Sensory Acceptability and Nutrient Stability in Micronutrient-Fortified Soymilk Prepared in Small-Scale Batch Processes" (2019)

Brigham Young University BYU ScholarsArchive

Theses and Dissertations

2019-04-01

Sensory Acceptability and Nutrient Stability in Micronutrient- Fortified Soymilk Prepared in Small-Scale Batch Processes

Dallin Max Hardy Brigham Young University

Follow this and additional works at: https://scholarsarchive.byu.edu/etd

BYU ScholarsArchive Citation Hardy, Dallin Max, "Sensory Acceptability and Nutrient Stability in Micronutrient-Fortified Soymilk Prepared in Small-Scale Batch Processes" (2019). Theses and Dissertations. 8284. https://scholarsarchive.byu.edu/etd/8284

This Thesis is brought to you for free and open access by BYU ScholarsArchive. It has been accepted for inclusion in Theses and Dissertations by an authorized administrator of BYU ScholarsArchive. For more information, please contact [email protected], [email protected]. Sensory Acceptability and Nutrient Stability in Micronutrient-Fortified

Soymilk Prepared in Small-Scale Batch Processes TITLE PAGE

Dallin Max Hardy

A thesis submitted to the faculty of Brigham Young University in partial fulfillment of the requirements for the degree of

Master of Science

Michael L. Dunn, Chair Oscar A. Pike Frost M. Steele

Department of Nutrition, Dietetics, and Food Science

Brigham Young University

Sensory Acceptability and Nutrient Stability in Micronutrient-Fortified Soymilk Prepared in Small-Scale Batch Processes

Dallin Max Hardy Department of Nutrition, Dietetics, and Food Science, BYU Master of Science

Fortified and unfortified soymilk were produced from the same production batches for comparative evaluation. Fortification included a comprehensive array of micronutrients of interest to community and humanitarian nutrition programs. The effects of time after fortification prior to cooling, cooling method, and light or dark refrigerated storage on the stability of 5 vitamins (vitamin A, vitamin C, thiamine, riboflavin, and folate) were investigated for both fortified and unfortified soymilk. Significant vitamin C loss (6%) and mild vitamin A isomerization occurred while soymilk was hot immediately following fortification. Cooling bottled soymilk in an ambient water bath or ice water bath made no difference in the levels of any of the vitamins measured. Significant loss of riboflavin (18%) and significant vitamin A isomerization to cis isomers other than 13-cis, resulting in loss of bioactivity, occurred during 12 days of light-exposed refrigerated storage. An increase of 13-cis isomer was observed in dark refrigerated storage but with no significant loss of vitamin A bioactivity. No significant degradation of any other vitamins occurred during 12 days of dark refrigerated storage. Sensory evaluation by a panel of youth and children revealed no significant preferences between fortified and unfortified soymilk excepting colour, for which property there was a slight preference for unfortified soymilk. Acceptable vitamin stability and sensory characteristics can be achieved in comprehensively fortified soymilk produced in small-scale batch processes with appropriate management of production and storage conditions.

Keywords: fortification, vitamin, isomerization, mineral, micronutrient, soy, soymilk, beverage, processing, storage, stability, sensory, Ecuador ACKNOWLEDGEMENTS

I’m grateful to the BYU Vitamin Analysis Lab analysts for running and re-running so many samples, including: Grace Park, Jason Kim, Seth Ellsworth, Fred Bassett, David Bae,

Jacob Foist, Topher McNiel, Josh Lehr, Nathan Camp, Mark Johnson, Eliza Lawrence, Erin

Hiatt, Parker Dunn, Garth Lee, Murphy Campbell, Brad Bae, and Muriel Johnson. I’d also like to thank Dr. Michelle Lloyd and all the BYU Sensory Analysis Lab employees for their capable assistance with the sensory panel administration and data analysis. I am especially grateful to

Jiping Zou for his masterful HPLC analysis skills, to Josh Lehr for heading up the sedimentation study, and to Jacob Foist for his generous and dedicated help with folate data analysis.

I’d like to thank Marilyn Nash, Bridget Owen, Scott Buchannan, and Gnoc at NSRL for their generous support; Sondra Heaston & Sheri Palmer of the BYU Nursing Program for the

Ecuador experiences; Carlos Teran, Soroya, and Yaneth, of DSM Colombia for the fortification premix; Hogar de Cristo and all the miracle workers serving there – especially Luis Tavara,

Veronica Rosales, Marianella Holguin, Maricruz, and Mariogenia; and Dr. Doug Heiner, whose generous donation made the Ecuador portion of the project possible.

I’m deeply grateful to my advising committee for their oversight and invaluable assistance – especially to Dr. Michael Dunn for the wonderful opportunities and experiences he has provided me, as well as for his patience with me, his faith in me, and his friendship. And I’d like to thank Hardy Nutritionals for giving me time off work to write, defend, and submit this thesis. Finally, I’d like to thank my wonderful wife, Heather, for breathing life and joy into every step of the journey with her unwavering support, encouragement, and love.

TABLE OF CONTENTS

TITLE PAGE ...... i

ABSTRACT ...... ii

ACKNOWLEDGEMENTS ...... iii

TABLE OF CONTENTS ...... iv

LIST OF TABLES ...... viii

LIST OF FIGURES ...... ix

Introduction ...... 1

Experimental Design ...... 5

Experiments 1 & 2: Delayed Cooling and Cooling Method ...... 6

Experiment 3: Light Exposure in Refrigerated Storage ...... 7

Materials & Methods ...... 8

Materials ...... 8

Methods ...... 9

Preparation of Fortification Premix ...... 9

Soymilk Processing ...... 9

Premix Incorporation...... 10

Packaging, Cooling, Transport, & Sampling ...... 11

Refrigerated Storage Stability Study ...... 12

Nutritional Analysis ...... 13

Sensory Evaluation ...... 17

Statistical Analysis ...... 18

iv Results ...... 19

Delayed Cooling ...... 19

Cooling Method ...... 22

Refrigerated Storage and Light Exposure ...... 23

Sensory Evaluation ...... 26

Discussion ...... 29

Conclusion ...... 34

References ...... 36

APPENDICES ...... 42

Appendix A: NSRL Soymilk Production & Study Design ...... 42

Appendix B: DSM Premix Information and Composition ...... 50

Appendix C: Detailed Analysis Methods ...... 52

Vitamin A Analysis ...... 52

Vitamin C Analysis ...... 55

Thiamine & Riboflavin Analysis (Combined Method) ...... 59

Total Folate Analysis ...... 63

Appendix D: Standard Curves ...... 90

Vitamin A ...... 90

Vitamin C ...... 93

Thiamine ...... 96

Riboflavin ...... 96

Folate ...... 96

Appendix E: BYU Sensory Panel Ballot (NSRL Soymilk) ...... 97

v Appendix F: NSRL Raw Data ...... 101

Vitamin A ...... 101

Vitamin C ...... 107

Thiamine ...... 112

Riboflavin ...... 120

Folate ...... 128

pH ...... 136

Appendix G: NSRL Raw Statistical Analysis Output ...... 137

Vitamin A ...... 137

Vitamin C ...... 152

Thiamine ...... 158

Riboflavin ...... 178

Folate ...... 198

Sensory Panel Demographic Question Results...... 218

Sensory Panel Hedonic Question Results ...... 223

Appendix H: Sedimentation Study ...... 238

Introduction...... 238

Experimental Design ...... 238

Materials & Methods ...... 239

Results ...... 244

Discussion ...... 258

Conclusions ...... 258

Appendix I: Hogar de Cristo soymilk production process summary ...... 259

vi Appendix J: Ecuador Sensory Ballot ...... 262

Appendix K: Ecuador Implementation Study ...... 266

Introduction...... 266

Experimental Design ...... 268

Materials & Methods ...... 268

Results ...... 273

Discussion ...... 282

Conclusions ...... 284

References ...... 285

Appendix L: Ecuador Raw Data ...... 286

Vitamin A ...... 286

Vitamin C ...... 287

Thiamine ...... 290

Riboflavin ...... 291

Folate ...... 293

Minerals (Iron & Zinc) ...... 294

Appendix M: Ecuador Sensory Panel Statistical Analysis Output ...... 295

Ecuador Sensory Panel – All complete ballots ...... 295

Ecuador Sensory Panel – All complete child ballots ...... 314

vii LIST OF TABLES

Table 1: Soymilk temperatures upon MSB addition...... 11

Table 2: Vitamin content of fortified soymilk held hot for 0, 7.5, and 15 minutes after MSB addition...... 21

Table 3: Vitamin content of unfortified soymilk held hot for 0, 7.5, and 15 minutes following

MSB addition...... 21

Table 4: Vitamin content of fortified soymilk cooled in ambient temperature water or in an ice water bath...... 22

Table 5: Vitamin content of unfortified soymilk cooled in ambient temperature water or in an ice water bath...... 22

Table 6: Panelist ratings of fortified (F) and unfortified (U) soymilk for sensory properties and likeability...... 28

Table 7: Panelist rankings of soymilk samples in order of preference...... 28

viii LIST OF FIGURES

Figure 1: Delayed cooling (“Hot Hold”) experimental design, and “Cooling Method”

experimental design for fortified and unfortified soymilk...... 6

Figure 2: “Light exposure” experimental design comparing light-exposed versus dark refrigerated storage...... 7

Figure 3: Thiamine, vitamin C, & folate levels in fortified (F) and unfortified (U) soymilk

during light (L) and dark (D) refrigerated storage...... 23

Figure 4: Riboflavin levels in soymilk stored in light (L) and dark (D) refrigerated storage...... 24

Figure 5: Retinol palmitate isomers and total bioactive vitamin A (sum of bioactivity-adjusted isomer quantities) as a percent of total vitamers measured during light (L) and dark (D) refrigeration...... 26

Figure 6: Panelist age distribution for sensory evaluation...... 27

ix Introduction

The advantages and disadvantages of directly fortifying foods and beverages for humanitarian

purposes may parallel the pros and cons of ready-to-use therapeutic foods (RUTFs). Ready-to-

use therapeutic, supplementary, and complementary foodstuffs can improve consumption

compliance, avoid contamination, and preserve product integrity in humanitarian and community

nutrition programs (Bisimwa 2012). Increased intake compliance and reduced malnutrition have

been achieved in children with moderate acute malnutrition status by using RUTFs, instead of

supplementary foods which require preparation (Nackers et al. 2010).

