that were consistent with the observed data (Figure 3.8).

133

3.3.3 Estimation of treatment effectiveness using individual-level data

Under the assumption that the overall force of infection was substantially reduced immediately after community-wide treatment, effectiveness of treatment was calculated for individuals as a comparison for the calculation using population data and the model.

Table 3.6 Estimation of treatment effectiveness from Na-yao 2002-04 data

Baseline survey Follow-up study Baseline Follow- Incidence rate Cumulative Cumulative Follow-up Results Prevalence up time (/100 person- Results Incidence Proportion Prevalence (%, 95% CI) (years) years) (%, 95% CI) (%, 95% CI) (%, 95% CI) 20.0 New negative 22.9 (16.5-23.4) Positive 1.19 (19.3-26.5) 2.9 Still positive (1.4-4.3) 20.7 (17.3-24.2) 23.1 17.8 New positive 77.1 22.1 (19.1-27.7) (14.6-21.4) Negative 1.19 (73.5-80.7) (17.6-27.3) 76.9 59.3 Still negative (72.7-80.9) (55.0-63.4)

Table 3.7 Estimation of treatment effectiveness from Na-ngam 2012-14 data

Baseline survey Follow-up study Baseline Follow- Incidence rate Cumulative Cumulative Follow-up Results Prevalence up time (/100 person- Results Incidence Proportion Prevalence (%, 95% CI) (years) years) (%, 95% CI) (%, 95% CI) (%, 95% CI) 8.2 New negative 8.9 (6.3-10.1) Positive 1.17 (6.8-10.8) 0.7 Still positive (0.1-1.2) 8.1 (6.2-10.0) 8.1 7.4 New positive 91.1 7.2 (6.1-10.2) (5.6-9.3) Negative 1.17 (89.2-93.1) (5.0-10.0) 91.9 83.7 Still negative (89.8-93.9) (81.1-86.3)

As reported in Table 3.6 and 3.7, the result estimated treatment effectiveness of praziquantel obtained from indirect calculation from fieldwork data. Considering the 2 areas by merging their results, the treatment effectiveness should have ranges between 88.7 – 92.2%. Additionally, the estimated treatment failure would be 12.7% (95% CI: 7.8 – 15.5) in Na-yao Area and 7.9% (95% CI: 1.6 – 10.6) Na-gnam Area. Tables 3.6 and 3.7 shows the calculated treatment effectiveness for Na- yao Area equalled 87.3% (95% CI: 84.5 - 92.2) and for Na-ngam Area equalled 92.1 % (95% CI: 88.7 – 98.4).

134

3.3.4 Basic reproductive number; R0

Table 3.8 Basic reproductive number

Study area R0 95% CI Na-yao 2002 2.74 (2.61 – 2.89) Na-ngam 2012 2.30 (2.23 – 2.40)

The relationship was investigated between the basic reproductive number and the duration of

infectiousness, within the content of previous results. R0 was estimated to be 2.74 (95% CI: 2.61 – 2.89) for Na-yao area and 2.30 (95% CI: 2.23 – 2.40) for Na-ngam area, assuming that the duration of infectiousness was a fixed ranged between 1.5 and 1.8.

R0 with infection dynamics

5 0

R 4

0 1 2 3 4 5 6 7 8 9 10 Infectious time (years)

Na-yao incidence 95% CI of incidence Na-ngam incidence 95% CI of incidence

Figure 3.9 Basic Figurereproductive STYLEREF number 1 \s 3.with SEQ various Figure duration \* ARABIC of infectiousness\s 1 9 Basic reproductive number with various duration of infectiousness This range was much lower than the range of values of R0 consistent with the duration of infectiousness from the reviewed literature (9, 13, 45, 70, 128). Due to various durations of

infectiousness from 1 to 10 years, the R0 slightly increased from the shorter duration of infectiousness and then rapidly rose after 8 years especially in the Na-yao study area. For Na-yao Figure STYLEREF 1 \s 3. SEQ Figure \* ARABIC \s 1 9 Basic reproductive number with various Area, the R0 ranged from 2.52 (95% CI: 2.44 – 2.62) to 6.10 (95% CI: 5.30 – 7.30) and in Na-ngam duration of infectiousness Area ranged from 2.21 (95% CI: 2.16 – 2.27) to 3.37 (95% CI: 2.98 – 3.86). The functional form that showed this dependency can be seen in Figure 3.9.

135

3.3.5 Estimating the effectiveness of the National Control Program using the model

Figure 3.10 FittingFigure model STYLEREF with prevalence 1 \s 3. SEQ from Figure National \* ARABIC Control \s Program. 1 10 Fitting model with prevalence from National Control Program. The prevalence of OV infection was plotted against study years from 1987 to 2001 during the National Control Program implementation presented in Figure 3.10. The blue shaded area represents the calculated prevalence from the approximate range of effectiveness.

The prevalence was 63.6% at the beginning of the program. After praziquantel treatment was provided, prevalence decreased to approximately 10% within 5 years and remained stable after 1992. Assuming the parameters setting was similar to current fieldwork and incidence was uniform through time, the incidence was estimated from the model fitted to the prevalence data.

The estimate of treatment effectiveness was used to calculate the likely impact of the National Control Program in reducing prevalence. The model could fit the data to estimate incidence rate. From Figure 3.10, model fitting indicated that the incidence rate should range from 9.8 – 47.3/100 person-years considering 80 – 100% treatment effectiveness. Estimation of treatment effectiveness was assessed though the fitted model with fieldwork data (Section 3.3.2) and indirect calculation from individual-level data (Section 3.3.3). The approximate range of effectiveness was 80 – 100% (maximum and minimum of shaded area shown in Figure 3.10).

136

3.4 Discussion

In this chapter, the duration of infectiousness was determined from model fitting. The point estimates of the duration of infectiousness were 1.51 years (95% CI: 1.15 - 2.22) to 1.81 years (95% CI: 1.21 - 1.84). Two methods were used to calculate the effectiveness of Praziquantel treatment. From the mathematical model, the effectiveness was estimated to be 80.1 – 92.5%. The effectiveness for individual-level obtained from indirect calculation from fieldwork data ranged from

88.7 – 92.2%. R0 was estimated to be 2.30 (95% CI: 2.23 – 2.40) to 2.74 (95% CI: 2.61 – 2.89) assuming that the duration of infectiousness was a fixed ranged between 1.51 and 1.81. The approximate range of treatment effectiveness was used to calculate the likely impact of the National Control Program in reducing prevalence. The model could fit the data to estimate incidence rate ranging from 9.8 – 47.3/100 person-years.

3.4.1 Infection dynamics and study limitation

The major concerns regarding model analysis are presented and discussed here. The method of model fitting was the fine tuning of one free parameter to visually find the optimal value of another. The analysis could have been more robust by using the maximum likelihood allowing the robust confidence intervals to be defined. For further work, future models will likely to incorporate more unknown parameters. However, because only one free parameter was tuned, the difference in point estimates between the results presented here and more statistically robust results should be small.

The infection dynamics of this system are more complicated than presented here. The actual biological life cycle of OV is complicated involving multiple intermediate hosts including snails and cyprinoid fish. The fish themselves contain more than 10 species and all are edible. Besides that, many animal species also serve as a reservoir hosts (2, 3, 34, 44, 45).

The force of infection was assumed to be constant. From the model, the force of infection value was fixed to the incidence rate obtained from the fieldwork. Therefore, the contact rate (β) varied over time. The contact rate itself posed some interesting issues. By definition, the contact rate is the interaction between infectious and susceptible individuals. OV infection is not transmitted directly from person to person. The infection is transmitted through the parasite life cycle within the host body and external environment. Therefore, susceptible individuals indirectly transmit the infection from infectious individuals by infective larvae; metacercariae, found in uncooked cyprinoid fish undergoing multiple steps of growth and development. Little is known about the contact rate for OV infection. When compared with other trematodes such as Schistosomiasis, the mode of infection still differs much (94, 121, 123). 137

Concerning the relationship within force of infection components, the interaction between each compartment, the contact rate and the population dynamics, it would be more realistic to think about the force of infection as a dynamic quantity. The input parameters were assumed to be constant over time including birth and death rates. The death rate should vary according to the host characteristics for each compartment. For example, chronic infection among individuals might suffer from pathological changes in their hepatobiliary system resulting in higher mortality rates than those of uninfected individuals (113).

Migration may have affected the transmission dynamics. The population pyramid structure shows the constricted pattern referring to the active population movement in the early to middle-age classes. The main reason obtained from the qualitative study revealed that temporary immigration occurs during off season. This setting could potentially have affected the infection dynamics as a result from the remaining proportion of the current population and might not truly reflect the population structure. As mentioned earlier, this would also result in an irregular age structure from statistical analysis.

The incidence obtained from the fieldwork was derived from the follow-up population, so only negative cases were followed up for incidence. The true incidence might have been underestimated because the high-risk population with infection were excluded from the follow-up population. The re-infection rate could be calculated using the indirect method from the data but would not be accurate as direct incidence from the study. As result, fieldwork incidence was used as a reference for the summary of force of infection.

3.4.2 Basic reproductive number

Parasite Study location R0 Onchocerca volvulus Sudan 5 – 35 Ascaris lumbricoides Iran 4 – 5 Necator americanus India 2 – 3 Schistosoma mansoni Brazil 1 – 2 Opisthorchis viverrini Thailand 2.21 – 2.52 The table is adapted from Roy M Anderson et al. (99)

From epidemiological data, the R0 from Na-yao was 2.74 (95% CI; 2.61 – 2.89) and for Na-ngam was

2.30 (95% CI; 2.23 – 2.40), compared with R0 from other flukes; S. mansoni was 1 - 2 and O. volvulus was 5 - 35.

138

3.4.3 Treatment effectiveness

This was the very first model attempting to assess the treatment effectiveness regarding the infection dynamics of OV infection. As stated earlier, the National Control Program has been applied to reduce the infection from 1984 – 2001 covering up to 40 provinces in Thailand with praziquantel regimen (26-28, 48). The prevalence decreased from 63.6% from 1984-87 to 9.4% in 2001. Then the program was switched to a passive strategy where the control program was no longer a national scale.

The model agreed with various treatment efficacies based on the incidence of the infection. The fitting incidence was assumed to be uniform and ranged from 9.8 – 47.3/100 person-years. As study areas dated after the National Control scheme, the result might reflect some similarity of epidemiological pattern of the infection between my fieldwork and the National data.

However, the incidence should be higher considering the initial prevalence started at 63.6% in 1987 and became endemic. In other surveys, the prevalence and incidence could reach up to 100% (40, 83). Hence, the estimated incidence was rather an average rate to fit the overall prevalence trend. Clearly, the incidence would vary according to the transmission factors as well. The model still needs a more flexible approach on parameters such as demographic factors or updated data on transmission dynamics.

Either pre-national deworming field trials or National Control Program focused on the cure from infection. The reduction of worm intensity was not taken into account because it did not reflect the achievement of infection control. Moreover, that the infection intensity decreased while diagnosis was still positive is not practically understandable; which might be reasonable for public perceptions.

3.4.4 Treatment failure

The estimated treatment failure rate of Na-yao Area was 12.7% (7.8 – 15.5, 95% CI) and Na-ngam Area was 7.9% (1.6 – 10.6, 95% CI) probably consistent with estimated treatment effectiveness (Table 3.9).

Originally, the treatment failure rate was not directly evaluated because only negative cases from the baseline survey were followed in the subsequent follow-up study. The remaining participants in the follow-up comprehensive survey did not completely match the original population from the baseline survey. However, the treatment failure rate could be calculated by indirect method from the fieldwork data assuming that all baseline positive cases received Praziquantel treatment and treatment failure cases remained infected until the follow-up study. Therefore, the obtained 139

treatment failure rate stated here is a crude estimation by indirect method and still far less accurate. Further fieldwork studies should include immediate posttreatment evaluation when treatment failure needs to be assessed.

Helminths are multicellular and complex-structure organisms. Their size is relatively much larger than microorganisms such as bacteria or viruses. As a result, they require effective defence mechanisms against immune systems to evade and remain viable within their host body. Assuming this, helminth infection is more likely a constant state where the host cannot produce adequate immunity against re-infection or further infection. Thus, the infection could be accumulated within a host producing a temporal relationship with respect to time-dependence.

The phenomenon does occur where the prevalence increases in the aging population as seen in many helminth infections. In terms of formulating a mathematical model, the age pattern could shape a more precise structure to a model considering differing levels of risk population.

However, this model did not consider transmission dynamics within parasite biology. Greater insight regarding transmission dynamics could potentially contribute various control strategy choices such as the role of intermediate host control or vaccination.

Much fieldwork on chemotherapy and mathematically-statistical analysis favour treatment strategies directed at target populations (28, 40, 83, 105). Potential risk behaviours and high-risk populations could be identified. Therefore, the population was less likely a random mix. Infection intensity is more likely to be associated with the intensity of uncooked fish consumption practice particularly local dishes such as Koi pla (70, 83). Moreover, qualitative aspects of OV infection indicate that the consumption pattern is strongly related to demographic influence such as agriculture-related jobs or alcohol-related social events (70, 114). Therefore, the contact pattern regarding demographic heterogeneity should be concerned.

140

3.5 Conclusion

Infection is a complex phenomenon involving multiple factors including infectious agents, hosts and the environment the so-called epidemiologic triangle. Understanding infection dynamics is crucial as it provides the strategy to battle the infectious disease; both treatment and prevention.

Mathematical modelling has become an effective tool to explore these complicated relationships. It could visualize the flow of infection dynamics. Some infectious diseases have been well-investigated such as schistosomiasis and malaria, leading to contributions of ongoing interventions with huge efforts.

OV infection caused by Opisthorchis viverrini is considered an NTD, and prevalent within Southeast Asian, where most countries are still developing. The limited resources result in difficulties and a constant struggle to sustain prevention and control. Ultimately, OV infection can cause a lethal malignancy: cholangiocarcinoma. Hence, the prominence of this infection should drive the health community to expand treatment and prevention measures while ensuring resources are used under effective strategies.

Mathematical modelling is among the many useful tools to extend the knowledge of infection dynamics. The data from the study project with additional data from previous works are able to formulate effective models and provide greater insight about how each compartment can be related.

141

Chapter 4 Evaluation of risk behaviours for OV infection as determined by a qualitative approach

Abstract

Central Thailand was considered a non-endemic area for OV infection. However, my fieldwork revealed that the prevalence and incidence were relatively high compared with the regional average data. I have hypothesized that the behavioural-psycho-social background of the study population played an important role in the high burden of the infection which quantitative approach may not fully describe such characteristics. As a result, the qualitative approach was used to highlight the potential of social components affecting the infection dynamics to gain more understanding of the risk behaviours and their contexts.

The qualitative approach including focus group discussion and in-depth interview was employed in Na-ngam area in 2012. Study participants were allocated to each group based on their stool examination result from baseline and follow-up study reflecting the current infection status as never-infected, previously-infected, newly-infected and re-infected within 14 months of study period. Local heath volunteers were also assessed as to their role in the community regarding the infection. Discussion themes were directed by predefined topics regarding the current quantitative data and reviewed literature to develop greater understanding of OV infection and the associated health risks. Framework analysis was mainly used to explore the association between factors from thematic content extracted from gathered data. Additionally, grounded theory was used when new ideas were generated from the analysis.

The results showed that Koi pla was still a popular dish in the community as the dish itself represented Northeastern culture. The cultural norm had been transferred from the ancestors to their descendants; the evidence could be visually observed that they still preserved their lifestyle despite the fact that they had already moved out from the northeast around 40 years ago.

Population dynamics also influenced the social aspects of OV infection. Some elders complained that discontinuing the consumption of Koi pla went against old traditions with respect to cultural norms and socialization. Higher education teaches about hygienic living conditions including OV infection transmission, so teenagers and young adults tended to modify their lifestyle including eating habits. Therefore, health education only might not be practical to motivate the behavioural change. More efforts are needed to support the transformation. Children are a potential key to pass knowledge to 142

their parents and school-based education programs can serve as a practical hub for knowledge distribution.

For rural settings, community leaders function as an important key in the social structure. As the local administration is managed by the central and regional authority, governmental staff might not be aware of local cultural norms. Community leaders could bridge this gap and help facilitate the governmental affairs. Local health volunteers play an important role for the communication and interaction among villagers and health authorities. They could act as community leaders and as role models for health interventions. Because they are locals, they could more easily gain trust and cooperation from the villagers that are a crucial part in maintaining the sustainability of health campaigns in the community.

Social influence showed a strong impact on infection dynamics of OV infection. Within the social component, the interaction of knowledge and culture transfer across generations provides more insight to bio-psycho-social aspects.

143

4.1 Introduction

Health is becoming a subject of interest for social science disciplines (129). From the quantitative research perspective, the questions for social aspect concerning health have become more important to understand the underlying process of human behaviours associated with disease.

In terms of infectious disease, 3 main factors contribute to the infection: host, agent and environment (130). From Figure 4.1, Chapters 2 and 3 described the infection dynamics regarding the quantitative approach including statistical and mathematical modelling; the results quantified the size of the problem such as prevalence and incidence of the infection or what was the risk of the infection. For my study, age and Koi pla consumption were significant risk factors for acquiring OV infection.

144

Intervention

Prevention Diagnosis and Treatment

Risk behaviours Opisthorchiasis

Social influence

Figure 4.1 Conceptual framework for qualitative approach

Consumption of Koi pla is a direct mode of transmission of OV infection (2). Therefore, it is directly involved with human behaviour. However, understanding the behaviour may require more tools to more completely understand the underlying causes.

A study in a Northeastern population (4) revealed that knowledge toward OV infection was at a good level (80 %). However, good knowledge did not reflect safe behaviours regarding consumption of uncooked fish. The pattern of consumption was also observed (70, 114, 131) for uncooked fish consumption; male was the predominant sex for Koi pla consumption and it was associated with alcoholic drinking. The change from active to passive strategies to prevent OV infection created some issues such as the role of primary prevention (28, 132). It was challenging to fill the missing gap with qualitative tools for a more comprehensive result.

My fieldwork was conducted in Central Thailand, which was considered a non-endemic area (11, 69). However, the prevalence and incidence found from the study were relatively high compared with regional average data (Sections 2.3.3.1 and 2.3.3.2 from Chapter 2). I have hypothesized that the 145

behavioural-psycho-social background of the study population played an important role involving a high burden of infection.

The quantitative approach may not fully describe such characteristics regarding the behavioural- psycho-social background. Therefore, the qualitative approach highlighted the potential of the social component, which has played a major role in infection dynamics to gain more understanding with regard to the “how” and “why” (129) of the concerning issues. The contribution of the social component could shape the risk behaviours and also influence the underlying process of intervention including prevention and control.

This Chapter aims to describe the social influence on infection dynamics with regard to the stated questions below.

 Why OV infection is still a burden in the area, and how does the community react to the problem?

 How do knowledge, attitudes and perceptions interact with OV infection and related risk behaviours?

 What is the interaction between OV infection and social components from the conceptual framework of OV infection (Figure 4.1)?

 What is the impact of the OV infection from individual to community perspective?

146

4.2 Methodology

The study design comprised 2 approaches including group and in-depth interviews.

4.2.1 Group interview

In the qualitative study, group interviews allowed researchers to investigate the study topic from more than one perspective. In the social sciences and urban planning, group interview allows researchers to study people in a more natural conversation pattern than typically occurs in a one-to- one interview. In combination with participant observation, they can be used for learning about groups and their patterns of interaction. An advantage is their fairly low cost compared with surveys, as one can get results relatively quickly and increase the sample size of a report by talking with several people at once.

Group interviews could be classified in subtypes depending on particular features. However, the fundamental concepts are based on gathering data from groups of people. Characteristics of participants and the aims of data implication would define its subtypes including consensus panels, focus group discussions, natural group discussion and community interviews (129).

Focus group discussion was the selected technique used in this study. The pattern of the focus group discussion was typically more formal than the general meeting with a more structural pattern. The participants were enrolled to meet the sampling criteria. They would discuss and provide opinions under the facilitator’s direction with predefined topics. However, the participants were allowed to express their views freely to seek a broad range of ideas on open-ended topics. Meeting places and times were usually scheduled. The conversation was stored in a recording device for later decoding and analysis.

Focus group discussions usually contained 8 – 12 participants to ensure the coverage of opinion expression. When fewer members attended, the meeting format would be replaced with small group interviews.

4.2.2 In-depth interview

In-depth interviews are face-to-face conversations between the researcher and participants. Generally, interview is the most common form of data collection in both quantitative and qualitative design. In this study, the in-depth interview was used to assess some interesting points that had emerged during focus group discussions or deviant cases reflecting alternative ideas from the discussed issues. 147

4.2.3 Study population

Qualitative study was conducted in the area of Na-ngam Area from 2012 – 14. Study participants were enrolled from the quantitative study where demographic data, questionnaire data and stool examination results were collected (Chapter 2).

Research participants were invited to a group conference or interview in a focus group discussion. The selection method was stratified purposive sampling. The enrolled participants were categorized in 5 groups based on stool examination results for OV infection. Demographic characteristics were considered the selection criteria based on homogeneity of the study population to avoid influential members leading or dominating the conference. However, the enrolment process was flexible and preliminary interview was given to demonstrate attitudes and opinions of the invited members.

Table 4.1 Selection criteria for focus group discussion

1 2 3 4 5 Never-infected Previously-infected Newly-infected Re-infected Local heath Baseline study 2012 Negative Positive Negative Positive volunteers Follow-up study 2014 Negative Negative Positive Positive

As shown in Table 4.1, the main criteria was based on stool examination results showing infection dynamics during the follow-up period reflecting the corresponding factors contributing to the infection. For the re-infected group, infected cases from baseline study were treated with praziquantel and assumed to be cured. If their follow-up stool examination result was positive, they were assumed to be re-infected during the follow-up period.

The local health volunteers who acted as fieldwork collaborators were categorized in separate groups to avoid them dominating the discussions.

For some unique cases, in-depth interviews were arranged for a specific case with interesting issues.

148

4.2.4 Study process

4.2.4.1 Focus group discussion 1. The stool examination result was analysed for classification criteria with respect to study results in 2012 and 2014. 2. The method of selection was purposive sampling. Joining the study was voluntary and written informed consent was given by the enrolled participants. The consent also included the permission for voice and image recording during the conversation. 3. Each participant was informed about their group, time and venue for the group discussion. They were notified again shortly before the appointment by any means of communication such as mobile phone or personal reminder. 4. On the conference day, an ice-breaking exercise was performed at the beginning of the meeting. The moderator would introduce him/herself and let each of the participants speak as an introduction to get to know each other and establish a friendly environment. Then participants were informed about the objectives and the process of the discussion. 5. The moderator managed the conversation by asking focused questions to meet the study objectives and maintain the flow of opinions expressed. The moderator maintained the structure of the conference while continuing the session in a relaxed atmosphere; the participants received equal opportunity to freely express their opinions. 6. Discussions focused on various aspects relating to OV infection and its risk factors. The flow of discussion followed the main topic themes listed below.  Knowledge o Basic knowledge of the infection; liver fluke life cycle, mode of transmission and infection, risk factors, diagnosis and treatment o Health-related consequences from the infection; role of carcinogens, cholangiocarcinoma  Attitude o Perception of OV infection and its consequences o Health concerns of the risk factors and the infection  Uncooked fish consumption behaviours o Situation of uncooked fish consumption in the community o Current pattern of consumption behaviours regarding social aspects of the community  Impact of the infection on the community o Health and social impacts from the individual to community levels 149

 Solutions o Roles of the National Control Program from the community perspective o Concerns regarding treatment and control o Roles of primary prevention o Accessibility to healthcare services 7. The moderator used open-end questions to contribute various expressions from the members. Closed-end questions were also used to clarify some interesting points. 8. The conversations were recorded using a voice recorder. The data collection proceeded continuously until the discussions were saturated when nothing new was being generated from the meeting. It usually took 30 – 45 minutes for each session. 4.2.4.2 In-depth interview

Some members were invited to face-to-face interviews. The conversations focused on specific points of interest raised during the focus group discussions.

4.2.5 Data processing and analysis

Data processing and analysis included multiple approaches to interpret physical conversations in systematic content, exploring and explaining the situation of OV infection in qualitative aspects.

Table 4.2 Analysis approach to qualitative data

Analysis approach Aim Thematic content analysis To identify and organize themes from data Framework analysis To explore content within data framework Grounded theory To develop theory from data

The methods used included direct quotations and selected words to consider actual local words used by the participants. Table 4.2 summarizes the analysis approaches to the collected and processed data as described below.

150

4.2.5.1 Thematic content analysis

The recorded conversations from each session were transcribed to text using a word processing program. Then text-based data were sorted and coded according to categories or common themes. This first step helped rearranging unorganized conversation data in systemic form of thematic data. The guidelines from the discussed topics are indicated below.

 Knowledge of and attitude towards the infection  Risk behaviours  Impact on the community  Solutions

In addition, relevant themes emerging from the analysis were taken into account.

4.2.5.2 Framework analysis

The main framework followed the OV infection model from Figure 4.1 highlighting the broad interaction of social influence to other compartments.

Thematic contents were categorized in a systematic chart and used to analyse of the thematic framework. The thematic framework was elaborated and expanded from the social influence compartment. The analysis included exploring and comparing data within this framework to explain the interaction between each factor regarding thematic contents.

4.2.5.3 Grounded theory This method was used when new ideas were encountered from the analysis. In addition to analysing contents within the framework in Section 4.2.3.2, new theories could be obtained from the data.

The qualitative analysis aimed to achieve the outcomes listed below.

1. Contributing conceptual definition from discussed topics.

2. Developing typologies and classifications.

3. Explaining the situation of the infection with consideration of bio-psycho-social aspects.

4. Exploring the association between attitudes, health behaviours and social influence of OV infection and its risk factors.

5. Generating new theories or ideas regarding the infection. 151

Since the discussion theme was directed by predefined topics regarding the current quantitative data and reviewed literature to develop an understanding of OV infection and the health risks, framework analysis was mainly used to explore the associations between factors. From the thematic content extracted from the gathered data, the analysing steps included those listed below.

 Data familiarization: data was carefully read and transcript to computer-based text.

 Thematic analysis: the thematic content analysis was preliminarily used to identify main topics; repetitive quotes, expression or statements were also summarized by frequency. Saturation was observed when little or no change was found in the codebook indicating that the thematic topics were extracted and well demonstrated.

 Indexing: data was revised, organized and summarized for analysis. Each theme was extracted from the content then coded and labelled by category.

 Charting: Summarized thematic data was rearranged to visualize the whole picture of the contents. The chart summary allowed the researcher to look through the range of data to sort, compare and organize the content systematically.

