Epidemiol. Infect. (2002), 128, 419–431. # 2002 Cambridge University Press DOI: 10.1017\S0950268802006933 Printed in the United Kingdom
Case-control studies of sporadic cryptosporidiosis in Melbourne and Adelaide, Australia
" " " # $ B. ROBERTSON , M.I. SINCLAIR *, A.B. FORBES , M. VEITCH , M. KIRK , % " & D. CUNLIFFE , J. WILLIS C.K. FAIRLEY
" Cooperatie Research Centre for Water Quality and Treatment at Department of Epidemiology and Preentie Medicine, Monash Uniersity, Alfred Hospital, Prahran, Victoria 3181, Australia # Microbiological Diagnostic Unit, Department of Microbiology and Immunology, Uniersity of Melbourne, Australia $ OzFoodNet, Australia New Zealand Food Authority % Department of Human Serices, South Australia & Department of Public Health, Uniersity of Melbourne, Australia
(Accepted 11 January 2002)
SUMMARY Few studies have assessed risk factors for sporadic cryptosporidiosis in industrialized countries, even though it may be numerically more common than outbreaks of disease. We carried out case-control studies assessing risk factors for sporadic disease in Melbourne and Adelaide, which have water supplies from different ends of the raw water spectrum. In addition to examining drinking water, we assessed several other exposures. 201 cases and 795 controls were recruited for Melbourne and 134 cases and 536 controls were recruited for Adelaide. Risk factors were similar for the two cities, with swimming in public pools and contact with a person with diarrhoea being most important. The consumption of plain tap water was not found to be associated with disease. This study emphasizes the need for regular public health messages to the public and swimming pool managers in an attempt to prevent sporadic cryptosporidiosis, as well as outbreaks of disease.
of epidemiological studies suggest that the incidence INTRODUCTION of gastroenteritis in developed nations is around 0n7 Outbreaks of cryptosporidiosis have been studied episodes per person per year [2–6]. On this basis it can extensively. The largest reported outbreak occurred in be estimated that in the United States during 1993 Milwaukee in 1993, when a drinking water supply waterborne gastroenteritis from recognized outbreaks contaminated by Cryptosporidium was estimated to accounted for less than 0n25% of all gastroenteritis, have infected 403000 people [1]. Because of drinking despite the occurrence of over 400000 cases of illness waters’ propensity to rapidly infect large numbers of from the Milwaukee Cryptosporidium outbreak [7]. It people the primary focus of international and Austra- is important to establish risk factors for sporadic lian research to date has been the cause of outbreaks cryptosporidiosis and examining the role of drinking of cryptosporidiosis rather than factors contributing water as a risk factor for sporadic disease is considered to sporadic disease. Sporadic disease however, may be particularly important. numerically more common than outbreaks. A number We aimed to examine risk factors for sporadic * Author for correspondence. cryptosporidiosis amongst the general communities of 420 B. Robertson and others both Melbourne and Adelaide, Australia. These two years old, 40–! 65 years old and & 65 years old. cities were chosen because they represent two extremes Telephone sampling using the Electronic White Pages in raw water quality and treatment in Australia. (EWP) residential telephone directory was used to Melbourne has high quality source water from highly recruit controls [10]. Prior to contacting a randomly protected catchments which undergoes only minimal generated telephone number an introductory letter treatment with chlorination [8], whilst Adelaide has was first sent to the household to improve par- poor quality source water from unprotected catch- ticipation rates [11]. Approximately 2 weeks after the ments which undergoes full conventional water treat- introductory letter, telephone calls to these numbers ment including coagulation, sedimentation, filtration were then made, and a database of willing potential and chlorination [9]. A case-control study design was controls was created in advance of cases over the used to examine risk factors for sporadic crypto- course of the study. sporidiosis. After a case was recruited and had the CATQ administered, random potential controls of the same sex and age bracket were automatically generated from this database. If a suitably matched control was METHODS available the CATQ was administered, but if not then Case-control studies were conducted in Melbourne a new randomly selected potential control was from June 1998 to May 2001 and in Adelaide from generated until a control was eventually recruited. November 1998 to May 2001. Ethical approval was Four matched controls were interviewed for each obtained from the local institutional review commit- case. We attempted to administer the control question- tees. Participants were recruited from the general naires within 3 working days of a case questionnaire. communities of both cities. Controls were directly Only one control was recruited from each household matched to cases according to age and sex, with four for each case, and the control was then retained in the controls matched to each case. database of potential controls for possible selection at We defined a case as having Cryptosporidium a later time. Controls were retained to prevent possible oocysts detected in a faecal specimen by an accredited selection bias by their exclusion from the database pathology laboratory, the onset of any diarrhoea or [12–14]. vomiting within 8 weeks before the administration of The CATQ inquired into demographic information, questionnaire, residence in a household with a fixed clinical details of the case’s illness, education level, telephone connection and the ability to speak English. employment, the consumption of tap water, the Participating pathology laboratories sent reports of consumption of particular food groups, recreational newly diagnosed cases of cryptosporidiosis to the water activities, the presence of immunological