While ready-to-consume fortified foods share many of the practical advantages of

RUTFs (Briend et al. 2015), at-home fortification programs may be plagued with adherence

problems similar to those reported for therapeutic foods requiring at-home preparation (Teshome et al. 2018). Compliance with point-of-use fortification may also be limited by caregiver- dependent variables. One study delivering micronutrient beverage powders in single dose one-a- day satchets to young Kenyan children at risk of micronutrient deficiency reported that only

60.6% of the children achieved 80% compliance or better for thirty days (at least 24 satchets), and adherence was predicted by parent age (Teshome et al. 2018). In another study, pregnant women were found to be more likely to adhere to taking tablets than using point-of-use micronutrient beverage powders (Suchdev et al. 2015). It may be that fortified beverages, such as soymilk, result in better compliance than point-of-use fortification sachets or fortified solid foods. Weidner et al. (2008) reported >96% consumption of both unfortified soymilk and soymilk fortified with plant sterols during an 8 week double-blind, randomized controlled trial in which the soymilk was consumed daily at home with instructions and follow-up only every 4

1 weeks. Although differences in sensory acceptability between the two treatments were not

evaluated, children were 91% compliant with a fiber-fortified beverage but only 77% compliant

with a fiber-fortified solid food snack in parallel arms of a double-blinded intervention study of

constipated Canadian children (Flogan and Dahl 2010).

As a ready-to-drink fortified beverage, packaged, fluid soy milks have an advantage over

dairy milks in that they are accessible to large numbers of people worldwide who are lactose

intolerant (Swagerty et al. 2002). Soymilk also has potential logistical advantages over animal milks because dry soybeans are more easily transported and can still produce quality soymilk after 3 months of non-refrigerated storage up to 25oC and 50% relative humidity (Lambrecht et

al. 1996). Furthermore, soymilk can be produced using simple, small-scale, mechanical

equipment in urban settings where raising milk-producing animals is not a viable option. With

the application of normal post-process refrigeration, soymilk can be produced in locations that

are proximate to consumers in both time and distance (Uzzan et al. 2007).

Soymilk continues to be proposed as a cost-effective alternative to less accessible animal

source foods for the prevention of kwashiorkor and marasmus in developing countries

(Mazumder and Begum 2016). But since the mid-1900’s the focus on protein and energy intake

in developing countries has expanded to include a strong emphasis on micronutrients (Allen

2003). Although mass fortification is more impactful and economical, large-scale (often

national) micronutrient fortification programs are slower to implement and they still do not reach

all members of the population, so special, targeted fortification can help reach specific sub-

groups (Wesley and Horton 2010). Given that entrepreneurs, charities, and international aid

organizations alike may have already sponsored “soy cows” (small-scale soymilk processing

equipment) in locations strategically situated for servicing certain malnourished populations

2 (Boomgarden 2016), fortification of soymilk through these existing distribution channels may be

a rapidly implementable and cost-effective approach to micronutrient deficiency control.

As a vehicle for micronutrient delivery, soymilk has demonstrated capability

organoleptically (Reilly et al. 2006), functionally (Tang et al. 2010), and therapeutically (Ho et

al. 2005). However, the emphasis is often on calcium fortification in order to present soymilk as

an alternative to dairy milk (Dharmasena and Capps 2014). Calcium bioavailability equivalent

to cow’s milk has been achieved in calcium-fortified soymilk (Zhao et al. 2005; Tang et al.

2010), and only 375 ml per day of calcium-fortified soymilk was effective in increasing bone

mineral density in adolescent Chinese schoolgirls (Ho et al. 2005). Furthermore, in mice, under

conditions of adequate folic acid intake, soy isoflavone introduction during adolescent growth

significantly improved adult bone health indicating that using soy as a delivery vehicle for calcium may have added benefits in children (Kaludjerovic and Ward 2013).

Calcium is far from the only nutrient of concern in micronutrient fortification programs, however, and sensory acceptability is likely as important to the consumption of a micronutrient- rich food as it is to any other food (Carrillo et al. 2012). In low-income countries worldwide, children are at risk of multiple micronutrient deficiencies; more than half of preschool children are anemic (suggesting iron and/or B-vitamin deficiency), millions of children are clinically or sub-clinically deficient in vitamin A, and most are likely deficient in zinc (Rivera 2003).

Unfortunately, iron and zinc are among the micronutrients most likely to introduce undesirable organoleptic changes due to fortification (Tripathi et al. 2011), and vitamin A may undergo significant degradation during processing and storage (Pinkaew at al. 2012). Calcium-fortified soymilk has been shown both to be well accepted by children (in a school lunch program) and to

be a nutritionally significant addition to their diet (Reilly et al. 2006), but it is hard to know how

3 the children’s acceptance of the soymilk might have differed had the study used soymilk fortified

with a more comprehensive array of micronutrients. In a meta-analysis of non-dairy beverage

interventions fortified with multiple micronutrients most were fruit-flavoured dry drink mix

powders and only one contained soy – as protein isolate (Aaron et al. 2015).

Concerns about the efficacy of soy products in reducing malnutrition include the presence

of anti-nutritional factors, and the absence of certain essential nutrients (de Pee and Bloem

2009). Fortunately, micronutrient fortification and proper processing of soymilk can resolve the

vast majority of such problems. For example, soy fiber may cause a decrease in zinc and folic

acid absorption not seen with fiber from other sources (Cossack and Rojhani 1992). The effect

on zinc is likely due to high levels of phytic acid, which is present in all cereal and legume seeds,

and has been known to reduce the bioavailability of iron and zinc in soy products (Hurrell et al.

2004). However, the effect on folic acid appears to be more generally polysaccharide-dependent

(i.e. including fiber). Fortunately, pressure-processing can reduce phytic acid levels in soymilk

to insignificant levels compared to raw soy (Torrezan et al. 2010), and fortification can also help

overcome mineral absorption problems (Hurrell 2004). As for folate, much of the fiber present

in raw soy – most of which is insoluble – is removed during milk production (Cossack and

Rojhani 1992; Giri and Mangaraj 2012). In folate-enriched soymilk, the polysaccharides which

remain (including low amounts of fiber) may confer sufficient benefit to offset the reduction in

bioavailability they may cause by protecting added folic acid from light, heat, and oxygen,

minimizing its loss to radiative, oxidative, and thermal degradation (Liu et al. 2012; Ding and

Yao 2013). Trypsin inhibitors in raw soy are also a potential concern, but pressure-cooking at

100oC for 20 minutes reduces trypsin inhibitor activity to residual levels of 13% (Yuan et al.

2008).

4 The processing conditions as well as the form and quantity of nutrients used in a fortification blend can significantly affect the sensory, nutritional, and functional properties of soymilk (Hurrell 2002; Giri and Mangaraj 2012; Anjum 2013). For example, flavour-destroying lipoxygenases must be inactivated. Conveniently, this can be achieved under blanch conditions of 80oC for 2 minutes, which is well within the pressure and temperature parameters necessary for dephytinization and trypsin inhibitor degradation (Yuan et al. 2008). Soy milks containing fortified calcium also experience settling problems, but this can be managed by using more appropriate calcium salts, carefully managing the pH, or by adding thickening agents to the beverage (Heaney and Rafferty 2006; Pathomrungsiyounggul et al. 2007; Li et al. 2011).

In short, soymilk is a promising nutrient delivery vehicle for malnourished populations, but the potential for micronutrient fortification to result in adverse sensory properties is compounded with increasing number and quantity of added nutrients. The aim of this work was to evaluate the sensory acceptability of soymilk fortified with a broad array of micronutrients produced in small batches befitting humanitarian and localized community nutrition programs, and to monitor the micronutrient stability in fortified soymilk having undergone different rates of post-production cooling and up to two weeks of light-exposed or dark refrigerated storage.

Experimental Design

Two batches of soymilk were processed from the same soybean source. Each batch of soymilk was split into two treatment lots, one of which received a premix of micronutrients, sugar, and salt, while the other received a premix of only sugar and salt. The soymilk was then subjected to

3 different experiments evaluating the effect on vitamin stability of delayed cooling (bulk soymilk), cooling in an ambient temperature water bath versus an ice-water bath (bottled

5 soymilk), and the effect of light exposure during refrigerated storage (bottled soymilk). All bottling, treatment, and sampling procedures were the same for both production batches.

Figure 1: Delayed cooling (“Hot Hold”) experimental design, and “Cooling Method” experimental design for fortified and unfortified soymilk.

Experiments 1 & 2: Delayed Cooling and Cooling Method

Analytical samples were taken from the bulk soymilk of each treatment lot (fortified and

unfortified) immediately after premix addition and again at 7-8 minutes and 15 minutes to

investigate vitamin stability under circumstances of delayed cooling (see Figure 1 “Hot Hold”).

Between minutes 7 and 8 following premix addition, four 473 ml clear plastic bottles were filled from each treatment lot (4 fortified, 4 unfortified) and assigned to one of 2 different cooling treatments in matched pairs of 2 bottles per treatment: ambient water bath or ice water bath.

Analytical samples were taken from each of the treatment bottles to investigate potential

6 differences in vitamin stability between the 2 cooling methods (see Figure 1 “Cooling Method”).

Fifteen minutes after premix addition, all the remaining soymilk (both fortified and unfortified)

from each production batch was bottled in 473 ml clear plastic bottles, cooled in an ice water bath, and refrigerated without light exposure until use in either the “light exposure” storage experiment or the sensory panel.

Figure 2: “Light exposure” experimental design comparing light-exposed versus dark refrigerated storage.

Experiment 3: Light Exposure in Refrigerated Storage

From the soymilk which had been bottled and cooled in the ice water bath 15 minutes after

fortification, 10 fortified samples and 10 unfortified samples from each production batch were

assigned to an experiment evaluating the effect of light exposure on vitamin stability during

7 refrigerated storage. Soymilk bottles were subjected to light exposure for 8 hours per day or

were kept in the dark during refrigerated storage for up to 12 days, resulting in parallel

treatments of light exposed versus dark refrigerated storage for the fortified and unfortified

soymilk (see Figure 2). Analytical samples were taken at baseline (0 days), 5 days, and 12 days

of storage. New, unopened soymilk bottles were sacrificed for analysis at all three sampling

times in order to avoid introducing error through opening, sampling, and re-closing the bottles.