 Mapping and interpretation: diagrams and tables were used to explore the relationship between concepts and typologies developed from the contents.

Additionally, grounded theory was a helping tool to explore new ideas and theories emerging from the contents. The method employed the deductive approach where data were moved back and forth between theory and data content.

Firstly, data was broken into smaller pieces of information or data fractures. Each entry was marked as open coding and intensively analysed line-by-line to explore the insight meaning to develop initial concepts of what was occurring from the situation.

Initial concepts provided further steps of analysis. Open coding can produce a list of concepts and then be reorganized into similar phenomena to explore properties and dimensions.

Properties comprised the characteristics or attributes of the phenomena and dimensions were the continua along which properties could be arranged. The method attempted to explain the current situation or reasoning the phenomena based on the generated ideas.

152

The next step included axial coding; fracture data were classified in categories to develop a coding paradigm entailing a set of questions about each code as described below.

 Condition: What conditions give rise to the categories?

 Context: What was the context of the issue?

 Interactional strategies: How to manage to interact or handle the phenomena?

 Consequences: What were the consequences from the strategy?

Finally, selective coding was used to analyse and compare to explore the most relevant concepts or core categories to explain most of what was occurring.

153

4.3 Results 4.3.1 Population characteristics The qualitative study was conducted in Na-ngam. Details of the study population and community background are described in Sections 1.6.1 and 1.6.2.

Figure 4.2 Population pyramid of Na-ngam 2012 prevalence data

Table 4.3 Age structure classified by generations of Na-ngam 2012 prevalence data

Age group N % Generation Years Children 0 – 12 47 10.2 Teenagers 13 – 17 16 3.5 Adults 18 - 60 305 65.9 Elderly ≥ 60 95 20.5

According to the Institute for Population and Social Research (IPSR), Mahidol University and the National Statistical Office of Thailand, age-class can be classified in 4 major generations based on bio-psycho-social aspects. From Figure 4.2, the population pyramid showed a constrictive pattern where age-class 10 – 29 years was relatively low compared with other classes. Table 4.3 indicates that majority of the population comprised adults. It was observed that the elderly population was about 20% of the total proportion. Referring to Table 2.5 from Chapter 2, the proportion of females and males was nearly equal (54.9% and 45.1%, respectively). Ages ranged from 2 – 87 years. Mean age was 43.3 ± 19.9 and median age was 47.0 ± 23.

154

Table 4.4 Characteristics of focus group discussion participants

Discussion group Characteristic Local health volunteer Never-infected Newly-infected Previously-infected No. of members 12 9 7 7 Gender Male 1 (8.3 %) 1 (11.1 %) 5 (71.4 %) 1 (14.3 %) Female 11 (91.7%) 8 (88.9 %) 2 (28.6 % ) 6 (85.7 %) Mean age (years) 39.8 49.8 62.8 56.0 Agricultural work Agricultural work Agricultural work Agricultural work Occupation Self-employed Self-employed Self-employed Student Educational level Primary school Primary school Primary school Primary school Average income 12000 5000 5000 5000 (baht/month)

The sessions were categorized in 4 groups; local health volunteers, never-infected, newly-infected and previously-infected groups (Table 4.4). Participants in the re-infected group were too few to form a group discussion, so they were allocated to in-depth interviews only. From a total of 35 members; 8 were male (22.9%) and 27 were female (77.1%). Females dominated in all groups except the newly-infected group. Overall mean age of participants was 51.1 years. Most of them were agriculture-related workers. The most common educational level was primary school and average income was approximately 6750 baht/month (135 GBP).

155

Table 4.5 Characteristic of in-depth interview participants

In-depth interview group Characteristic Re-infected Previously-infected Newly-infected No. of participants 3 2 1 Gender Male 1 (33.3 %) 2 (100 %) 0 Female 2 (66.7 %) 0 1 (100 %) Mean age (years) 59.7 38.5 57.0 Agricultural work Unemployed Self-employed Occupation Self-employed Student Unemployed Educational level Primary school Primary school Primary school Average income 8300 2500 4000 (baht/month)

Some members who made interesting contributions to the group discussions were invited to join in- depth interviews to provide more details of the unique issues. Re-infected cases were also invited to be interviewed to provide information about the recurrence of the disease. As shown in Table 4.5, 6 members enrolled for the interview reorganized in 3 groups, i.e., Re-infected, control and newly- infected groups. Males and females were equal in proportion. However, previously-infected and newly-infected groups comprised only one sex. Total mean age was 51.7 years. The major occupation was self-employed. The main educational level was primary school and average income was 4933 baht/month (100 GBP).

156

4.3.2 Thematic contents 4.3.2.1 Situation of the OV infection in the community

Table 4.6 Situation of the OV infection in the community

Themes Concerning topics Emerging issues  Acknowledgement of the  Cultural value and social  Local health volunteers were infection influence of uncooked fish regarded as respected  Uncooked fish consumption consumption members of community was identified as health risk  Misunderstanding of uncooked fish menus

Health campaign based on the National Control Program was provided to the community including health education, diagnosis and treatment by the collaboration between research team and local health volunteers.

Local health volunteers played an important role to communicate and interact among villagers and local health authorities. They gained respect from the villagers as important key members of the community.

As shown in Table 4.6, most participants acknowledged that OV infection was still a health burden to the community. They perceived that the infection was strongly associated with consumption behaviours and chronic infection could result in serious outcome. However, misunderstanding and misconception still remained in some villagers. The potential of social and cultural influence also impacted on attitudes and perceptions of the infection. High-risk populations and risk factors continued to be identified. Moreover, the prevalence and incidence of OV infection were still relatively high compared with the baseline population.

Most villagers knew that some of their popular fish dishes were uncooked. Some were still confused that all uncooked dished could cause OV infection while some of them knew that only fresh fish material could lead to infection.

Koi pla (chopped raw fish salad) is a popular dish that has been consistently identified as a major risk factor of the infection. Although the community learned that it could lead to infection, the dish was still valued as a Northeastern culture norm. Daily consumption might have decreased; Koi pla consumption was still popular during festive events, friend or family meeting and male social drinking.

157

Older age showed strong attachment to the traditional value of Northeastern culture; “Most of the elders still eat raw fish; they came from the Northeastern region” said one 48-year-old woman from the newly-infected group mentioning the strong attachment between the elders and their ancestors who originally lived in Northeastern Thailand and later migrated to the current area.

4.3.2.2 Current situation of consumption behaviours

Table 4.7 Current situation of consumption behaviours

Themes Concerning topics Emerging issues  Popularity of uncooked fish  Subpopulations were  Some uncooked fish menus consumption identified for risk behaviours were the way of life  Eating pattern among  Burden in behavioural  Eating habit as family generation modification tradition

The current situation of consumption behaviours is shown in Table 4.7; adult participants acknowledged that instantly-prepared uncooked fish could cause the infection. Some of them knew that Koi pla dish was a risk factor, so they avoided eating Koi pla but sought a substitute dish. Some of them perceived that all kinds of uncooked fish dishes led to infection, so they chose to continue their eating behaviours unchanged because modifying a familiar lifestyle was too complicated.

Moreover, some dishes represented a traditional value, especially Koi pla, which they always consumed during festive events. Some elderly members stated that,

“I’ve been living with this food for a long time; it’s a part of my life. It’s valuable. But children don’t eat it anymore”

“When I have family meetings, if they serve you (Koi pla) you should eat it, or else you are insulting them” said a 51-year-old woman from the newly-infected group.

Although uncooked fish dishes varied in this community, Koi pla and Pla ra were the most popular dishes villagers frequently consumed in both sexes and all age groups.

“Everybody eats raw fish; it’s very rare to see someone who doesn’t” said one member of the re- infected group. However, consumption of Koi pla decreased in females and at younger ages, “As I recognize, it’s once a year, twice as a maximum.” said one 40-year-old woman describing having Koi pla. 158

Koi pla was shown to be strongly related to multiple aspects as their way of life. It was considered a food frequently eaten with alcoholic drinks especially in male social drinking.

“Koi pla was served in a drinking party among male friends. Koi pla was the best when paired with alcohol”, one 70-year-old man from the newly-infected group mentioned about the relationship between alcohol drinking and uncooked fish eating habit. “It’s a socialization process,” said one 61- year-old from the newly-infected group.

“If I have to quit Koi pla, I have to quit alcohol, which is a highly unlikely thing for me to want to do,” said one 48-year-old man from the re-infected group.

Some of them stated that eating Koi pla was related to agriculture-related occupations: “Farmers always eat Koi pla, fish is easily caught from the river when they go to a paddy field” said one 40- year-old woman from the never-infected group.

On the contrary, Pla ra consumption was popular at all ages and both sexes. Pla ra was used as a seasoning, food ingredient or even consumed as a main dish. The roles of Pla ra are various. When mentioning uncooked fish dishes, Pla ra would be the very first dish they recognized. “I can’t live without Pla ra,” said one 48-year-old woman from the previously-infected group.

Some families whose parents still consumed Koi pla kept their children away from it unless cooked. Moreover, if children and teenagers consumed Koi pla, they were always from families with uncooked fish eating habits. “I want to eat it,” one 7-year-old boy mentioned about Koi pla, who had parents who regularly consumed Koi pla and he was a newly-infected case.

Some villagers stated that uncooked fish consumption was an old tradition. It might be difficult to avoid or stop eating due to the strong cultural attachment. However, it began to gradually fade out from the younger generation as one 64-year-old women from the never-infected group said, “They don’t even recognize Pla ra; they just enjoy hotdogs.”

The teenage group said that they did not like Koi pla. They thought the preparation method and food appearance looked disgusting and unacceptable. Moreover, the fishy taste did not match their more urbanized lifestyle. Younger generations loved to eat instantly prepared foods such as instant noodles or ready-to-eat meals. Some thought it was fashionable to eat what was advertised on the television.

“Teenagers do not have much uncooked food. They enjoy instant noodles to anything easy to prepare such as fried eggs” 159

“They like cooked food, whatever kind of preparation; fried, grilled or roasted”

Health education provided more insights about the threat of uncooked fish consumption. The roles of local health volunteers and health campaigns influenced the behavioural patterns of consumption.

“The newer generation never eats raw fish; they were educated from their school,” said one 64-year- old woman from the never-infected group.

4.3.2.3 Knowledge of the infection

Table 4.8 Knowledge of the infection

Themes Concerning topics Emerging issues  Causes of parasitic infection  Inaccurate knowledge leads  Misinterpretation come from many routes to risk behaviours  Information overload  Individual parasite needs its  Lack of visible evidence of  Healthy lifestyle does not own mode of infection parasite affect awareness guarantee healthy life

Table 4.8 shows that participants knew that uncooked Cyprinoid fish contained metacercariae, which could lead to infection. However, some of them were confused about the life cycle of other parasites;

“When I go out barefooted, the worms penetrate my skin and go up to infect the liver” “They go through your nail once you chop fish and leave the utensil unwashed”

Some of them perceived that the mode of infection was through the faecal-oral route, the common intestinal helminth life cycle.

“It’s not about eating raw fish; it’s about fresh vegetables. Don’t you see eating raw fish was obviously reduced nowadays but people still become infected? I’m sure it was contaminated with vegetables,” said one 51-year-old woman from the newly-infected group.

Some villagers perceived that fish from clean water sources was free of metacercariae.

“The water body was clean without aquatic weeds or chemical substances, so the fish was also clean from parasites,” said one 70-year-old from the newly-infected group. 160

They also knew that Koi pla was a risk factor for OV infection. Most of them understood that dressing uncooked fish meat with acidic agents such as lime juice or spirits only changed its colour, it was not completely cooked and metacercariae were still viable. However, one member from the re-infected group had never heard about OV infection.

“I knew nothing until the doctors came. It’s very usual around here; fish are chopped and mixed with ingredients, tasty,” one 48-year-old man admitted.

Teenagers showed better knowledge of OV infection. They knew the important part of the life cycle and were able to explain the mode of infection and transmission. However, most of their knowledge was acquired from textbooks or internet. They stated that they only knew of the parasite without experiencing it which might have affected their awareness. The adult and elder groups recognized that OV infection could lead to bile duct cancer but some believed it was a result of agricultural insecticide contaminated with fish meat or immunocompromised host as evidenced below.

“I don’t see any association between the infection and cancer. In fact, chemical toxins do. I see many people who never smoke, drink or eat raw fish but still die of cancer. I think it depends on each individual. Of course these stuffs have toxins, but if you were strong enough you’ll be fine. It doesn’t matter what you take into your body. If your immunity was weak you’ll get cancer even you eat vegetables,” One 70-year-old-man discussed about the cause of cancer. A 62-year-old man from newly-infected group also supported that; “Some people have a very healthy lifestyle. Sadly they died at 40 or 50 years from cancer. It doesn’t make any sense and it’s proven that they’re related” The villagers acknowledged the transmission process of the parasite. They knew that once stool contaminated with parasite eggs were released in a natural water source, the eggs hatched and continued its life cycle.

Some of them mentioned that public excretion could be avoided and in fact most of them used toilets on a regular basis, but some factors could not prevent disease transmission. “I saw the toilet pumping service car dumped stool in the river,” one 70-year-old man mentioned about how the authorities managed the public toilet pumping service.

161

4.3.2.4 Perception of OV infection and its consequences

Table 4.9 Perception of OV infection and its consequences

Themes Concerning topics Emerging issues  Cooked food was considered  The infection was easy to  Stigmatization from being hygienic threat infected  Doctors were responsible for  Primary prevention was not a  School was an information peoples’ health primary concern hub

The villagers perceived that OV infection was the major problem to the community (Table 4.9). They showed better attitudes towards OV infection. Although some of them might have misunderstood about the life cycle, they agreed that cooked food was more hygienic and could prevent not only OV infection but other intestinal parasites as well. Their consumption behaviours might be difficult to observe in practice, but parents were trying not to let their children consume Koi pla; “Cooked food was good, vegetables need to be washed as well”

After receiving stool examination results, some villagers decreased their uncooked fish consumption as they mentioned, “Though I’ve never seen the worms, the result makes me stop eating raw fish”. Even when the result was negative, some were still concerned about other’s results or they were afraid that if they chose to maintain risk behaviours. One day the result might turn to be positive, “I see doctors collect stool and give medicine; I think it’s time to stop eating raw fish. I don’t want to be one of the infected”

Some were concerned about possible side effects of taking praziquantel. They were afraid that if they were infected, they would have to take the medication again, “It didn’t feel good after taking the medicine,” said one 48-year-old from the re-infected group.

The younger generation had better accessibility to information. The school taught them about basic hygiene. Most households could not afford the internet service and IT infrastructure was still under development. However, internet was available at school and internet cafés in community centres. Some of them spent time during break or after class looking for more information about OV infection, “I saw the pictures of worms on the internet,” said one school-age participant.

As the villagers perceived that OV infection was a disease, they preferred to focus on treatment rather than prevention. In addition, some of them learned that treatment for the infection was simple and feasible so they were willing to wait for the healthcare worker as they thought it should 162

be the physician’s responsibility to take care of them; “Only the doctors know the disease, so it was better to leave it in their hands” said one 70-year-old man.

4.3.2.5 Health concern of risk factors and the infection

Table 4.10 Health concern of risk factors and the infection

Themes Concerning topics Emerging issues  Infection threats the  Asymptomatic manifestation  The term “liver fluke” was a community affects health awareness misnomer since the liver was  Each generations approach to  Health concerns do not not directly involved disease threat in different associate with eating habit  Concern was primarily raised ways by the doctors

Table 4.10 reveals the health concerns of risk factors and the infection, high prevalence and incidence of OV infection triggered concerns about the health consequences to the community. However, they suggested that the main clinical manifestation was asymptomatic and presented symptoms were also nonspecific. They understood that OV infection caused abdominal distension, ascites and jaundice and were actually the symptoms of cholangiocarcinoma or liver diseases. They also knew that OV infection could be cured by anthelminthic medication. Most of them acknowledged the health threat of the infection. Because they were physically healthy, health awareness seemed to decrease due to the disease’s natural history.

All groups expressed that they were asked for stool examination because the doctors (research team) thought it was important. Practically, nobody ever asked for a stool examination at the local healthcare facility, “I consider what the doctor concerns”

However, some thought to look after the health was the doctor’s responsibility. “I’ll do whatever they say. They said I have the parasite, so they gave me a drug.” said one 70-year- old man.

Most villagers were not sure about the relationship between OV infection and cholangiocarcinoma. It was noticeable that the villagers also expressed various opinions toward the parasite and infection with respect to their age; teenage, adult or elderly.

Teenagers and adults knew that chronic infection could lead to cancer, but they did not know exactly what kind of cancer resulted from OV infection. Moreover, the name “liver fluke” (also called in Thai) confused the villagers and that the pathology occurred in liver tissue. Some believed that consuming 163

Koi pla with alcohol could accelerate the process of liver cirrhosis. Elderly members thought OV infection was not related to cholangiocarcinoma. They observed that many were still healthy despite the fact that they regularly consumed uncooked fish. Although all age groups realized the medical importance of OV infection, younger aged members were concerned about live worms living in their body while they were less aware of being sick from the infection. Male youths started to consume uncooked fish in adolescence to imitate their friends. Small children might be fed uncooked fish by their parents.

Adults and the elderly paid more attention to infection from the parasite. Lastly, the elderly members were concerned about both parasite and the infection. However, some admitted that they could not avoid eating uncooked fish despite knowing the unhealthy consequences.

The villagers stated that uncooked fish consumption needed to be slowly reduced. Abruptly discontinuing consumption seemed impractical. For Koi pla consumption, quantity and frequency were decreased over time except in the elder group that stated it was difficult to give up their way of life. Moreover, some of them mentioned that contributing to a major change was not worthwhile because they only had a few years left in their life. “I’ve been living so far and I all have to die anyway, why not just enjoy life while it lasts,” said one 83-year-old woman from the re-infected group.

164

4.3.2.6 Diagnosis and treatment

Table 4.11 Diagnosis and treatment

Themes Concerning topics Emerging issues  OV infection was curable  Praziquantel was not  Treatment causes more available in the local health symptoms than infection facilities  Misunderstanding of anthelmintic drug

From Table 4.11, standard treatment with praziquantel was provided by the research team under the project after the stool was examined. OV infection cases received their diagnosis with praziquantel and other parasitic infections also received corresponding medication.

When they were aware of being infected, two approaches allowed them to receive medication. Firstly, they waited for the stool examination result from the research team who regularly visited them. They would definitely receive praziquantel with this method, “Doctors come very often. If I got infected, then they’ll give me drugs,” said one 47-year-old woman in the local volunteer group.

Secondly, they also perceived that anthelminthic medication was available at the community healthcare centre or they could directly purchase over-the-counter medicine from the pharmacy for their convenience. Unfortunately, those anthelminthic drugs were for intestinal helminths such as albendazole or mebendazole. Praziquantel was not available in the community pharmacies unless villagers went to the district hospital which was 1 hour away.

Additionally, some thought that a single drug could cure all parasitic infections including OV infection. Especially for those who currently consumed uncooked fish, some purchased medicine from a drugstore and misunderstood that it could cure liver flukes so they returned to resume their eating behaviours, “It’s simple, just go to the drug store,” Said one 61-year-old from the newly- infected group. After receiving praziquantel, some felt that the infection was truly eradicated from their body, “After I took your drug, all worms were killed. So I feel I’m stronger,” said one 60-year-old woman from the previously-infected group.

However, some who used to take medicine still recognized praziquantel from its side effects, “The drug was so strong, I vomited until nothing was left in my stomach,” said one 46-year-old from the local volunteer group. 165

Some stated that a home remedy was available to cure parasitic infection, “From the old days, I mixed salt with coconut cream to treat parasites”

4.3.2.7 Role of National Control Program

Table 4.12 Role of National Control Program

Themes Concerning topics Emerging issues  National control program  Avoidance of all uncooked  Primary prevention needs generally improves heath fish menus was not practical more specific strategy  Only fresh-prepared uncooked fish can cause the infection

As shown in Table 4.12, the participants stated that they had better knowledge after they received health education. Consumption of Koi pla decreased, even some could not definitely quit but could reduce in terms of frequency and amount of consumption.

“The doctors come and talk about it every year. I see the health campaign posters every day, it keeps reminding me not to eat Koi pla,” said one 54-year-old from the previously-infected group.

Because the development of cholangiocarcinoma is a long process and mainly asymptomatic, so the cancer was not a major concern to the villagers after being infected with OV infection. In fact, the prevention was simple. Avoidance of uncooked fish did not require sophisticated medical intervention. Moreover, the treatment was also effective and affordable. At this point, some of them admitted that they did not pay much attention to cholangiocarcinoma, “I can seek healthcare whenever I think I’m getting the infection,” said one member of the adult group.

The National Control Program has focused on the control of uncooked fish consumption, which would interrupt the infection process; therefore, it could potentially prevent the occurrence of cholangiocarcinoma from OV infection. However, the consumption behaviours were strongly attached to the local culture, “There were many raw fish dishes, I can’t avoid all of them,” some mentioned about uncooked fish dishes. The control program suggested that all uncooked fish dishes should be avoided.

The group discussion agreed that when a particular dish was identified as a risk factor, it would be easier and more promising for behavioural modification. 166

Children had a better understanding about the knowledge of infection. Some adults and elderly felt more comfortable and hesitated less to be educated from their offspring. Children served as an effective medium for transferring knowledge. Additionally, they did not want their children to practice the same habit.

“It was a good strategy. They’re ashamed if they do not behave as role models,” said one 38-year-old female from the adult group.

Most villagers had toilets in their house. Some of them installed toilets on their farm. Otherwise they would excrete in the field and cover with soil. Just a few people excreted directly in the natural water resource.

“Toilets are everywhere, nowadays. Those who can’t find one use a spade to cover it with soil,” one 55-year-old local health volunteer mentioned about how to excrete in the paddy field without toilet. “I would hold it until I get home,” said one 18-year-old from the never-infected group.

167

4.4 Discussion

Intervention Prevention Diagnosis and Treatment

B C

A G F Risk behaviours Opisthorchiasis

D E

Social influence

Figure 4.3 Conceptual framework of opisthorchiasis

Modern education H Information-based

Children

Knowledge Teenagers transfer Culture transfer

I J Adults

Elderly Traditional-belief Traditional-based

Figure 4.4 Social component integrated to conceptual framework

168

4.4.1 Social component integrated in the main conceptual framework

Figure 4.3 illustrates the comprehensive conceptual framework with regard to social components. The infection dynamics (A) have been discussed from previous Chapters (2 and 3) with rigorous statistical analysis and mathematical modelling to describe the epidemiology of OV infection and its risk factors with regard to the interventions (B and C).

As illustrated in Figure 4.4, the social component has played a major role in influencing the infection dynamics. Thematic contents provide various outcomes regarding bio-psycho-social aspects; the impact of social components to risk factors (D) and also the roles of intervention (F and G). OV infection itself impacts the population as well (E).

I have elaborated on the social context from the baseline framework (Figure 4.3) to explore the behavioural-psycho-social aspects of the infection. The main theme is based on age structure (H) where I have observed that the process of knowledge (I) and culture (J) transfer shows some interesting and unique characteristics.

Regarding the overall picture, basic knowledge of OV infection requires modern education usually provided by formal resources such as schools. Some traditional beliefs still play important roles concerning knowledge of infection. Knowledge and attitudes towards OV infection leads to the choice of practicing risk behaviours and also the perception of prevention and control. The pattern of knowledge transfer is likely to start with children and teenagers when their parents agree to let them have modern education from school (I). Younger generations can access more informative sources such as the internet, and therefore, can pass modern knowledge to the older generation, i.e., their parents and the elderly.

However, the culture transfer seems to show a confronting direction (J) where the traditional-beliefs of the local Northeastern culture still have a strong influence on attitudes and perceptions toward risk behaviours passed on from previous generations. This complex relationship shows that younger generations with modern education tend to have multiple sources of information and start to justify the risks and benefits from practicing risk behaviours.

The interaction between modern information and traditional beliefs can be observed during transitional periods among teenagers and early adults when self-esteem and independence starts to develop.

169

4.4.2 Knowledge, attitude and perception of OV infection contributed by age structure

Table 4.13 Knowledge, attitude and perception of OV infection contributed by age structure

 Knowledge was school-based o Health education and hygiene lesson were given at the school. Young o Some of the younger generation tended to have higher educational children levels than their parents; therefore, school was merely a knowledge hub for children.  Knowledge from school was also transferred to parents by children.

 Role models were important o Family o School-based: teachers and friends Teenagers o Social trend: internet and social media  Sought behaviours for information and tended to believe by evidence-based data.

 Direct evidence triggers concerns such as fatal cases from Adults Cholangiocarcinoma.  Adults hesitated to listen to their children less than non-family members.

 Age was a natural barrier to learning new knowledge. Elders  Trust was given doctors and medical personnel.

As shown in Figure 4.4, the interaction between the process of knowledge and culture transfer could describe the method of how knowledge and attitudes could contribute to the practice of consumption.

Knowledge of OV infection comprised several parts: liver fluke life cycle, mode of infection and transmission, health consequences and diagnosis and treatment. Results indicated that participants perceived information and understood the infection concerning different aspects (Table 4.13).

170

It would be impractical to gain full insights in the infection considering that other diseases need to of equal concern. However, some vital issues were important to raise concerns leading to prevent risk behaviours as described below.  Mode of infection was consumption of uncooked fish. The specific dishes were instantly prepared such as Koi pla.  The parasites grow and live in the body for many years.  An infected person may not be aware of the condition because of the asymptomatic features, but chronic infection leads to cancer of the bile duct (Cholangiocarcinoma) was and is highly fatal.  Risk of infection maybe higher when practicing risk behaviours, so avoiding uncooked fish consumption was preferred as a primary prevention.

All age groups acknowledged that the term “parasite” referred to a kind of pathogen and caused illness. Children and teenagers are formally educated about parasitic infections in school on a regular basis as core content in hygiene lessons including liver fluke infection. Additionally, OV infection was also introduced to the community by the research team. As a part of the community medicine course, medical students provide health education emphasizing OV infection in the community including community meetings, public relation campaigns and also local radio broadcasts.

School plays a major role as an information hub equipped with academic resources such as teachers, books and internet access. Therefore, the majority of health knowledge among young children and teenagers is school-based.

Teenager behaviours are also motivated by social influence; they are naturally sensitive to information with respect to physical and psychological development during their puberty. Therefore, they are more likely to follow social trends including eating habits. The modern lifestyle leads them to choose more urbanized food. However, teenagers have come to the point that they are starting to choose independently. In terms of eating habits, they might choose to consume uncooked fish when they perceived that their previous generation was a role-model. From their information seeking behaviours, they might have to adjust between knowledge they had about OV infection and how they imitated others or developed life-long eating habits.