im- Departments of Human Services in Victoria and pairment, the consumption of regular medication, South Australia and these were forwarded to the contact with persons who may pose a risk of Department of Epidemiology and Preventive Medi- cryptosporidiosis, animal contact, rural or overseas cine (DEPM). With the permission of the case’s travel and exposure to child-care or breast feeding. attending doctor, cases were recruited by telephone To isolate apparently sporadic cases, clusters of calls. A maximum of nine telephone calls were made cases (and their matched controls) were removed from to recruit individual cases. On contacting the case, or the database for analysis. The removal of clusters was their main care-giver for children less than 12 years done by firstly identifying the symptom onset date of old, the interviewer requested their consent to par- primary cases and then removing subsequent cases ticipate. If consent was given, the computer assisted who had their symptom onset within a 14 day period telephone questionnaire (CATQ) was administered if they either: resided within the same household, immediately or at a mutually convenient time. attended the same swimming pool or attended the We defined a control as not having diarrhoea or same childcare centre. vomiting in the 2 weeks before the onset of the matching case’s illness, residence in a household with a fixed telephone connection and the ability to speak Data analysis English. Controls were matched to cases by sex and the following age brackets: ! 1 year old, 1–! 5 years The power of the study was calculated using drinking old, 5–! 12 years old, 12–! 18 years old, 18–! 40 water as the most important exposure to be assessed. Sporadic cryptosporidiosis 421
For Melbourne with 201 cases and 795 controls the related exposures were examined from the same group study had 80% power, at 2-sided 5% significance the most important single exposure based on pub- level, to detect a trend in odds ratio (OR) of 1n22 per lished evidence was chosen. quartile increase in the exposure [15]. For Adelaide with 134 cases and 536 controls the study had 80% RESULTS power, at 2-sided 5% significance level, to detect a Melbourne trend in OR of 1n28 per quartile increase in the exposure. There were 271 cases reported in Melbourne during Detectable OR (with 80% power) for binary the study period. Of those, 201 (74n2%) had the exposures in the Melbourne study ranged from 1n6 for CATQ administered and were included in the analysis an exposure with 50% prevalence among controls to (Table 1). Twenty-four (8n9%) had the CATQ 1n9 for 10% prevalence. The corresponding detectable administered but were retrospectively excluded from OR for Adelaide were 1n8 for 50% prevalence to 2n2 the analysis because they were found to form disease for 10% prevalence. clusters, 14 (5n2%) had the CATQ administered but Data were analysed using Stata [16]. The standard were retrospectively excluded because it was admini- analysis method for matched case-control studies is stered more than 8 weeks after their symptom onset, conditional logistic regression when the matched sets 7(2n6%) cases refused to participate, 4 (1n5%) doctors are independent. However, in this study controls from refused permission, 11 (4n1%) where case notification the same household could be selected at multiple times to the DEPM failed, 3 (1n1%) cases were unavailable, for different cases, thereby inducing a correlation in 1(0n4%) did not have a fixed telephone connection, 1 the exposures across different matched sets, which (0n4%) had an incorrect telephone number recorded needed to be accounted for in the analysis. Standard and 5 (1n8%) could not be contacted within 8 weeks of conditional logistic regression was formulated as a symptom onset. stratified Cox proportional hazard regression mode, Of the residential telephone numbers from which with the matched sets as strata [17]. This allowed an the control sample was generated, letters were sent to extension to accommodate the clustering effect of 5691 households. Of these, 3191 were contacted by individuals from the same household using the telephone and were eligible to participate in the study, ‘cluster’ option in the ‘stcox’ command in Stata. In with 2177 (68n2%) agreeing to participate. Of those this way the Cox proportional hazards model was households agreeing to participate, 4322 individual used to calculate crude and adjusted odds ratios for potential controls were recruited for the database, the various exposures examined, together with 95% with 795 CATQ administered to controls and included confidence intervals (CI). Linear trend per unit being in the analysis, with 90 controls having conducted measured (e.g. per glass of water consumed) or per more than one CATQ. For the 2500 households that category being measured (e.g. per two glass categories were either not contactable or ineligible to participate, of water consumed) was also calculated, together with 54 (0n9%) letters were returned as the address was 95% CI. Linear trend was used to evaluate a incomplete, 230 (4n0%) had the maximum of 9 dose–response relationship when this was relevant. telephone calls exhausted without contact, 36 (1n0%) All exposures could not be adjusted for, because were ineligible because of a household member with a many were highly correlated, making such an analysis chronic gastrointestinal illness, 302 (5n3%) had insuffi- numerically unstable. Because of the importance of cient English, 968 (17n0%) had an incorrect phone examining plain drinking water as a risk factor for number recorded, 126 (2n2%) returned the letter as sporadic cryptosporidiosis, this exposure was the there was a new resident, 773 (13n6%) had not had all main focus of the adjusted analysis. Individual 9 calls exhausted but calling was ceased because confounding was sought by adjusting plain water sufficient numbers had been recruited and 11 (0n2%) against each exposure separately. Biologically plaus- were unavailable. ible associations between exposures and plain drinking The median age for recruited cases was 11 years