Materials & Methods

Materials

A vitamin-mineral premix was obtained from DSM Nutritional Products, Colombia, S.A. The premix was formulated to meet governmental nutrient content guidelines for the national school feeding program (“Programa de Alimentación Escolar” PAE)) of Ecuador. The premix was intended to deliver 60% of Ecuador’s daily recommended values (“ingesta diaria recomendada”

(IDR)) for Vitamin A, Vitamin D, Vitamin E, Vitamin C, Thiamine, Riboflavin, Vitamin B6,

Niacin, Folate, Vitamin B12, Iron, Zinc, Copper, and Selenium, and 20% of the IDR for calcium

per 350 ml portion of fortified beverage. Corn maltodextrin served as the diluent and carrier in

the premix. The chemical forms of the fortificants added to the premix were retinol palmitate

(250,000 IU/g), Vitamin D3 (cholecalciferol), Vitamin E (dl-alpha tocopherol), thiamine mononitrate, riboflavin, pyridoxine hydrochloride, sodium ascorbate, folic acid, niacinamide, vitamin B12 (cyanocobalamin), tricalcium phosphate, copper gluconate, ferric pyrophosphate, sodium selenite, and zinc sulphate.

Soybeans of the Emerge-389 variety from the 2012 crop were supplied by Clarkson

8 Grain Company, Inc., Cerro Gordo IL, USA. Granulated white sugar, non-iodized salt, and cooling ice were obtained from the local supermarket. Municipal water was used for all batches.

Methods

Preparation of Fortification Premix

Fortified and an unfortified micronutrient/sweetener blends (MSB) were prepared 2 days prior to soymilk processing by weighing out the amounts of salt (7.57g), sugar (804.41g), and vitamin- mineral premix (15.96g, fortified MSB only) needed to fortify a single batch of soymilk. All

MSB components were combined in a 1 gallon plastic storage bag with ample headspace, and shaken in a vigorous circular motion for three minutes to allow ample mixing. After mixing, bulk MSB was kept in light-impermeable packaging and stored and transported at ambient temperature to the processing facility.

Soymilk Processing

Two batches of soy milk were prepared at the National Soybean Research Laboratory (NSRL), located at the University of Illinois Champaign-Urbana, using the facility’s standard processing procedures. For each batch of soymilk, 2 kg of dry soybeans were soaked overnight in excess water under refrigeration. Immediately prior to processing the beans were drained, rinsed, and added to the pressure cooker (Soya Cow Machine, Model No.: SC-20, SSP Private Ltd.,

Faridabad, India). Water was added to the cooker at a ratio of 8 parts water to 1 part soybeans

(based on the dry soybean weight recorded prior to soaking). The cooker was then closed and steam was injected (with venting until the air was fully displaced with steam – approximately 12 seconds), whereupon the beans were ground for 2 minutes. Manually controlled steam

9 pressurization continued until the target temperature (105oC) had been maintained for 2 minutes

(a total of 24 minutes for batch 1 and 35 minutes for batch 2). Pressure remained steady at approximately 12 psi (82.7 kPa) for batch 1 and 14 psi (96.5 kPa) for batch 2 for the majority of the processing time. Following processing the steam valve was shut off and the milk was discharged through a filter into a 5 gallon plastic pail.

Premix Incorporation

Immediately after collection, each batch of soymilk was divided into 2 approximately equal lots in tared 5-gallon plastic pails and the weight of soymilk in each pail was recorded. The weight of previously prepared MSB was then adjusted as necessary to deliver the following target inputs per kg of soymilk: salt, 0.8 g; white sugar, 80 g, DSM fortification premix 1.687 g (fortified soymilk only). Temperatures were recorded with a thermocouple probe immediately prior to premix addition, whereupon an appropriate weight of fortified MSB was added to one pail

(768.30 g for batch 1, 773.90 g for batch 2) while the other pail received an appropriate amount of unfortified MSB to serve as a sweetened, unfortified control (753.02 g for batch 1, 758.94 g for batch 2). The fortified and the unfortified premixes were vigorously and simultaneously stirred into each respective treatment lot with a large wire whisk for 1 minute. The hot, fortified soymilk was held for 1 additional minute after stirring prior to filling sample bottles, to provide at least 2 full minutes of pasteurization time, post MSB addition. This delay would accommodate any cooling rate and simulate allowance for potential variation in final temperatures at different processing facilities.

10 Table 1: Soymilk temperatures upon MSB addition.

MSB addition Soymilk Treatment rate Temperature Batch Lot (g/kg) (oC) Unfortified 79.8 80.0 1 Fortified 94.6 82.0 Unfortified 85.8 78.6 2 Fortified 87.5 81.3

Packaging, Cooling, Transport, & Sampling

Delayed cooling or“Hot Hold” Samples: The temperature of the soymilk gradually dropped at a rate of roughly 1oC per minute in the bulk soymilk following MSB addition. To monitor vitamin degradation in the hot soymilk prior to cooling, triplicate analytical samples were taken from bulk soymilk (both fortified and unfortified lots) in 60 ml glass bottles immediately following the

2 minute whisking and pasteurization period, and then again 7.5 and 15 minutes later. These samples were immediately cooled in an ice water bath for 30 minutes before transferring to a freezer. Frozen samples were shipped overnight to Brigham Young University (Provo, Utah), and stored at -80oC until analysis.

“Cooling Method” Samples: Immediately after packaging, the 473 ml PETE bottles (6.35 cm diameter) of both fortified and unfortified soymilk were divided between 2 cooling treatment baths (ambient vs. ice water), where they were allowed to cool for 30 min. The ambient water bath consisted of 38 L of water at about 26oC prior to sample addition. For both batches the ambient water bath remained below 29.2oC for the entire cooling process. The ice water bath was prepared anew for each batch using 19 L of water and 10 kg of ice. Additional ice was added during cooling, and for both batches the ice water bath measured below 3.5oC prior to

11 sample addition and remained below 7oC for the duration of the cooling process. Fully cooled

samples were refrigerated at NSRL and shipped on ice to Brigham Young University for

analysis. Upon receipt, triplicate analytical samples were drawn from each of the two 473 ml

PETE bottles representing every combination of nutrient status and cooling treatment and stored

in 60 ml glass bottles at -80oC until analysis.

“Light Exposure” Storage Study Samples: From each batch 10 samples of fortified soymilk and

10 samples of unfortified soymilk in 473 ml PETE bottles (6.35 cm diameter) were placed in the

large ice-water bath (described above) 15 minutes after fortification and pasteurization was

complete. After 35-40 minutes of cooling, the samples were removed to refrigeration.

Samples for Sensory Analysis: All remaining soymilk was packaged in 473 ml PETE bottles

(6.35 cm diameter), cooled using ice-water, and refrigerated without light exposure for sensory analysis.

Refrigerated Storage Stability Study

After refrigerated shipment to BYU, the soymilk bottles were placed in a walk-in refrigerator kept at 1.1oC. Half of the fortified samples and half of the unfortified samples from each batch

were exposed to 8 hours per day of direct fluorescent lighting during refrigerated storage, while

all the other samples were stored adjacent in a light-protected box.

Light intensity ranged from 1076 -1281 lux at the upper shoulder of the sample bottles,

from 1012-1141 lux at bottle center, and from 969-1066 lux at the base of the sample bottles

depending on bottle placement. Light-exposed sample bottles were rotated in position each day

12 when the light was turned on so as to experience approximately equal light exposure for the

duration of refrigerated storage.

On day 0, analytical samples were taken in triplicate from 2 fortified and 2 unfortified bottles from each production batch and frozen at -80oC in 60 ml glass bottles until analysis. On

days 5 and 12 of refrigerated storage, triplicate analytical samples were drawn from 2 of the 473

ml bottles representing each of the 4 storage treatment groups of each batch (fortified or

unfortified, light or dark storage) and stored in 60 ml glass bottles at -80oC until analysis. In

total, the samples taken represented both fortified and unfortified samples from each batch which

had been stored with or without light exposure for 0, 5 or 12 days.

Nutritional Analysis

All analyses were performed under subdued, yellow, uv-filtered lighting, and all sample replicates were analysed by several analysts who had demonstrated their proficiency with the analysis methods prior to analysing the study samples.

Thiamine & Riboflavin: Thiamine and riboflavin were measured using AOAC method 953.17

(AOAC 2012), with some modifications. Five millilitres of unfortified soymilk or 1 ml of fortified soymilk was pipetted into a tared 150 ml Erlenmeyer flask and the weight recorded.

Forty millilitres of 0.1 N HCl was then added and the sample was swirled for 1 minute before

adjusting the pH to 4.5 ± 0.05 with 2.5 M sodium acetate. Next, 500 mg taka-diastase from

Aspergillus oryzae (100U/mg) was added and the flask was swirled, covered with foil, and

incubated without agitation at 37°C for 18 hours. After incubation, the soy suspension was

filtered through Whatman #541 filter paper into a 100 ml volumetric flask using a glass funnel.

After triple rinsing the Erlenmeyer flask and the soy residue with deionized water, the volumetric

13 flask was brought to volume, capped, and repeatedly inverted to ensure homogeneity.

Approximately 1.5 ml of this solution was collected with a 2 ml syringe, filtered through a 0.2

μm membrane into an amber HPLC vial, crimp capped, and refrigerated for same-day riboflavin

analysis. Ten millilitres of the remaining filtrate was then added to a large centrifuge tube along

with 2.5 g of NaCl and swirled for 2 minutes or until the salt dissolved. While gently swirling, 3

ml of oxidizing reagent was added without it touching the sides of the tube (1 ml of 3%

potassium ferricyanide brought to 25 ml in a volumetric flask with 15% NaOH). After 5

seconds, 15 ml of isobutanol was added and the sample was immediately shaken vigorously for

20 seconds and then for an additional 2 minutes prior to centrifugation. Samples were centrifuged at 1200xG for 4 minutes and 1 ml of supernatant was filtered through a 0.2 μm membrane into an amber HPLC vial, crimp capped, and refrigerated for same-day thiamine analysis (as thiochrome).

Thiamine and riboflavin were quantified against an external standard curve using HPLC separation (Agilent Technologies, Inc., Santa Clara, CA) and fluorometric detection. Ten microliter sample injections were eluted isocratically using a methanol-sodium acetate (0.05 M) mobile phase (30:70 v/v) and a flow rate of 1 ml/min. The stationary phase was an olctadecyl silane column (Luna C8(2), 150 mm x 4.6 mm, 5 μm particle size, Phenomenex Inc., Torrance,

CA). Excitation and emission wavelengths were, respectively: 422 nm and 522 nm for riboflavin; 366 nm and 435 nm for thiochrome.