Teenagers and adults also required evidence-based data. For instance, they asked the research team as they could be shown actual liver flukes or eggs. Data presentation from health professionals also focused their attention on the infection. 171

The elderly had barriers for learning new knowledge, especially when new knowledge conflicted with their beliefs. This issue also occurred among adult as they did not pay much attention to the messages from youth. However, data analysis showed that age barriers could be minimized. Knowledge transfer could be performed more smoothly by a family member. Adults showed less hesitation when the information was provided by their offspring. This connection provided the mechanism for the transfer of school-based knowledge to the community by school children as a medium creating a practical strategy for knowledge distribution.

4.4.3 Attitude and perception towards risk behaviours

An earlier qualitative study in Na-yao area from 2007 – 09 (70) indicated that the National Control Program strategy may have some issues regarding the prevention of uncooked fish consumption campaign (26-28). The participants found that avoiding all uncooked fish dishes was impractical as indicated by the guidelines provided, so they were more likely to continue the same eating habits (70) resulting in continued Pla ra consumption. Pla ra was widely used as a main ingredient for local dishes, so the indirect impact from unspecified uncooked fish avoidance campaign resulted in continued Koi pla consumption.

For liver fluke infection, the mode of infection was simple and straightforward. However, eating practices could result in different approaches with respect to attitudes, perceptions and socioeconomic aspects. From the study, patterns of uncooked fish consumption varied across many factors. Age group had a distinctive pattern observed from the analysis.

Children’s food was mainly prepared by their parents Therefore, the attitudes towards health risk was difficult to determine because children’s eating behaviours were strongly dependent on the parents’.

Teenagers were the first to start deciding what to eat. They were sensitive to information such as social media or TV and they were more likely to become a trend follower. Urbanization has suggested more modern food choices but eating behaviours could proceed in two ways as described below.

 They may decide to have more modern food while rejecting local dishes as outdated.  They still consumed uncooked fish dishes.

Practical strategies suggest that teenagers could try both approaches. Trial and error could be a good strategy during physical/mental development in puberty. 172

Adult represented an independent generation; therefore, they had confidence to decide independently. The underlying knowledge incorporated with personal experience could motivate the choices of what to eat or not eat. This study revealed an important practical point where participants perceived that heath risk was controllable and an occasional exception was acceptable. Strong evidence may be needed to modify existing beliefs often misunderstandings.

The lifestyle was an important factor to be more greatly exposed to health risks. Fish were caught and prepared instantly during agricultural fieldwork. Men also consumed alcohol at social events.

A relationship was established between states of dependence and independence referring to adult- and childhood; while they contributed their own decisions, most of them forbade their children from consuming uncooked fish. However, in high-risk families, all family members consumed Koi pla on a regular basis.

The elderly exhibited strong attachment to local tradition and culture. Long life expectancy has allowed some excuses seem logical and rational to them. They have been maintaining a long-term lifestyle without any health problems relating to parasitic infection, so they did not wish to change or modify anything. They wished to enjoy their remaining years with a peaceful life.

Knowledge, attitudes and perceptions across generations raised some interesting issues. Common characteristics of a particular age group indicated points for further intervention.

Table 4.14 Common characteristics regarding knowledge, attitude and perception across generations

Common characteristics

Children Teenager Adult Elder  Family tradition  Transitional period  Knowledge misunderstanding  Personal belief

Although children were mainly educated in school, they may be fed uncooked fish by their parents. Food and utensils accidentally contaminated with raw fish might also cause the infection. Additionally, some families have shown a strong preference for uncooked fish consumption affecting children’s attitudes in a long run.

As seen in Table 4.14, teenagers were in a transitional period and became more independent when selecting food. Therefore, they tended to absorb and process a great deal of information. Information overload may interfere with how they would comprehend the data. Imitation among 173

friends occasionally occurred due to biologically impulsive behaviours. Concern could arise when they chose to pursue agricultural careers contributing a higher chance to contact risk behaviours and at risk environments.

Adults and the elderly revealed a misunderstanding through knowledge and personal beliefs. Solid evidence could provide a stronger impact on their perception such as evidence of toxic waste, polluted natural water bodies and causes of illness. Moreover, some elderly believed that they have been living long enough without harmful diseases confirmed by their healthy lifestyle. Regarding cultural norms, some of them believed that serious illness was a matter of destiny or Karma.

174

4.4.4 Impact of OV infection on health, treatment and prevention

Table 4.15 Impact of OV infection on health, treatment and prevention

Young  Health was under the parent’s supervision. children  OV infection was demonstrated at school.

 Healthy lifestyle also incorporates social influence; peer pressure, social media and urbanization. Teenagers  OV infection was intimidating, the parasite was disgusting. Local uncooked fish dishes were not trendy.

 Physical health must be maintained to work every day.  Some risk behaviours were considered a reward or treat after hard work. o Alcohol consumption Adults o Uncooked fish consumption  OV infection was asymptomatic; others diseases were considered more harmful. Moreover, it did not obviously affect physical health.

 Healthcare was more likely a defensive strategy. o Doctors were responsible for their health. Elders  Their long life has proved that they have a healthy lifestyle.  OV infection was intimidating, but they couldn’t change their way of life.  They felt that some medical treatments were more harmful than the disease.

Table 4.15 describes the impact of OV infection on health, treatment and prevention across generations.

Young children and teenagers are in a growing phase of life. They were generally considered at a healthy age with energetic activity. Therefore, physical health was not a primary concern. Young children’s health is under parental supervision. Most of them were in education institutions; they were taught hygiene lessons beginning in school including basic hygienic practices generally indicating that uncooked food was unhealthy.

Teenagers are just starting to learn that pathogens can cause disease. Moreover, uncooked food was not considered trendy. Teenagers tended to pay more attention to their physical body; they became more sensitive to their image and also body transformation. After seeing actual specimens, then 175

knowing that raw fish meat contained live metacercariae that actually resided in their bodies made them feel revolted.

Adult showed more concern to physical function than general appearance. They had to maintain physical fitness to perform work almost every day. Moreover, some risk behaviours were considered rewards after a long and hard day at work such as social drinking group or consuming Koi pla considered a tasty dish along with alcohol.

Many chronic diseases begin at adult age. Some conditions such as hypertension or diabetes have asymptomatic manifestations at the early phase. OV infection shares this feature and affects health awareness. In fact, they were more likely to be concerned about what would immediately compromise their fitness such as myalgia or trauma.

The elderly thought that healthcare should be under a doctors’ supervision. Their ideas represented a passive defensive strategy where they only sought medical care after contracting illness. They also justified the efforts compared with the benefits for behavioural modification. The benefits from primary prevention may not be visible because the disease is prevented rather than when an illness is cured.

However, the issue involves self-conflicts. They also perceived that taking praziquantel as a treatment for OV infection caused more illness than being infected and remaining asymptomatic. Feeling nausea and dizziness were unfavourable outcomes.

The reason may lead to the point where they think that medical care should only involve medical equipment and healthcare workers such as seeing doctors in the hospital or receiving medicine. As a result, primary prevention represented a separate activity rather than involving a part of healthcare. Because prevention was less prioritized, treatment became a crucial part for health issues.

Lack of motivation was an important issue for both prevention and treatment. Moreover, even when they understood fatalities resulted from Cholangiocarcinoma, it could be a result from chronic infection occurring many years ago. Considering their remaining years of age, they calculated having cancer now was unlikely.

176

4.5 Conclusion

The qualitative approach including focus group discussions and in-depth interviews revealed that uncooked fish consumption was decreased in terms of quantity and frequency. However, some people still consumed uncooked fish regularly. Uncooked fish dishes such as Koi pla are still popular. From a psychosocial aspect, the dish itself represents Northeastern culture.

Population dynamics also influenced the social aspects of OV infection. Regarding the quantitative study results, the infection increased among older ages; the elderly generation recently migrated from Northeastern Thailand. Moreover, traditional beliefs were transferred from their ancestors to their descendants. The evidence could be visually observed while they still preserved their lifestyle despite the fact that they already left their origins.

Additionally, Koi pla was considered a tasty dish, which many people considered an important factor to continue consumption. However, the generation gap also revealed that the younger generation was likely to absorb modernized culture and become more urbanized.

Some elderly complained that refusing to eat local fish dishes might go against old traditions with respect to cultural norms and socialization. Higher education teaches about hygienic conditions including OV infection, so teenagers and young adults tended to modify their lifestyle including eating habits.

Therefore, health education alone would not be sufficient to motivate the behavioural changes required. More efforts are needed to support the transformation. Children represent a potential key to pass knowledge to their parents and schools are practical hubs for knowledge distribution.

From rural settings, community leaders are important keys in the social structure. As the local administration is managed by the central and regional authority, governmental staff might not be aware of local cultural norms. Community leaders could bridge this gap and help facilitate the government’s strategies.

Medical doctors and healthcare personnel are also considered community leaders looking after villagers’ health. People respect and trust doctors when they are instructed and advised regarding their health. However, doctors are not permanently stationed at local health centres. They regularly visit the community once a week mostly for chronic disease appointments. In Thailand, the shortage of human resources is still a major problem in rural areas. 177

Local health volunteers played important roles for communicating and interacting among villagers and health authorities. They could act as community leaders and as role models for health interventions. Because they are locals, they could more easily gain trust and cooperation from the villagers, a crucial part to maintain the sustainability of health campaigns in the community (70, 114, 132).

Social influence plays an important role for shaping the infection dynamics of OV infection. Within the social components (Figure 4.3), the interaction of knowledge and culture transferred across generations provided more insight to bio-psycho-social dynamics (Figure 4.4).

In Chapter 5, community intervention employing a mixed-method study as a comprehensive tool revealed the potential use of the application of quantitative and qualitative approaches with regard to the host-agent-environment aspect of the infection dynamics.

178

Chapter 5 Community-based intervention: PRECEDE-PROCEED model framework for controlling OV infection

Abstract

OV infection is a liver fluke infection caused by Opisthorchis viverrini. The important mode of infection is consumption of uncooked freshwater fish containing infective metacercariae. OV infection mainly features asymptomatic manifestations but chronic infection could lead to bile duct cancer or Cholangiocarcinoma. Over decades, the National Control Program has targeted endemic areas aiming to suppress prevalence down to 10%.

Current fieldwork conducted in 4 areas of Ta-Kadarn District located in rural central Thailand indicated that the prevalence was higher than the regional average and incidence was also high considering the studied areas were outside the endemic Northeastern Thailand. Follow-up studies have report significant association of Koi pla consumption with increased risk of OV infection and a qualitative approach has revealed the potential influence of bio-psycho-social factors contributing to the practice of Koi pla consumption related to Northeastern culture where the community founders originally moved from.

The PRECEDE-PROCEED model framework was applied to the community of Tung-heang to systematically engage the problem with the aim of reducing the incidence of OV infection through behavioural modification by decreasing Koi pla consumption. The community-based intervention aimed to reduce risk behaviours was developed using a mixed-method study design. Community trials were conducted to assess the effectiveness of the intervention on incidence of OV infection and consumption of Koi pla. The distribution and risk factors of OV infection were obtained by the cohort design and qualitative approach using up-to-date information.

Community trial was followed by an open-label, nonrandomized community trial design. The intervention group participated in the community by mutual agreement to discontinue Koi pla consumption. After the 17-month study period, the intervention group and control group were diagnosed for OV infection by stool examination and assessed for risk behaviour by questionnaire.

The results indicated that community intervention could significantly reduce incidence of OV infection by 63% (95% CI: 7 – 85), p = 0.03 and reduce Koi pla consumption by 46% (95% CI: 1 - 71), p = 0.04. For the cohort analysis, Koi pla consumption could increase risk of OV infection with IRR = 2.52 (95% CI: 1.07 – 5.92) when adjusted for age group. Additionally, social and cultural attributes were potential factors leading to Koi pla consumption. 179

Consumption behaviours were complex involving traditional beliefs, attitudes and diverse cultural backgrounds. Following the PRECEDE-PROCEED framework, community participatory action including community-derived intervention was considered adaptive and practically suited to villagers’ lifestyles. Local health volunteers were originally community members and could play an important role as a reinforcing factor.

From the fieldwork study, the mixed-method design provided comprehensive approaches to assess OV infection in various aspects. The PRECEDE-PROCEED model framework also provided a comprehensive approach for intervention/evaluation planning. The project initiative provided more insight of community management which was expanded to a larger scale in which the community intervention package could be developed and applied for public health implementation on a national scale.

180

5.1 Introduction

The prevalence of OV has been estimated across Thailand. The record from the 2001 national survey shows various distributions of infection. The prevalence was high in the North (19.3%) and the Northeast (15.7%) while the prevalence was 3.8% in the central Thailand and 0% in the South. Overall, the national prevalence was roughly 10%. The endemic areas were limited to local areas with local habitats for intermediate hosts and traditional eating habits (27, 28).

The first national survey in the 1980s revealed the prevalence of OV infection was 63.6%. Therefore, an OV control program was started from the 6th 5-year National Public Health Development Plan (1987 – 1991) aimed to control the infection as described in Section 1.4.1.2. The national prevalence declined to 9.4% in 2001. Recently, the control activity has focused on areas in the North and Northeast where the infection has remained endemic (28).

Infection dynamics of OV infection involve multiple factors. Consumption behaviours were complex activities, which could be influenced by many factors such as traditions or cultural norms (1, 5, 15, 29). The qualitative approach showed that bio-psycho-social aspects played an important role for the consumption pattern. Apart from the qualitative components, OV infection itself shows a dynamic picture of prevalence and incidence (Chapters 2 and 3). I have observed that the prevalence and incidence were not linearly associated where the incidence obtained from fieldwork was significantly higher than calculated incidence from prevalence-age linear function (Section 3.1.2 from Chapter 3). Moreover, re-infection could play an important role for a chronic infection picture and shape a sustained prevalence (Section 3.4.1 from Chapter 3).

Based on the previous situation when the national prevalence was high from 1984 – 1987; 63.6% (28), the National Control Program was designed to mainly focus on secondary prevention along with treatment of infected cases (26, 28) from1987 - 2001. In the present day, this strategy should be re-evaluated to deal with emerging cases or re-infection.

A Planning model has been proposed to systematically engage health burden (133-135). The PRECEDE-PROCEED model (PPM) framework (133, 136) is a planning model which organizes the method of examining root cause of the problem and sets up the goals for reducing the burden with the planned intervention. The PPM has been used in some parasitic infection cases such as the plan for controlling Taenia solium in Nepal (137) and school-based health promotion for helminth control in KwaZulu-Natal (138). Both studies used PPM to assess and develop community intervention with regard to social and environmental factors with consideration of feasibility and compatibility with current policy. 181

In the macroscopic level, the PPM framework was used to approach the challenges posed by OV infection transmission at the community level while reflecting the Thailand rural community perspective.

5.1.1 Overview of PRECEDE-PROCEED model

The Figure 5.1 Diagram of PRECEDE – PROCEED model

PRECEDE-PROCEED planning model was a framework for the public health planning aim to reduce disease burdens at the macroscopic level (133, 136, 139). To establish an effective intervention, many factors were considered from individual to health policy levels. Thus, the PPM was very much an ecological approach. The relationship between individual characteristics and environmental and social contexts were incorporated in the framework.

Figure 5.1 was adapted from Crosby R. (133) based on a primary framework developed by Green L.W and Kreuter, M.W (136). Basically, heath behaviors were considered complex and dynamic (140). It was challenging for public health practitioners to engage heath issues in the community. To approach complicated community-based problems, local assessment was needed to identify specific problems fitting in a particular setting. From Figure 5.1, the PRECEDE-PROCEED diagram was used to analyze factors while intervention could be planned and applied at the same time using appropriate theory for practical guidelines. Two key aspects to form a framework were planning and evaluation.

From the diagram, PPM involves 9 stages of health planning and evaluation. The framework begins with largest goal, ultimately judged by the outcome evaluation after the relevant factors were assessed and the intervention was implemented. The process was performed step by step. 182

PRECEDE stands for Predisposing, Reinforcing, and Enabling Constructs in Educational/environmental Diagnosis and Evaluation (steps 1 to 4). In general, public health issues normally start with a disease or health threat. To generate a health plan, the root cause was traced back to the source of the problems where relevant factors were assessed by a multidisciplinary approach. Health-related objectives were created to set the ultimate goals for the intervention. The objectives were usually measurable and time-limited, which was practical, to be later evaluated for the outcome.

PROCEED stands for Policy, Regulatory, and Organizational Constructs in Educational and Environmental Development (steps 5 to 9). After the planning stage was carefully assessed, an intervention based on health theory was used to modify health-related risk factor in this stage. The intervention package was implemented in the community and the evaluation process finally assessed the outcome of the intervention such as the reduction of disease burden. The vital key was the evaluation process, which monitored the intervention program to ensure the fidelity of the plan.

183

Intervention

Prevention Diagnosis and Treatment

Evaluation phase

Risk behaviours Opisthorchiasis Planning phase

Social influence

Figure 5.2 Conceptual framework for community intervention study

Based on the OV infection conceptual framework from Figure 5.2, the planning process involved assessing data regarding the infection dynamics described below.

 Social diagnosis assessed the burden of disease and the strengths and weaknesses of the community setting to engage the problem.  Epidemiological study assessed the burden the disease and determined risk factors contributing to the infection.  Administrative and policy assessment evaluated the roles involved in the current management of OV infection, which will help in guiding the design of intervention with regard to feasibility and resources.

Collected data will guide the specification of objectives to reduce disease burden and its associated risk factors. The planning process will establish community intervention for achieving the objectives. 184

The evaluation process involves the implementation of the intervention, which requires an appropriate study design. Intervention aimed to achieve the objectives involving risk factor modification to eventually reduce OV infection.

185

5.2 Methodology 5.2.1 Overview of study methods

[A] Cross-sectional conducted in 2010

[B] Study outcome was explained to the community

[C] Initiation of community intervention to reduce opisthorchiasis

[D] Planning process: PRECEDE

 [Step 1] Evaluate burden of the opisthorchiasis  [Step 2] Set key objective to reduce opisthorchiasis  [Step 3] Set sub-objective to reduce risk behaviour contributing to opisthorchiasis  [Step 4] Design of community intervention to achieve objective and sub-objective

[E] Evaluation process: PROCEED

 [Step 5] Evaluate opisthorchiasis-related policy  [Step 6] Implement community intervention  [Step 7] Evaluate the flow of study  [Step 8] Evaluate the outcome of sub-objective  [Step 9] Evaluate the final outcome of objective

[F] Inform and discuss the study result to the community

Figure 5.3 Summary of study flowchart 186

Figure 5.3 summarizes the entire process of this Chapter. The study was conducted in Tung-heang Area. In 2010, the baseline survey was performed for OV infection for which prevalence was 11.42% (95% CI: 8.69 – 14.63) [A].

After the baseline survey, the fieldwork team provided information to the community [B]. The community perceived the burden of the problem and realized this was a priority issue for the community, which initiated the idea of reducing OV infection by community-based solutions.

Members of Tung-heang agreed to initiate the community-based intervention [C]. The whole Tung- heang Area also agreed to join a follow-up study for evaluating the incidence of OV infection.

Tung-heang community members brainstormed for a method of controlling OV infection based on behavioural modification by mutual agreement. The fieldwork team agreed to participate in this project by providing diagnosis at the end of follow-up time and provide data management to evaluate the effectiveness of the intervention [E]. In the final process, the research team analyse data and transfer information to the community [F].

The PRECEDE-PROCEED model was used to holistically approach the problem, the process of intervention and study outcome. A planning model was created following the key steps. The pilot community-based intervention was then initiated.

5.2.2 PRECEDE – PROCEED model establishment

The study flow followed the 9 steps in the PPM framework as presented in Figures 5.3 [D] and [E].

PRECEDE (planning phase)

5.2.2.1 Step 1 Social diagnosis

This step investigated the burden of OV infection, strengths and weaknesses of the community setting and community acknowledgement of the health issues.

Initially, the baseline prevalence from the study survey in 2010 was provided to the community. Additionally, risk factors obtained from adjacent study areas; published data (69, 70) and fieldwork data (Chapter 2), indicated that Koi pla consumption was a major risk factor for acquiring the infection.

The community perceived the burden of the problem and realized this was a priority issue. Social diagnosis was a description of community characteristics and a relationship with the targeted disease; OV infection. Social contexts were evaluated by qualitative approach including focus group 187

discussion and in-depth interview. Therefore, the qualitative approach was used to gain social- related data (129).

These key questions were provided and discussed to assess the impact of the infection.

1. How do community members acknowledge OV infection?  Basic knowledge of OV infection; risks factors, modes of transmission, diagnosis and treatment  Consequences of chronic infection; role of carcinogens, Cholangiocarcinoma 2. Attitudes toward the infection and its subsequent outcome  Perceptions of becoming infected  Burden of problem in psycho-social aspect. 3. Uncooked fish eating behaviours  Current situation of eating habit  Precipitating factors to promote uncooked fish consumption  Precipitating factors to discontinue uncooked fish consumption 4. How does the community react to the problem?  Community concern of the problem  How will they prioritize the situation?  How will they manage a solution? In this circumstance, community leaders played an important role as a respectful key person and also acted as the coordinators between the research team and community members.

Community leaders comprised of administrative and spiritual leaders such as head of the village, local intellectuals and monks. Local governmental officers who work in the area also served as key persons organizing governmental-related affairs in the community. For healthcare aspects, community leaders worked as healthcare personnel at the local health facility and as local heath volunteers.

The Community Advisory Board (CAB) was a key part of the entire PPM framework. The members of the CAB were chosen from community leaders and community members. As a part of community participation, the CAB coordinated communication with the community members following the PPM framework: gathering and exchanging information. Therefore, the CAB was an integral part of the entire PPM and the maintenance of the CAB was required to coordinate communication between community members and the research team.

188

5.2.2.2 Step 2 Epidemiological diagnosis

This step emphasized creating the objective for final outcomes of the program. In general, the aim was the reduction of disease burden or increased rate of health promotion campaign. The objective was usually measurable and time-limited and was later evaluated. Additionally, available field data aided the planner to set the ultimate goal for the intervention program.

The primary objective was to reduce incidence of OV infection by the end of study with a 17-month follow-up period.

5.2.2.3 Step 3 Behavioural and environmental diagnosis

This step identified potential behavioural and environmental factors influencing the burden of disease. These factors may increase risks for acquiring disease or prevent effective prevention and control.

Then the identified factors were applied as sub-objectives directly leading to planning for the intervention. The sub-objectives were also set as measurable goals for example, reducing the prevalence of risk behaviours contributing to the infection or modifying some environmental factors which naturally supported the pathogen. From the framework, the behavioural and environmental modification ultimately reduced the disease burden.

The sub-objective was to reduce risk factors for acquiring OV infection, i.e., consumption of Koi pla, by the end of study with 17-month follow-up period.

5.2.2.4 Step 4 Educational and ecological diagnosis

In this step, an intervention plan was developed to achieve the sub-objectives proposed in step 3. This was a vital step to manage either behavioural or environmental modification through a health intervention.

The PPM framework also provided a systemic strategy to assist the planner to develop an intervention plan. Each sub-objective required a separate plan to fulfil each goal. At this point, three subsequent factors were considered (133, 136).

Predisposing factors

Predisposing factors were originated at the cognitive level through attitudes and beliefs. From individual to community levels, perception, knowledge and health beliefs toward the disease were important factors shaping attitudes for health concerns. They could be derived from either internal 189

or external sources, for example, passing traditional beliefs from the previous generation or health education provided by health authorities. In general, modification of predisposing factors was achieved by educational campaigns.

Reinforcing factors

According to the framework, the planned intervention should contribute to behavioural or environmental modification. After intervention was applied, reinforcing factors assured the sustainability of the program effectiveness. This was a crucial step as normally public health interventions were only performed in a point or period of time. Repeating desired behaviours as routine is a challenge for health planners to maintain the sustainability of the intervention program in the community.

Enabling factors

Enabling factors make it easier or possible for the behaviour to occur such as required skills to perform the intervention, and an affordable price for the medical campaign or even convenient transportation to the health facility. More support will provide more cooperation from the community. Skill-related enabling factors and environmental factors should be concerned to develop suitable community-based intervention.

Behavioural science theory was widely used to support the intervention planning. The PPM framework was not attached to any specific theory. In fact, each designated intervention had different objectives. Therefore, health theory was selected based on the proposed intervention developed in this step.

190

The role of behavioural science theory in the developing of community intervention

In general, health intervention was perceived as an action aimed to reduce health burden (141). Many studies suggest that health intervention will be more effective if the design is based on theoretical foundation. Many theories have been used to construct health interventions. Therefore, the development and implementation of the behavioural science theory should be based on the conceptual framework and objectives of the health program. The core fundamental concepts regarding the success of intervention was to comprehensively understand the risk behaviour and its context (128, 139, 140, 142).

From steps 2 and 3, objectives and sub-objective, goals were set to reduce incidence of OV infection by modifying one of the potential risk factors; consumption of Koi pla. Two theories served as an underlying framework for the community intervention development.

Health belief model (HBM)

The HBM is widely used in public health programs (143, 144) as a core principle of how an individual reacts to a health problem and justifies the action to reduce the risk of a given problem (140). The HBM explains that the individuals will justify the risk behaviour once they perceive appropriate information on the severity of the health problem and the feasibility and obstacles to modify risk behaviour (128). The theory emphasizes self-efficacy where the individuals will recognize and modify their risk behaviours by themselves.

The HBM was effective against diseases with long asymptomatic periods requiring the accumulation of risk behaviour such as the study of reducing risk factors for cardiovascular disease (145). The intervention developed under the HMB framework aims to reduce the health risk to reduce or prevent the subsequent disease as a primary prevention.

Therefore, the concept was consistent with the OV infection framework where the risk behaviour was a major key contributing to the infection. Moreover, OV infection presents mainly an asymptomatic period with silent development of subsequent Cholangiocarcinoma. Information about OV infection including prevalence and incidence, consequence of chronic infection and results from fieldwork study indicated the risk from consuming uncooked fish were transferred to the community.

Self-efficacy comes from the individual to community levels. As each member perceives the threat of OV infection, they might still have the problem regarding the method of behavioural modification. 191

Then community perception and awareness plays an important role as it ensures the community members that the goal for handling OV infection problem will be carried on in the same direction.

Social cognitive theory (SCT)

From the framework, SCT expanded the behavioural modification regarding self-efficacy factors by proposing that personal factors also interacted with environmental factors (128). Health behaviour was not only influenced by personal experience but also by observing other actions, which was called the reciprocal model (146).

Community intervention developed by the members of community themselves can play a major role as the activity derives from the collective opinions and are easier and more accessible to the local norm. Self-efficacy was still a major concept for SCT with additional concerns of social perspective to push the intervention forward as a whole community.