water, and exposures and cryptosporidiosis were (range 0–81 years) and for controls 11 years (range considered, particularly those that were statistically 0–83 years). One hundred and seven cases (53n2%) significant. Where possible there needed to be suffi- were male (Table 1). cient numbers of participants in the exposed group to Cryptosporidiosis was not associated with the achieve numerical stability. Finally, when a number of consumption of plain tap water. Neither the crude nor 422 B. Robertson and others † 99) 04) 15) † † † n n n 0) 0) 0) 9) 1) 3) 0) 7) 0) 5) 0) 8) 6) n n n n n n n n n n n n n 70–0 92–1 84–1 n n n 03 3–1 4–1 3–1 4–0 4–2 6–4 4–3 7–1 7–2 47 85 5–1 8–2 7–1 7–1 n n n n n n n n n n n n n n n n 0 0 0 82 (0 6(0 6(0 6(0 0 6(0 0 9(0 0 6(0 0 1(0 0 1(0 2(0 0 98 (0 98 (0 9(0 2(0 2(0 0 1(0 0 l l l n n n n n n n n n n n n n n n n n n n n n n n n 0 0 0 0 P P P Crude OR*, 95% CI 5%) 0 1%) 4%) 0%) 1 0%) 0%) 1 9%) 1 6%) 1 1%) 1 3%) 1 7%) 1 0%) 1 2%) 0 3%) 1 8%) 1 7%) 1 3%) 1 7%) 1 n n n n n n n n n n n n n n n n n n 1%) 0 4%) 1 0%) 1 n n n 7%) 83 (15 9%) 199 (37 4%) 88 (16 0%) 166 (31 0%) 370 (69 1%) 166 (31 1%) 229 (90 7%) 241 (95 7%) 237 (94 8%) 232 (43 1%) 186 (34 2%) 118 (22 7%) 103 (19 6%) 98 (18 9%) 117 (21 8%) 218 (40 2%) 318 (59 8%) 218 (40 n n n n n n n n n n n n n n n n n n 0%) 23(9 4%) 11(4 4%) 15(6 n n n 44 (32 79 (59 Adelaide † 98) 11)28) 0 0 † † n n n 1) 17 (12 8) 1) 18 (13 9) 9) 5 (8 5) 4 (6 6) 4 (6 7) 56 (41 2) 51 (38 5) 21 (15 5) 29 (21 1) 32 (23 9) 82 (61 n n n n n n n n n n n n n 73–0 99–1 96–1 n n n 02 4–1 4–0 4–1 5–0 5–2 7–3 3–1 7–1 5–1 15 10 0–2 6–1 9–2 9–1 n n n n n n n n n n n n n n n n 0 0 0 85 (0 7(0 6(0 7(0 0 55 (41 6(0 0 55 (41 2(0 0 58 (92 5(0 0 59 (93 7(0 0 59 (93 1(0 7(0 0 27 (20 05 (0 11 (0 6(1 9(0 4(0 0 52 (38 3(0 0 52 (38 l l l n n n n n n n n n n n n n n n n n n n n n n n n 0 0 0 P P P Crude OR*, 95% CI Cases Controls 1%) 0 1%) 8%) 0 0%) 1 0%) 0%) 1 9%) 1 4%) 1 6%) 1 0%) 1 2%) 0 7%) 1 3%) 1 2%) 0 3%) 1 3%) 1 7%) 1 3%) 1 8%) 67 (50%) 268 (50%) n n n n n n n n n n n n n n n n n n n 1%) 1 6%) 1 4%) 0 n n n ) CI ( als 9%) 136 (17 4%) 279 (35 9%) 110 (13 8%) 270 (34 2%) 525 (66 8%) 270 (34 0%) 377 (95 0%) 375 (95 0%) 360 (91 0%) 366 (46 5%) 304 (38 5%) 125 (15 9%) 177 (22 9%) 216 (27 8%) 225 (28 4%) 177 (22 6%) 618 (77 4%) 177 (22 2%) 428 (53 n n n n n n n n n n n n n n n n n n n 0%) 16(4 0%) 18(4 0%) 33(8 n n n Melbourne Cases Controls day) (four categories) \ Crude analysis with 95% confidence inter 00 42 (20 00 64 (31 n n 00 58 (28 n 4 01–4 01–2 category increase category increase (four categories) n n Linear trend per one Daily 32 (15 Weekly 55 (27 Monthly 26 (12 None 88 (43 Yes 113 (56 No 88 (43 Yes 5 (5 No 95 (95 Yes 7 (7 No 93 (93 Yes 6 (6 No 94 (94 University 106 (53 Trade qualification 61 (30 Linear trend per one glass increaseLinear trend per one0–12 years or schooling 33 (16 1 1 " 2 0 None 37 (18 Yes 164 (81 No 37 (18 Uncooked carrots Uncooked carrots Work with animals Work in child-care Work with sewage Education level Plain water (glasses Plain water Table 1. NumberMedian age in yearsMale (range) 11 (0–81) 11 (0–83) 201 107 (53 795 10 (0–83) 11 (0–86) 134 536 Sporadic cryptosporidiosis 423 ] † † † † † 22) 45) † † † † † † n n 3) 8) 4) n n n 9) 3) 8) 2) 3) 1) 5) 1) 4) 4) 6) 0) erleaf n n n n n n n n n n n n 04–1 05–1 n n 4–0 2–5 5–3 8–13 4–3 7–20 004 9–3 6–2 5–2 9–2 019 3–5 2–3 0–3 9–2 2–12 n n n n n n n n n n n n n n n n n 0 0 6(0 0 4(2 0 4(1 0 1(3 0 2(1 0 5(2 0 12 (1 8(0 2(0 1(0 0 3(0 0 24 (1 7(1 9(0 0 9(1 0 3(0 0 9(1 0 l l n n n n n n n n n n n n n n n n n n n n n n n n n n n n n 0 3 2 7 2 7 1 1 2 1 3 P P continued o [ 8%) 2%) 1 6%) 1 0%) 0%) 1 8%) 1 7%) 3%) 1 1%) 1 5%) 1 5%) 1 5%) 1 0%) 1 0%) 1 3%) 1 7%) 1 n n n n n n n n n n n n n n n n 4%) 3%) 9%) 3%) 1 1%) 1 0%) 1 2%) 9%) 0 0%) 8%) n n n n n n n n n n 7%) 374 (69 3%) 162 (30 1%) 45(8 9%) 491 (91 6%) 59 (11 4%) 476 (89 2%) 12(2 9%) 522 (97 6%) 100 (18 4%) 436 (81 0%) 530 (99 2%) 39(7 9%) 421 (78 1%) 115 (21 9%) 421 (78 0%) 509 (95 0%) 509 (95 6%) 125 (23 4%) 411 (76 0%) 531 (99%) 1 n n n n n n n n n n n n n n n n n n n n 0%) 5 (0 7%) 49(9 2%) 27(5 5%) 17(3 5%) 10(1 0%) 27(5 0%) 4 (0 n n n n n n n 8(6 7(5 2(1 80 (59 31 (23 29 (21 20 (15 42 (31 15 (11 35 (26 10 (7 12 (9 37 (27 † † † † 24) 30) † † † † † † † † † † † n n 5) 2) n n 8) 2) 1) 3) 5) 3) 13 (9 3) 4) 7) 1) 4) 4) 5) 4 (3 n n n n n n n n n n n n n 12–1 08–1 n n 4–0 4–3 6–3 2–12 2–2 5–55 0001 5–5 9–2 7–4 8–3 0001 4–3 3–4 5–3 8–3 7–5 n n n n n n n n n n n n n n n n n 0 0 6(0 0 54 (40 1(1 0 103 (76 2(1 0 105 (78 3(4 0 112 (84 6(1 0 91 (68 0(3 0 126 (94 18 (1 7(2 4(0 7(1 0 99 (73 5(1 0 99 (73 2(1 18 (1 3(1 0 122 (91 3(1 0 122 (91 5(1 0 97 (72 0(0 0 130 (97 l l n n n n n n n n n n n n n n n n n n n n n n n n n n n n n 0 2 2 7 1 1 3 2 2 2 1 2 2 2 P P 14 9%) 1%) 1 7%) 1 9%) 1%) 1 2%) 1 0%) 0%) 1 6%) 1 7%) 1 2%) 1 8%) 2%) 1 3%) 1 7%) 3%) 1 8%) 2%) 1 7%) 1 n n n n n n n n n n n n n n n n n n n 3%) 8%) 4%) 1%) 0%) 5%) 2%) 3%) 2 n n n n n n n n 8%) 484 (60 3%) 311 (39 5%) 66(8 5%) 726 (91 4%) 102 (12 6%) 691 (87 5%) 22(2 5%) 769 (97 9%) 175 (22 2%) 620 (78 0%) 792 (99 4%) 64(8 4%) 101 (12 9%) 56(7 2%) 574 (72 8%) 221 (27 2%) 574 (72 4%) 52(6 1%) 710 (89 9%) 85 (10 1%) 710 (89 8%) 245 (30 2%) 550 (69 5%) 784 (98 n n n n n n n n n n n n n n n n n n n n n n n n 0%) 3 (0 5%) 33(4 5%) 10(1 n n n 6 yr at home attending childcare 6 yr at home with diarrhoea 5 yr at home with diarrhoea ! ! " 2 days 41 (20 1 day 23 (11 Yes 100 (49 No 101 (50 Yes 31 (15 No 169 (84 Yes 49 (24 No 152 (75 Yes 31 (15 No 169 (84 Yes 62 (30 No 139 (69 Yes 8 (4 Linear trend per 1 day increase No 193 (96 " 2 days 25 (12 1 day 26 (12 No days 109 (54 Yes 92 (45 Linear trend per 1 day increase No 109 (54 " 1 day 15 (7 No days 163 (81 Yes 38 (18 No 163 (81 Yes 98 (48 No 103 (51 Yes 5 (2 No 194 (97 Animal contact at home Persons Children Children Children at home in nappies Immune system illness Separate days in incubation period using public adults’ pool Swimming in public adults’ pool Separate days in incubation period using public toddlers’ pool Swimming in public toddlers’ pool Swimming in public pool