Folate: Folate was measured using the AOAC trienzyme extraction method 2004.05 (AOAC

2012) with slight alterations, including many from Chapman et al. (2010). Rat plasma with anticoagulant factors provided the folate conjugase enzyme (0.1 ml, male, non-sterile, with lithium and heparin; Pel-Freeze Biologicals, Catalog #36161-2, Rogers AR). The folic acid

14 working standard was brought to a concentration of 1 μg/ml, instead of the 10 ng/ml concentration used in the original method. Following filtration through Whatman 2V filter paper, samples were diluted in a volumetric flask with deionized, autoclaved water by a factor appropriate to approximately equalize the folate concentration between the folic acid standard and all samples.

The 96-well microtiter plate microbiological assay (L. casei subsp. Rhamnosus, ATCC

#7469) of Tamura (1990) was used with minor modifications. Inoculum was maintained by weekly transfers into fresh lactobacilli broth, followed by 24 hours of incubation at 37°C and then refrigeration until use. Prior to plating the samples, cultures were transferred to depletion media, prepared from lactobacilli broth and folic acid casei media following the method of Chen and Eitenmiller (2007). Optical density was read using a FLUOstar OPTIMA microtiter plate reader. Data analysis was performed in excel by comparing optical densities of sample wells from within the linear range of the spectrophotometer with the linear range of the folic acid standard from the same plate (as confirmed by a HorRat value of between 0.3 and 1.3).

Vitamin A: Vitamin A was measured as all-trans retinol palmitate and the 2 degradation isomers with the highest bioactivity, 13-cis and 11-cis retinol palmitate. Bioactivity was calculated using the values reported by Weiser and Somorjai (1992). Vitamin A was reported in micrograms of retinol activity equivalents (mcg RAE), and was calculated using formula 1, below. The bioactive portion recovered was also reported as a percent of all vitamers as per formula 2. The all-trans, 13-cis, and 11-cis isomers were also reported as a wt/wt percent of all vitamers without adjusting for bioactivity.

all-trans + (0.73*13-cis) + (0.34*11-cis) (1)

15 {[all-trans + (0.73*13-cis) + (0.34*11-cis)] / [all-trans + 13-cis + 11-cis]}*100% (2)

Five millilitres of soymilk sample was pipetted into a tared 50 ml centrifuge tube and the weight recorded, followed by digestion as per AOAC Official Method 2012.10 (AOAC 2016).

Samples were then centrifuged for 10 minutes (1200xG) and a 3 or 5 ml syringe was used to filter the supernatant through a 0.45 µm PETE filter into an amber HPLC vial. Vials were filled as full as possible to minimize oxygenated headspace, capped, and refrigerated for same-day analysis.

Vitamin A was quantified against an external standard curve using HPLC separation

(Agilent Technologies, Inc., Santa Clara, CA) and diode array detection. Sample injections of 20

µL were eluted as per the gradient elution cycle outlined by McMahon et al. (2013) except that a flow rate of 1 ml/min was necessary to obtain full baseline resolution of the 13-cis and 11-cis isomer peaks from each other and from the retinol palmitate peak. Mobile phase A was pure hexane, and mobile phase B was hexane-methyl-t-butyl ether (75:25 v/v). The stationary phase was amino-propyl silane bonded to ZORBAX SIL (Zorbax NH2 column, 4.6 x 150 mm, 5 μm particle size, Agilent Technologies, Inc., Santa Clara, CA). The detection wavelengths were set at 10 and 325 nm, with reference wavelengths of 100 nm and 400 nm.

Vitamin C: Vitamin C was analysed using a method revised from Wang et al. (1988) with modifications based on Visser’s commentary of solvent effects (2012) and the work of Chase et al. (1993). One millilitre of soymilk sample was pipetted into a tared 25 ml volumetric flask and the weight recorded before bringing to volume with an acidic digestion solvent (2% analytical grade trichloroacetic acid (TCA), 0.2% dithiothreitol (DTT), and 0.05% metaphosphoric acid).

Six small teflon-coated magnetic stir bars were then added and the flask was capped with a stopper while the sample was mixed on a stir plate at high speed for 10 minutes. Samples were

16 then transferred to 50 ml plastic centrifuge tubes and centrifuged for 5 minutes at 1600 g. Using

a 3 ml syringe, supernatant was filtered through a 0.45 μm membrane into a crimp-capped amber

HPLC vial and immediately analysed. Any sample extracts which had not undergone HPLC analysis within 4 hours of extraction were discarded and the entire analysis was repeated anew.

Vitamin C was quantified against an external standard curve using HPLC separation

(Agilent Technologies, Inc., Santa Clara, CA) and diode array detection. Sample injections of 10

µL were eluted isocratically with a sodium acetate mobile phase (0.5 M, pH 4) and a flow rate of

1 ml/min through a reverse stationary phase (Synergi™ Hydro-pro C-18 column, 250 x 4.6 mm,

5 μm particle size, Phenomenex Inc., Torrence, CA). The detection wavelength was 254 nm, with reference wavelengths of 100 and 360 nm.

Sensory Evaluation

With approval of the university’s Institutional Review board, and with parental consent children ages 7 through 17 were recruited by BYU’s Sensory Analysis Laboratory, and the potential panelists were screened for soy allergies and their willingness to try soymilk. Panelists were rewarded for their participation with their choice from a selection of candy upon completion of the sensory evaluation, and their parents were remunerated at the price of $15 per child or youth who participated in the panel.

Throughout the panel, soymilk was stored on ice in pitchers, and about 75 ml of fortified and unfortified soymilk were poured into 4 fl. oz. (119 ml) clear plastic cups which had been previously labelled with a 3-digit blinding code. Each panelist was presented with 2 soymilk samples (1 fortified and 1 unfortified), side-by-side on the serving tray. Sample presentation from left to right was randomized, and BYU Sensory Lab workers ensured that the amount of

17 soymilk in each cup was visually similar. A cracker and water were provided to cleanse the

palate between samples.

Data was gathered via a digital ballot presented on a computer screen in individually

partitioned booths where samples were presented on trays via a pass-through compartment. The demographic information gathered consisted of age, gender, and whether they liked soymilk, disliked soymilk, or were not sure if they liked or disliked soymilk. Participants then answered 6 questions assessing overall impression, flavour, colour, smell, mouthfeel, and aftertaste.

Panelists were also asked to rank the samples in order of preference by selecting which of the two samples they liked best, and a final question asked how likely they would be to completely drink a full cup of each sample (fortified and unfortified soymilk) in a school cafeteria setting.

Organoleptic properties were rated for each sample using a 7-point hedonic scale with the

following descriptors: really bad, bad, just a little bad, maybe good or maybe bad, just a little

good, good, really good. The final ‘willingness to consume’ question was scored for each

sample using a 5 point hedonic rating scale with the following descriptors: definitely would not

drink all of it, probably would not drink all of it, maybe drink – maybe not drink all of it,

probably would drink all of it, definitely would drink all of it. Undergraduate BYU Sensory Lab

workers and parents were available to clarify the ballot questions to the participants.

Statistical Analysis

Vitamins: Fortified and unfortified samples taken from hot soymilk 2, 7.5, and 15 minutes after

premix addition were analyzed for thiamine, riboflavin, folate, vitamin A, and vitamin C content

using mixed models ANOVA (α=0.05), blocking by sample bottle and production batch. Paired t-tests with post-hoc Tukey-Kramer adjustment (α=0.05) were also performed to investigate potential differences between sampling times. In order to account for time and temperature

18 dependent matrix effects which interfered with vitamin A extraction, isomer data were analyzed

as a percent of total retinyl palmitate vitamers.

Statistical analysis for potential differences between cooling treatments for thiamine,

riboflavin, folate, vitamin A, and vitamin C content was performed on both fortified and

unfortified samples using a paired t-test with post-hoc Tukey adjustment (α=0.05).

The effects of light exposure and dark refrigerated storage on thiamine, riboflavin, folate,

vitamin A, and vitamin C levels for 0, 5 and 12 days were investigated using mixed models

ANOVA (α=0.05), blocking by sample bottle and production batch, for fortified and unfortified samples. Paired t-tests with post-hoc Tukey-Kramer adjustment (α=0.05) were also performed between all combinations of treatments and sampling times in light-exposed and dark refrigerated storage.

Sensory Evaluation: Data from participants who reported a dislike of soymilk were included in the final analysis, because the sole exclusion criteria were soy allergy and an unwillingness to try soymilk. Statistical analysis of hedonic scale questions was carried out using one way ANOVA with post-hoc Tukey’s HSD (α = 0.05). The question asking panelists to rank the samples in order of preference was analyzed using Friedman analysis of rank. Ballot presentation and statistical data analysis was performed using Compusense 5 software (Compusense Inc., Guelph

ON, Canada).

Results

Delayed Cooling

Table 2 and Table 3 show the levels of vitamin A, vitamin C, thiamine, riboflavin, and folate

19 after 0, 7.5, and 15 minutes in hot fortified and unfortified soymilk. Delayed cooling for up to 15 minutes after premix addition (~20 minutes after exiting the “soy cow”) did not affect riboflavin, thiamine, or folate levels in fortified or unfortified samples (lowest p-value=0.7386). Vitamin C loss in uncooled soymilk was significant 15 minutes after fortification (6% loss, p=0.0132).

The reliability of vitamin A recovery was questionable in the hot soymilk after fortification. However, proportional isomer content, reported as quantity of each isomer relative to the sum of all retinol palmitate vitamers, standardized comparisons across recoveries and sampling times. Statistically significant degradation of the 100% bioactive all-trans isomer (2% loss, p=0.0138), primarily to the 11-cis isomer (34% bioactivity), resulted in a slight reduction in bioactive vitamin A relative to the sum of all vitamers (1% loss, p=0.0144).