More efforts were provided to the intervention planning to increase self-efficacy by three strategies described below.

 Setting precise and achievable objectives; clear targets can be more easily approached. Avoiding confusion with simple objectives was more appropriate to engage the problem in a larger scale such as the community level. Problems cannot be solved in every respect at once. Therefore, the objectives from steps 2 and 3 were clear for reducing the incidence of OV infection by reducing Koi pla consumption.  Applying specific instructions on behavioural modification, which was more practical to conduct within the same direction. Health education was also provided to the community to build up self-efficacy to engage in the intervention.  Reinforcement: the factor contributing the continuity and sustainability of behavioural modification. In general, monitoring system were an effective method for sustaining individual awareness to the ongoing intervention.

192

Initiation of community intervention

From the PPM framework, the intervention was initiated on the basis of community participation with the SCT strategy. Following the HBM, community was concerned about the problem and agreed to take action to lower OV infection incidence emphasizing Koi pla consumption. The intervention was developed based on community mutual agreement (Table 5.1). The research team agreed to evaluate the intervention outcome by providing diagnosis of OV infection at the end of study as a result of behavioural modification and observed the intervention process along with the CAB.

Table 5.1 Community mutual agreement

Community mutual agreement • Joining agreement was voluntary. After being informed, participants agreed to sign in consent form to confirm their intention. The participants must be 18 years of age and over. • Participants should follow guidelines to avoid risk of acquiring OV infection: o Targeted strategy emphasize on Koi pla consumption (risk was identified through statistical modelling and qualitative approach). o Avoid eating Koi pla (chopped raw fish salad) o Maintain good hygiene of the kitchen to prevent food contaminated with raw fish scraps. • Defecate in hygienic toilets to prevent transmission. • Following the programme schedule, participants should send their stool to fieldwork team for OV diagnosis at the end of study.

The intervention group was occasionally reminded by CAB as a monitoring system to promote the sustainability of intervention.

Those who did not participate in the intervention group were invited to join the study to provide a stool specimen to evaluate the incidence as a control group and risk factor analysis. Once a stool examination diagnostic result was available, it was delivered to participants individually. At this step, it was their decision whether to provide their own result to other community members. The fieldwork team only took part in overall result analysis and did not participate in any subsequent activity from the agreement. Incidence of OV infection in the intervention group was compared with the control group to assess effectiveness.

193

PROCEED (evaluation phase)

5.2.2.5 Step 5 Administrative and policy assessment

This was the most challenging step in the entire process, which assessed the available resources to implement the intervention in the community. Careful and informative assessment will aid the successfulness of the intervention establishment.

The objectives of the assessment can be divided into two categories explained below.

Health education

Health education serves as a long-standing scheme in public health campaigns. Basically, health education efforts will focus on predisposing and reinforcing factors. Knowledge could be widely provided to the community through various sources. Social norms could be reshaped to serve as a method of reinforcement for healthy practices. For enabling factors, better health knowledge also enables the community to improve the performance for performing the intervention.

Policy, regulation and organisation structures

The health policy involves all aspects of healthcare serving as an infrastructure to the community. The medical facility was the primary source for the community members to seek for healthcare. Thus, the role of existing health policy needed to be assessed to achieve the research goal.

An intervention plan might be already compatible with the current health policy. Otherwise, the planner can consider altering the plan or modifying the policy to finally suit the feasibility and fidelity of the program. An assessment can also aid the planner to seek political support to aid the policy establishment.

The focus group discussions and interviews with participants provided information about the health policy from the local healthcare administrators in the community.

Qualitative data could provide two-way opinions from the healthcare worker viewpoint who manage to follow National guidelines for OV infection and the community’s expression to the healthcare roles. It could also assess the feasibility of the policy to the intervention. Furthermore, the health policy should support when the intervention needs to be self-sustained and solely managed by the community in the future.

194

5.2.2.6 Step 6 Implementation

This step was one of the crucial steps for the planning model. To achieve the goal of the intervention, planners should consider the role of program evaluation. While the health program was running through timeframe in the population, the evaluation plan should be also assessed from the beginning to evaluate all important aspects of the program comprising three parts: process evaluation, impact evaluation and outcome evaluation.

419 Baseline population from 2010 cross-sectional study with negative result for opisthorchiasis

357 Age ≥ 18 years 62 Age < 18 years

[A] Enrolled to community intervention study

Not consented to intervention study

Subject allocation

[B] Enrolled for study of risk factors

Intervention group Control group

Figure 5.5 Figure 5.6

Figure 5.4 Summary of fieldwork flowchart

195

As shown in Figure 5.4, negative cases from the baseline survey were invited to join the follow-up study in 2012 to measure the incidence of OV infection. The eligible subjects were enrolled i the community-based intervention program. Based on 2010 baseline survey, total eligible subjects for a follow-up study totaled 419. The studies were divided in community intervention [A] and risk factors studies [B].

The whole study was conducted in Tung-heang Area. Individuals with 18 years of age and over were enrolled in the community intervention study and then allocated to intervention and control groups. The remaining subjects from the intervention group including ≥18 years-old control group, those who did not consent to community intervention and the remaining <18 years-old individuals were enrolled to provide stool specimens and questionnaires for the risk factors study.

The community intervention followed the protocols shown in Figure 5.5 and risk factors study was undertaken as described in Figure 5.6.

Both community intervention and risk factors studies had the same end-point at 17 months of follow-up period. The demographic data and risk behaviours were assessed through questionnaires. The additional qualitative study was used to provide in-depth information regarding bio-psycho- social aspects.

5.2.2.7 Fieldwork methodology

The study aimed to evaluate the community intervention based on fieldwork data and explore disease epidemiology using multimodality approaches. The study design was a mixed-method design described below.

 The quantitative study aimed to use statistical analysis to capture epidemiological data. o Open-label, nonrandomized community trial for community intervention study o Retrospective cohort study for distribution and risk factors of OV infection  The qualitative study aimed to explain the cause of risk factors and evaluate the impact of community-based intervention in bio-psycho-social aspects.

196

Mixed-method approach

 Eligible subjects were invited to participate in the study. The inclusion criteria are stated below.  Resided in Tung-heang Village  Participated in the epidemiological survey in 2010 where OV infection was diagnosed by stool examination and negative for OV infection  Consented to participate in the study project  Participants in intervention group were ≥18 years of age. Excluded participants from this criterion were still eligible for the risk factors study.  Able to communicate with the fieldwork team in answering the questionnaire or joining the discussion groups.  From October 2010, participants were divided in two study groups: intervention and control. The intervention group formed a community-based intervention called a community mutual agreement; they volunteered to avoid risk behaviours aimed to reduce the incidence of OV infection.  The follow-up time was 17 months and the study finished in March 2012. The fieldwork team evaluated the intervention outcome by measuring incidence of infection compared with intervention with the control group.  Epidemiological data were assessed to explore the situation of OV infection along with updating disease distribution and risk factors.  Social contexts were evaluated by qualitative approaches, i.e., focus group discussions and in- depth interviews.

197

419 subjects with negative result for opisthorchiasis from 2010 cross-sectional study

Excluded subjects for

 Age < 18 years  Not consented

Eligible subjects allocated to the study group

Subjects allocated to intervention group Subjects allocated to control group

Follow-up at 17 months

Completed study

 Excluded for incomplete questionnaire  Final analysis

Figure 5.5 Community intervention flowchart

As shown in Figure 5.5, the study design was an open-label, nonrandomized community trial. Enrolled subjects were at least 18 years of age. Enrolled subjects were allocated in the intervention and control groups. Based on community mutual agreement protocols, the participants in the intervention group avoided eating Koi pla. Therefore, the allocation process was not randomized because consenting participants volunteered for the behavioural modification.

The intervention group followed protocols from the community mutual agreement (Table 5.1). The intervention group were reminded about the intervention protocols individually every 1 – 2 months by the CAB. No records were made with regard to their consumption behaviours in both study 198

groups. The study outcome was assessed through stool examination and questionnaires for the incidence of OV infection and risk factors after 17 months of follow-up period.

419 subjects with negative result for opisthorchiasis from 2010 cross-sectional study

Excluded subjects for

 Age ≥ 18 years and allocated to intervention group from intervention study  Not consented to risk factor study

Enrolled subjects for risk factor study

 Age < 18 years  Age ≥ 18 years and allocated to control group from intervention study  Excluded from intervention study

Follow-up at 17 months

Completed study

 Excluded for incomplete questionnaire

 Final analysis

Figure 5.6 Distribution and risk factors study flowchart

As shown in Figure 5.6, the study of distribution and risk factors of OV infection followed the retrospective cohort design. From 419 baseline subjects, participants enrolled in the intervention group in community trial were excluded.

Enrolled subjects for the study included subjects age <18 years. Subjects with age ≥18 years were allocated to the control group from community trial, and those who were excluded from the community trial study. 199

After 17 months of follow-up, study participants provided stool specimen to diagnose OV infection and complete questionnaires to assess risk behaviours regarding the infection within the follow-up period.

Outcome evaluation

Based on standard fieldwork protocols explained in Section 2.2.2 from Chapter 2, incidence of OV infection was estimated from stool examination data. Demographic data and risk factors were collected by standardized questionnaires.

Stool collection and processing

Based on Section 2.2.4 protocols, each participant received a survey package including questionnaire and a sealable container for collecting a stool specimen at the end of follow-up time

Each participant received their personal stool examination result; reporting any detected intestinal pathogenic parasite infections such as helminth and protozoa infection, and health education based on National Health guidelines. Infected participants received proper medical treatment according to their pathogen(s).

Questionnaires

Each participant was required to complete a standardized questionnaire to assess basic demographic data and risk behaviours for acquiring the infection. The questionnaire had been validated and used in earlier studies in the same study area. Basic demographic data included sex, age and occupation. Risk behaviours were evaluated by assessing uncooked-fish eating behaviour patterns (69, 70, 98). Uncooked fish preparations were classified in two major groups; Koi pla and Pla ra as described in Section 2.2.3

Data analysis

Each participant was matched to questionnaire and specimen using a code-embedded identification system. Stool examiners were blinded to the individual who provided the specimens. The data management unit matched all corresponding data (questionnaire and stool examination results) for data analysis. Incomplete questionnaires with missing data on relevant exposures; sex, age, occupation and consumption behaviours were excluded to perform risk factor analysis. 200

Descriptive statistics were used to describe data; outcome was reported as frequency and percentage. Statistical analysis was reported with a significance level of 0.05 and 95% confidence interval.

For the community trial study, the results were based on two objectives illustrated below.

 Primary objective: to reduce the incidence of OV infection by the end of study with 17 months follow-up. Incidence rate of OV infection between intervention and control group were compared using incidence rate ratio. p – value <0.05 indicating a significant reduction of OV infection from the intervention group.  Sub-objective: to reduce of Koi pla consumption by the end of study with 17 months of follow- up. Incidence of Koi pla consumption between intervention and control groups was compared using relative risk. p – value <0.05 indicating a significant reduction of Koi pla consumption from intervention group.

For the distribution and risk factors study, distribution of OV infection was reported as incidence rate (/100 person-years). Univariable analysis was initially performed to examine the association of OV infection of each exposure of interest and providing crude association. Exposures included: sex, age, occupation and uncooked fish consumption behaviours. The association between univariable risk factors and the infection was quantified using Pearson’s chi-square. Multivariable analysis was performed using regression modelling to adjust risk factors. Risk factors were adjusted for age based on previous works (69, 70). Factors with p value <0.2 from univariable analysis were also considered for inclusion in the multivariable model. Instant fish preparation (Koi pla) consumption was recognized as risk factor from previous works in this project (69, 70). However, extensively fermented fish (Pla ra) were considered to be included in the model to control for confounders based on the data that they were also consumed on a regular basis. Poisson’s regression models were developed to study incidence outcome with risk factors and with results reported as incidence rate ratios.

201

Qualitative study

Research participants were invited to a group conference or interview in a focus group discussion, and in-depth interviews. The qualitative study design was based on the study protocols from Section 4.2 in Chapter 4 comprising group (Section 4.2.1) and in-depth interviews (Section 4.2.2). The qualitative approach aimed to achieve the following objectives.

 Evaluate the impact of community-based intervention

 Evaluate the risk-behaviour relationship of OV infection with respect to bio-psycho-social aspects

The impact of community intervention was assessed through the intervention group. The risk- behaviours’ relationships were evaluated from participants in the distribution and risk factors study.

The interview sessions were flexible according to the number of participants, schedule and environment. The method of participant selection was purposive sampling.

Table 5.2 Selection criteria for focus group discussion

1 2 3 4 5 Koi pla consumption Baseline study 2010 + + - - CAB Follow-up study 2012 - + - +

As shown in Table 5.2, the selection criteria were based on patterns of Koi pla consumption. The pattern was assessed from the 2010 baseline study whether the subjects reported consuming Koi pla or not, and then the 2012 follow-up study when they still reported consuming Koi pla or not. Group 5 involved the CAB which was evaluated for additional administrative and policy assessment from Section 5.2.2.5. The moderator organized the meeting and managed the questions about the relationship of uncooked fish consumption behaviour and the community intervention program. The themes of questions are described in Section 5.2.2.1 (step 1 for social evaluation).

202

Data processing

The recorded conversations from each session were transcribed to text. Then text-based data were sorted and coded according to categories. Repetitive quotes, expression or statements were also summarized in frequency. Data were revised, organized and summarized for analysis. The methods included content analysis, direct quotations and selected words to consider actual local words used by the participants.

Analysis approach included thematic content analysis, framework analysis and grounded theory. The analysis process is described in Section 4.2.3 from Chapter 4.

Ethical considerations

The project was approved by the Ethics Committee of the Medical Department of the Royal Thai Army. Informed consent was obtained from enrolled participants or parents of young participants following standard protocols. Those who were positive for intestinal parasitic infection received appropriate anthelmintic treatment.

203

5.2.2.8 Step 7 Process evaluation

This step was to monitor the flow of the intervention. During the course of the health program, the protocols might not have been accurately followed. Therefore, the process evaluation was acquired to ensure the fidelity of the program blueprints. The process included continuous monitoring and correction which was conducted when the program was nearly completed.

Table 5.3 Review of study process and method of assessment

Step Task Method of assessment 1 Social diagnosis  Data from baseline cross-sectional study  Qualitative approach 2 Epidemiological diagnosis Data from baseline cross-sectional study PRECEDE 3 Behavioural and Data from baseline cross-sectional study (Planning phase) environmental diagnosis 4 Educational and ecological  Qualitative approach diagnosis  Community intervention establishment 5 Administrative and policy  Interview with community advisory board assessment  Review of literatures PROCEED 6 Implementation Summary of study process (Evaluation phase) 7 Process evaluation Summary of study response 8 Impact evaluation Follow-up fieldwork data 9 Outcome evaluation Follow-up fieldwork data

As shown in Table 5.3, the process was monitored by the CAB. The research team observed the process evaluation by occasionally attending their monthly meetings. At the end of study, the research team provided stool examinations for the study participants and conducted analysis. The process of public relations and specimen collection require the cooperation from the CAB.

5.2.2.9 Step 8 Impact evaluation

This step assessed goals from the behavioural and environmental sub-objectives developed in step 3. Impact evaluation determined immediate outcome or risk factors of the disease burden. For example, the planner could rapidly measure the reduction of health-risk behaviours along the implementation of the program or notice that some environmental factors had been modified to decrease health burden. While the planner could instantly measure the success of these sub- objectives outcome, the disease burden might need a longer time to show the effect from the 204

intervention. From the framework, risk factors were controlled, in other words, the behavioural and environmental modifications theoretically reduced the disease burden.

From social diagnosis in step 1 (Section 5.2.2.1), Koi pla was identified as a potential risk factor for OV infection. Therefore reducing consumption was assumed to reduce the incidence of the infection.

5.2.2.10 Step 9 Outcome evaluation

This step measured the ultimate goal of the program by the objectives created in step 2 (Section 5.2.2.2). The final outcome was usually a certain health burden.

However, the success of impact evaluation in step 8 does not definitely guarantee a successful outcome, or at least show immediate outcome in the program timeframe. In fact, it was more likely to reduce the disease burden according to theory. Planners should also consider other factors which could unavoidably interfere with the study’s proposed risk-outcome relationship such as genetic variability within the population, socioeconomic factors, pre-existing underlying health conditions or spatial factors. Moreover, some disease burdens need long-term intervention. Therefore, the sustainability of the program was another key factor considered.

Thus, the success of outcome evaluation was the ultimate achievement for the health program. However, the planner could evaluate its performance in various aspects. For example, a decline in risk behaviours or health-related risk environments was an immediate and measurable outcome while health burden reduction could be a long-term endpoint.

The final outcome was measuring the incidence of OV infection in 2012 at the end of intervention program and comparing with the intervention group with the control group.

Additionally, the intervention was also evaluated by a qualitative approach. The stakeholders were the subjects who participated in the intervention, which directly impacted their lifestyle and the community leaders who were influential to community policy and villager’s mentality.

205

5.3 Results

Results from PRECEDE-PROCEED model process

The community intervention program was processed through the PPM framework which followed key steps.

5.3.1 Step 1 Social diagnosis

Epidemiological data were successfully communicated to the community; previous baseline prevalence of OV infection and its risk factors, leading to the establishment of community-based intervention. Additional data were obtained by qualitative approach indicating social context of infection in the community.

5.3.1.1 Demographic data

Figure 5.7 Population pyramid of Tung-heang 2012 data

As illustrated Figure 5.7, the population pyramid revealed the structure of Tung-heang population based on age-class and sex. The 2012 data refer to follow-up population from the 2010 baseline study; therefore, the pyramid reflected only the population with negative results for OV infection from the previous prevalence survey. The structure shows a similar pattern to other study areas from Figure 2.6 in Chapter 2. Age-class 20 – 29 years was relatively constricted compared with other age-classes. 206

Additionally, the age structure of this follow-up population needed to be compared with the baseline population because the individuals identified as infected at baseline were excluded.

The population from the 2012 data was allocated to the two studies; community intervention and risk factor. The subpopulation structure will be presented in the final section.

Table 5.4 Population characteristics of Tung-heang area

Characteristic N % Sex Female 180 54.7 Male 149 45.3 Age profile Mean ± S.D. 39.3 ± 21.3 Median ± IQR 42.0 ± 37.0 Min – max 2 – 88 Age group (in years, as quartiles) 0 – 19 77 25.8 20 – 42 73 24.4 43 – 56 77 25.8 ≥ 57 72 24.1 Occupation Unemployed 79 24.0 Others 33 10.0 Agriculture 193 58.7

As shown in Table 5.4 shows that the study population comprised more females than males. The mean age was 39.3 ± 21.3 years and median age was 42.0 ± 37.0 years. The most common occupation was agriculture-related work.

5.3.1.2 Characteristics of the study populations from qualitative approach

The mean age of the focus group discussion participants was 28 years for females (range: 22 - 65 years) and 38 years for males (range: 30-58 years). All were literate. Most had primary education and the highest education they obtained was junior high school. The majority of discussion participants’ occupations were agriculture.

207

From the designated groups according to Table 5.2, no subjects were found to meet criterion 4; new Koi pla consumption subjects. The results were categorized in two main domains shown below.

1. Precipitating factors to promote Koi pla consumption

2. Precipitating factors to discontinue Koi pla consumption 5.3.1.3 Precipitating factors for promoting uncooked fish consumption

Knowledge of OV infection

At first, a misunderstanding was observed about the threat of eating uncooked fish. Koi pla has been recognized as a popular dish for a long time. They were still confused about the link between eating raw fish and becoming infected with OV infection. In fact, the majority of infected cases were asymptomatic. Their ancestors consumed Koi pla for long time and they looked healthy. Therefore, they had no reason to go against what their ancestors did as evidenced by the following statements.

“They didn’t teach about it in school, I just followed the elders. They said it’s safe”

“They saw them (the elders) as a model; they seemed alright, so should I”

Some of them were still confused about the source of infection. They were not sure which of the foods they ate caused liver fluke infection. One of the villagers unconvincingly stated, “It (liver fluke eggs) may be in shrimps and I ate chopped raw shrimps salad so I got infected. Water vegetables in the pond can also have liver fluke eggs.”

Some of them were not sure about the route of transmission. They thought that eating any raw food, skin penetration or drinking water can cause the infection. The following quotes serve as examples.

“The liver flukes penetrate my skin by not wearing shoes.”

“I can get infected by drinking water.”

“Any type of raw food has liver flukes.”

208

Cultural context

They insisted that the Northeastern people like to eat raw food. It was their lifestyle to eat raw fish, which was very hard to change. Some of them insisted that this tradition was acceptable.

“I think environment forces us to eat this way. I eat whatever my parents do. It’s inevitable.”

“It’s more than traditional; it’s a part of my lives,”

“It’s my living style that I eat simple dishes in the rice field. Doctors wake up and cook something to eat. Farmers wake up and go to the field and find something to eat.”

Fish flavour

In the harvest season (dry season), Koi pla was consumed more than other seasons due to the fact that cyprinoid fish in this season were more delicious. One of infected male insisted, “In the harvest season, fish was delicious.”

Even though they knew about the threat of eating Koi pla, they just could not deny that they enjoyed eating it. Some of them voiced the following.

“It was delicious, just simple like that.”

“Better than cooked (laughing)”

“I knew what could happen, but if I die I won’t have a chance to eat it, right?”

Socialisation

Children and adult females chose to consume Koi pla less often than adult males. Usually, adult females cooked food at home and families had their meals together. Although females did not prepare uncooked food at home, males who wanted to eat uncooked fish would prepare it themselves. Koi pla was not consumed daily. However, it was consumed more frequently when alcohol was consumed, as evidenced by the following quotes.

“When you drink, it would be a great side dish” “When I got drunk, I just let myself go.”

Some of them believed that mixing spirit beverage (40% alcohol) and lime in the raw fish dish could kill the liver flukes but some of them, especially females, disagreed.

209

5.3.1.4 Precipitating factors to discontinue Koi pla consumption

Health education campaign

They were asked where they could obtain the information about liver flukes. Some of them thankfully expressed the following quotes.

“I just knew about liver flukes when the doctor (fieldwork team) came in 2010.”

“Doctors teach us; they showed us scary pictures (of cholangiocarcinoma patient). What would happen to me if I get that?”

Impact of OV infection to the community

After the community was educated concerning OV infection, some of them learned that several patients in the village suffered from cholangiocarcinoma. They expressed the following.

“My father passed away with this disease. He asked me for the last time, eat cooked fish.” “I saw him suffering from cancer, it was really painful”

Emotional effect

Fear of death seemed to be an important factor for them to initiate health concerns and the mutual agreement as they mentioned below.

“It’s so scary” “OV was scaring me, and the agreement keeps reminding me what I should do.” “Nobody forced me to do it (agreement), but it’s about time to change.”

5.3.2 Step 2 Epidemiological diagnosis

Based on primary objective, the study period was 17 months starting in October 2011 and finished in March 2013. The outcome will be reported in step 9 (Section 5.3.9) as outcome evaluation.

5.3.3 Step 3 Behavioural and environmental diagnosis

Up-to-date incidence and risk factors were estimated from the distribution and risk factors study. The impact of intervention on Koi pla consumption were reported in step 8 (Section 5.3.8) as impact evaluation.

210

5.3.4 Step 4 Educational and ecological diagnosis

To achieve behavioural modification and reduce OV infection, three potential factors were explored to enhance strategic plans.

Predisposing factors:

The current National Control Program, which has been underway for almost 30 years, focused on prevalence reduction. The main strategies for the control program included stool examination and treatment of infected cases, a health education campaign to promote cooked fish consumption and improved hygienic defecation through environmental sanitation development. The strategy already emphasized knowledge to control the infection.

Enabling factors:

When the Nation Control Program was still active (before 2001), the strategy was active case detection and treatment. After reaching the goal (prevalence <10%), the service was switched to a passive strategy.

However, the disease is mainly asymptomatic. Some of cases would suffer from bile duct cancer from chronic infection. An active strategy was still needed for communities with endemic infection, but a program similar to the National Control Program required too much human resources and financial support. However, preventing OV infection would prevent bile duct cancer. Therefore, intervention in prevention and/or treatment form was mandatory.

Reinforcing factors:

The study aimed to identify reinforcing factors which encouraged infected communities to maintain safe eating habits on the basis that knowledge about OV infection was available (predisposing factors), and also health facilities were accessible (enabling factors).

Along with community perceptions, the fieldwork team urged Tung-heang villagers to think over the burden of the problem. They agreed to initiate the participatory agreement control risk behaviours, which would result in decreased incidence of OV infection.

The community-derived solution might be an important reinforcing factor for the consistency of the program in the community.

211

5.3.5 Step 5 Administrative and policy assessment

The national OV infection control program began in 1950 with small groups of high risk areas. After many studies and trials, large-scale control began from 1984-1987 with nationwide prevalence as high as 63.6% (28). The control activities involved mobile stool examination teams, promotion of cooked fish consumption and hygienic defecation, and finally, mass anthelminthic chemotherapy. In 2001, the prevalence was 9.6% (27, 28). Recently, control programs were targeted rather than national in scale. The activities were prioritized by the degree of the problem in each area. Cross sectional studies could not identify the newly infected cases so the effectiveness of cooked-fish consumption promotion campaigns could not be clearly evaluated.

As mentioned before, programs similar to the National Control Program required too much human resources and financial support for the current passive strategy. Although health expenses were fully covered by the National Universal Coverage welfare, the local health authority was not equipped with laboratory and medication to handle OV infection. The regional secondary health care, the district hospital, equipped to provide full support, was 60 km away.

Multiple studies conducted in Tarkradarn Subdistrict including Na-yao and Tung-heang Areas contributed field data to gain insights into community-based situations. From understanding the relationship between host, agent and environment of OV infection, community-based intervention was designed and initiated. Relevant studies in this area indicated the role of primary prevention. Focus group discussions also gained information directly from the community members. Access to health facility

The local health centre could not provide any diagnostic services for the villagers. The community members also were not concerned with OV infection because they presented no symptoms.

Praziquantel, the drug of choice, was not available in the local health centres and local pharmacies. Seeking for treatment was difficult. They expressed the following opinions.

“There was no liver fluke drug in the market; there was only general anthelmintic drug.”

212

Attitude toward diagnosis and treatment

The community approach of the fieldwork team provided them more accessibility to diagnosis and treatment of OV infection. Some of them realized the threat and urged others to avoid contracting the infection, as evidenced below.

“It must be really important, that’s why doctors come and do lots of work for helping us, and I should do something about this.”

“Kids love doctors (medical students), they keep reminding us what doctors have said”

They also knew that praziquantel was effective against the infection so they did not question taking the medicine.