Unpasteurized milk products 424 B. Robertson and others † † † † † † 9) 6) n n 0) 6) 1) 4) 7) 7) 9) 0) 1) n n n n n n n n n 7–3 9–33 1–14 2–4 1–2 0–6 2–0 4–0 4–0 5–1 5–1 n n n n n n n n n n n 4(0 0 0 0(1 0 6(2 0 4(1 0 5(0 0 5(1 0 4(0 0 5(0 0 6(0 0 7(0 0 7(0 0 n n n n n n n n n n n n n n n n n n n n n n n 8 5 2 0 0 0 Crude OR*, 95% CI 4%) 1 6%) 1 7%) – 3%) 1 6%) 1 1%) 1 5%) 1 2%) 1 5%) 1 8%) 3%) 1 3%) 7%) 1 2%) 8%) 1 3%) 0 7%) 1 7%) 0 3%) 1 n n n n n n n n n n n n n n n n n n n 4%) 9%) 5%) 8%) 0 5%) 2 n n n n n 0%) 67 (89 n 0%) 77 (48 0%) 82 (51 0%) 534 (99 8%) 531 (99 8%) 517 (96 5%) 521 (97 3%) 528 (98 7%) 138 (25 3%) 398 (74 1%) 291 (54 9%) 245 (45 5%) 338 (63 5%) 197 (36 6%) 189 (35 4%) 347 (64 0%) 277 (51 0%) 259 (48 n n n n n n n n n n n n n n n n n 0%) 2 (0 2%) 5 (0 2%) 19(3 5%) 15(2 7%) 8 (1 n n n n n 4(3 7(5 2(1 5(3 11 (8 51 (38 37 (27 59 (44 Adelaide † † † † † † † † 1) 6) n n 4) 22 (55 1) 0 (0%) 8 (10 0) 6) 4) 3) 17 (12 8) 1) 69 (51 0) 0) n n n n n n n n n n 4–1 2–2 4–26 8–15 1–3 2–4 7–5 7–1 4–0 6–1 5–1 5–1 n n n n n n n n n n n n 7(0 0 18 (45 6(0 0 22 (100 9(6 0 130 (97 6(4 0 127 (94 8(1 0 123 (91 3(1 0 132 (98 1(1 0 129 (96 9(0 0 117 (87 6(0 0 83 (61 8(0 0 65 (48 7(0 0 97 (72 7(0 0 75 (56 n n n n n n n n n n n n n n n n n n n n n n n n 8 1 2 3 0 0 0 12 Crude OR*, 95% CI Cases Controls 5%) 0 5%) 1 9%) 0 1%) 1 7%) 1 2%) 1 3%) 1 7%) 1 0%) 1 6%) 0 4%) 1 2%) 8%) 1 8%) 0 2%) 1 2%) 8%) 1 0%) 0%) 1 n n n n n n n n n n n n n n n n n n n 3%) 8%) 7%) 3%) 0%) n n n n n 05 level. 3%) 135 (51 7%) 127 (48 1%) 99 (86 4%) 10(1 6%) 785 (98 4%) 14(1 6%) 781 (98 6%) 749 (94 0%) 776 (97 5%) 770 (97 4%) 171 (21 6%) 622 (78 8%) 375 (47 2%) 420 (52 8%) 435 (54 3%) 359 (45 9%) 208 (26 1%) 587 (73 3%) 350 (44 7%) 445 (56 n n n n n n n n n n n n n n n n n n n n n 9%) 16 (13 5%) 45(5 0%) 18(2 5%) 24(3 0 n n n n % P Melbourne Cases Controls 6 years old 3 years old ! ! .) cont Yes 29 (45 No 35 (54 Yes 4 (8 No 41 (91 Yes 27 (13 No 174 (86 Yes 27 (13 No 174 (86 Yes 19 (9 No 182 (90 Yes 10 (5 No 191 (95 Yes 17 (8 No 184 (91 Yes 41 (20 No 160 (79 Yes 70 (34 No 131 (65 Yes 102 (50 No 99 (49 Yes 40 (19 No 161 (80 Yes 73 (36 No 128 (63 Bold type indicates statistically significant at Participant in child-care if Participant breast fed if Consume unboiled water, ice cubes or salad overseas Travel overseas Drink unboiled water from a country river, lake or dam within Australia Lamb contact away from home Calf contact away from home Cat contact away from home Dog contact away from home Animal contact away from home, e.g. pet stores, farms, zoos Cat contact at home Table 1 ( Dog contact at home * Adjusted for matching† factors (age, sex) and family clustering. Sporadic cryptosporidiosis 425
Table 2. Adjusted analysis with 95% CI*
Melbourne Adelaide
Plain water No 1n01n0 Yes 1n3(0n9–2n1) 1n0(0n7–1n6) Pain water (glasses\day) (four categories) None 1n01n0 0n01–2n00 1n4(0n9–2n4) 1n1(0n7–2n0) 2n01–4n00 1n0(0n6–1n7) 1n2(0n7–2n0) " 4n00 1n5(0n9–2n6) 0n8(0n4–1n3) Linear trend per one glass increase 1n04 (0n98–1n10) 0n96 (0n90–1n02) P l 0n25 P l 0n21 Linear trend per one category increase (four categories) 1n10 (0n95–1n28) 0n95 (0n80–1n12 P l 0n21 P l 0n52 Educational level 0–12 years or schooling 1n01n0 Trade qualification 0n7(0n4–1n3) 1n6(0n9–2n9) University 1n0(0n6–1n6) 1n2(0n7–2n1) Uncooked carrots No 1n01n0 Yes 0n6(0n4–0n9)† 0n6(0n4–0n9)† Swimming in public pool No 1n01n0 Yes 2n7(1n9–3n8)† 1n2(0n8–1n9) Children ! 6 yr at home with diarrhoea No 1n01n0 Yes 7n4(4n0–13n8)† 8n6(4n8–15n6)† Persons " 5 yr at home with diarrhoea No 1n01n0 Yes 1n8(1n1–2n9)† 3n7(2n2–6n2)† Animal contact at home No 1n01n0 Yes 0n6(0n4–0n8)† 0n6(0n4–0n9)† Calf contact away from home No 1n01n0 Yes 2n9(1n5–5n7)† 5n1(1n5–17n3)† Drink unboiled water from a country river, lake or dam within Australia No 1n01n0 Yes 1n5(0n8–2n7) 3n1(1n5–6n5)†
* Adjusted for matching factors (age, sex), family clustering, plain drinking water (as dichotomous variable). † Bold type indicates statistically significant at P % 0n05 level. adjusted OR for drinking any plain tap water dam within rural Australia, overseas travel and the compared to no plain tap water was statistically consumption of unboiled water, ice-cubes or salad significant. Similarly, there was no significant dose– overseas. Statistically significant protective factors response relationship using a linear trend (Table 1). included the consumption of uncooked carrots, For the crude analysis, statistically significant risk contact with any animal at home, dog contact at and factors associated with sporadic cryptosporidiosis away from home, and cat contact at home. included swimming in public pools (both toddler’s After performing an adjusted analysis on a limited and adult’s), immune system impairment, having number of exposures (Table 2), swimming in public children at home in nappies, having contact with pools, contact with persons with diarrhoea, and calf children or adults with diarrhoea and having children contact away from home all remained statistically at home attending childcare. Other risk factors significant risk factors, while the consumption of included calf and lamb contact away from home, the unboiled rural water from a river, lake or dam within consumption of unboiled water from a river, lake or Australia was no longer significant. Contact with any 426 B. Robertson and others
Table 3. Clinical features of cases
Clinical feature Melbourne Adelaide
Duration of illness (days) Median (range) 21 (1–100) 15 (2–120) Mean (range) 22n4 (1–100) 19n4 (2–120) Hospital admission for & one nights 14 (7n0%) 16 (11n9%) Median number of nights (range) 2 (1–6) 2 (1–4) Visits to the doctor Median number of total visits (range) 2 (0–16) 2 (0–8) Median number of visits before the visit at which the faecal 1 (1–5) 1 (1–6) specimen was collected (range) Symptoms Diarrhoea present 198 (98n5%) 132 (98n5%) Median stools per day when worst (range) 9n5 (1–40) 7 (1–40) Mean stools per day when worst (range) 11n3 (1–40) 10n3 (1–40) Watery motions 195 (98n5%) 129 (96n3%) Bloody motions 17 (8n5%) 8 (6n0%) Medication to stop diarrhoea 62 (31n3%) 31 (23n5%) Vomiting 111 (55n2%) 95 (70n9%) Abdominal pain 182 (90n5%) 121 (91n7%) Felt hot 81 (40n3%) 51 (38n1%) Time lost from study or work if participant 6 years old or more 101 (77n6%) 59 (44n0%) Median number of days lost (range) 5 (1–21) 5 (1–77) Time lost from study or work by family member other than case 76 (37n8%) 52 (38n8%) Median number of days lost (range) 5 (1–28) 5 (1–77)