20 Table 2: Vitamin content of fortified soymilk held hot for 0, 7.5, and 15 minutes after MSB addition.

0 min* 7.5 min 15 min (per 100g) SE (per 100g) SE (per 100g) SE Units p-values** Vitamin A† 113.22 ± 10.35 146.56 ± 10.35 147.74 ± 10.35 mcg RAE p=0.1106, p=0.9964, p=0.0978 % All-trans 94.42 a ± 0.29 93.85 a,b ± 0.29 92.93 b ± 0.29 % (wt/wt) p=0.3891, p=0.1169, p=0.0138 % 13-cis 5.58 a ± 0.13 5.71 a ± 0.13 5.95 a ± 0.13 % (wt/wt) p=0.7762, p=0.4115, p=0.1576 % 11-cis 0.00 a ± 0.23 0.44 a,b ± 0.23 1.12 b ± 0.23 % (wt/wt) p=0.4087, p=0.1511, p=0.0190 % Bioactive‡ 98.49 a ± 0.17 98.17 a,b ± 0.17 97.65 b ± 0.17 % (wt/wt) p=0.3835, p=0.1239, p=0.0144 Vitamin C 12.50 a ± 0.14 12.04 a,b ± 0.14 11.80 b ± 0.14 mg p=0.1045, p=0.4900, p=0.0132 Thiamine 0.324 a ± 0.015 0.312 a ± 0.015 0.328 a ± 0.015 mg p=0.8382, p=0.7386, p=0.9825 Riboflavin 0.221 a ± 0.012 0.220 a ± 0.012 0.218 a ± 0.012 mg p=0.9992, p=0.9891, p=0.9827 Folate 60.20 a ± 1.27 60.44 a ± 1.27 60.54 a ± 1.27 mcg p=0.9900, p=0.9985, p=0.9810 *Describes samples taken immediately after the premix addition and incorporation process was complete. **P-values for, respectively, the pair-wise comparison of sampling times 0 min & 7.5 min, 7.5 min & 15 min, and 0 min & 15 min. †Data not reliable due to variable recovery between sampling times. Relative isomer content, reported as a percent of all vitamers measured, standardizes the isomer data across different analytical recoveries. ‡ Proportion of recovered retinol palmitate with biological vitamin A activity, defined as {[All-trans + (0.73*13-cis) + (0.34*11-cis)] / sum of all vitamers} * 100%

Table 3: Vitamin content of unfortified soymilk held hot for 0, 7.5, and 15 minutes following MSB addition.

0 min* 7.5 min 15 min (per 100g) SE (per 100g) SE (per 100g) SE Units p-values** Thiamine 0.045 a ± 0.003 0.043 a ± 0.003 0.046 a ± 0.003 mg p=0.8584, p=0.8068, p=0.9944 Riboflavin 0.018 a ± 0.002 0.017 a ± 0.002 0.019 a ± 0.002 mg p=0.9905, p=0.8816, p=0.9353 Folate 15.12 a ± 1.21 15.63 a ± 1.21 15.76 a ± 1.24 mcg p=0.9532, p=0.9960, p=0.9287 Vitamin A and Vitamin C were not detected in unfortified samples. *Describes samples taken immediately after the premix addition and incorporation process was complete. **P-values for, respectively, the pair-wise comparison of sampling times 0 min & 7.5 min, 7.5 min & 15 min, and 0 min & 15 min.

21 Cooling Method

No differences were found for any of the vitamins analyzed (riboflavin, thiamine, folate, vitamin

A, and vitamin C) between samples cooled in an ice water bath or in an ambient temperature water bath. This was true for both fortified and unfortified soymilk (see Table 4 and Table 5).

Table 4: Vitamin content of fortified soymilk cooled in ambient temperature water or in an ice water bath.

Ambient Ice (per 100g (per 100g WWB) SE WWB) SE Units p-value Vitamin A 152.76 a ± 7.68 157.08 a ± 7.68 mcg RAE p=0.7049 % All-trans 93.53 a ± 0.20 93.42 a ± 0.20 % (wt/wt) p=0.6897 % 13-cis 5.57 a ± 0.07 5.60 a ± 0.07 % (wt/wt) p=0.7665 % 11-cis 0.90 a ± 0.19 0.99 a ± 0.19 % (wt/wt) p=0.7516 % Bioactive* 97.90 a ± 0.13 97.84 a ± 0.13 % (wt/wt) p=0.7227 Vitamin C 12.03 a ± 0.45 11.66 a ± 0.45 mg p=0.5730 Thiamine 0.327 a ± 0.011 0.335 a ± 0.011 mg p=0.6226 Riboflavin 0.226 a ± 0.009 0.223 a ± 0.009 mg p=0.8449 Folate 60.85 a ± 1.17 61.49 a ± 1.14 mcg p=0.7127 *Proportion of recovered retinyl palmitate with biological vitamin A activity, defined as: {[All-trans + (0.73*13-cis) + (0.34*11-cis)] / sum of all vitamers} * 100%

Table 5: Vitamin content of unfortified soymilk cooled in ambient temperature water or in an ice water bath.

Ambient Ice (per 100g) SE (per 100g) SE Units p-value Thiamine 0.043 a ± 0.003 0.046 a ± 0.003 mg p=0.4467 Riboflavin 0.018 a ± 0.002 0.019 a ± 0.002 mg p=0.6833 Folate 16.32 a ± 0.62 16.65 a ± 0.57 mcg p=0.7094 Vitamin A and vitamin C were not detected in unfortified samples.

22 Refrigerated Storage and Light Exposure

In both fortified and unfortified soymilk 12 days of refrigerated storage had no statistically

significant effect on thiamine, folate, and Vitamin C levels, regardless of light exposure (see

Figure 3). From baseline to 12 days of refrigerated storage mean thiamine levels decreased by

5% from 0.321 mg to 0.306 mg per 100 g of soymilk, mean vitamin C levels decreased by18% from 10.94 mg to 8.93 mg per 100 g of soymilk, and mean folate levels decreased by 5% from

59.50 mg to 56.67 mg per 100 g of soymilk in fortified, light-exposed samples, but the changes were not significant (p=0.9147, p=0.0986, and p=0.0809, respectively). The treatment

comparison between light and dark refrigerated storage did not approach significance for any of

these vitamins at 5 days or 12 days of storage (lowest p-value p=0.4739).

Thiamine Vitamin C Folate

0.40 12 70 0.35 60 10 0.30 50 0.25 8 40 U,D 0.20 6 30 U,L mg/100g 0.15 mg/100g 4 mcg/100g F,D 0.10 20 0.05 2 10 F,L 0.00 0 0 -1 0 5 11 12 -1 0 5 11 12 -1 0 5 11 12 Sampling Day Sampling Day Sampling Day

Data points and error bars are means and 95% confidence intervals, respectively.

Figure 3: Thiamine, vitamin C, & folate levels in fortified (F) and unfortified (U) soymilk during light (L) and dark (D) refrigerated storage.

23 Degradation of riboflavin in fortified, light-exposed refrigerated storage was significant from baseline to 12 days, with a mean decrease from 0.232 mg to 0.190 mg per 100 g of soymilk

(18% loss, p=0.0059) as shown in Figure 4. However, native riboflavin levels in unfortified samples were unchanged after 12 days of light exposure (p=0.4065). In contrast, riboflavin loss in dark refrigerated storage was not significant at 12 days for fortified or unfortified samples (p values of 0.6587 and 0.6919 respectively). The difference in mean riboflavin content between light and dark storage in fortified soymilk was marked after 12 days but was not statistically significant (p=0.0817).

Riboflavin 0.300

0.250 a a a 0.200 a b F,D 0.150 F,L mg/100g 0.100 U,D U,L 0.050

0.000 -1 0 5 1112 Sampling Day Data points and error bars are means and 95% confidence intervals, respectively. Significant riboflavin loss occurred in light-exposed storage, from 232 ± 0.014 mg/100g at 0 days (a) to 190 ± 0.015 mg/100g at 12 days (b); p=0.0059.

Figure 4: Riboflavin levels in soymilk stored in light (L) and dark (D) refrigerated storage.

In light-exposed fortified soymilk a 15% decrease in mean vitamin A (from 200.99 mcg

RAE/100 g to 170.81 mcg RAE/100 g) was pronounced but not statistically significant after 12

days of refrigerated storage (p=0.0840). The comparison between light and dark-stored samples

24 at 12 days was also not statistically significant (p=0.1136). Retinoid vitamin A was not detected

in unfortified soymilk.

In proportional terms, however, changes in the vitamin A isomer profile were significant,

as shown in Figure 5. All-trans isomer content, as a percent of all vitamers measured,

significantly decreased between 0 days to 5 days, 5 days to 12 days, and 0 days to 12 days in

light-exposed refrigerated storage (5%, 4%, and 9% loss, respectively; all p-values <0.0001).

The treatment effect was also very significant between light and dark storage at both 5 days and

12 days (4% and 8% difference, respectively; both p-values <0.0001). In light-exposed

refrigeration, isomerization favoured the 11-cis product, which increased from time zero by

nearly 5-fold at 5 days and by more than 8 times by 12 days relative to all measured isomers

(both p-values <0.0001) while the 13-cis isomer remained unchanged. The increase of 11-cis

isomer in light-exposed samples was significant compared to dark-stored samples at 5 days and

12 days (p<0.0001 for both comparisons). By contrast, formation of 11-cis isomer in dark

refrigerated storage was not significant even after 12 days (p=0.2562), while 13-cis isomer

content increased significantly after 12 days of dark storage (8% increase relative to all vitamers;

p=0.0186). Thirteen-cis isomer formation in light-exposed storage was not significant (p>0.8000

for 0-5 and 5-12 day comparisons, p=0.2440 from 0-12 days), and the difference between light

and dark treatments was not significant for the 13-cis isomer at either duration of storage

(p=0.5730 and p=0.5980 at 5 and 12 days, respectively). The net result of all isomerization, as measured by the bioactivity-adjusted sum relative to the non-adjusted sum of all vitamers reported, was a significant loss of vitamin A activity from baseline in light-exposed storage at both 5 days (3%) and 12 days (5%), which was also significantly different from dark storage

(p<0.0001 for all comparisons).

25 All-trans isomer & 13-cis & 11-cis Bioactive vitamin A isomers 10 100 ** a a‡ a ‡ 9 c 8 b 95 7 a a,a b † † c 6 a a a a 13-cis: D 90 Bioactive: D 5 b b ** 13-cis: L Bioactive: L 4 % wt/wt 11-cis: D % wt/wt c All-trans: D 3 85 2 11-cis: L All-trans: L 1 a a a 80 0 -1 00 5 11 12 -1 0 00 5 11 12 12 Sampling Day Sampling Day

Data points and error bars are means and 95% confidence intervals, respectively. Bioactive vitamin A decreased significantly in light-exposed storage from 97.85 ± 0.20 % at 0 days (a) to 94.95 ± 0.21 % at 5 days (b) and 92.66 ± 0.20 % at 12 days (c); differences between light and dark treatments at 5 days and 12 days were also significant (‡); all p-values <0.0001. All-trans vitamin A decreased significantly in light-exposed storage from 93.41 ± 0.39 % at 0 days (a) to 88.94 ± 0.40 % at 5 days (b) and 85.39 ± 0.39 % at 12 days (c); differences between light and dark treatments at 5 days and 12 days were also significant (†); all p-values <0.0001. 13-cis isomer increased significantly from 5.65 ± 0.18 % at 0 days (a) to 6.08 ± 0.18 % at 12 days (b), p=0.0186. 11-cis isomer increased significantly from 0.94 ± 0.26 % at 0 days (a) to 5.28 ± 0.27 % at 5 days (b) and 8.71 ± 0.25 % at 12 days(c); differences between light and dark treatments at 5 days and 12 days was also significant (**); all p-values <0.0001.