Role of community leaders

In the rural setting, a community leader is a vital key to drive and influence the community’s activity. It could be either an administrative or spiritual leader; head of village, monk or spiritual leader. The discussion sessions approached community leaders as important stakeholders to drive the OV control strategy forward; as discussed below.

“The reduction of OV infection (idea of community intervention) Yes, I were talking about it every monthly meeting” “This was a very important problem, it needs immediate response” “I always told them (the villagers), if I cannot make it, how could I say anything (to the doctors; fieldwork team)”

213

5.3.6 Step 6 Implementation

419 Baseline population from 2010 cross-sectional study with negative result for opisthorchiasis

357 Age ≥ 18 years 62 Age < 18 years

295 Enrolled to community intervention study 62 Not consented to intervention study

Subject allocation 304 Enrolled for study of risk factors

115 Intervention group 180 Control group

Figure 5.9 Figure 5.10

Figure 5.8 Implementation of fieldwork

Figure 5.8 summarizes the fieldwork under the PPM framework. From 419 eligible subjects with negative results of OV infection from the 2010 study, 357 subjects were ≥18 years of age and eligible to enrol in the community trial study. Enrolled subjects were 295 which 62 subjects did not consent to participate in the community trial. After allocation, 115 subjects allocated in the intervention group and 180 subjects belonged to the control group. The disproportion between intervention and control groups was the result of the enrolment process when joining the intervention was voluntary. 214

A total of 62 subjects with age <18 were enrolled in the distribution and risk factors study. The control group from the community trial; that did not receive any intervention, also enrolled in the risk factors study. Additionally, subjects who did not consent to the community trial were still invited to join the risk factors study again as if they were willing to participate in the follow-up study. The enrolled subjects for distribution and risk factors study totalled 304.

5.3.7 Step 7 Process evaluation 5.3.7.1 Summary of study flow and response rate

419 subjects with negative result for opisthorchiasis from 2010 cross-sectional study

124 Excluded

 62 Age < 18 years  62 Not consented

295 Eligible and allocated to the study group

115 Allocated to intervention group 180 Allocated to control group

Follow-up at 17 months Follow-up at 17 months

 1 Withdrew: died or severe illness  6 Withdrew: died or severe illness  29 Loss to follow-up/unable to contact  66 Loss to follow-up/unable to contact

93 Completed study 131 Completed study

 8 Excluded for incomplete  23 Excluded for incomplete

questionnaire questionnaire  85 Included in final analysis  108 Included in final analysis

Figure 5.9 Summarized flowchart for community trial study

215

Table 5.5 Response for community trial study

Subjects from 2010 baseline 419 survey Excluded from the study Age < 18 years 62 (14.8 %) Not consented 62 (14.8 %) Eligible subjects 295 (70.4 %) Intervention group Control group Total p N (%) N (%) N (%) Subjects allocated to study group 115 180 295 Withdrew: died or severe illness 1 (0.9) 6 (3.3) 0.18 7 (2.4) Loss to follow-up/unable to 21 (25.2) 43 (36.7) 0.25 87 (29.5) contact Completed study 93 (80.9) 131 (72.8) 0.11 224 (75.9) Excluded for incomplete questionnaire 8 (7.0) 23 (12.8) 0.11 31 (10.5) Eligible for final analysis 85 (73.9) 108 (60.0) 0.01 193 (65.4)

Figure 5.9 provides details of study flow for the community trial study. According to Table 5.5, eligible subjects totaled 295 or 70.4% of the baseline population. Withdrew and lost to follow-up subjects did not significantly differ between the two study groups. However, after excluding incomplete questionnaires, which were not eligible for data analysis, the final subjects significantly differed between both groups. The percentage of response rate in the intervention group was significantly higher than in the control group (73.9% and 60.0%, respectively, p = 0.01) 216

Table 5.6 Population characteristics of community trial study

Intervention Control Total Characteristic p N (%) N (%) N (%) Sex Female 44 (51.8) 61 (56.5) 105 (54.4) 0.51 Male 41 (48.2) 47 (43.5) 88 (45.6) Age profile Mean ± S.D. 46.4 ± 15.9 51.3 ± 13.0 49.1 ± 14.5 0.02 Median ± IQR 49.0 ± 26.0 50.0 ± 19 49.0 ± 22.0 Min – max 18 - 88 19 - 84 18 - 88 Age group (in years, as quartiles) 18 – 38 30 (35.3) 20 (18.5) 50 (25.9) 0.02 39 – 49 14 (16.5) 33 (30.6) 47 (24.4) 50 – 60 25 (29.4) 29 (26.9) 54 (28.0) ≥ 61 16 (18.8) 26 (24.1) 42 (21.8) Occupation Unemployed 7 (8.5) 4 (4.0) 11 (6.1) 0.41 Others 12 (14.6) 13 (13.1) 25 (13.8) Agriculture 63(76.8) 82 (82.8) 145 (80.1)

As shown in Table 5.6, the proportion of male and female subjects was not significant for the intervention and control groups. Mean age and age-group significantly differed between the two study groups which meant age was significantly higher in the control group compared with the intervention group. Occupation showed no difference in both study groups.

217

419 subjects with negative result for opisthorchiasis from 2010 cross-sectional study

Excluded 115 subjects ≥ 18 years and allocated to intervention group in intervention study

304 Eligible subjects for risk factor study

 62 Age < 18 years  62 Not consented to intervention study  180 Age ≥ 18 years and allocated to control group from intervention study

248 Enrolled to risk factor study

 56 Not consented

Follow-up at 17 months

 6 Withdrew: died or severe illness  43 Loss to follow-up/unable to contact

199 Completed study

 25 Excluded for incomplete questionnaire  174 Included in final analysis

Figure 5.10 Summarized flowchart for distribution and risk factors study

218

Table 5.7 Response for distribution and risk factors study

N % Subjects from 2010 baseline survey 419 Excluded for subjected allocated to intervention group 115 27.4 Eligible subjects for risk factor study 304 72.6 Excluded to risk factor study Not consented 56 13.4 Enrolled subjects 248 Withdrew: died or severe illness 6 2.4 Loss to follow-up/unable to contact 56 22.6 Completed study 199 80.2 Excluded for incomplete questionnaire 25 10.1 Eligible for final analysis 174 70.2

As described in Table 5.7, 115 subjects (27.4%) were initially excluded for allocating to the intervention group in the community trial study resulting in 304 subjects being eligible for this study. As illustrated in Figure 5.10, 62 subjects with age <18 years, 62 subjects who did not consent to the community trial and 180 subjects with age ≥18 years and allocated to the control group from community trial, were enrolled in the study.

After enrolment, 56 subjects or 13.4% of eligible subjects did not consent to participate. Therefore, 248 subjects joined the study. At the end of study, the number of subjects eligible for the risk factors analysis totalled 174 (or 70.2%).

219

Table 5.8 Population characteristics of distribution and risk factors study

Characteristic N % Sex Female 98 56.3 Male 76 43.7 Age profile Mean ± S.D. 41.4 ± 21.2 Median ± IQR 45.0 ± 30.0 Min – max 2 - 84 Age group (in years, as quartiles) 0 – 27 44 25.3 28 – 45 43 24.7 46 – 57 44 25.3 ≥ 58 43 24.7 Occupation Unemployed 38 21.8 Others 15 8.6 Agriculture 109 62.6

Table 5.8 shows the population characteristics from risk factor study. The age-group was categorized into quartiles. There were more females than males. Agriculture-related work was a common occupation. The mean age was 41.4 ± 21.2 years, median age was 45.0 ± 30.0 years and age ranges from 2 to 84 years old.

5.3.7.2 Uncooked fish consumption from distribution and risk factors study

Table 5.9 Distribution of uncooked fish consumption

Fish menus N % Chopped raw fish salad (Koi pla) No 143 82.2 Yes 31 17.8 Extensively fermented fish (Pla ra) No 90 51.7 Yes 84 48.3

As shown in Table 5.9, the consumption of uncooked fish showed that Pla ra was consumed in approximately half of the study population and Koi pla was consumed by 17.8% of the study participants. 220

Table 5.10 Univariable and multivariable analysis of population characteristics and uncooked fish consumption behaviours

Koi pla Pla ra Characteristic Crude RR Adjusted RR Crude RR Adjusted RR (95% CI) (95% CI) (95% CI) (95% CI) Sex Female 1 1 1 1 Male 1.21 (0.60 – 2.45) 1.09 (0.51 – 2.34) 1.06 (0.69 – 1.64) 0.95 (0.61 – 1.50) Age group (years, as quartiles) 0 - 27 1 1 1 1 28 - 45 5.12 (1.12 – 23.35) 8.97 (0.89 – 90.85) 2.47 (1.26 – 4.85) 3.37 (1.03 – 11.11) 46 - 57 5.50 (1.22 – 24.81) 8.58 (0.86 – 86.02) 2.00 (1.00 – 4.00) 2.57 (0.77 – 8.54) ≥ 58 4.09 (0.87 – 19.27) 5.42 (0.66 – 44.60) 1.62 (0.79 – 3.34) 2.05 (0.68 – 6.24) Occupation Unemployed 1 1 1 1 Others 0.84 (0.09 – 8.12) 0.17 (0.01 – 2.66) 1.06 (0.37 – 3.00) 0.48 (0.12 – 1.84) Agriculture 2.67 (0.80 – 8.90) 0.52 (0.78 – 3.43) 1.78 (0.95 – 3.29) 0.76 (0.25 – 2.29)

Table 5.10 reports the univariable and multivariable analysis of population characteristics and uncooked fish consumption behaviours. The rationale for selection of factors was discussed from Section 2.3.2 in Chapter 2 and shown in Tables 2.8 and 2.9.

From univariable analysis, age-groups 28 – 45 and 46 – 57 years significantly increased the risk of consumption of both Koi pla and Pla ra compared with the reference age-group (0-27 years).

However, none of population demographic characteristics; sex, age-group and occupation, showed any significant association in the multivariable analysis.

221

5.3.7.3 Incidence of OV infection

Tung-heang 2012 cohort study 35 30 25

year wirh year 95% CI)wirh 20 -

0 Incidence (/100 (/100 person Incidence 0 - 27 28 - 45 46 - 57 ≥ 58 Age group

Figure 5.11 Incidence of opisthorchiasis distributed by age-group

Incidence rate of OV infection was 9.31 /100 person-years (95% CI: 5.90 – 13.97) from the distribution and risk factor study (Figure 5.10).

From Figure 5.11, the quartiles age-group indicated that incidence was highest in the 3rd quartile; age-group 46 – 57 years with incidence of 16.01 /100 person-years (95% CI: 7.68 – 29.43). (Supplement Table 5-9)

222

5.3.8 Step 8 Impact evaluation

5.3.8.1 Evaluation of risk factors for acquiring OV infection

Table 5.11 Univariable and multivariable analysis of risk factors for acquiring OV infection

Crude IRR Adjusted IRR Characteristic P p (95% CI) (95% CI) Sex (female as reference) Male 1.41 (0.62 – 3.19) 0.41 Age group (years, as quartiles) 0 - 27 1 1 28 - 45 1.53 (0.26 – 9.19) 0.64 1.21 (0.20 – 7.41) 0.84 46 - 57 5.00 (1.10 – 22.82) 0.04 3.87 (0.82 – 18.20) 0.09 ≥ 58 4.09 (0.87 – 19.27) 0.08 3.41 (0.71 – 16.32) 0.12 Occupation Unemployed 1 Others 5.07 (0.93 – 27.66) 0.1 Agriculture 2.27 (0.51 – 10.04) 0.28 Fish menus Koi pla Yes 2.97 (1.28 – 6.85) 0.01 2.52 (1.07 – 5.92) 0.03 Pla ra Yes 1.67 (0.72 – 3.85) 0.23

Table 5.11 shows univariable and multivariable analysis of risk factors for acquiring OV infection from the follow-up study. From univariable analysis, the relative risk was not significant for males, occupation and Pla ra consumption compared with the reference. Age-group showed that age 28 – 45 years had a significantly higher risk of infection (IRR = 5.00, 95% CI: 1.10 – 22.82) compared with reference age-group (0 – 27 years).

Based on Section 2.3.4.2 from Chapter 2, the final multivariable model selected two significant covariates obtained from current fieldwork study: age and Koi pla consumption. Adjusted IRR for Koi pla was 2.52 (95% CI: 1.07 – 5.92) and age-group showed no significance for increasing risk compared with the reference.

. 223

5.3.8.2 Evaluation of impact of community intervention on Koi pla consumption

Result showed that community intervention significantly reduced the incidence of consumption of Koi pla within the study period; RR = 0.54 (95% CI: 0.29 – 0.99). The intervention could reduce the consumption by 46% (95% CI: 1 – 71). (Supplement Table S-10)

The qualitative approach from social diagnosis in Section 5.3.1 also provided additional evidence of how psycho-social factors could influence Koi pla consumption patterns

5.3.9 Step 9 Outcome evaluation

5.3.9.1 Evaluation of impact of community intervention on incidence of OV infection

At the end of study, the incidence of OV infection in the intervention group was 5.13/100 person- years (95% CI: 1.88 – 11.17) compared with the control group 13.69/100 person-years (95% CI: 8.48 – 20.93). The IRR was 0.37 (95% CI: 0.15 – 0.93) and statistically significant. Thus, the intervention yielded a 63% (95% CI: 7 – 85) reduction in the incidence of infections. (Supplement Table S-11)

224

5.3.9.2 Qualitative approach for evaluation of community intervention

The impact of intervention was also evaluated by the qualitative approach. Most of them expressed that the agreement was a good policy. They were keeping in mind that they needed to avoid those risks of infection; consuming raw fish especially Koi pla, adequate sanitation or how to manage kitchen waste. They perceived that the aim of agreement was for their own health. Some expressed the opinions below.

“It keeps us alert, to be aware of eating raw fish. It’s good for you, not anyone else. I think it’s a good project. Everyone recognizes it and so do I. That’s the way I’ve tried to control myself. ”

“I mean yes, I volunteered. Of course sometime it makes me feel uncomfortable. But in the end it’s good for me at all.”

“I think it’s about the process. You commit to do something, something very important. Not just for yourself, also community. Look at your family, your neighbors, everyone concerned about it (agreement). It’s about how to change or modify your life. It’s challenging.”

“It’s like you make a promise and you want to keep it. If you don’t want to keep it at the first place, you wouldn’t commit anything like this (laugh)”

225

5.4 Discussion 5.4.1 Study design and response

The Tung-heang fieldwork comprised two studies: a community trial and follow-up study. The community trial assessed the effectiveness of community-based intervention using the PRECEDE- PROCEED framework and the follow-up study assessed the relationship of risk factors and OV infection.

Considering the study design, the method resembled that in the other two areas; Na-ngam and Na- yao Villages. After exclusions for unmatched and incomplete questionnaires, the final response rate was 70.2%, significantly higher (p <0.001) than the Na-ngam (53.1 %) and Na-yao study (36.1 %). It was assumed that incorporating the control group from the community trial would result in a higher response rate. However, the study population was considered low: 248 enrolled subjects and 174 subjects at the end of study due to the size of Tung-heang community.

The Tung-heang cohort design highlighted the same issues as found in the other two areas (Section 2.4.1 from Chapter 2). Since the consumption of Koi pla was collected retrospectively by questionnaires, potential bias included recall bias and social desirability bias (147, 148).

Recall bias can occur when the study subjects have to recall a prior exposure differently depending on the disease status (148, 149). Recall bias is a concerned issue for case-control design when the disease outcome is addressed first and then the corresponding exposures are assessed retrospectively from past memory (124). Reporting of exposures might be overestimated or underestimated depend on participants’ concern, for example, cases might pay more attention on assessing risk factors they believe would be the cause of their illness. The impact of recall bias can result in differential misclassification.

The study design used in this chapter included a community trial and follow-up study. The outcome was prospectively determined when stool specimens were examined at the end of the study for OV infection diagnosis. However, the uncooked fish consumption behaviors regarded as disease exposure were determined retrospectively. The study participants provided the details of their consumption habits by answering the self-administered questionnaire at the same time as providing stool specimens. From the study method, the data collection method was designed concerning the recall bias issue as described below.

 Stool specimen and questionnaires were collected at the same time at the end of study period when fieldwork was conducted. Therefore, the stool examination process was ready for 226

reporting the result to the participants after collecting the questionnaire so they were not aware of their infection status during the questionnaire survey. Moreover, the study participants comprised a follow-up population and were preliminary negative for OV infection from the previous baseline cross-sectional study. Therefore, they were less likely to be influenced by the stool examination result for giving details of consumption behaviors.  Based on the questionnaires, the study participants needed to provide details whether they ever consumed Koi pla or Pla ra during the study period. The quantity of food and frequency of consumption were not determined, which could reduce the confusion to provide recalled details. However, survey questions about sensitive issues could lead to misreporting of the variables of interest resulting in social desirability bias.

Generally, individuals tended to preserve their image within the social norm (150). They would try to avoid the negative impact from social interaction such as negative criticism, feeling of shame or embarrassment resulting in a state of insecurity and loss of self-confidence. Psychologically, the defense mechanisms will respond to maintain social desirability including lying or manipulating data in any way to avoid such negative feedback (151). In the research field, social desirability bias will be likely to appear when the study issues are considered sensitive such as sexual, racial or religious- related issues (150). Therefore, misreporting the variable of interest could impact the result of the study. Psychological studies indicate that misreporting on sensitive questions in a social desirable way is a result from voluntary process under responder’s consciousness (152, 153). The responders are more likely to answer the question with a self-controlled process than a subconsciously autonomic mental response. Therefore, data collection method could be designed to reduce this bias.

In this study, the association of Koi pla consumption and OV infection was communicated to the community when Koi pla was identified as a potential health threat. The data collecting method aimed to reduce recall bias by reporting study results after collecting questionnaires. However, the undesired outcome of Koi pla consumption might still affect the answers provided in the questionnaire for consumption behavior history.

From qualitative data from Chapter 4, Koi pla consumption was still a common practice in the community with regard to cultural background. The study participants acknowledged Koi pla as a health risk from the research team. Therefore, some of them might provide the answer in a desirable way to the research team as they perceived that the team comprised healthcare personnel. 227

Considering social desirability bias, the community trial protocols were designed to reduce the bias by the methods stated below.

 The confidentiality of the data of study participants was preserved through the entire process. The questionnaire and stool specimen were processed under a code-embedded system. The research protocol administrator, specimen examiner and data analyzer were different persons. The information was used with regard to the whole group. Names and personal data were not individually identified in the research. Stool result was reported to each participant in person with a sealed envelope.  For the qualitative approach, each group in the focus-group discussion was categorized based on same Koi pla consumption pattern to observe the direction of opinions from each particular eating practice. The homogeneity in each group would let the participants express their opinions more freely. For in-depth interview, a special session was conducted in a private setting with the responder’s consent. The interviewer always introduced the session with the academic benefits from the participant’s cooperation. The provided answers would contribute scientific benefit and would not interfere with regular healthcare. Moreover, they would not lose any benefit when they were unwilling to join the study.  The data of Koi pla consumption was collected at the end of the study. Study participants were informed that their identity remained anonymous during data analysis. The self-administered questionnaire would decrease concern about admitting to Koi pla consumption (154, 155).

The given principle was also applied to all part of the fieldwork involving data collection on uncooked fish consumption.

The issues of recall bias and social desirability bias could be improved in future studies. The prospective data collection on variables of interest could reduce the effort to recall memory. To avoid answering sensitive questions about uncooked fish consumption, a food diary could be proposed to collect daily food intake in a more general way, which could reduce the frustration of answering any specific food item (154). The method could also collect the amount and frequency of a particular food type as well.

The stool examination comprised three methods to improve sensitivity. However, the microscopically-based methods might reduce specificity because minute intestinal fluke (MIF) eggs were very similar to liver fluke eggs. Moreover, co-infection of MIF and OV (156) have been reported in Thailand. However, MIF was not found in this region by PCR technique (46). Based on cohort methods, only negative cases from the baseline study were enrolled in the follow-up study. 228

Therefore, the incidence might have been underestimated due to the high-risk group; previously infected cases were not enrolled in the study.

The community trial was conducted to assess the community-based intervention based on open- label, nonrandomized method due to the design of the intervention. The intervention was a community mutual agreement where the study participants agreed to follow instructions emphasizing behavioral modification with respect to risk factors. Therefore, the participants consented and volunteered to join the intervention. Thus, the allocation method was not randomized. The CAB coordinated between study participants and research team. The control group only agreed that they would provide a stool specimen and answer a questionnaire at the end of the follow-up period. Therefore, they were eligible for the study of risk factors.

Considering the intervention protocol and method for study allocation, the design was open-label because the study group and investigator were not blinded. However, the specimen was blinded to examiners during stool examination process to avoid bias toward reporting results. Age-structure significantly differed between study groups (Table 5.6) which meant age for the control group was higher than the intervention group. Subjects that withdrew or were lost to follow-up and incomplete questionnaire subjects were not significant in the two groups. However, the final response rate was significantly higher in the intervention group. Probably, the lower loss to follow-up and incomplete questionnaire proportion in the intervention group resulted because the study participants volunteered; therefore, they might be more likely to provide cooperation to the study.

An interesting issue arose for allocating the study population in the qualitative approach with regard to patterns of Koi pla consumption. Within the 17 months of study period, no individuals, who never consumed Koi pla before, choose to have it during study period. Hence, the factor to promote Koi pla consumption was obtained through individuals who sustained their Koi pla consumption behaviors.

5.4.2 Situation of OV infection

Two surveys were conducted on OV infection in Tung-heang area in 2008 and 2010. However, no risk factors were assessed. From the fieldwork database, the prevalence of OV infection was 10.38% (95% CI: 7.85 – 13.39) in 2008 and 11.42% (95% CI: 8.69 – 14.63) in 2010. However, the data were not yet published. Risk factors were evaluated in this follow-up study. Koi pla consumption was consistently identified as a risk factor for contributing OV infection in adjacent areas including Na- ngam, Na-yao and Na-isarn (Table 2.10 – 2.12 from Chapter 2). 229

5.4.2.1 Situation of Koi pla consumption

As shown in Table 5.9, Koi pla was consumed by 17.8% of the follow-up population. Based on the questionnaire, individuals were asked whether they consumed Koi pla at least once during the follow-up period to evaluate the relationship of Koi pla consumption and OV infection. It was observed that the rate of consumption was lower than the other 4 areas (Section 2.3.2 from Chapter 2); 39.8 – 44.1 %. However, Koi pla consumption was 25.9% in the control group from the community trial.

Univariable analysis showed that age-group significantly increased the risk of Koi pla consumption compared with the reference age (Table 5.10). However, multivariable analysis did not show any significant population characteristics on Koi pla consumption. However, the confidence interval for each age-group was wide, which could have resulted from the small sample size (174 subjects from Table 5.7).

The qualitative approach could provide more details about Koi pla consumption patterns. Classifying consumption patterns in 5 different groups (Table 5.2) provided more insights of how the behavior would change over time with respect to underlying causes.

Precipitating factors for continuing Koi pla consumption emphasized the cultural context of the behavior. Consumption of Koi pla itself did not provide any health benefit and the dish was not a regular dish for daily living. The potential of Koi pla was an underlying core of local norms and traditions. Male always consumed Koi pla in terms of socialization, while females accepted it as a norm and children learned that was pathway to growth in the community.

However, eating habits were dynamic and could be potentially influenced by many factors. Many individuals stopped eating Koi pla during the study period. The interesting issue was that initiation of community-based intervention derived from the members of the community. The mutual agreement might have acted as a new type of social norm for the study participants, to be discussed in Section 5.4.3.1.

5.4.2.2 Incidence of OV infection

From the follow-up study, the incidence of OV infection was 9.31/100 person-years (95% CI: 5.90 – 13.97), higher than both national and regional incidence (11, 30, 69, 70). Compared with other follow-up studies in this area, the incidence of OV infection was 7.98/100 person-years (95% CI: 5.49 – 11.20) in the Na-yao 2013 study and 6.80/100 person-years (95% CI: 4.68 – 9.54) in the Na-ngam 2014 study. Tung-heang incidence was significantly higher than other two study areas (p <0.05). 230

Multivariable analysis indicated that Koi pla consumption significantly increased the risk of infection with IRR = 2.52 (95% CI: 1.28 – 6.85) when adjusted for age-group (Table 5.11). Age-group showed no significant association with the infection. Considering the age-group pattern in the Na-yao 2013 and Na-ngam 2014 follow-up studies, no age-pattern was observed as well.

From the qualitative approach, newly-infected cases were likely derived from individuals with constant Koi pla consumption. From the study method, the questionnaire aimed to determine Koi pla consumption during the study period. Therefore, long-term consumption was not assessed using this method. Based on assumptions from Chapter 3, OV infection needed sustained re-infection to maintain its endemicity. According to the mode of infection, uncooked fish should be constantly consumed as well. Study on factors shaping consumption patterns could play an important role for the understanding of infection dynamic with regard to risk behaviors.

5.4.3 Community intervention 5.4.3.1 Impact of intervention on Koi pla consumption

From the National Control Program (28), the strategy to interrupt transmission of the infection was avoiding uncooked fish consumption. The problem was that the villagers could not avoid all types of dishes, especially extensively fermented fish (Pla ra), which was the main ingredient for I-sarn (northeastern) dishes. Pla ra was regarded as a cultural symbol for Thailand (69). Individuals might interpret the campaign to reduce the consumption of unspecific uncooked fish as saying they have to stop eating Pla ra as well as Koi pla. Quantitative data revealed that Pla ra was more popular than Koi pla in terms of incidence of consumption (48.3% and 17.8%, respectively). Many studies showed that metacercariae, the infective larva, could not survive in some particular environments. It definitely survives in fresh fish meat, so consuming Koi pla would biologically increase risk infection with metacercariae, while metacercariae degenerate after two days of preservation. For extensive Pla ra, the fish is preserved in a highly concentrated salt solution for 3 – 6 months, which was not suitable for the metacercariae to survive (87, 88) . The results show that of the fish dishes considered, only Koi pla was an independent risk factor for acquiring OV infection. This could lead to the design of a focused control strategy, especially combined with the bio-psycho-social data from the qualitative analysis.

To deal with the situation, the available quantitative data may not be sufficient to establish an effective health policy to control OV infection. The problems involve every aspect not only the incidence rate and the risk factors obtained from the study, but also important bio-psycho-social aspects. 231

Based on the PPM framework, the reasons why Koi pla was still popular; attitudes, knowledge, or even ideas of conducting a solution were then obtained from the qualitative approach based on the community’s own thinking. The framework allowed us to acknowledge the problems in various aspects. Results showed that both quantitative and qualitative data were consistent; Koi pla consumption was a main risk factor for contributing to OV infection. Study outcomes highlighted that particular uncooked fish dishes were traditional and regarded as tasty. Changing or modifying eating habits probably pose a large burden for their lives. Some perceived that despite consuming uncooked fish dishes for long periods, they were still healthy and sometimes stool examination came with negative results even when they were classified as a high-risk person for acquiring the infection. Thus, little incentive existed to change their eating habits.