animal at home and the consumption of uncooked by another family member was a median of 5 days carrots all remained statistically significant protective (range 1–28 days). factors. The clinical details of cases are described in Table 3. The duration of the illness was a mean of 22n4 days Adelaide (range 1–100 days). Fourteen (7n0%) cases required hospital admission for 1 night or more, with a median There were 173 cases reported in Adelaide during the stay of 2 nights (range 1–6 nights). Cases visited a study period. Of those, 134 (77n5%) had the CATQ doctor a median of 2 times (range 0–16 visits). The administered and were included in the analysis (Table faecal specimen was usually collected on the first visit 1). Twenty-two (12n7%) had the CATQ administered to the doctor (median 1, range 1–5 visits). Diarrhoea but were retrospectively excluded from the analysis was present in 198 cases (98n5%). The number of because they were found to form disease clusters, 5 stools passed per day when the diarrhoea was at its (2n9%) had the CATQ administered but were retro- worst was a median of 9n5 and mean of 11n3 (range spectively excluded because it was administered 1–40 stools\day). One hundred and ninety-five (98n5 beyond 8 weeks of their symptom onset, 1 (0n6%) case %) reported watery motions and 17 (8n5%) bloody refused to participate, 2 (1n2%) doctors refused motions. Sixty-two (31n3%) reported the use of permission, 3 (1n7%) cases were unavailable, 2 (1n2%) medication to stop their diarrhoea. For other symp- didn’t have a fixed telephone connection, 1 (0n6%) toms, 111 (55n2%) reported vomiting, 182 (90n5%) had an incorrect telephone number, 1 (0n6%) could reported abdominal pain and 81 (40n3%) reported not be contacted within 8 weeks of symptom onset feeling hot. Of those aged 6 years old or more, 101 and 2 (1n2%) had insufficient English. (77n6%) lost time from study or work. The time lost Of the residential telephone numbers from which from study or work was a median of 5 days (range the control sample was generated, letters were sent to 1–21 days). Of all cases, 76 (37n8%) had another 4492 households. Of these, 2558 were contacted by family member lose time from study or work so they telephone and eligible to participate in the study, with could be cared for. The time lost from study or work 1766 (69n0%) agreeing to participate. Of those Sporadic cryptosporidiosis 427 households agreeing to participate, 3362 individual The clinical details of cases are described in Table 3. potential controls were recruited for the database, The duration of the illness was a mean of 19n4 days with 536 CATQ administered to controls and included (range 2–120 days). Sixteen (11n9%) cases required in the analysis, with 49 controls having conducted hospital admission for 1 night or more, with a median more than one CATQ. For the 1934 households that stay of 2 nights (range 1–4 nights). Cases visited a were either not contactable or ineligible to participate, doctor a median of 2 times (range 0–8 visits). The 49 (1n1%) letters were returned as the address was faecal specimen was usually collected on the first visit incomplete, 159 (3n5%) had the maximum of 9 to the doctor (median 1, range 1–6 visits). Diarrhoea telephone calls exhausted without contact, 23 (0n5%) was present in 132 cases (98n5%). The number of were ineligible because of a household member with a stools passed per day when the diarrhoea was at its chronic gastrointestinal illness, 192 (4n3%) had insuffi- worst was a median of 7 and a mean of 10n3 (range cient English, 696 (15n5%) had an incorrect phone 1–40 stools\day). One hundred and twenty-nine (96n3 number recorded, 198 (4n4%) returned the letter as %) reported watery motions and 8 (6n0%) bloody there was a new resident, 613 (13n6%) had not had all motions. Thirty-one (23n5%) reported the use of 9 calls exhausted but calling was ceased because medication to stop their diarrhoea. For other symp- sufficient numbers had been recruited and 4 (0n1%) toms, 95 (70n9%) reported vomiting, 121 (91n7%) were unavailable. reported abdominal pain and 51 (38n1%) reported The median age of recruited cases was 10 years feeling hot. Of those aged 6 years old or more, 59 (range 0–83 years) and for controls 11 years (range (44n0%) lost time from study or work. The time lost 0–86 years). Sixty-seven cases (50n0%) were male from study or work was a median of 5 days (range (Table 1). 1–77 days). Of all cases, 52 (38n8%) had another Cryptosporidiosis was not associated with the family member lose time from study or work so they consumption of plain tap water. Neither the crude nor could be cared for. The time lost from study or work adjusted odds ratio for drinking any plain tap water by another family member was a median of 5 days compared to no plain tap water were statistically (range 1–77 days). significant. Similarly, there was no significant dose– response relationship using a linear trend (Table 1). For the crude analysis, statistically significant risk DISCUSSION factors associated with sporadic cryptosporidiosis included the consumption of unpasteurized milk This is one of the first case-control studies to examine products, swimming in a public toddlers’ pool, risk factors for sporadic cryptosporidiosis in an immune system impairment, having children at home industrialized country. Other studies have focused on in nappies, having contact with persons with diarrhoea outbreaks of disease. Drinking water was not associ- and having children at home attending childcare. ated with sporadic disease which is consistent with Other risk factors included the consumption of findings from a randomized clinical trial recently unboiled water from a river, lake or dam