Figure 5: Retinol palmitate isomers and total bioactive vitamin A (sum of bioactivity-adjusted isomer quantities) as a percent of total vitamers measured during light (L) and dark (D) refrigeration.

Sensory Evaluation

A total of 57 children and youth well distributed across the recruited age range participated in the

sensory panel (see Figure 6), consisting of 27 girls (47.4%) and 30 boys (52.6%). Only 23

participants (40.4%) reported having had soymilk before. Most panelists (n=46, 80.7%) were

not sure if they liked soymilk or not, and only 2 panelists (3.5%) claimed to dislike soymilk

26 before tasting the samples.

10 8 6 4

Panelists 2 Number of 0 7 8 9 10 11 12 13 14 15 16 17 Panelist Age (yrs) Figure 6: Panelist age distribution for sensory evaluation.

As shown in Table 6, no difference was found between fortified and unfortified samples in overall likeability (the first evaluation made), and 79% of responses were positive for both the fortified and unfortified soymilk (i.e. “just a little good”, “good”, or “really good”). The median rating was “good” for both samples. When asked in the final question, “If you were served a full cup of this sample in your school cafeteria, would you DRINK all of it or not?” 24 (fortified) and

21 (unfortified) of the 57 panelists responded that they “probably” or “definitely” would drink all of the soymilk, while 24 and 20 panelists (for the fortified and unfortified soymilk respectively) said they “probably” or “definitely” would not drink all of it. Again, the responses were not different between fortified and unfortified samples.

27 Table 6: Panelist ratings of fortified (F) and unfortified (U) soymilk for sensory properties and likeability.

Attribute Sample Mean SD p-value U 5.2 a 1.38 Overall Likability 0.8295 F 5.2 a 1.57 U 5.0 a 1.61 Flavour 0.5630 F 5.1 a 1.56 U 6.0 a 1.02 Colour b 0.0065 F 5.6 1.22 U 5.1 a 1.22 Smell a 0.4009 F 5.0 1.44 U 5.6 a 1.42 Mouthfeel a 0.1848 F 5.8 1.49 U 4.5 a 1.81 Aftertaste 0.4509 F 4.6 a 1.78 U 3.0 a 1.15 Willingness to Consume* 0.7843 F 3.0 a 1.27 Organoleptic properties were rated using the following 7-point hedonic scale: 1-really bad, 2-bad, 3-just a little bad, 4-maybe good or maybe bad, 5-just a little good, 6-good, 7-really good. *The 'willingness to consume' question was rated using the following 5-point hedonic scale: 1-definitely would not drink all of it, 2-probably would not drink all of it, 3-maybe would, maybe would not drink all of it, 4-probably would drink all of it, 5-definitely would drink all of it.

No preferences between fortified and unfortified soymilk emerged for all of the

organoleptic properties excepting colour (see Table 6), for which panelists indicated a slight

preference for the unfortified soymilk (0.4 points, p = 0.0065). Mean scores for all of the

organoleptic properties evaluated were positive, with colour getting the highest average score

and aftertaste scoring the lowest, regardless of fortification status. When ranking the samples in

order of preference (see Table 7), neither sample was preferred over the other by the panellists

(p=0.354).

Table 7: Panelist rankings of soymilk samples in order of preference.

Ranked as 1st Ranked as 2nd Preference Preference # of # of Significantly Different Sample panelists (%) panelists (%) than Sample Fortified 32 (56.1) 25 (43.9) a Unfortified 25 (43.9) 32 (56.1) a No difference between the samples at the 5% level, p=0.354.

28 Discussion

Micronutrient fortified soymilk appears well liked by children and youth, and no preferences for

unfortified soymilk over fortified soymilk were exhibited by the panelists for overall

acceptability and any of the organoleptic properties evaluated except colour. More importantly,

willingness to consume the soymilk was not negatively affected by fortification; the panelists

seemed equally likely to consider consuming a full serving as not. This result is encouragingly

consistent with the results of the sensory study by Reilly et al. (2006), in which, after 4 weeks of

menu availability alongside diary milk, 22.3% of students in 3 different Miami, Florida,

elementary schools voluntarily chose soymilk and consumed an average of 58% of each carton.

Given that in neither study children were likely to be driven to choose soymilk due to excessive

hunger and, in the case of Reilly et al. (2006), excellent alternatives were available, even higher

consumption may reasonably be expected in a humanitarian or community nutrition setting.

The fortified soymilk had a distinctly yellow hue (likely due to the riboflavin in the

fortification premix), which made the unfortified soymilk look lighter and whiter in the side-by-

side panel comparison. However, while the preference for the colour of unfortified soymilk over

fortified soymilk was statistically significant, the practical significance of this preference is

questionable. In fact, the colour of soymilk, fortified or unfortified, seems to be one of its most

appealing properties, because colour received the highest mean score of all the organoleptic

properties evaluated and the standard deviation was relatively small. (No panelists ranked the

colour of either soymilk sample as “bad” or “very bad,” while both fortified and unfortified

samples were rated as “bad” or “very bad” by one or more panelists for every other organoleptic

property evaluated.) Given that 64.9% of the panelists rated the fortified soymilk colour either

“good” or “really good,” it is clear that the colour of the fortified soymilk was actually quite

29 acceptable and that the colour of the fortified sample was viewed negatively only in contrast to the unfortified soymilk – a comparison perhaps less likely to be made in a real-world setting of soymilk consumption.

The only other organoleptic characteristic even remotely approaching statistical differentiation between the two samples was mouthfeel (p = 0.185), with the average score for fortified soymilk being slightly more positive than for the unfortified samples. The fortified samples were slightly thicker, possibly due to the corn starch diluent which made up nearly 25% of the fortification premix by weight. Another likely explanation for the slight increase in thickness was the potential interaction of added minerals, such as calcium, with the soy proteins.

Saeidy et al. (2014) reported increased viscosity in soymilk due to calcium fortification, and used microencapsulation and chelation of calcium to help reduce viscosity. Given that the slight increase in viscosity was not significant and did not negatively affect sensory scores, it does not appear necessary to alter the fortificant to preserve the texture of unfortified soymilk.

The fact that there was no difference in thiamine, riboflavin, and folate levels between soymilk cooled immediately after fortification and soymilk left hot for fifteen minutes prior to cooling is evidence that delivery of these three nutrients is robust even in less than ideal processing conditions. Impressively, the vitamin A used in our experiment (water-dispersible;

250,000 IU/g) was only slightly less stable prior to cooling. It is possible that the water- solubilizing matrix used to deliver the retinol palmitate protected the vitamin A from degradation

– to the point of making it difficult to fully extract for analysis until it had been exposed to temperatures above 60°C for 15 minutes, cooled, and entered into storage. Conversion to less bioactive isomers did occur after the fortified soymilk had been left hot for 15 minutes.

However, the degradation did not appear to be practically significant, and increases in 13-cis

30 isomer reported by others during heat treatment of foodstuffs (Kurzer 2013) were not observed

here. The proportion of 11-cis isomer relative to all isomers significantly increased 15 minutes

after premix addition, but less than 2% of the all-trans isomer and only 1% of the bioactive

portion were lost. This suggests that, at least for the type of vitamin A ingredient used in the

fortification premix, the majority of vitamin A activity can be preserved in our soymilk

production process even without expedited bottling and cooling after fortification.

A significant portion of the fortified Vitamin C was lost due to heat exposure. Most of the loss seemed to occur at the highest temperatures immediately after fortification (two thirds of the total average decrease between premix addition and 15 minutes post-fortification occurred within the first 7.5 minutes), suggesting that the degradation would be difficult to avoid even if considerable effort were made to rapidly bottle and cool the product. Sharma and Lal (2005) found that added microencapsulated vitamin C in pasteurized (30 min. at 63°C) and sterilized (15 min. at 121°C) buffalo milk decreased by 12% and 44%, respectively, indicating that time- temperature relationships are significant with vitamin C degradation. Given that the sodium ascorbate used in the fortification premix was not protected by microencapsulation, the observed loss of vitamin C is not surprising. Despite the loss, however, fortification of soymilk in the present study still increased the vitamin C content of soymilk from an undetectable level in unfortified soymilk to more than 10 mg/100 ml in the fortified product prior to storage. Given that the dosages involved are still well below established tolerable upper levels of acceptable intake and that much of the loss of vitamin C occurs quickly enough as to be difficult to fully avoid with rapid cooling, increased vitamin C overage in the premix (e.g. from 50% to 100%) to account for loss in the hot soymilk after fortification would be acceptable and warranted.

31 Vitamin degradation in refrigerated storage varies by vitamin and can be significant

depending on the extent of light-exposure. Thiamine was the most stable vitamin of those

analyzed, with no degradation detected after 12 days in dark or light-exposed refrigeration.

Native riboflavin in unfortified samples was also resistant to degradationin light-exposed refrigerated storage, but fortified riboflavin was not. Added folate and vitamin C (the former apparently quite resilient and the latter quite vulnerable to thermal degradation) both appeared mildly susceptible to degradation in light-exposed refrigeration. Okwunodulu and Iwe (2015) also reported slow but steady loss of vitamin C during storage of fortified, sprouted soymilk.

Although statistical significance was not reached within 12 days in this study (α=0.05), degradation of these two vitamins may very well have been significant under conditions of stronger lighting, more than just eight hours of light exposure per day, or longer duration of storage. In a lighted, refrigerated display case of a supermarket with long operating hours, for example, a faster rate of degradation than was measured here is quite plausible, emphasizing the importance of reducing the time to consumption of the fortified beverage. Folate-polysaccharide molecular interactions were not investigated in this study and the carbohydrate content

(including fiber) was not analyzed, but the folate stability observed during both prolonged heat exposure and light-exposed storage may in part be due to the protective effect of polysaccharides

(including any fiber which may have persisted beyond the filtration process) against light, heat, and oxygen reported by Liu et al. (2012) and Ding and Yao (2013).