Poor knowledge or insufficient information of disease risk and routes of transmission were obstacles for behavior change among the study participants. They required adequate knowledge from easily understandable messages to clarify the risk of acquiring OV infection. Social and traditional cultural factors were important to explain Koi pla consumption. The risk and benefit from eating Koi pla could be demonstrated as they should eventually perceive problems themselves and if successful, a sustainable solution would come from the community itself.

The intervention program should have had a substantial impact on reducing Koi pla consumption, because it did not target uncooked fish consumption generally but used the precise term “Koi pla” in the community (Table 5.1). Therefore, they would not be confused with other foods, such as Pla ra. This made modification of their eating behaviours much easier.

The impact evaluation from step 8 (Section 5.3.8.2) indicated that the intervention could significantly reduce Koi pla consumption by 46% (p = 0.04) within the study period. The outcome highlights the potential effectiveness of the primary prevention strategy. However, the non-random allocation process should be concerned as it could potentially produce selection bias.

Educational and ecological diagnosis from step 4 (Section 5.3.4) indicated that reinforcing factors were a key point for the community to maintain consistency of the intervention. A community- derived solution could play an important role to reinforce the community agreement to reduce OV infection. The CAB, along with the role of local health volunteers, acted as a key role for implementing the intervention at the community level with minimal assistance from outsiders such as the research team. However, outcomes of intervention still need evaluation. Microscopic-based stool examination requires technical skill, but can be conducted in an ambulatory setting or local health facility. 232

5.4.3.2 Impact of intervention on incidence of OV infection

The outcome evaluation from Section 5.3.9.1 reported that the intervention could significantly reduce incidence of OV infection by 63% (95 % CI: 7 – 85, p = 0.03) within the study period.

Moreover, the administrative and policy assessment in step 5 (Section 5.3.5) identified some interesting issues regarding the control strategy. Improving diagnosis and treatment at the local health centre should be implemented under the national strategy because so many people originating from Northeastern Thailand have dispersed around the country like in this community. Stocking praziquantel at the local pharmacy might encourage the community members to buy it themselves and take it whenever they consumed uncooked fish.

The intervention group from community trials showed positive attitudes toward the intervention. Some of them mentioned that even though they could not abruptly stop eating Koi pla, the consumption eventually became less frequent. Further studies may need to assess frequency of consumption. Moreover, the burden of OV infection could be evaluated by infection intensity as well to monitor the decline of severity in addition to curing the infection (as described in Section 2.4.2.3 from Chapter 2).

233

5.4.4 Implication of PRECEDE – PROCEED model framework

Intervention

Prevention Diagnosis and Treatment

5 4

8 6 9 Risk behaviours Opisthorchiasis 3 2

7 1 1

Social influence

Figure 5.12 PRECEDE – PROCEED model integrated to conceptual framework

Figure 5.12 shows how the PPM framework integrated the conceptual framework of the study. With the aid of mixed-method study designs, quantitative and qualitative fieldwork provided results in various aspects highlighting that complex structure regarding infection dynamic could be holistically approached by systematic framework as described below.

 Step 1: evaluate the burden of OV infection through the prevalence and incidence of the infection. Qualitative approach visualized the social influence affecting infection dynamics. This was a vital step to identify the root cause of the problem.  Step 2: setup the objective to reduce OV infection. From the framework, reducing the prevalence of OV infection was directly involved with diagnosis and treatment, which is considered a secondary prevention. Considering the reduction of incidence of the infection, the intervention should be applied through risk behaviours to interrupt the mode of infection. By emphasizing primary prevention, the objective was to reduce incidence of infection. 234

 Step 3: setup the sub-objective to reduce risk behaviours. From studies conducted in this area, Koi pla consumption was identified as potential risk behaviour (Chapter 2). The qualitative approach (Chapter 3) also suggests that the intervention should specify precise action such as “reduced Koi pla consumption” (Table 5.1) to better modify behaviour.  Step 4: setup community intervention (Table 5.1) focusing on reduction of Koi pla consumption, which could eventually reduce the incidence of OV infection. The crucial step of this process was that the intervention was developed and initiated by the community. Moreover, the community members also hold responsibility to manage the flow of intervention with the support of the research team for the study design and outcome evaluation. The concept of community intervention was based on three important factors: predisposing, reinforcing and enabling factors.  Step 5: current strategy and resource with respect to National policy were assessed to evaluate the feasibility and compatibility of community intervention. The designated intervention should be compatible with existing policy and also adapted to community characteristics. This was a challenging step as the intervention could be applied to the community. From step 4, the aim of reducing Koi pla consumption was based on the National Control guidelines (28), but the details have been modified to achieve more specific goals; precise intervention was also supported by the evidence-based mixed-method study conducted in multiple areas.  Step 6: the fieldwork was designed and conducted to achieve objectives and sub-objectives with the mixed-method design. The community trial was used to evaluate the community intervention for reducing incidence of OV infection and Koi pla consumption. The follow-up study served as an additional study to evaluate the distribution and risk factor for up-to-date information. The qualitative approach could also access data in many steps to describe the details of given issues.  Step 7: the flow of the whole study was summarized (Table 5.3) to ensure the completeness of fieldwork. The CAB was a vital key for managing the study flow. The initiation of intervention and cooperation from community members reflected the success of community-based action against the health problem.  Step 8: the impact of intervention on Koi pla consumption was evaluated from the community trial. Up-to-date results of risk factors contributing the infection were obtained from the follow- up study.  Step 9: the impact of intervention on the incidence of OV infection was evaluated from the community trial. 235

Following the PPM framework, community participatory action could play an important role as a reinforcing factor as the intervention was developed within the community environment which was considered adaptive and practically suited to the villagers’ lifestyles when consumption behaviours were complex involving traditional beliefs, attitudes and cultural backgrounds. To initiate behavioural modification, strict law enforcement may not be available and probably not suitable for such a public health campaign.

It appears that the National Control Program effectively reduced OV infection prevalence (26-28), but failed to suppress the incidence of disease sustainably (69, 70). Seeking cases and providing treatment were more likely an effective secondary prevention. The PPM framework allowed us to perceive and understand the weak points of the public health campaign. Reinforcing factors were crucial to sustain health campaign continuity.

236

5.5 Conclusion

Although this project was a pilot study to assess the community in small scale, the study process and results provided insights into community management to support expansion into a larger scale in which CIP (Community Intervention Package) could be developed and applied on a national scale.

Providing adequate knowledge of OV infection and improving health care facilities would offer the best alternatives for disease control at the community level. Community participatory action was considered important for self-sustained activity. In a resource-limited setting, primary prevention was promising for a public health campaign.

237

Chapter 6 Discussion

6.1 Overview

This project aimed to comprehensively study the infection dynamics of OV infection using multiple tools including statistical model, mathematical model and qualitative approach to explore the infection in various aspects.

Study results then could be used to develop community interventions using the framework of the planning model. The planning model was based on integrated outcomes from the comprehensive study with respect to a rural community setting which reflects the common characteristics of rural Thailand.

Intervention

Prevention Diagnosis and Treatment

Evaluation

[Host] [Agent]

Risk factors Opisthorchiasis Planning

[Environment]

Social influence

Figure 6.1 Integrated conceptual framework from comprehensive study

As shown in Figure 6.1, the conceptual framework was integrated with comprehensive results from the study with respect to the epidemiological triangle relationships of host, agent and environment. 238

Study outcomes would aid the developing of intervention planning and lead to design and implementation of interventions in the evaluation phase. The project explored the fundamental knowledge and background of OV infection and built up consolidated information to formulate a community intervention package.

From the planning phase, the relationship between risk factors, OV infection and social influences were explored reflecting the host-agent-environment interaction.

The study of host referred to the population characteristics and factors determining the risk behaviours, which could lead to acquiring the infection. The population structure was visualized by a population pyramid to examine the pattern of age with respect to age-class.

The results from Chapter 2 (Figure 2.7) showed that the population pyramid from 4 study areas was considered a constricted pattern with the relatively small proportion of age-class 20 – 29 years. The qualitative approach supported the underlying reasons that many individuals in their 20s temporarily move out to seek jobs in an urban setting during the off-harvest season. This is a common characteristic of Thai rural villages where the primary jobs are agriculture-related with part-time jobs during off-season (70).

The study of agent referred to OV infection caused by Opisthorchis viverrini. This project assessed the epidemiology of the infection including prevalence and incidence by extensive fieldwork. The infection dynamics were explored through a mathematical model highlighting the complicated relationship between prevalence and incidence.

Environmental factors were considered to have important roles in shaping the infection dynamics. OV infection was known to be locally specific to each area (43). This project emphasized the study of the social factors which influenced the risk behaviours. The study examined the advantages and limitations of the National Health Policy with regard to controlling OV infection through health campaigns proposed by health authorities for the community. The Long-term National Control Program has successfully reduced the prevalence of OV infection as national goal. A new challenging issue has emerged from this project when the study site selected was a non-endemic area with high incidence of OV infection.

239

The interaction between factors was assessed by particular study methods described below.

6.1.1 Host and agent relationship

In Chapter 2, prevalence and incidence of OV infection were extensively examined through cross- sectional and follow-up studies within 3 study areas. Regarding risk behaviour; Koi pla consumption was determined to increase the risk of acquiring infection and also increase the intensity of the infection. For infection intensity, the age-prevalence-intensity relationship indicated that the age- structure component within the intensity and the level of parasite aggregation interacted to where the intensity tended to increase and the parasites were more equally spread based on the assumption that incidence was uniform at all ages. Over-dispersion was observed for the infection intensity, and a two-part model was introduced to handle over-dispersed and zero-excess data.

From Chapter 3, mathematical modelling was also formulated to explore the relationship between prevalence and incidence of the infection. The model was fitted to serial prevalence obtained from the Na-yao area to project the longitudinal trend of infection dynamics. The treatment effectiveness of praziquantel was also estimated from both mathematical models; 80.1 – 92.5% and indirect calculation from individual-level fieldwork data; 88.7 – 92.2 %. Additionally, the model was fitted to the prevalence trend of OV infection under the National Control Program.

From Chapter 4, the qualitative approach was conducted to explore the social dynamics which provided important underlying reasons behind the practice of risk behaviours with regard to uncooked fish consumption. The social component was considered an important part for infection dynamics with the influence of knowledge, attitudes and perceptions.

6.1.2 Host and environment relationship

The qualitative approach from Chapter 4 described the relationship between host and environment in terms of risk behaviours and social influence with group discussion conducted in Chapters 4 and 5 based on the age generation and pattern of Koi pla consumption.

Age structure, which was classified in age generations, played an important role for the underlying attitudes and thinking processes toward OV infection and its risk behaviours. The qualitative study investigated how the knowledge transfer and cultural transfer interacted within particular generations.

From Chapter 2, the impact of sex, age-group and occupation on uncooked fish consumption; Koi pla and Pla ra were evaluated by statistical modelling. Male sex and increasing age-group significantly 240

increased the risk for Koi pla consumption and agriculture-related work also increased the risk of Pla ra consumption.

The underlying social-influence details were additionally explained by focus group discussions and in-depth interviews. Male-associated environments might have influenced the practice of Koi pla consumption; fish were usually caught and instantly prepared during their agricultural fieldwork. Male social drinking was also associated with Koi pla consumption by qualitative approach. Age was considered a potential factor; older age population showed a picture of cultural attachment in which Koi pla represented the original Northeastern tradition.

Pla ra provided different perspectives of uncooked fish consumption. Biologically, metacercariae could not survive in Pla ra because of its high salt concentration but the pattern of consumption practice could influence other uncooked fish dishes. Pla ra was considered the main ingredient in most local dishes; therefore, any campaigns to avoid all uncooked fish consumption might be too broad to target the actual risk behaviour, and therefore, it might be problematic for behavioural modification.

6.1.3 Agent and environment relationship

OV infection itself had an impact on the community which can be evaluated from Chapter 4. The study results were transferred to the community which triggered some concerns and lead to subsequent action as described in Chapter 6 for the community intervention.

Community background and social structure precipitate and limit the burden of OV infection as described in Chapter 4.

6.1.4 Development of planning model

Figure 6.1 briefly describes the interaction of host-agent-environment discussed in Section 6.1.1 to 6.1.3. A planning model integrated the study results from Chapters 2 – 4 to engage the disease burden at the community level under PPM.

The community trial initiated the community intervention package based on the fieldwork data obtained from both quantitative and qualitative studies. The comprehensive data supported the development of a planning model with community participation and considered the importance of background especially population characteristics and social components.

The intervention was built based on community perspectives reflecting on the current situation of OV infection. As the community members perceived and acknowledged that OV infection was a 241

burden, they set up a goal to reduce the incidence of OV infection using behavioural modification to reduce Koi pla consumption. The planning phase was followed by the evaluation phase where the intervention was implemented and the outcomes were evaluated.

The PPM also assessed the advantages and disadvantages of the current health policy to justify and adapt the proposed intervention to be more feasible and practical in the community setting.

Therefore, Chapter 5 provides an integrated method to use comprehensive data obtained from fieldwork and analysed with rigorous statistical methods, then applied the community intervention package to the actual public health setting. 242

6.2 Summary of study findings

6.2.1 Epidemiology of Opisthorchis viverrini infection: distribution and risk factors

Chapter 2 highlighted the potential of OV infection epidemiology concerning distribution and risk factors, within the rural community setting regarded as a majority part of Thailand. Regarding the well-preserved Northeastern environment, cultural background could play a major role in continuing risk behaviours, which could potentially maintain a higher prevalence in this non-endemic area.

This Chapter aimed to explore the epidemiology of OV infection for better understanding of the infection in various aspects. The main findings showed that prevalence and incidence were relatively high considering the study area was non-endemic. The prevalence of OV infection was 7.85% (95% CI: 4.96 – 10.75) in Na-isarn 2013 and 9.29% (95% CI: 6.64 – 11.93) in the Na-ngam 2012 study. The incidence rate was 7.98/100 person-years (95% CI: 5.49 – 11.20) in Na-yao 2013 and 6.80/100 person-years (95% CI: 4.68 – 9.54) in the Na-ngam 2014 study. Compared with the regional average, the prevalence of OV infection in central Thailand was (3.8%) (28) and the incidence has never been officially published before in this region (9). The population demography is linked to the Northeastern background, their ancestors’ origin.

Consumption of Koi pla was determined to present a potential risk factor for acquiring the infection. Data from the two follow-up studies and additional supplemental data from another two follow-up studies (69, 70) provided rational supporting evidence for the causal relationship between Koi pla and OV infection.

Infection intensity showed an over-dispersed distribution of egg count outputs suggesting the use of negative binomial distribution for count outcomes. The age-prevalence-intensity relationship indicated that an age-structure component within the intensity and level of parasite aggregation. Assuming incidence was uniform at all ages, the intensity tended to increase and the parasites were more equally distributed with increasing age within the population.

The compound model including zero-inflated-based models and hurdle-based models was introduced to handle the zero excess and over-dispersed data. The model provided both infection probability and infection intensity, which was applicable for the data interpretation. The model result reported the risk of being infected from identified factors and also the level of infection intensity. From the study, the negative binomial logit hurdle model (NBLH) was the model of choice indicating increasing risk of OV infection from Koi pla consumption as observed in prevalence and cohort data analysis. A trend was observed among age groups in NBLH model where the risk of 243

acquiring infection increased with age when compared with the reference age group. Moreover, consumption of Koi pla and the 4th quartile age-group also significantly increased infection intensity.

6.2.2 Infection dynamics of OV infection

Chapter 3 showed that re-infection played an important role for sustaining the chronic infection picture and shaping a sustained prevalence. Praziquantel was estimated to show the same effectiveness as previous fieldworks at 80 – 95 % (26, 40, 49).

From Chapter 2, the incidence of OV infection can be estimated from catalytic function of prevalence data using linear regression modelling. However, the incidence directly estimated from the cohort fieldwork was significantly higher than the fitted incidence obtained from the regression model. Therefore, the infection dynamics would need a more complex framework with respect to host – agent – environment.

The inconsistency of the given relationship was further explored in Chapter 3. My aim was to explore the infection dynamics with regard to complicated relationships between prevalence and incidence from the fieldwork using a more complex framework which emphasizing the human population dynamics. Chapter 3 also evaluated the effectiveness of praziquantel treatment on infection dynamics and applying the findings to the National Control Program data.

The model was based on the classic susceptible – infectious – recovered (SIR) model. The main extension was to add a second susceptible class so there are primary susceptible individuals (S1) who will be infected for the first time. Infected individuals will be naturally recovered for some period and become secondary susceptible (S2) again for the re-infection with reduced infectivity. The model assumed that infected individuals who recover because of praziquantel treatment will gain longer immunity before potentially becoming re-infected. The force of infection is assumed to be uniform and the recovery rate depends on the parasite life expectancy.

The model showed some key data from the fieldwork. The model assumes that the infection is in the endemic phase where Na-yao 2002 prevalence is 22.9 % (95% CI: 19.3 – 26.5) and Na-ngam 2012 is 9.3 % (95% CI: 6.9 – 11.6). I estimated that the average duration of infectiousness was in the range

1.51 – 1.81 years. Basic reproductive numbers (R0) were estimated from the model. Finally, the model is used to assess the effectiveness of the National Control Program. R0 for Na-yao is 2.74 (95% CI: 2.61 – 2.89) and for Na-ngam is 2.30 (95% CI: 2.23 – 2.40).

244

The effectiveness of praziquantel treatment is estimated from Na-yao study where multiple surveys have been conducted. The treatment effectiveness, as cure rate of praziquantel is 80.1 – 92.5 % from the mathematical model. An alternative calculation of from individual-level fieldwork data provided a similar estimate of 88.7 – 92.2 % of treatment effectiveness.

The model, fitted to the prevalence data from the National Control Program from 1984 – 2001, revealed that the incidence was estimated to be 9.8 – 47.3/100 person-years by assuming that incidence was uniform with praziquantel effectiveness of 80 – 100%.

6.2.3 Evaluation of OV infection and risk behaviours using qualitative approach

Chapter 4 showed that social influence was a potential component for infection dynamics of OV infection. The interaction of knowledge and culture transfer across generations provided more insights to the bio-psycho-social dynamics.

The infection dynamics of OV infection were assessed through the qualitative approach using focus group discussions and in-depth interviews. Discussion themes were directed by pre-defined topics regarding the current quantitative data and reviewed literature to develop the understanding of OV infection and related health risks. Framework analysis was mainly used to explore the association between factors from thematic content extracted from gathered data. Additionally, grounded theory (129) was used when new ideas were generated from the analysis.

The prevention campaign, based on the National Control Program, targets uncooked fish consumption broadly. The locals have found some difficulty in avoiding all uncooked fish dishes and as a result often maintained the same eating habits. The outcome suggested that the campaign should emphasize the risks associated with specific items and should be supported by evidence- based data.

Koi pla remained a popular dish in the community as the dish itself represents Northeastern culture. The cultural norm had been transferred from their ancestors who originally moved from the Northeastern region around 4 decades ago. Consumption of Koi pla, which was not a regular eating habit for central Thais, reflected a potential cultural attachment.

From Chapter 2, the NBLH model indicated the potential risk of the increasing age-group for acquiring OV infection. The qualitative approach revealed that age-generation also influenced the process of knowledge transfer and cultural transfer in the matter related with OV infection and risk behaviours. Knowledge of OV infection can be easily distributed to younger generations by formal education and social networks when they can pass this knowledge to older generations through 245

their family relationship. However, the knowledge-transfer scheme will confront the cultural transfer, which was conversely passed from older generations. This interaction has an important impact on the risk behaviours; therefore, it was crucial for the intervention strategy.

For rural settings, the community leader is an important key in the social structure. As centralized authority might not be aware of local cultural norms, community leaders can help in filling this gap in terms of communication. Local health volunteers played an important role for the interaction between villagers and health authorities. They could act as community leaders and as facilitators for the health intervention. Because they were locals, they could more easily gain trust and cooperation from the villagers, which were a crucial part to maintain the sustainability of health campaigns in the community. 246

6.2.4 Community-based intervention: PRECEDE-PROCEED model framework for controlling OV infection

Chapter 5 showed that the mixed-method study design provided comprehensive approaches to assess OV infection in various aspects. The PRECEDE-PROCEED model framework also provided a comprehensive approach for the intervention/evaluation planning. The project initiative provided more insights of community management, which aimed to expand to a larger scale in which the community intervention package could be developed and applied for public health implementation on a national scale.

From Chapters 2 – 4, extensive infection dynamics data were obtained from statistical modelling, mathematical modelling and also the qualitative approach. In Chapter 5, the aim was to construct an intervention in the community using a public health framework supported by study data.

The PPM framework was applied in the community to systematically engage the problem. The framework consisted of two major phases; planning and evaluation. The planning phase aimed to setup goals to reduce the incidence of OV infection through behaviour modification by decreasing Koi pla consumption. The community intervention was developed through efforts of community participation concerning local and environmental contexts. From the framework, the intervention was initiated on the basis of the health belief model and social cognitive theory. Members of the community agreed to take action to lower OV infection incidence emphasizing Koi pla consumption. The evaluation phase involved a process to monitor the intervention and evaluate the study outcome. Study outcome was measured by a mixed-method study design. An open-label, nonrandomized community trial was conducted to assess the effectiveness of intervention on incidence of OV infection and consumption of Koi pla.

The study flow was maintained by the community advisory board (CAB) formed by community leaders with the support of the research team. The participatory action revealed a potential self- maintenance capability for the community to sustain a community-derived health campaign.

The results indicated that community intervention could significantly reduce the incidence of OV infection by 63% (95% CI: 7 – 85), p = 0.03 and reduce Koi pla consumption by 46% (95% CI: 1 - 71), p = 0.04. Additional cohort analysis showed that Koi pla consumption could increase risk of OV infection with IRR = 2.52 (95% CI: 1.07 – 5.92) when adjusted for age-group. Additionally, social and cultural attributes were potential factors leading to Koi pla consumption.

247

6.3 Public health implication

The comprehensive epidemiological study provided up-to-date data with the aim that study results will be useful for public health policy.

6.3.1 Primary prevention

Considering primary prevention, study result provided support for public health campaign planning. Following the National Control Program guideline, health education was integrated as a strategic plan for community approach (5, 15). Providing health education was practical and did not require complex tools to conduct. However, a further study might be needed in order to explore the effectiveness of health education.

As I can observe from qualitative study, the method of knowledge distribution should be modified. The younger generation might gain benefit from social network campaigns. School-age children and young adults were very effective targets; they were capable of learning new things, transfer to their family and feasible for behavioural modification. Elders were more likely to trust medical doctors. However, local health care put more effort into providing diagnosis and treatment. I suggested that local health volunteers could play an important role to provide health education especially for older age groups when they gain more trust from being a member of healthcare personnel.

School acts as a community centre and perceived as a knowledge hub. A school-based prevention of schistosomiasis had been initiated in Tanzania and Nigeria using participatory action research (157, 158) using school as a centre for distributing intervention. Moreover, PPM framework had been used to assess the feasibility of school as an entry point for implementing the intervention of helminth control to the community in Kwazulu-Natal (138).

In this study, the role of community centre extended to the formation of CAB where the community members have shown a potential to initiate and sustain health campaign. In OV infection context, the infection was regarded as a locally-specific where the social and environmental could play a major role shaping the infection dynamics. The intervention designed for each area might be different in details, but the fundamental concept can be consolidated. The roles of predisposing, enabling and reinforcing factors from PPM were major concerns for developing particular health campaigns. The process of planning and evaluation within PPM indicated that health burden can be systematically approached with community participation.

248

6.3.2 Secondary prevention

Based on national policy, the management for OV infection switched to a passive strategy where the activity mainly focused on diagnosis and treatment (28) requiring a hospital-based setting.

However, providing routine stool examination and treatment may encounter both administrative and technical problems. The process needs a hospital-based setting was and is not yet available in the community healthcare system. Although the equipment for microscopic stool examination was convenient for an ambulatory setting, the examination method needs technically skilled health personnel to identify OV eggs and distinguish them from other parasites. The most reliable method needs more laboratory work such as formalin-ethyl acetate concentration technique or molecular technique was and is not feasible for rural settings (106). The follow-up (70) study in 2007 - 09 suggested that diagnostic methods and equipment should be improved in this community. This recommendation was not yet followed up and the local health centre was still equipped with same instruments and stool examination was under the research team support, completely. Skill training for local health staff was potentially possible, but the process would require time and government support while short-term resolutions may be considered. A nationwide survey may be too costly for a developing country. Thus, to effectively deploy resources, a screening program is preferable in high risk areas. The strategy should be capable of identifying high-risk populations especially outside the endemic areas where OV infection might be less concerned.

Mass chemotherapy was also considered a strategy to decrease the disease burden as a secondary prevention. Praziquantel was the drug of choice and provides high efficacy and fewer side effects than other anthelmintic drugs. A recent study in an endemic area in Thailand reported that previous use of praziquantel may increase risk of re-infection if the participants lacked adequate knowledge of prevention as only treatment will not prevent subsequent infection as they still continue consuming uncooked fish (6). The side effects would come from the drug itself and also a result of the immune system’s reaction to dying parasites (159) including major side effects such as drowsiness, dizziness or mild abdominal distress. The drug usage should be justified and provided for confirmed cases to avoid unnecessary collateral health risks. Moreover, high incidence would require more than effective secondary prevention for a control strategy.

249

6.4 Future work

Future work can be extended to various aspects. Within 5 years, the project will be expanded in the surrounding areas where the fieldwork has not yet been conducted. Therefore, future research could collect epidemiological data in untreated populations and also conduct a clinical trial to assess the effectiveness of treatment and intervention. Effectiveness was indirectly evaluated through mathematical modelling in this study, but clinical trials can provide more accurate data. Moreover, side effects and compliance can be assessed.

Clinical trials have been conducted in the Mekong region especially in Lao PDR where the efficacy of praziquantel and other drugs against OV infection has been evaluated (49, 53). In Thailand, the efficacy of praziquantel against OV infection was directly studied from 1980 – 1990 in the period of the National Control Program implementation (27, 28, 105). Clinical trials could be conducted in this study area to re-evaluate the effectiveness of praziquantel. Moreover, newer drugs such as tribendimidine could be assessed.