within rural conducted in Melbourne [6]. This study examined Australia, overseas travel and the consumption of gastroenteritis rates in families receiving either plain unboiled water, ice-cubes or salad overseas. Stat- or filtered tap water and found no evidence of istically significant protective factors included the waterborne gastrointestinal disease. We found that consumption of uncooked carrots, contact with any the risk factors for sporadic infection were generally animal at or away from home, as well as dog and cat similar for both Melbourne and Adelaide despite contact away from home. vastly different water supplies. These included risk After performing an adjusted analysis on a limited factors commonly associated with outbreaks of number of exposures (Table 2), contact with children disease, most importantly, swimming in public pools and adults with diarrhoea, and the consumption of and person-to-person transmission. These results unboiled rural water within Australia remained suggest that drinking water is unlikely to cause a statistically significant risk factors, whilst calf contact significant amount of sporadic cryptosporidiosis in away from home became statistically significant. major Australian cities. Contact with any animal at home and the con- There were several specific exposures within the sumption of uncooked carrots remained statistically swimming related group that were associated with significant protective factors. sporadic cryptosporidiosis. The most consistent fea- 428 B. Robertson and others ture for the two cities was an association with public Unpasteurized milk products such as cheese were toddlers’ pools, which can be used by large numbers significantly associated with sporadic cryptosporidio- of children. This source of recreational water activity sis for Adelaide but not for Melbourne. The small was a biologically plausible risk factor for crypto- numbers in the exposed groups make it difficult to sporidiosis, considering its faecal-oral transmission make deductions about differences between the two route, the chlorine resistant nature of oocysts and the cities because of the limited statistical power. Unpas- generally poorer hygiene habits of younger children teurized milk products are biologically plausible as a which increase the risk of a faecal accident. In risk factor and have been associated with cryptospori- addition, public pools have been associated with diosis previously [28, 29]. numerous previous outbreaks [18], including several The consumption of unboiled water from a rural outbreaks in Australian cities [19–22]. river, lake or dam within Australia was a statistically Person-to-person transmission was a risk factor for significant risk factor for Adelaide only. It is a sporadic cryptosporidiosis in both Melbourne and biologically plausible risk factor since the microbio- Adelaide. All types of exposures examined were logical quality of country water sources is highly associated with disease. The strongest risk factor variable in Australia. These water sources are not within this group of exposures was having young necessarily maintained in accord with drinking water children at home with diarrhoea. This is not surprising quality guidelines, and may have livestock grazing and is biologically plausible considering the route of along their banks, with such animals having been transmission. Furthermore, it is well recognized as a implicated in the contamination of water sources source of secondary transmission according to pre- overseas [30]. vious research [22–24]. Immune system illness, overseas travel and the Calf and lamb contact away home were found to be consumption of unboiled water, ice-cubes and salad statistically significant risk factors in Melbourne, overseas were strongly associated risk factors for whilst only the former risk factor significant in sporadic cryptosporidiosis. Although biologically Adelaide. Calves are reservoirs for Cryptosporidium plausible and having been associated with cryptospori- oocysts and contact with these animals has been diosis previously, their OR are also likely to be associated with cryptosporidiosis previously [25, 26]. inflated due to the ascertainment bias of cases. This In contrast, various types of animal contact at home may be because overseas travellers who present to were found to be protective factors. Other investi- their doctor with diarrhoea are more likely to have a gators have suggested companion animals pose a faecal specimen requested, than those with diarrhoea small risk [27]. A possible explanation for these who haven’t travelled overseas. Secondly, Melbourne findings could be that animals kept at home shed and Adelaide pathology laboratories are more likely oocysts to which occupants are regularly exposed, to test for Cryptosporidium oocysts in an overseas which could confer active immunity. Adding to the traveller (unpublished data). complexity is that some types of animal contact away We did not identify plain tap water as a risk factor from home were also found to be protective, whereas for sporadic cryptosporidiosis in either Melbourne or presumably this type of contact is more infrequent. Adelaide. The case-control studies allowed the op- This area requires further attention to be more fully portunity to evaluate water supplies from different understood. ends of the water quality spectrum. Melbourne has Several food-related exposures were examined in- high quality source water from protected catchments cluding some meats (sausage meat, hamburgers, offal), that undergoes very little treatment, whereas Ade- lettuce and other leafy vegetables, uncooked berries, laide’s has poor quality source water that undergoes uncooked mushrooms and uncooked carrots, but extensive treatment. Although drinking water has none were identified as risk factors and uncooked been associated with many past outbreaks of crypto- carrots were associated with a protective effect against sporidiosis, Melbourne’s drinking water supply was sporadic cryptosporidiosis. The protective effect of not considered to be a likely source of