The loss of riboflavin due to light exposure was significant under the study conditions, and the loss of added vitamin A nearly reached statistical significance. In fact, the decrease (on

average) by a considerable 15% of retinol activity equivalents in 12 days of refrigeration is arguably practically significant in spite of the lack of statistical significance. Statistical

32 significance may very well have been reached under conditions of stronger lighting, more light exposure per day, or less variation in analytical recovery.

In terms of each isomer as a percent of total retinol palmitate vitamers measured, vitamin

A degradation was very statistically significant – showing very clearly that substantial vitamin A degradation was occurring. Not only was there a significant decrease in the proportion of fully bioactive all-trans retinol palmitate during light exposure, but the 11-cis isomer with inferior bioactivity (34%) was produced preferentially over the more bioactive (73%) 13-cis isomer.

This result confirms reports by Jung et al. (1998) and Chen et al. (1996) that the formation of isomers other than 13-cis (all of which have lower bioactivity) predominate under conditions of light-exposed storage while 13-cis isomer formation is favoured in the dark. As noted by Panfili et al. (2008) there is some error inherent to the interpretation of vitamin A degradation when all isomers are not measured, but because all of the isomers that were not measured exhibit lower bioactivities still than those that were, this observation hardly changes the practical significance of the observation that all-trans retinol palmitate was being degraded in light-exposed storage. In summary, our results suggest that significant loss of vitamin A and riboflavin could occur within

5-10 days with even moderate, non-continuous light exposure, and loss of vitamin C and folate may also become significant in a setting of high-intensity, long-duration light exposure such as might occur in a well-lit store display.

Given the degradative trends observed in light-exposed refrigerated storage for 4 of the 5 vitamins measured, strategies for reducing nutrient loss is of interest if broadly fortified soymilk must be exposed to light in storage or consumer display. Work by Jung et al. (1998) and Kim et al. (2000) suggests that fortification with vitamin C and/or an array of other vitamins is protective against vitamin A isomerization. But this strategy is already exhausted in our

33 fortification setting due to the comprehensive nature of the micronutrient premix, and it may not

even be effective in the presence of fortified minerals, in particular iron and zinc (Pinkaew et al.

2012). Given that these two minerals are among the nutrients most critical for at-risk

populations, removing them from the fortificant in an attempt to reduce vitamin loss may not be

desirable. For all the vitamins (except for folate) that trended toward degradation in light-

exposed storage with or without statistical significance, a large portion of the degradation

seemed to occur within the first 5 days, suggesting that managing inventory so as to decrease, as

much as possible, the time between production and consumption may preferable. Light-reducing

packaging may be one of the most promising options. For instance, Bianchi et al. (2015)

demonstrated that sufficient light-protective additives in HDPE plastic bottles were able to

protect soymilk against sensory changes for up to 15 days and riboflavin degradation for up to 29

days compared to clear plastic controls. Obviously, however, avoiding light exposure between

processing and the point of consumption is ideal. In spite of significant 13-cis isomer production

in dark refrigerated storage, no significant loss of bioactive vitamin A occurred, and no

significant loss of any of the other vitamins measured occurred in light-protected storage either.

Conclusion

Micronutrient enrichment of soymilk appears feasible in small scale batch production settings,

such as might be operated by small businesses and localized humanitarian agencies. The demonstrated possibility of delivering to children and youth substantial quantities of a large number of essential micronutrients in just a small bottle of soymilk with practically no compromise in organoleptic acceptability, confirms that fortified soymilk could potentially

34 significantly impact micronutrient intake as a solo medium. Based on the sensory panel results, willingness to consume soymilk may be the biggest obstacle in supplementing a child’s diet with micronutrients through this means. It is hard to say how sensory results with children in the

United States might compare to those for children in less-developed countries, but aesthetic packaging and appetite-conducive distribution settings may enhance humanitarian and community nutrition programs. Additionally, for better or worse, the populations most in need of supplemental nutrition may be more willing to consume soymilk if it were offered to them.

In terms of the stability of fortified nutrients, the fact that ambient temperature water baths were no worse than ice water baths and the fact that many of the vitamins were resilient in the face of delayed cooling bode well for the slow, non-mechanized, and often minimally-staffed settings of small-scale batch processes which may lack the facilities and manpower required for rapid bottling and cooling. However, increased overage of vitamin C is recommended to reduce thermal loss during processing. The discovery that none of the vitamins evaluated degraded significantly during 12 days of dark refrigerated storage allows for a moderate refrigerated shelf life that could extend the impact of each production batch and allow for the carrying of inventory, as long as light-protection is sufficiently complete. Since the more light exposure is permitted, the sooner consumption would need to occur to get maximum delivery of fortified nutrients, strategies such as light-proof secondary packaging (as in this study), dark refrigeration until consumption, or primary packaging with light-protective additives are recommended.

35 References

Aaron GJ, Dror DK, Yang Z. 2015. Multiple-micronutrient fortified non-dairy beverage interventions reduce the risk of anemia and iron deficiency in school-aged children in low-middle income countries: a systematic review and meta-analysis. Nutrients. 7(5):3847-3868. Allen LH. 2003. Interventions for micronutrient deficiency control in developing countries: past, present and future. J Nutr. 133:3875S–3878S. Anjum K, ChikkeGowda RK, Bhagya S. 2013. Functional and nutritional evaluation of supplementary food formulations. J Food Sci Technol. 50(2):309-16. [AOAC] AOAC International. 2012. Official methods of analysis of AOAC International. 19th ed. Gaithersburg (MD). AOAC International. Official method 953.17. [AOAC] AOAC International. 2012. Official methods of analysis of AOAC International. 19th ed. Gaithersburg (MD). AOAC International. Official method 2004.05. [AOAC] AOAC International. 2016. Official methods of analysis of AOAC International. 19th ed. Gaithersburg (MD). AOAC International. Official method 2012.10. Bianchi LM, Duncan SE, Webster JB, Johnson DS, Chang HH, Marcy JE, OKeefe SF. 2015. Protecting soymilk flavor and nutrients from photodegradation. Food Sci Nutri. 3(4):319- 330. Bisimwa G, Owino VO, Bahwere P, Dramaix M, Donnen P, Dibari F, Collins S. 2012. Randomized controlled trial of the effectiveness of a soybean-maize-sorghum-based ready-to-use complementary food paste on infant growth in South Kivu, Democratic Republic of Congo. Am J Clin Nutr. 95(5):1157-64. Boomgarden D. 2016. In northern Ghana, it’s the soycows that are boosting protein in kids’ diets. Frontlines (online) [Internet]. [Last updated: October 13, 2017; cited: February 8, 2019]; 2016 (July/August). Available from: https://www.usaid.gov/news- information/frontlines/july-august-2016/northern-ghana-soycows-are-boosting-protein Briend A, Akomo P, Bahwere P, de Pee S, Dibari F, Golden MH, Manary M, Ryan K. 2015. Developing food supplements for moderately malnourished children: lessons learned from ready-to-use therapeutic foods. Food Nutr Bull. 36(1):S53-S58.

36 Carrillo E, Varela P, Fiszman S. 2012. Effects of food package information and sensory characteristics on the perception of healthiness and the acceptability of enriched biscuits. Food Res Int. 48:209-216. Chapman JS, Steele FM, Eggett DL, Johnston NP, Dunn ML. 2010. Stability of native folate and added folic acid in micronutrient-fortified corn masa and tortillas. Cereal Chem. 87(5):434-438. Chase GW, Landen WO Jr, Soliman AG, Eitenmiller RR. 1993. Method modification for liquid chromatographic determination of thiamine, riboflavin and pyridoxine in medical foods. J AOAC Int. 76(6):1276-80. Chen HE, Peng HY, Chen BH. 1996. Stability of carotenoids and vitamin A during storage of carrot juice. Food Chem. 57(4):497-503. Chen L, Eitenmiller RR. 2007. Single laboratory method performance evaluation for the analysis of total food folate by trienzyme extraction and microplate assay. J Food Sci. 72:C243- C247. Cossack ZT, Rojhani A. 1992. Fiber in enteral formulas - effects of sugar-beet versus soy fiber on zinc and folic-acid absorption in human-subjects. Clin Nutri. 11(1):35-8. de Pee S, Bloem MW. 2009. Current and potential role of specially formulated foods and food supplements for preventing malnutrition among 6-to 23-month-old children and for treating moderate malnutrition among 6-to 59-month-old children. Food Nutri Bull. 30(3):S434-S63. Dharmasena S, Capps O Jr. 2014. Unraveling demand for dairy-alternative beverages in the United States: the case of soymilk. Agric Resour Econ Rev. 43(1):140-157. Ding X, Yao P. 2013. Soy protein/soy polysaccharide complex nanogels: folic acid loading, protection, and controlled delivery. Langmuir. 29(27):8636-44. Flogan C, Dahl W. 2010. Effects of fiber-fortified foods on children with constipation: potential improved stool frequency and decreased energy intake. Infant Child Adolesc Nutr. 2(5):312–317. Giri SK, Mangaraj S. 2012. Processing influences on composition and quality attributes of soymilk and its powder. Food Eng Rev. 4(3):149-64. Heaney RP, Rafferty K. 2006. The settling problem in calcium-fortified soybean drinks. J Am Diet Assoc. 106(11):1755-3; author reply.

37 Ho SC, Guldan G, Woo J, Yu R, Tse M, Sham A, Cheng J. 2005. A prospective study of the effects of 1-year calcium-fortified soy milk supplementation on dietary calcium intake and bone health in Chinese adolescent girls aged 14 to 16. Osteoporos Int. 16(12):1907- 16. Hurrell RF. 2002. Fortification: overcoming technical and practical barriers. J Nutr. 132(4):806S-12S. Hurrell RF. 2004. Phytic acid degradation as a means of improving iron absorption. Int J Vitam Nutr Res. 74(6):445-52. Hurrell RF, Lynch S, Bothwell T, Cori H, Glahn R, Hertrampf E, Kratky Z, Miller D, Rodenstein M, Streekstra H, Teucher B, Turner E, Yeung CK, Zimmermann MB. 2004. Enhancing the absorption of fortification iron - a SUSTAIN task force report. Int J Vitam Nutr Res. 74(6):387-401. Jung MY, Lee KH, Kim SY. 1998. Retinyl palmitate isomers in skim milk during light storage as affected by ascorbic acid. J Food Sci. 63(4):597-600. Kaludjerovic J, Ward WE. 2013. Adequate but not supplemental folic acid combined with soy isoflavones during early life improves bone health at adulthood in male mice. J Nutri Biochem. 24(10):1691-6. Kim YS, Strand E, Dickmann R, Warthesen J. 2000. Degradation of vitamin A palmitate in corn flakes during storage. J Food Sci. 65(7):1216-1219. Kurzer AB. 2013. The protective effect of antioxidants on vitamin A stability in nonfat dry milk during thermally accelerated storage [master’s thesis]. [Provo, (UT)]. Brigham Young University. Lambrecht HS, Nielsen SS, Liska BJ, Nielsen NC. 1996. Effect of soybean storage on tofu and soymilk production. J Food Qual. 19:189-202. Li ZG, Yue XY, Liu YJ, Zhang YX. 2011. Effect of calcium salts and tri-sodium citrate on certain properties of calcium-fortified soy milk. Afr J Biotechnol. 10(52):10698-705. Liu Y, Tomiuk S, Rozoy E, Simard S, Bazinet L, Green T, Kitts DD. 2012. Thermal oxidation studies on reduced folate, L-5-methyltetrahydrofolic acid (L-5-MTHF) and strategies for stabilization using food matrices. J Food Sci. 77(2):C236-C43. Mazumder AR, Begum AA. 2016. Soymilk as source of nutrient for malnourished population of developing country: a review. Int J Adv Sci Tech Res. 6(5):192-203.