Household and public sanitation have been essentially improved where the villagers use their toilets when they were home on a regular basis. Defecation at the paddy field might remain a major problem for transmitting eggs to the environment. Considering the OV biological life cycle involving multiple intermediate hosts, interruption of uncooked fish consumption seemed a more practical intervention compared with biological control of OV (9). Many studies have indicated that reservoir hosts such as pets and livestock might have played an important role in sustained OV transmissions (160, 161).

The completeness of fieldwork data could provide more insight about the infection dynamics. The epidemiology of the direct OV life cycle including the intermediate and reservoir hosts can provide more insights of transmission dynamics. Further models will consider population dynamics such as non-random mixing with regard to age pattern and contact pattern and density dependence should be further incorporated in the model.

Bayesian geostatistical modelling was introduced to explore the epidemiology of Schistosomiasis (100) and OV infection (14). The Bayesian approach provides an effort for handling uncertainty. The spatial geostatistical model under the Bayesian formulation was used to build spatial correlation on the distribution of random effects. The model helps in creating risk maps and is capable of predicting the infection in non-survey areas, which was useful for targeting the intervention or health campaign. 250

OV infection was known to be locally-specific (14). Within adjacent areas, the infection dynamics (and thus the prevalence) can vary considerably. The spatial geostatistical model was a tool to explore the potential of spatial property, which influences the infection and its risk factors. A study in southern Lao PDR indicated that strong spatial heterogeneity was found among study areas. The environmental factors played an important role shaping the infection dynamics (51) by incorporating spatial factors in the disease model to provide more accuracy to the results and also improve cost- effectiveness in targeting areas with interventions.

Compound models were important for handling over-dispersed data. The hurdle model provided a rational interpretation of infection data with regard to diagnostic method. A zero-inflated model had been developed with the Bayesian framework (100) for mapping Schistosomiasis transmission. Along with geostatistical modelling, the Bayesian geostatistical zero-inflated regression model showed a better fit than the spatial negative binomial model. Future research will involve collecting spatial data from fieldwork to create OV infection risk maps based on the Bayesian framework. The compound models (zero-inflated and hurdle models) will be incorporated in the mapping model. The result will aid the health authorities to design and revise public health campaigns.

Diagnostic methods used in fieldwork can result in uncertainties and inaccurate reporting of infection. The latent class (LC) model was used to compare the accuracy of 5 diagnostic measures for Schistosoma haematobium infection in Ghana (162) providing more accurate sensitivity and specificity. The comparison results in an ability to identify the most accurate test among multiple methods when the gold standard test was not yet identified. Furthermore, more accurate diagnostic results can be further applied to more accurate reporting of infection prevalence, which was eventually substantial for epidemiological studies.

For OV infection, the considered “gold standard” can refer to direct identification of Opisthorchis viverrini eggs in stool specimen or bile. The morphology of these eggs was distinct from other helminths. However, the eggs were very similar to minute intestinal flukes (MIF). Reports of co- infections of OV infection and MIF were made along the Mekong basin (156). Therefore, prevalence can be overestimated due to the morphological similarity.

The serological techniques were constantly improved for accuracy over time (15). However, a cross- reaction problem remains for other flukes such as Clonorchis sinensis. Moreover, the differentiation of previous and current infections can be problematic due to the long half-life of the immune response (163). PCR, a molecular technique, provides nearly 100% specificity and ability to distinguish Opisthorchis viverrini, Clonorchis sinensis and Opisthorchis felineus (164, 165). However, 251

both serological and molecular technique may not be practical for ambulatory settings for rural fieldwork in Thailand. The microscopic-based diagnosis was still feasible for this research setting.

Therefore, LC models could be used to validate the diagnostic test in the absence of a gold standard method in the field. Future research will use the LC model for a test validation and re-assess the prevalence of OV infection.

Finally, the qualitative study showed a potential result with respect to infection dynamics. The contribution of the planning model also encourages community participation to engage the health problem. Further fieldwork will then be incorporated with the qualitative approach so the problem can be carefully assessed in all aspects. The community intervention package will be introduced to health policy as a product of rigorous research.

252

6.5 Conclusion

The overall outcome was valuable for Thailand National Health Policy and epidemiological data will provide the basis for further rigorous academic research. The study highlights the epidemiological view of OV infection in the rural community setting, the majority part of Thailand. Primary data from the field study could be useful for extensive epidemiological tools such as mathematical and statistical modelling to gain more insights into the complexity of OV biology and disease risk relationship providing greater understanding of the infection dynamics and parasite-host relationship. Disease surveillance might be useful for the high endemic area of OV infection where the high risk population could be primarily identified by the history of Koi pla consumption. The qualitative approach provides more understanding of the disease dynamics in bio-psycho-social aspects. The National Control Program should be maintained as its core principle is fundamental but the strategy needs to be adapted for social dynamics, which have changed over time especially among different generations. Koi pla consumption was clearly identified as a potential risk factor for acquiring OV infection, which must be interpreted as a primary prevention and should be a prioritized major concern. For secondary prevention, selective treatment for infected cases was shown to be effective where the stool examination proved positive. The role of mass chemotherapy should be re-evaluated and justified for effectiveness. Re-infection should be evaluated as well to determine the magnitude of the OV infection burden. Up-to-date information on disease dynamics contributes a potential impact on public health and further ongoing extensive research in this field is essential to provide effective public health management.

253

References

1. Conlan JV, Sripa B, Attwood S, Newton PN. A review of parasitic zoonoses in a changing Southeast Asia. Vet Parasitol. 2011;182(1):22-40. 2. Keiser J, Utzinger J. Food-borne trematodiases. Clin Microbiol Rev. 2009;22(3):466-83. 3. Chai JY, Darwin Murrell K, Lymbery AJ. Fish-borne parasitic zoonoses: status and issues. Int J Parasitol. 2005;35(11-12):1233-54. 4. Macpherson CN. Human behaviour and the epidemiology of parasitic zoonoses. Int J Parasitol. 2005;35(11-12):1319-31. 5. Sithithaworn P, Andrews RH, Nguyen VD, Wongsaroj T, Sinuon M, Odermatt P, et al. The current status of opisthorchiasis and clonorchiasis in the Mekong Basin. Parasitol Int. 2012;61(1):10- 6. 6. Saengsawang P, Promthet S, Bradshaw P. Infection with Opisthorchis viverrini and use of praziquantel among a working-age population in northeast Thailand. Asian Pac J Cancer Prev. 2013;14(5):2963-6. 7. Yong TS, Shin EH, Chai JY, Sohn WM, Eom KS, Lee DM, et al. High prevalence of Opisthorchis viverrini infection in a riparian population in Takeo Province, Cambodia. Korean J Parasitol. 2012;50(2):173-6. 8. Saengsawang P, Promthet S, Bradshaw P. Prevalence of OV infection in Yasothon Province, Northeast Thailand. Asian Pac J Cancer Prev. 2012;13(7):3399-402. 9. Sithithaworn P, Haswell-Elkins M. Epidemiology of Opisthorchis viverrini. Acta Trop. 2003;88(3):187-94. 10. Kaewpitoon N, Kaewpitoon SJ, Pengsaa P. Opisthorchiasis in Thailand: review and current status. World J Gastroenterol. 2008;14(15):2297-302. 11. Andrews RH, Sithithaworn P, Petney TN. Opisthorchis viverrini: an underestimated parasite in world health. Trends Parasitol. 2008;24(11):497-501. 12. Sripa B. Pathobiology of opisthorchiasis: an update. Acta Trop. 2003;88(3):209-20. 13. Kaewkes S. Taxonomy and biology of liver flukes. Acta Trop. 2003;88(3):177-86. 14. Furst T, Duthaler U, Sripa B, Utzinger J, Keiser J. Trematode infections: liver and lung flukes. Infectious disease clinics of North America. 2012;26(2):399-419. 15. Sripa B, Bethony JM, Sithithaworn P, Kaewkes S, Mairiang E, Loukas A, et al. Opisthorchiasis and Opisthorchis-associated cholangiocarcinoma in Thailand and Laos. Acta Trop. 2011;120 Suppl 1:S158-68. 16. Kaewpitoon N, Kaewpitoon SJ, Pengsaa P, Sripa B. Opisthorchis viverrini: the carcinogenic human liver fluke. World J Gastroenterol. 2008;14(5):666-74. 17. Sripa B, Kaewkes S, Sithithaworn P, Mairiang E, Laha T, Smout M, et al. Liver fluke induces cholangiocarcinoma. PLoS Med. 2007;4(7):e201. 18. Songserm N, Promthet S, Sithithaworn P, Pientong C, Ekalaksananan T, Chopjitt P, et al. Risk factors for cholangiocarcinoma in high-risk area of Thailand: Role of lifestyle, diet and methylenetetrahydrofolate reductase polymorphisms. Cancer Epidemiol. 2011. 19. Vennervald BJ, Polman K. Helminths and malignancy. Parasite Immunol. 2009;31(11):686-96. 20. Sripa B, Pairojkul C. Cholangiocarcinoma: lessons from Thailand. Curr Opin Gastroenterol. 2008;24(3):349-56. 21. Wongratanacheewin S, Sermswan RW, Sirisinha S. Immunology and molecular biology of Opisthorchis viverrini infection. Acta Trop. 2003;88(3):195-207. 22. Smout MJ, Sripa B, Laha T, Mulvenna J, Gasser RB, Young ND, et al. Infection with the carcinogenic human liver fluke, Opisthorchis viverrini. Mol Biosyst. 2011;7(5):1367-75. 23. Fried B, Reddy A, Mayer D. Helminths in human carcinogenesis. Cancer Lett. 2011;305(2):239-49. 254

24. Poomphakwaen K, Promthet S, Kamsa-Ard S, Vatanasapt P, Chaveepojnkamjorn W, Klaewkla J, et al. Risk factors for cholangiocarcinoma in Khon Kaen, Thailand: a nested case-control study. Asian Pac J Cancer Prev. 2009;10(2):251-8. 25. Lim JH. Liver flukes: the malady neglected. Korean J Radiol. 2011;12(3):269-79. 26. Jongsuksuntigul P, Imsomboon T. The impact of a decade long opisthorchiasis control program in northeastern Thailand. Southeast Asian J Trop Med Public Health. 1997;28(3):551-7. 27. Jongsuksuntigul P, Imsomboon T. Epidemiology of opisthorchiasis and national control program in Thailand. Southeast Asian J Trop Med Public Health. 1998;29(2):327-32. 28. Jongsuksuntigul P, Imsomboon T. Opisthorchiasis control in Thailand. Acta Trop. 2003;88(3):229-32. 29. Grundy-Warr C, Andrews RH, Sithithaworn P, Petney TN, Sripa B, Laithavewat L, et al. Raw attitudes, wetland cultures, life-cycles: socio-cultural dynamics relating to Opisthorchis viverrini in the Mekong Basin. Parasitol Int. 2012;61(1):65-70. 30. Upatham ES, Viyanant V. Opisthorchis viverrini and opisthorchiasis: a historical review and future perspective. Acta Trop. 2003;88(3):171-6. 31. Schuster RK. Opisthorchiidosis--a review. Infect Disord Drug Targets. 2010;10(5):402-15. 32. Brockelman WY, Upatham ES, Viyanant V, Ardsungnoen S, Chantanawat R. Field studies on the transmission of the human liver fluke, Opisthorchis viverrini, in northeast Thailand: population changes of the snail intermediate host. Int J Parasitol. 1986;16(5):545-52. 33. Haas W, Granzer M, Brockelman CR. Opisthorchis viverrini: finding and recognition of the fish host by the cercariae. Exp Parasitol. 1990;71(4):422-31. 34. Petney T, Sithithaworn P, Andrews R, Kiatsopit N, Tesana S, Grundy-Warr C, et al. The ecology of the Bithynia first intermediate hosts of Opisthorchis viverrini. Parasitol Int. 2012;61(1):38- 45. 35. Saijuntha W, Sithithaworn P, Wongkham S, Laha T, Pipitgool V, Tesana S, et al. Evidence of a species complex within the food-borne trematode Opisthorchis viverrini and possible co-evolution with their first intermediate hosts. Int J Parasitol. 2007;37(6):695-703. 36. Touch S, Komalamisra C, Radomyos P, Waikagul J. Discovery of Opisthorchis viverrini metacercariae in freshwater fish in southern Cambodia. Acta Trop. 2009;111(2):108-13. 37. Ngern-klun R, Sukontason KL, Tesana S, Sripakdee D, Irvine KN, Sukontason K. Field investigation of Bithynia funiculata, intermediate host of Opisthorchis viverrini in northern Thailand. Southeast Asian J Trop Med Public Health. 2006;37(4):662-72. 38. Vichasri S, Viyanant V, Upatham ES. Opisthorchis viverrini : intensity and rates of infection in cyprinoid fish from an endemic focus in Northeast Thailand. Southeast Asian J Trop Med Public Health. 1982;13(1):138-41. 39. Sadun EH. Studies on Opisthorchis viverrini in Thailand. American journal of hygiene. 1955;62(2):81-115. 40. Upatham ES, Viyanant V, Brockelman WY, Kurathong S, Lee P, Kraengraeng R. Rate of re- infection by Opisthorchis viverrini in an endemic northeast Thai community after chemotherapy. Int J Parasitol. 1988;18(5):643-9. 41. Ghouri YA, Mian I, Blechacz B. Cancer review: Cholangiocarcinoma. Journal of carcinogenesis. 2015;14:1. 42. Torgerson PR, Macpherson CN. The socioeconomic burden of parasitic zoonoses: global trends. Vet Parasitol. 2011;182(1):79-95. 43. Furst T, Keiser J, Utzinger J. Global burden of human food-borne trematodiasis: a systematic review and meta-analysis. Lancet Infect Dis. 2012;12(3):210-21. 44. Rim HJ, Sohn WM, Yong TS, Eom KS, Chai JY, Min DY, et al. Fishborne trematode metacercariae in Luang Prabang, Khammouane, and Saravane Province, Lao PDR. Korean J Parasitol. 2013;51(1):107-14. 45. Harinasuta C, Harinasuta T. Opisthorchis viverrini: life cycle, intermediate hosts, transmission to man and geographical distribution in Thailand. Arzneimittel-Forschung. 1984;34(9B):1164-7. 255

46. Traub RJ, Macaranas J, Mungthin M, Leelayoova S, Cribb T, Murrell KD, et al. A new PCR- based approach indicates the range of Clonorchis sinensis now extends to Central Thailand. PLoS Negl Trop Dis. 2009;3(1):e367. 47. Prasongwatana J, Laummaunwai P, Boonmars T, Pinlaor S. Viable metacercariae of Opisthorchis viverrini in northeastern Thai cyprinid fish dishes--as part of a rational program for control of O. viverrini-associated cholangiocarcinoma. Parasitol Res. 2013;112(3):1323-7. 48. Jongsuksuntigul P, Manatrakul D, Wongsaroj T, Krisanamara P, Sawatdimongkol S, Wongsaroj S. Evaluation of the Helminthiasis Control Program in Thailand at the End of the 8th Health Development Plan, 2001. J Trop Med Parasitol. 2003;26:18-45. 49. Lovis L, Mak TK, Phongluxa K, Aye Soukhathammavong P, Vonghachack Y, Keiser J, et al. Efficacy of praziquantel against Schistosoma mekongi and Opisthorchis viverrini: a randomized, single-blinded dose-comparison trial. PLoS Negl Trop Dis. 2012;6(7):e1726. 50. Glinz D, Silue KD, Knopp S, Lohourignon LK, Yao KP, Steinmann P, et al. Comparing diagnostic accuracy of Kato-Katz, Koga agar plate, ether-concentration, and FLOTAC for Schistosoma mansoni and soil-transmitted helminths. PLoS Negl Trop Dis. 2010;4(7):e754. 51. Forrer A, Sayasone S, Vounatsou P, Vonghachack Y, Bouakhasith D, Vogt S, et al. Spatial distribution of, and risk factors for, Opisthorchis viverrini infection in southern Lao PDR. PLoS Negl Trop Dis. 2012;6(2):e1481. 52. Kobayashi J, Vannachone B, Sato Y, Manivong K, Nambanya S, Inthakone S. An epidemiological study on Opisthorchis viverrini infection in Lao villages. Southeast Asian J Trop Med Public Health. 2000;31(1):128-32. 53. Soukhathammavong P, Odermatt P, Sayasone S, Vonghachack Y, Vounatsou P, Hatz C, et al. Efficacy and safety of mefloquine, artesunate, mefloquine-artesunate, tribendimidine, and praziquantel in patients with Opisthorchis viverrini: a randomised, exploratory, open-label, phase 2 trial. Lancet Infect Dis. 2011;11(2):110-8. 54. De NV, Murrell KD, Cong le D, Cam PD, Chau le V, Toan ND, et al. The food-borne trematode zoonoses of Vietnam. Southeast Asian J Trop Med Public Health. 2003;34 Suppl 1:12-34. 55. Sripa B, Kaewkes S, Intapan PM, Maleewong W, Brindley PJ. Food-borne trematodiases in Southeast Asia epidemiology, pathology, clinical manifestation and control. Adv Parasitol. 2010;72:305-50. 56. Kerr A. Intestinal parasites in northern Siam. Trans Soc Trop Med. 1916;9:82-9. 57. Mongkolintra S, Leelapanmetha P, Sornmani S, Fungladda W, Siripornpitak S, Siriwichai P, et al. Encouraging self-reliance in the fight against liver fluke. World health forum. 1988;9(3):341-4. 58. Kurathong S, Lerdverasirikul P, Wongpaitoon V, Pramoolsinsap C, Upatham ES. Opisthorchis viverrini infection in rural and urban communities in northeast Thailand. Trans R Soc Trop Med Hyg. 1987;81(3):411-4. 59. Tesana S, Sithithaworn P, Prasongwatana J, Kaewkes S, Pipitgool V, Pientong C. Influence of water current on the distribution of Opisthorchis viverrini infection in northeastern villages of Thailand. Southeast Asian J Trop Med Public Health. 1991;22(1):93-8. 60. Sithithaworn P, Tesana S, Pipitgool V, Kaewkes S, Pairojkul C, Sripa B, et al. Relationship between faecal egg count and worm burden of Opisthorchis viverrini in human autopsy cases. Parasitology. 1991;102 Pt 2:277-81. 61. Maleewong W, Intapan P, Wongwajana S, Sitthithaworn P, Pipitgool V, Wongkham C, et al. Prevalence and intensity of Opisthorchis viverrini in rural community near the Mekong River on the Thai-Laos border in northeast Thailand. J Med Assoc Thai. 1992;75(4):231-5. 62. Radomyos B, Wongsaroj T, Wilairatana P, Radomyos P, Praevanich R, Meesomboon V, et al. Opisthorchiasis and intestinal fluke infections in northern Thailand. Southeast Asian J Trop Med Public Health. 1998;29(1):123-7. 63. Radomyos P, Radomyos B, Tungtrongchitr A. Multi-infection with helminths in adults from northeast Thailand as determined by post-treatment fecal examination of adult worms. Trop Med Parasitol. 1994;45(2):133-5. 256

64. Sriamporn S, Pisani P, Pipitgool V, Suwanrungruang K, Kamsa-ard S, Parkin DM. Prevalence of Opisthorchis viverrini infection and incidence of cholangiocarcinoma in Khon Kaen, Northeast Thailand. Trop Med Int Health. 2004;9(5):588-94. 65. Waree P, Polseela P, Pannarunothai S, Pipitgool V. The present situation of paragonimiasis in endemic area in Phitsanulok Province. Southeast Asian J Trop Med Public Health. 2001;32 Suppl 2:51-4. 66. Wiwanitkit V, Suwansaksri J, Chaiyakhun Y. High prevalence of Fasciolopsis buski in an endemic area of liver fluke infection in Thailand. MedGenMed : Medscape general medicine. 2002;4(3):6. 67. Waikagul J, Dekumyoy P, Chaichana K, Thairungroje Anantapruti M, Komalamisra C, Kitikoon V. Serodiagnosis of human opisthorchiasis using cocktail and electroeluted Bithynia snail antigens. Parasitol Int. 2002;51(3):237-47. 68. Saksirisampant W, Wiwanitkit V, Akrabovorn P, Nuchprayoon S. Parasitic infections in Thai workers that pursue overseas employment: the need for a screening program. Southeast Asian J Trop Med Public Health. 2002;33 Suppl 3:110-2. 69. Rangsin R, Mungthin M, Taamasri P, Mongklon S, Aimpun P, Naaglor T, et al. Incidence and risk factors of Opisthorchis viverrini infections in a rural community in Thailand. Am J Trop Med Hyg. 2009;81(1):152-5. 70. Suwannahitatorn P, Klomjit S, Naaglor T, Taamasri P, Rangsin R, Leelayoova S, et al. A follow- up study of Opisthorchis viverrini infection after the implementation of control program in a rural community, central Thailand. Parasit Vectors. 2013;6:188. 71. Tungtrongchitr A, Chiworaporn C, Praewanich R, Radomyos P, Boitano JJ. The potential usefulness of the modified Kato thick smear technique in the detection of intestinal sarcocystosis during field surveys. Southeast Asian J Trop Med Public Health. 2007;38(2):232-8. 72. Kaewpitoon SJ, Rujirakul R, Ueng-Arporn N, Matrakool L, Namwichaisiriku N, Churproong S, et al. Community-based cross-sectional study of carcinogenic human liver fluke in elderly from Surin province, Thailand. Asian Pac J Cancer Prev. 2012;13(9):4285-8. 73. Kaewpitoon SJ, Rujirakul R, Kaewpitoon N. Prevalence of Opisthorchis viverrini infection in Nakhon Ratchasima province, Northeast Thailand. Asian Pac J Cancer Prev. 2012;13(10):5245-9. 74. Hira PR, Al-Enizi AA, Al-Kandari S, Behbehani K. Opisthorchiasis in Kuwait: first report of infections in Thai migrant workers in the Arabian Gulf. Annales de la Societe belge de medecine tropicale. 1987;67(4):363-8. 75. Peng HW, Chao HL, Fan PC. Imported Opisthorchis viverrini and parasite infections from Thai labourers in Taiwan. J Helminthol. 1993;67(2):102-6. 76. Cheng HS, Shieh YH. Investigation on subclinical aspects related to intestinal parasitic infections among Thai laborers in Taipei. Journal of travel medicine. 2000;7(6):319-24. 77. Lo CT, Lee KM. Intestinal parasites among the Southeast Asian laborers in Taiwan during 1993-1994. Zhonghua yi xue za zhi = Chinese medical journal; Free China ed. 1996;57(6):401-4. 78. Greenberg Z, Giladi L, Bashary A, Zahavi H. [Prevalence of intestinal parasites among Thais in Israel]. Harefuah. 1994;126(9):507-9, 63. 79. Wang LC. Parasitic infections among Southeast Asian labourers in Taiwan: a long-term study. Epidemiol Infect. 1998;120(1):81-6. 80. Wang LC. Changing patterns in intestinal parasitic infections among Southeast Asian laborers in Taiwan. Parasitol Res. 2004;92(1):18-21. 81. Maneeboonyang W, Charusabha C, Prommongkol S, Puangsa-art S, Thanyavanich N, Buchachart K. Status of intestinal helminth infections among thai troops working along Thai- Myanmar border, Suanphung, Ratchaburi Thailand. J Trop Med Parasitol. 2004;27:59-64. 82. Upatham ES, Brockelman WY, Viyanant V, Lee P, Kaengraeng R, Prayoonwiwat B. Incidence of endemic Opisthorchis viverrini infection in a village in northeast Thailand. Am J Trop Med Hyg. 1985;34(5):903-6. 257

83. Brockelman WY, Upatham ES, Viyanant V, Hirunraks A. Measurement of incidence of the human liver fluke, Opisthorchis viverrini, in northeast Thailand. Trans R Soc Trop Med Hyg. 1987;81(2):327-35. 84. Sornmani S, Schelp FP, Vivatanasesth P, Patihatakorn W, Impand P, Sitabutra P, et al. A pilot project for controlling O. viverrini infection in Nong Wai, Northeast Thailand, by applying praziquantel and other measures. Arzneimittel-Forschung. 1984;34(9B):1231-4. 85. Ramsay RJ, Sithithaworn P, Prociv P, Moorhouse DE, Methaphat C. Density-dependent fecundity of Opisthorchis viverrini in humans, based on faecal recovery of flukes. Trans R Soc Trop Med Hyg. 1989;83(2):241-2. 86. Haswell-Elkins MR, Elkins DB, Sithithaworn P, Treesarawat P, Kaewkes S. Distribution patterns of Opisthorchis viverrini within a human community. Parasitology. 1991;103 Pt 1:97-101. 87. Sukontason K, Methanitikorn R, Piangjai S, Choochote W. Viability of metacercariae in northern Thai traditional foods. Southeast Asian J Trop Med Public Health. 1998;29(4):714-6. 88. Wiwanitkit V, Nithiuthai S, Suwansaksri J. Motility of minute intestinal fluke, Haplorchinae spp, metacercariae in fish dishes prepared by different uncooked methods. MedGenMed : Medscape general medicine. 2002;4(1):8. 89. Kruatrachue M, Chitramvong YP, Upatham ES, Vichasri S, Viyanant V. Effects of physico- chemical factors on the infection of hamsters by metacercariae of Opisthorchis viverrini. Southeast Asian J Trop Med Public Health. 1982;13(4):614-7. 90. Upatham ES, Viyanant V, Kurathong S, Rojborwonwitaya J, Brockelman WY, Ardsungnoen S, et al. Relationship between prevalence and intensity of Opisthorchis viverrini infection, and clinical symptoms and signs in a rural community in north-east Thailand. Bull World Health Organ. 1984;62(3):451-61. 91. Haswell-Elkins MR, Mairiang E, Mairiang P, Chaiyakum J, Chamadol N, Loapaiboon V, et al. Cross-sectional study of Opisthorchis viverrini infection and cholangiocarcinoma in communities within a high-risk area in northeast Thailand. Int J Cancer. 1994;59(4):505-9. 92. Songserm N, Promthet S, Sithithaworn P, Pientong C, Ekalaksananan T, Chopjitt P, et al. Risk factors for cholangiocarcinoma in high-risk area of Thailand: role of lifestyle, diet and methylenetetrahydrofolate reductase polymorphisms. Cancer Epidemiol. 2012;36(2):e89-94. 93. Bundy DA, Cooper ES, Thompson DE, Anderson RM, Didier JM. Age-related prevalence and intensity of Trichuris trichiura infection in a St. Lucian community. Trans R Soc Trop Med Hyg. 1987;81(1):85-94. 94. Bundy DA, Kan SP, Rose R. Age-related prevalence, intensity and frequency distribution of gastrointestinal helminth infection in urban slum children from Kuala Lumpur, Malaysia. Trans R Soc Trop Med Hyg. 1988;82(2):289-94. 95. Chipeta MG, Ngwira B, Kazembe LN. Analysis of Schistosomiasis haematobium infection prevalence and intensity in Chikhwawa, Malawi: an application of a two part model. PLoS Negl Trop Dis. 2013;7(3):e2131. 96. Saffari SE, Adnan R, Greene W. Hurdle negative binomial regression model with right censored count data. SORT 2012, vol 36, núm 2. 2012. 97. Hu MC, Pavlicova M, Nunes EV. Zero-inflated and hurdle models of count data with extra zeros: examples from an HIV-risk reduction intervention trial. The American journal of drug and alcohol abuse. 2011;37(5):367-75. 98. Dang TC, Yajima A, Nguyen VK, Montresor A. Prevalence, intensity and risk factors for clonorchiasis and possible use of questionnaires to detect individuals at risk in northern Vietnam. Trans R Soc Trop Med Hyg. 2008;102(12):1263-8. 99. Anderson RM, May RM. Population dynamics of human helminth infections: control by chemotherapy. Nature. 1982;297(5867):557-63. 100. Vounatsou P, Raso G, Tanner M, N'Goran E K, Utzinger J. Bayesian geostatistical modelling for mapping schistosomiasis transmission. Parasitology. 2009;136(13):1695-705. 258