disease prior to regularly consumed raw vegetables has been identified this study because of its highly protected catchment in unpublished outbreaks in the United Kingdom areas, with no public access or farming. In contrast, according to Casemore [27]. The reasons behind the Adelaide’s drinking water supply was considered to protective effect of only raw carrots remain specu- pose a greater risk of being associated with disease, lative. because of its poor quality source water with few Sporadic cryptosporidiosis 429 limitations on public access or farming. The results of estimates was possible for cases who suspected that a this study however, suggest that the water treatment particular exposure was responsible for their illness. processes currently in place in Adelaide are adequate Such a situation was only likely to occur on a large to prevent waterborne sporadic cryptosporidiosis. scale if publicity was heightened at the time the study The type of symptoms experienced by cases for was conducted, which linked cryptosporidiosis to the Melbourne and Adelaide were typical of those exposure in question. In July to September 1998 reported by other investigators [31, 32], with the vast Sydney’s drinking water supply received substantial majority of participants reporting watery diarrhoea local and international media attention when Crypto- and abdominal pain. Symptoms were often severe and sporidium and Giardia oocysts were reportedly detec- protracted, with a mean duration of 22n4 days (range ted, with three boil water advisories issued to the 1n0–100n0 days) and 19n4 days (range 2n0–120n0 days) public. This adverse publicity therefore adds further for Melbourne and Adelaide respectively. The mean weight to the study finding that cryptosporidiosis is number of stools passed per day when the diarrhoea not associated with the consumption of tap water. In was at its worst was high; 11n3 (range 1n0–40n0 stools) early 1998 prior to the commencement of the case- and 10n3 stools (range 1n0–40n0 stools) for Melbourne control studies, several Melbourne swimming pools and Adelaide respectively. were closed due outbreaks of cryptosporidiosis con- The severity of disease reported by cases in this nected with the pools in question [19]. Considering study was attributable to the selective effect of the most Melbourne cases were recruited at the end of passive surveillance system by which cases were 2000 and the early part of 2001 it is unlikely that recruited using laboratory notification. Such cases are biased reporting ensued. Furthermore, there were no likely to represent only a small fraction of all swimming pool closures in Adelaide over the course of cryptosporidiosis cases in the community as supported the study yet swimming in a public toddlers’ pool was by overseas research [33], since only those with still found to be a statistically significant risk factor. relatively severe symptoms are likely to seek medical The large numbers of exposures examined meant attention and have a faecal specimen examined. In that some of the factors associated with crypto- contrast, studies in volunteer subjects and outbreak sporidiosis might have been due to chance. To prevent investigations with active case finding have reported ‘data trawling’ however, we only asked about ex- less severe clinical symptoms and a shorter duration posures that were biologically plausible, or exposures of disease. In the volunteer study by DuPont [31] the that had been previously associated with crypto- mean duration of illness was 3n1 days (range 2n4–3n6 sporidiosis. days), with a mean of 6n4 unformed stools per day A weakness of this study was that due to its (range 4n0–11n0 stools) when the diarrhoea was at its relatively small size it did not have the statistical worst. In the Milwaukee outbreak [1] people who had power to detect associations for uncommon exposures clinical cryptosporidiosis which was not laboratory or exposures with small OR. For example, working confirmed had a mean duration of illness of 4n5 days with sewage may be a significant risk factor but due to (range 1n0–38n0 days), and a mean of 7n7 unformed only 6% of Adelaide controls reporting this exposure stools per day (range 1n0–60n0 stools) when the an OR of at least 1n9 would have needed to have been diarrhoea was at its worst. For the 285 laboratory detected. confirmed Milwaukee cases the disease was more In summary, this study suggests that drinking water severe with a mean duration of 12n0 days (range is unlikely to be a major cause of sporadic crypto- 1n0–55n0 days), with a mean of 19n0 unformed stools sporidiosis in metropolitan cities of Australia. It per day (range 1n0–90n0 stools) when the diarrhoea emphasizes three areas of public health importance. was at its worst. Firstly, public swimming pools are an important A methodological limitation of case-control studies potential source of large numbers of cryptosporidiosis is recall bias of exposures, particularly when reliant on cases. This therefore reinforces the need for ap- human memory [34]. Self-reported estimates for propriate and regular public health messages to the exposures investigated such as drinking water intake public and swimming pool managers, in an attempt to were likely to result in random rather than systematic prevent both sporadic disease and outbreaks. It also error, which would have the effect of underestimating reinforces the need for close surveillance of swimming the association between the exposure in question and pools by health departments to minimize outbreaks. cryptosporidiosis. However, biased over-reporting of Secondly, education of the public about general 430 B. Robertson and others hygiene measures to limit person-to-person trans- gastrointestinal health affects of drinking water quality. mission needs to be undertaken. Thirdly, due to the Environ Hlth Perspect 2001; 109: 773–8. severity of illness in immunocompromised people, this 7. Kramer MH, Herwaldt BL, Craun GF, Calderon RL, Juranek DD. Surveillance for waterborne-disease out- group should be cautioned about public swimming breaks – United States, 1993–4. MMWR 1996; 45: pool attendance, as well as being made aware of other 1–33. important risk factors identified from this study. 