101. Walker M, Hall A, Anderson RM, Basanez MG. Density-dependent effects on the weight of female Ascaris lumbricoides infections of humans and its impact on patterns of egg production. Parasit Vectors. 2009;2(1):11. 102. Zeileis A, Kleiber C, Jackman S. Regression Models for Count Data in R. 2008. 2008;27(8):25. 103. McCullagh P, Nelder JA. Generalized Linear Models, Second Edition: Taylor & Francis; 1989. 104. Bergquist R, Johansen MV, Utzinger J. Diagnostic dilemmas in helminthology: what tools to use and when? Trends Parasitol. 2009;25(4):151-6. 105. Saowakontha S, Pipitgool V, Pariyanonda S, Tesana S, Rojsathaporn K, Intarakhao C. Field trials in the control of Opisthorchis viverrini with an integrated programme in endemic areas of northeast Thailand. Parasitology. 1993;106 ( Pt 3):283-8. 106. Keiser J, Duthaler U, Utzinger J. Update on the diagnosis and treatment of food-borne trematode infections. Curr Opin Infect Dis. 2010;23(5):513-20. 107. Truant AL, Elliott SH, Kelly MT, Smith JH. Comparison of formalin-ethyl ether sedimentation, formalin-ethyl acetate sedimentation, and zinc sulfate flotation techniques for detection of intestinal parasites. J Clin Microbiol. 1981;13(5):882-4. 108. Qian MB, Yap P, Yang YC, Liang H, Jiang ZH, Li W, et al. Accuracy of the Kato-Katz method and formalin-ether concentration technique for the diagnosis of Clonorchis sinensis, and implication for assessing drug efficacy. Parasit Vectors. 2013;6(1):314. 109. Tesana S, Srisawangwonk T, Kaewkes S, Sithithaworn P, Kanla P, Arunyanart C. Eggshell morphology of the small eggs of human trematodes in Thailand. Southeast Asian J Trop Med Public Health. 1991;22(4):631-6. 110. Lovis L, Mak TK, Phongluxa K, Soukhathammavong P, Sayasone S, Akkhavong K, et al. PCR Diagnosis of Opisthorchis viverrini and Haplorchis taichui Infections in a Lao Community in an area of endemicity and comparison of diagnostic methods for parasitological field surveys. J Clin Microbiol. 2009;47(5):1517-23. 111. Upatham ES, Viyanant V, Kurathong S, Brockelman WY, Menaruchi A, Saowakontha S, et al. Morbidity in relation to intensity of infection in Opisthorchiasis viverrini: study of a community in Khon Kaen, Thailand. Am J Trop Med Hyg. 1982;31(6):1156-63. 112. Anderson RM, May RM. Population biology of infectious diseases: Part I. Nature. 1979;280(5721):361-7. 113. May RM, Anderson RM. Population biology of infectious diseases: Part II. Nature. 1979;280(5722):455-61. 114. Strandgaard H, Johansen MV, Aagaard-Hansen J, Petlueng P, Ornbjerg N. Local perceptions and practices in regard to opisthorchiasis in two villages in Lao PDR. Southeast Asian J Trop Med Public Health. 2008;39(1):19-26. 115. Brooker S, Alexander N, Geiger S, Moyeed RA, Stander J, Fleming F, et al. Contrasting patterns in the small-scale heterogeneity of human helminth infections in urban and rural environments in Brazil. Int J Parasitol. 2006;36(10-11):1143-51. 116. Anderson RM, May RM. Infectious Diseases of Humans: Dynamics and Control: OUP Oxford; 1992. 117. Grassly NC, Fraser C. Mathematical models of infectious disease transmission. Nat Rev Microbiol. 2008;6(6):477-87. 118. Dobson AP. The population biology of parasite-induced changes in host behavior. Q Rev Biol. 1988;63(2):139-65. 119. Furst T, Silue KD, Ouattara M, N'Goran DN, Adiossan LG, N'Guessan Y, et al. Schistosomiasis, soil-transmitted helminthiasis, and sociodemographic factors influence quality of life of adults in Cote d'Ivoire. PLoS Negl Trop Dis. 2012;6(10):e1855. 120. Feng Z, Eppert A, Milner FA, Minchella DJ. Estimation of parameters governing the transmission dynamics of schistosomes. Applied Mathematics Letters. 2004;17(10):1105-12. 121. Barbour AD. Modeling the transmission of schistosomiasis: an introductory view. Am J Trop Med Hyg. 1996;55(5 Suppl):135-43. 259

122. Chiyaka ET, Magombedze G, Mutimbu L. Modelling within host parasite dynamics of schistosomiasis. Comput Math Methods Med. 2010;11(3):255-80. 123. Woolhouse ME. On the application of mathematical models of schistosome transmission dynamics. I. Natural transmission. Acta Trop. 1991;49(4):241-70. 124. Gordis L. Epidemiology: Elsevier Health Sciences; 2013. 125. Akai PS, Pungpak S, Kitikoon V, Bunnag D, Befus AD. Possible protective immunity in human opisthorchiasis. Parasite Immunol. 1994;16(6):279-88. 126. Haswell-Elkins MR, Sithithaworn P, Mairiang E, Elkins DB, Wongratanacheewin S, Kaewkes S, et al. Immune responsiveness and parasite-specific antibody levels in human hepatobiliary disease associated with Opisthorchis viverrini infection. Clin Exp Immunol. 1991;84(2):213-8. 127. Wongratanacheewin S, Charupatana C, Bunnag D, Sirisinha S. Effect of praziquantel treatment on antibody levels and lymphoproliferative responses in patients with opisthorchiasis. Southeast Asian J Trop Med Public Health. 1988;19(1):109-16. 128. Glanz K, Rimer BK, Viswanath K. Health Behavior and Health Education: Theory, Research, and Practice: Wiley; 2008. 129. Green J, Thorogood N. Qualitative Methods for Health Research: SAGE Publications; 2004. 130. Giesecke J. Modern Infectious Disease Epidemiology, Second Edition: Taylor & Francis; 2001. 131. Petney TN. Environmental, cultural and social changes and their influence on parasite infections. Int J Parasitol. 2001;31(9):919-32. 132. Kaewpitoon N, Kaewpitoon SJ, Pengsaa P, Pilasri C. Knowledge, attitude and practice related to liver fluke infection in northeast Thailand. World J Gastroenterol. 2007;13(12):1837-40. 133. Crosby R, Noar SM. What is a planning model? An introduction to PRECEDE-PROCEED. Journal of public health dentistry. 2011;71 Suppl 1:S7-15. 134. Li Y, Cao J, Lin H, Li D, Wang Y, He J. Community health needs assessment with precede- proceed model: a mixed methods study. BMC health services research. 2009;9:181. 135. Mazur J, Szumska-Olczak A. [Health promotion indicators - selected model approaches]. Medycyna wieku rozwojowego. 2000;4(4 Suppl 1):53-66. 136. Green LW, Kreuter MW. Health Program Planning: An Educational and Ecological Approach: McGraw-Hill; 2004. 137. Joshi DD, Poudyal PM, Jimba M, Mishra PN, Neave LA, Maharjan M. Controlling Taenia solium in Nepal using the PRECEDE-PROCEED model. Southeast Asian J Trop Med Public Health. 2001;32 Suppl 2:94-7. 138. Taylor M, Coovadia HM, Kvalsvig JD, Jinabhai CC, Reddy P. Helminth control as an entry point for health-promoting schools in KwaZulu-Natal. South African medical journal = Suid-Afrikaanse tydskrif vir geneeskunde. 1999;89(3):273-9. 139. Goodson P. Theory in Health Promotion Research and Practice: Thinking Outside the Box: Jones & Bartlett Learning; 2010. 140. Conner M, Norman P. Predicting Health Behaviour: McGraw-Hill Education; 2005. 141. Glanz K, Bishop DB. The role of behavioral science theory in development and implementation of public health interventions. Annual review of public health. 2010;31:399-418. 142. Glanz K, Rimer BK, Institute NC. Theory at a glance: a guide for health promotion practice: U.S. Dept. of Health and Human Services, Public Health Service, National Institutes of Health, National Cancer Institute; 1997. 143. Hochbaum GM. Public Participation in Medical Screening Programs: A Socio-psychological Study: U.S. Department of Health, Education, and Welfare, Public Health Service, Bureau of State Services, Division of Special Health Services, Tuberculosis Program; 1958. 144. Rosenstock IM. The Health Belief Model and Preventive Health Behavior. Health Education & Behavior. 1974;2(4):354-86. 145. Will JC, Farris RP, Sanders CG, Stockmyer CK, Finkelstein EA. Health promotion interventions for disadvantaged women: overview of the WISEWOMAN projects. J Womens Health (Larchmt). 2004;13(5):484-502. 260

146. Bandura A. Social foundations of thought and action: a social cognitive theory: Prentice-Hall; 1986. 147. Hebert JR, Clemow L, Pbert L, Ockene IS, Ockene JK. Social desirability bias in dietary self- report may compromise the validity of dietary intake measures. Int J Epidemiol. 1995;24(2):389-98. 148. Holmberg L, Ohlander EM, Byers T, Zack M, Wolk A, Bruce A, et al. A search for recall bias in a case-control study of diet and breast cancer. Int J Epidemiol. 1996;25(2):235-44. 149. Friedenreich CM, Howe GR, Miller AB. An investigation of recall bias in the reporting of past food intake among breast cancer cases and controls. Annals of epidemiology. 1991;1(5):439-53. 150. Krumpal I. Determinants of social desirability bias in sensitive surveys: a literature review. Qual Quant. 2013;47(4):2025-47. 151. DePaulo BM, Kashy DA, Kirkendol SE, Wyer MM, Epstein JA. Lying in everyday life. Journal of personality and social psychology. 1996;70(5):979-95. 152. Holtgraves T. Social desirability and self-reports: testing models of socially desirable responding. Personality & social psychology bulletin. 2004;30(2):161-72. 153. Holtgraves T, Eck J, Lasky B. Face Management, Question Wording, and Social Desirability1. Journal of Applied Social Psychology. 1997;27(18):1650-71. 154. Bradburn NM, Sudman S, Blair E. Improving interview method and questionnaire design: Jossey-Bass; 1979. 155. Marsden PV, Wright JD. Handbook of Survey Research: Emerald; 2010. 156. Wongsawad C, Phalee A, Noikong W, Chuboon S, Nithikathkul C. Co-infection with Opisthorchis viverrini and Haplorchis taichui detected by human fecal examination in Chomtong district, Chiang Mai Province, Thailand. Parasitol Int. 2012;61(1):56-9. 157. Olaseha IO, Sridhar MK. Participatory action research: community diagnosis and intervention in controlling urinary schistosomiasis in an urban community in Ibadan, Nigeria. International quarterly of community health education. 2005;24(2):153-60. 158. Freudenthal S, Ahlberg BM, Mtweve S, Nyindo P, Poggensee G, Krantz I. School-based prevention of schistosomiasis: initiating a participatory action research project in northern Tanzania. Acta Trop. 2006;100(1-2):79-87. 159. Chai JY. Praziquantel treatment in trematode and cestode infections: an update. Infection & chemotherapy. 2013;45(1):32-43. 160. Aunpromma S, Tangkawattana P, Papirom P, Kanjampa P, Tesana S, Sripa B, et al. High prevalence of Opisthorchis viverrini infection in reservoir hosts in four districts of Khon Kaen Province, an opisthorchiasis endemic area of Thailand. Parasitol Int. 2012;61(1):60-4. 161. Enes JE, Wages AJ, Malone JB, Tesana S. Prevalence of Opisthorchis viverrini infection in the canine and feline hosts in three villages, Khon Kaen Province, northeastern Thailand. Southeast Asian J Trop Med Public Health. 2010;41(1):36-42. 162. Koukounari A, Webster JP, Donnelly CA, Bray BC, Naples J, Bosompem K, et al. Sensitivities and specificities of diagnostic tests and infection prevalence of Schistosoma haematobium estimated from data on adults in villages northwest of Accra, Ghana. Am J Trop Med Hyg. 2009;80(3):435-41. 163. Thammapalerd N, Tharavanij S, Nacapunchai D, Bunnag D, Radomyos P, Prasertsiriroj V. Detection of antibodies against Opisthorchis viverrini in patients before and after treatment with praziquantel. Southeast Asian J Trop Med Public Health. 1988;19(1):101-8. 164. Le TH, Van De N, Blair D, Sithithaworn P, McManus DP. Clonorchis sinensis and Opisthorchis viverrini: development of a mitochondrial-based multiplex PCR for their identification and discrimination. Exp Parasitol. 2006;112(2):109-14. 165. Pauly A, Schuster R, Steuber S. Molecular characterization and differentiation of opisthorchiid trematodes of the species Opisthorchis felineus (Rivolta, 1884) and Metorchis bilis (Braun, 1790) using polymerase chain reaction. Parasitol Res. 2003;90(5):409-14.

261

Appendices

Appendix 1 Supplement tables

Chapter 2

Table S-1 Uncooked fish consumption behaviors of cross-sectional study

Na-isarn Na-ngam Total Characteristic P N (%) N (%) N (%) Fish menus Chopped raw fish salad (Koi pla) No 196 (59.2) 263 (56.8) 459 (57.8) 0.49 Yes 135 (40.8) 200 (43.2) 335 (42.2) Extensively fermented fish (Pla ra) No 44 (13.3) 98 (21.2) 142 (17.9) 0.004 Yes 287 (86.7) 365 (78.8) 365 (82.1)

Table S-2 Uncooked fish consumption behaviors of follow-up study

Na-yao Na-ngam Total Characteristic P N (%) N (%) N (%) Fish menus Chopped raw fish salad (Koi pla) No 185 (55.9) 250 (60.2) 435 (58.3) 0.23 Yes 146 (44.1) 165 (39.8) 311 (41.7) Extensively fermented fish (Pla ra) No 105 (31.7) 63 (15.2) 168 (22.5) <0.001 Yes 226 (68.3) 352 (84.8) 578 (77.5)

262

Table S-3 Prevalence of OV infection

Prevalence (%)

(95%, CI) Na-isarn Na-ngam Total p Age group (as quartiles, median age) 0 – 36 3.41 0 – 35 5.13 0 – 35 4.04 Q1 (26) (0.71 – 9.64) (16) (1.90 – 10.83) (17) (1.76 – 7.81) 37 – 44 3.85 36 – 46 8.40 36 – 45 5.91 Q2 (41) (0.08 – 10.83) (41) (4.10 – 14.91) (41) (3.09 – 10.10) 45 – 56 11.36 47 – 58 9.49 46 – 57 11.17 Q3 (49) (5.59 – 19.91) (52) (4.83 – 16.34) (51) (7.21 – 16.28) ≥ 57 12.99 ≥ 59 14.41 ≥ 58 13.90 Q4 (64) (6.40 – 22.59) (65) (8.47 – 22.35) (64) (9.29 – 19.70) 7.85 9.29 8.69 Median age 44 46 45 0.48 (4.96 – 10.75) (6.64 – 11.93) (6.82 – 10.87)

Table S-4 Prevalence between completed and incomplete questionnaires group

Prevalence (%) (95% CI) Na-isarn p Na-ngam p Total p Completed questionnaire No 3.57 8.56 7.48 (1.28 – 5.87) 0.03 (5.85 – 11.27) 0.71 (5.19 – 9.77) 0.44 Yes 7.85 9.29 8.69 (4.96 – 10.75) (6.64 – 11.93) (6.82 – 10.87)

263

Table S-5 Incidence of OV infection

Incidence (/100 person-years)

(95%, CI) Na-yao Na-ngam Total p Age group (as quartiles, median age) Q1 0 – 38 4.82 0 – 34 3.29 0 – 36 4.85 (15) (1.56 – 11.25) (12) (0.90 – 8.41) (13) (2.42 – 8.67) 39 – 50 8.37 35 – 47 11.69 37 – 48 8.75 Q2 (45) (3.83 – 15.59) (42) (6.68 – 18.98) (43) (5.34 – 13.51) 51 – 60 9.20 48 – 57 4.70 49 – 59 7.80 Q3 (55) (4.41 – 16.91) (52) (1.52 – 10.96) (54) (4.62 – 12.32 ≥ 61 9.60 ≥ 58 6.64 ≥ 60 7.99 Q4 (69) (4.39 – 18.22) (64) (2.87 – 130.8) (66) (4.65 – 12.79) 7.98 6.80 7.34 Median age 50 47 48 0.52 (5.49 – 11.20) (4.68 – 9.54) (5.68 – 9.34)

Table S-6 Incidence between completed and incomplete questionnaires group

Incidence (/100 person-years) (95% CI) Na-yao p Na-ngam p Total p Completed questionnaire No 13.91 10.22 11.51 (6.01 – 27.41) 0.18 (5.10 – 18.29) 0.25 (6.93 – 18.00) 0.095 Yes 7.98 6.80 7.34 (5.49 – 11.20) (4.68 – 9.54) (5.68 – 9.34)

264

Table S-7 Univariable analysis of risk factors for acquiring OV infection of cross-sectional study

Na-isarn Na-ngam Characteristic Crude OR (95% CI) p Crude OR (95% CI) p Sex (female as reference) Male 1.01 (0.45 – 2.26) 0.98 1.80 (0.95 – 3.40) 0.07 Age group (as quartiles) Q1 as reference 0 – 36 1 0 – 35 1 Q2 37 – 44 1.13 (0.22 – 5.79) 0.88 36 – 46 1.70 (0.60 – 4.83) 0.32 Q3 45 – 56 3.63 (0.96 – 13.68) 0.057 47 – 58 1.94 (0.69 – 5.42) 0.21 Q4 ≥ 57 4.23 (1.12 – 15.98) 0.034 ≥ 59 3.12 (1.72 – 8.28) 0.023 Occupation (unemployed as reference) Others 0.54 (0.09 – 3.39) 0.51 0.52 (0.14 – 1.84) 0.31 Agriculture 1.48 (0.33 – 6.64) 0.61 1.37 (0.46 – 4.09) 0.57 Koi pla consumption Yes 2.95 (1.26 – 6.90) 0.013 2.69 (1.40 – 5.19) 0.003 Pla ra consumption Yes 1.92 (0.44 – 8.41) 0.39 1.02 (0.47 – 2.20) 0.97

Table S-8 Univariable analysis of risk factors for acquiring OV infection of follow-up study

Na-yao Na-ngam Characteristic Crude IRR (95% CI) P Crude IRR (95% CI) p Sex (female as reference) Male 1.36 (0.69 – 2.70) 0.38 1.30 (0.65 – 2.56) 0.46 Age group (as quartiles) Q1 as reference 0 – 38 1 0 – 34 1 Q2 39 – 50 1.74 (0.58 – 5.18) 0.32 35 – 47 3.56 (1.19 – 10.66) 0.023 Q3 51 – 60 1.91 (0.65 – 5.58) 0.24 48 – 57 1.43 (0.83 – 5.32) 0.60 Q4 ≥ 60 1.99 (0.67 – 5.94) 0.22 ≥ 57 2.02 (0.61 – 6.71) 0.25 Occupation (unemployed as reference) Others 0.68 (0.14 – 3.20) 0.63 0.74 (0.20 – 2.73) 0.65 Agriculture 0.91 (0.22 – 3.87) 0.90 1.08 (0.32 – 3.63) 0.90

Koi pla consumption Yes 2.53 (1.23 – 5.23) 0.012 1.61 (0.81 – 3.18) 0.17 Pla ra consumption Yes 0.93 (0.45 – 1.92) 0.84 1.79 (0.55 – 5.86) 0.34

265

Chapter 5

Table S-9 Incidence of OV infection

Incidence rate (/100 person-years) 95% CI Age group (as quartiles, median age)

0 – 27 3.20 0.39 – 11.56 (9) 28 – 45 4.91 1.01 – 14.36 (38) 46 – 57 16.01 7.68 – 29.43 (50) 46 – 57 13.10 5.66 – 25.82 (64) Median age 9.31 5.90 – 13.97 (45)

Table S-10 Impact of community intervention on Koi pla consumption

Incidence (%) of Study group Koi pla consumption RR (95% CI) p (95% CI) 14.12 Intervention 0.54 (0.29 – 0.99) 0.04 (7.51 – 23.36) 25.93 Control (17.97 – 35.24)

Table S-11 Impact of community intervention on incidence of OV infection

Incidence rate Study group (/100 person-years, 95% IRR (95% CI) p CI) 5.13 Intervention 0.37 (0.15 – 0.93) 0.03 (1.88 – 11.17) 13.69 Control (8.48 – 20.93)

266

Appendix 2 Fieldwork questionnaire

The following questionnaire was mainly used for collecting individual data for parasitic infection assessment. The Thai version was translated and adapted from this original questionnaire.

QUESTIONNAIRE RURAL PART ONE – INDIVIDUAL DATA (Circle appropriate answer, tick a box or fill in the blanks)

(CODE)

Name individual: ______Address: ______Date: ______Name of interviewer: ______Part 1 Characterization If individual is a child less than 12 years of age, please ask parents to answer questions on behalf of the child where questions are applicable

1. Code 2. Gender 1. Male 2. Female 3. Age of individual: ______Years 4. Religion 1. Buddhist 2. Christian 3. Islamic 4. Other

5. Level of education (If child under the age of 12 years, then maternal level of education: 1. Less than primary 2. Primary 3. Secondary 1 4. Secondary 2 5. Tertiary or College

6. Occupation: 1. Farmer 2. Business 3. Laborer 4. Civil servant 5. Others = 7

8. Per month household income (Baht) 1. < 5000 2. 5,000-10,000 3. 10,000-15,000 267

4. 15,000-20,000 5. > 20,000

9. Were you diagnosed with liver fluke infection within the past year? 1. No 2. Yes 10. If yes, were you treated with anthelmintic? 1. No 2. Yes 11. If yes, did you complete the medication? 1. No 2. Yes

Part II Health behavior To feed your children, have you been done these behaviors? 2. Yes 1. No 12. Feeding raw fish 13. Use the spoon contaminated with raw fish 14. Feeding with contaminated hands 15. Using contaminated utensils

16. Have you ever eaten Koi pla ? 1. No (go to No. 11) 2. Yes 17. If yes, have you been eaten in the past year? 1. No 2. Yes 18. If yes, how often have you been eaten? ………………..times/month What kind of fish did you use for Koi pla? No. 19-24 = type of fish Where do you acquire your fish from? 25. Pond in the village 26. Reservoir in the village 27. Regular market in the village 28. Weekend market 29. Other ______

When you acquire fish, what did you do? 30. Eat straight away 31. Eat at home 32. Preserve it 33. Store it, how = 34……………

35. When do you usually have Koi Pla 268

1. all year 2. Feb-Apr 3. Apr-Oct 4. Oct-Feb 5. other,…………………………..

What is the usual occasion for eating Koi Pla? 36. At home (for a normal meal) 37. Party / Celebrations 38. Drinking with friends 39. During working in the field 40. Other, specify______

Why do you have Koi pla? 41. Habit 42. Tasty 43. Easy to prepare 44. Just to be like others

What other raw, undercooked / fermented fish dish do you consume? _45, 47, 49, 51, 53, 55, 57, 59, 61 = raw food dish, 46, 48, 50, 52, 54, 56, 58, 60, 62 = monthly consumption 63. Do you wash your hands after preparing raw fish? 1. Never 2. Sometimes 3. At all times

64. After preparing raw fish, do you wash utensils before preparing other dishes? 1. Never 2. Sometimes 3. At all times

65. When cleaning raw fish, do you feed the remains to dogs and cats or leave them in a place where dogs and cats can find them? 1. No 2. Yes

66. What lavatory (toilet) facilities do you utilize? 1. No toilet, defecate outside 2. using toilet sometimes 269

3. always using toilet

67. If you go to work in the field, where do you defecate? 1. In the field 2. Near water reservoir 3. Nearest Toilet 4. Other = 68, ………………………..

69. Do you drink bottled or tap water? 1. Tap 2. Rain 3. Well 4. Bottle 5. Pond 6. Other,…70…………………..

71. How do you clean the water? 1. Never 2. Boil 3. Filter 4. Other,…= 72……………

73. How often do you wear footwear when you are outdoors? 1. Never 2. Rarely 3. Sometimes 4. All the time

74. Do you bathe or swim in nearby ponds or reservoirs? 1. No 2. Yes 75. How would you categorize your general health status? 1. Very Healthy 2. Healthy 3. Sick 4. Very sick

Have you suffered from any of the following symptoms in the past one week? 76. Upper abdominal pain or discomfort 77. Gurgling abdomen 270

78. Diarrhoea (3 bouts of watery stool / day or one bout of dysentery ) 79. Nausea 80. Indigestion 81. Malaise or general weakness 82. Anorexia 83. Weight loss (10% within past month)

Do you consume raw, undercooked or fermented 84. frog or 85. snake meat or 86. grasshoppers? 1. No 2. Yes 87. How often do you take de-worming medication? Frequency Anthelmintic name, if known 1. At least every 6 months 2. At least once a year 3. Occasionally 4. Rarely 0. Never 6. Unsure

88. Do you own a dog or regularly feed a dog? 1. Yes 0. No

89. Do you slaughter animals at home? 1. Yes 0. No 90. Type of animal that you slaughter

91. How often do you come in close contact with dogs (petting, playing or feeding)? 1. Daily 2. 2-3 times a week 3. Once a week 4. Occasionally (1-3 times a month) 5. Rarely 6. Never

92. Are you aware that internal parasites can be transmitted between dogs and humans? 1. No 2. Yes

271

Appendix 3 Ethical approval document

272