8. Melbourne Water Corporation – Website fhttp:\\ www.melbwater.com.au\g. 2001. 9. South Australian Water Corporation – Website fhttp:\\www.sawater.com.au\g. 2001. ACKNOWLEDGEMENTS 10. Australia on Disc CDROM. Dependable Database Data P\L, Sydney, NSW, Australia, 1998. We thank Professor Louis Pilotto for helpful com- 11. Robertson B, Sinclair M, Forbes A, Kirk M, Fairley ments and Pam Lightbody, Colin Fee, Geoff Aldred, CK. The effect of an introductory letter on participation Richard Hayes, Geoff Simmons, Hana Kazda, Ann rates using telephone recruitment. Aust NZ J Publ Hlth Brown and Judy Hankin who provided much valued 2000; 24: 552. 12. Wacholder S, McLaughlin JK, Silverman DT, Mandel research assistance. We also acknowledge the organiz- JS. Selection of controls in case-control studies 1. ations that funded this research which include the Principles. Am J Epidemiol 1992; 135: 1019–28. Cooperative Research Centre for Water Quality and 13. Wacholder S, Silverman DT, McLaughlin J, Mandel Treatment, Melbourne Water Corporation, Water JS. Selection of controls in case-control studies II. Services Association of Australia, South East Water, Types of controls. Am J Epidemiol 1992; 135: 1029–41. Yarra Valley Water and City West Water. We thank 14. Wacholder S, Silverman DT, McLaughlin JK, Mandel JS. Selection of controls in case-control studies III. the South Australian Water Corporation for provid- Design options. Am J Epidemiol 1992; 135: 1042–50. ing some technical support. We also thank the 15. Egret SIZ – Version 1.1. Statistics and Epidemiology general practitioners and pathology laboratories Research Corp: Seattle, WA, USA, 1993. which participated in this research, particularly Dore- 16. Stata version 5.0 for Windows, Stata Corporation, vitch, Gribbles, Hitech, Melbourne Pathology and the College Station, Texas, USA, 1999. Institute of Medical and Veterinary Science. Our final 17. Prentice R, Breslow N. Retrospective studies and failure time models. Biometrika 1978; 63: 153–8. thanks go to each of the subjects who participated in 18. Barwick RS, Levy DA, Craun GF, Beach MJ, Calderon this research. RL. Surveillance for waterborne-disease outbreaks – United States, 1997–1998. MMWR 2000; 49: 1–35. 19. Hellard ME, Sinclair MI, Black J, et al. An outbreak of cryptosporidiosis at an urban swimming pool: why are REFERENCES such outbreaks difficult to detect. Aust NZ J Publ Hlth 2000; 24: 272–5. 1. MacKenzie WR, Hoxie NJ, Proctor ME, et al. A 20. Lemmon JM, McAnulty J, Bawden-Smith J. Outbreak massive outbreak in Milwaukee of Cryptosporidium of cryptosporidiosis linked to an indoor swimming infection transmitted through the public water supply. pool. Med J Aust 1996; 165: 613–6. N Engl J Med 1994; 331: 161–7. 21. Stafford R, Neville G, Towner C, McCall B. A 2. Fox JP, Elveback LR, Spigland I, Frothingham TE, community outbreak of cryptosporidium infection Stevens DA, Huger M. The Virus Watch Program: a associated with a swimming pool complex. Commun continuing surveillance of viral infections in metro- Dis Intell 2000; 24: 236–9. politan New York families. Am J Epidemiol 1966; 83: 22. Puech MC, Lesjak M, Heron L. A statewide outbreak 389–412. of cryptosporidiosis in New South Wales associated 3. Payment P, Richardson L, Siemiatycki J, Dewar R, with swimming at public pools. Epidemiol Infect 2001; Edwardes M, Franco E. A randomized trial to evaluate 126: 389–96. the risk of gastrointestinal disease due to consumption 23. Hannah J, Riordan T. Case to case spread of crypto- of drinking water meeting current microbiological sporidiosis; evidence from a day nursery outbreak. Publ standards. Am J Publ Hlth 1991; 81: 703–8. Hlth 1988; 102: 539–44. 4. Monto AS, Koopman JS. The Tecumseh study. XI. 24. Brown EAE, Casemore DP, Gerken A, Greatorex IF. Occurrence of acute enteric illness in the community. Cryptosporidiosis in Great Yarmouth – the investiga- Am J Epidemiol 1980; 112: 323–33. tion of an outbreak. Publ Hlth 1989; 103: 3–9. 5. Mead PS, Slutsker L, Dietz V, et al. Food-related illness 25. Reif JS, Wimmer L, Smith JA, Dargatz DA, Cheney and death in the United States. Emerg Infect Dis 1999; JM. Human cryptosporidiosis associated with an 5: 607–25. epizootic in calves. Am J Publ Hlth 1989; 79: 1528–30. 6. Hellard ME, Sinclair MI, Forbes AB, Fairley CK. A 26. Miron D, Kenes J, Dagan R. Calves as a source of an randomized, blinded, controlled trial investigating the outbreak of cryptosporidiosis among young children in Sporadic cryptosporidiosis 431
an agricultural closed community. Pediatr Infect Dis J 31. DuPont HL, Chappell CL, Sterling CR, Okhuysen PC, 1991; 10: 438–41. Rose JB, Jakubowski W. The infectivity of Crypto- 27. Casemore DP, Wright SE, Coop RL. Cryptospori- sporidium parum in healthy volunteers. N Engl J Med diosis – human and animal epidemiology. In: Fayer R, 1995; 332: 855–9. ed. Cryptosporidium and cryptosporidiosis. New York, 32. MacKenzie WR, Schell WL, Blair KA, Addiss DG, London, Tokyo: CRC Press, 1997: 65–92. Peterson DE. Massive outbreak of waterborne Crypto- 28. Thomson MA, Benson JWT, Wright PA. Two year sporidium infection in Milwaukee, Wisconsin: recur- study of Cryptosporidium infection. Arch Dis Child rence of illness and risk of secondary transmission. Clin 1987; 62: 559–63. Infect Dis 1995; 21: 57–62. 29. White DG, Nichols GL, Mansfield VA, Burden P, 33. Wheeler JG, Sethi D, Cowden JM, et al. Study of Tanner EI, Rowbotham TJ. Cryptosporidiosis in infectious intestinal disease in England: rates in the England and Wales: prevalence and clinical epidemio- community, presenting to general practice, and reported logical features. BMJ 1990; 300: 774–7. to national surveillance. BMJ 1999; 318: 1046–50. 30. Craun GF, Hubbs SA, Frost F, Calderon RL, Via SH. 34. Schlesselman JJ, Stolley PD. Case-control studies: Waterborne outbreaks of cryptosporidiosis. J AWWA design, conduct, analysis. New York: Oxford Uni- 1998; 90: 81–91. versity Press, 1982: 354.