38 McMahon A, Christiansen S, Shine L, Loi C, Dowell D. 2013. Simultaneous determination of 13-cis and all-trans vitamin A palmitate (retinyl palmitate), vitamin A acetate (retinyl acetate), and total vitamin E (α-tocopherol and DL- α-tocopherol acetate) in infant formula and adult nutritionals by normal phase HPLC: first action 2012.10. J AOAC Int. 96(5):1073-81. Nackers F, Broillet F, Oumarou D, Djibo A, Gaboulaud V, Guerin PJ, Rusch B, Grais RF, Captier V. 2010. Effectiveness of ready-to-use therapeutic food compared to a corn/soy- blend-based pre-mix for the treatment of childhood moderate acute malnutrition in Niger. J Trop Pediatr. 56(6):407-13. Okwunodulu I, Iwe MO. 2015. Micronutrient evaluation of fortified soymilk from sprouted whole soybean for complementary feeding using response surface methodology. J Mol Stud Med Res. 01(01):16-25. Panfili G, Fratianni A, Di Criscio T, Gammariello D, Sorrentino E. 2008. Influence of microorganisms on retinol isomerization in milk. J Dairy Res. 75(1):37-43. Pathomrungsiyounggul P, Grandison AS, Lewis MJ. 2007. Effects of calcium chloride and sodium hexametaphosphate on certain chemical and physical properties of soymilk. J Food Sci. 72(8):E428-E34. Pinkaew S, Wegmuller R, Hurrell R. 2012. Vitamin A stability in triple fortified extruded, artificial rice grains containing iron, zinc and vitamin A. Int J Food Sci Technol. 47:2212–2220. Reilly JK, Lanou AJ, Barnard ND, Seidl K, Green AA. 2006. Acceptability of soymilk as a calcium-rich beverage in elementary school children. J Am Diet Assoc. 106(4):590-3. Rivera JA, Hotz C, Gonzalez-Cossio T, Neufeld L, Garcia-Guerra A. 2003. The Effect of Micronutrient Deficiencies on Child Growth: A Review of Results from Community- Based Supplementation Trials. J Nutr. 133:4010S–4020S. Saeidy S, Keramat J, Nasirpour A. 2014. Physicochemical properties of calcium-fortified soymilk with microencapsulated and chelated calcium salt. Eur Food Res Technol. 238(1):105-112. Sharma R, Lal D. 2005. Fortification of milk with microencapsulated vitamin C and its thermal stability. J Food Sci Technol-Mysore. 42(2):191-194.

39 Suchdev PS, Peña‐Rosas JP, De‐Regil LM. 2015. Multiple micronutrient powders for home (point‐of‐use) fortification of foods in pregnant women. Cochrane Database Syst Rev. 2015(6):CD011158. doi:10.1002/14651858.CD011158.pub2. Swagerty DL, Walling AD, Klein RM. 2002. Lactose Intolerance. Am Fam Physician. 65(9):1845-1850. Tamura T. 1990. Microbiological assay of folates. In: Contemporary issues in clinical nutrition. Vol 13. Picciano MF, Stokstad ELR, Gregory JF III, editors. Wiley-Liss. New York (NY). pp 121-138. Tang AL, Walker KZ, Wilcox G, Strauss BJ, Ashton JF, Stojanovska L. 2010. Calcium absorption in Australian osteopenic post-menopausal women: an acute comparative study of fortified soymilk to cows' milk. Asia Pac J Clin Nutr. 19(2):243-9. Teshome EM, Oriaro VS, Andango PEA, Prentice AM, and Verhoef H. 2018. Adherence to home fortification with micronutrient powders in Kenyan pre-school children: self- reporting and sachet counts compared to an electronic monitoring device. BMC Public Health. 18:205. Torrezan R, Frazier RA, Cristianini M. 2010. Effect of isostatic high pressure treatment on phytate and trypsin inhibitor content of soy protein. Boletim Do Centro De Pesquisa De Processamento De Alimentos. 28(2):179-86. Tripathi B, Ravi R, Prakesh M, Platel K. 2011. Sensory acceptability of iron-fortified millet products. Int J Food Sci Nutri. 62(6): 651–659. Uzzan M, Nechrebeki J, Labuza TP. 2007. Thermal and storage stability of nutraceuticals in a milk beverage dietary supplement. J Food Sci. 72(3):E109-E14. Visser FR. 2012. Some effects of replacement of metaphosphoric acid/acetic acid solvent system with trichloroacetic acid in microfluorometric determination of vitamin C. J AOAC Int. 67(5):1020-1022. Wang XY, Liao ML, Hung TH, and Seib PA. 1988. Liquid chromatographic determination of L- ascorbic 2-polyphosphate in fish feeds by enzymatic release of L-ascorbate. J Assoc Off Anal Chem. 71(6):1158-1161. Weidner C, Krempf M, Bard J-M, Cazaubiel M, and Bell D. 2008. Cholesterol lowering effect of a soy drink enriched with plant sterols in a French population with moderate hypercholesterolemia. Lipids Health Dis. 7:35.

40 Weiser H, Somorjai G. 1992. Bioactivity of cis and dicis isomers of vitamin A esters. Int J Vitam Nutr Res. 62(3):201-208. Wesley AS, Horton S. 2011. Economics of food fortification. In: Nutrients, dietary supplements, and nutriceuticals. Gerald J, Watson R, Preedy V, editors. Nutrition and health. New York City (NY): Humana Press. p. 31-40. Yuan S, Chang SKC, Liu Z, Xu B. 2008. Elimination of trypsin inhibitor activity and beany flavor in soy milk by consecutive blanching and ultrahigh-temperature (UHT) processing. J Agric Food Chem. 56(17):7957-63. Zhao Y, Martin BR, Weaver CM. 2005. Calcium bioavailability of calcium carbonate fortified soymilk is equivalent to cow's milk in young women. J Nutri. 135(10):2379-82.

41 APPENDICES

Appendix A: NSRL Soymilk Production & Study Design

Preparing Treatment Premixes (BYU, July 27, 2014) Based on mechanical cow batch volume gathered from a preliminary visit to the National Soybean Research Laboratory – NSRL (University of Illinois, Champaign-Urbana) - the fortification premixes were weighed out at BYU using a Mettler Toledo scale into gallon Ziploc bags and mixed by shaking in a vigorous swirling motion for three minutes - the bag containing a significant amount of air. The premixes were intended to approximate the main additives at the quantities used in Ecuador Hogar de Cristo soymilk production and were pre-prepared to facilitate rapid and efficient fortification during production. Immediately after weighing the soymilk from each batch into its two treatment portions, the appropriate amount of premix was weighed by subtracting the calculated amount of excess premix weight from the gallon Ziploc bag amount using Mettler Toledo scale. The premixes were whisked vigorously into the soymilk, with whisking commenced as the premix was poured slowly and steadily into the liquid. Premix was prepared to be adequate for batch sizes of 5 gallons (18.927 L) and targets were as follows: Salt: 0.8 g/L Sugar: 85 g/L DSM Vitamin-Mineral Premix: 1.687 g/kg (1 kg was assumed equal to 1 L) Unfortified Premix addition target: 38.92 g/lb soymilk Fortified Premix addition target: 39.687 g/lb

BATCH 1 (α) Unfortified Premix Salt: 7.57 g Sugar: 804.41 g Total: 811.98

BATCH 1 (α) Fortified Premix Salt: 7.57 g Sugar: 804.41 g DSM Vitamin-Mineral Premix: 15.96 g Total: 827.94

BATCH 2 (β) Unfortified Premix Salt: 7.57 g Sugar: 804.41 g Total: 811.98

BATCH 1 (β) Unfortified Premix Salt: 7.57 g Sugar: 804.41 g DSM Vitamin-Mineral Premix: 15.96 g Total: 827.94

42 Preparing Dry Beans (NSRL, July 29, 2014) 2 Kg dry beans in excess water per batch Soaked the beans overnight under refrigeration (started soak at 2:15 pm on 7/28)

Processing Beans Drained and rinsed beans: Batch 1 - 9:00a.m. Batch 2 – about 1:00 p.m. (using a back-up batch of beans we had soaked just in case…) because the 11:00 am batch failed to cook properly due to an air bubble in the pressure cooker that prevented proper pressurization and temperature control.

Added beans to cooker with 35.2 lbs of water (dry bean to water ratio of 1:8) Closed cooker and turned on steam. Vented for about 10 seconds to remove air and replace it with steam.

Closed valves and ground beans for 2 minutes.

Steam pressurization continued until bean slurry reached 105oC. This took 22 mins for batch 1(reached about 12 psi); 33 mins for batch 2. Batch 2 eventually maintained ~14 psi. The gauge did reach 20 psi at one point and some pressure had to be released from a release valve on the lid of the unit with some loss of soy milk. Scott Buchannan thought that an air bubble might have remained in the pressure cooker due to incomplete evacuation of the pot with steam. The air could have explained the pressure irregularities and difficulties coming to temperature which persisted until use of the release valve solved the problem.

Once 105oC was reached, the timer started, and the beans were “pasteurized” (held at this temperature) for 2 mins.

Steam turned off and milk discharged through filter funnel, which collected residual okara in effluent. Milk was collected in two 5 gallon plastic pails.

BATCH 1 (α):

Batch 1 Exit Temp =82oC: Bucket 1 = 80oC, Bucket 2 = 82oC. Total weight of soy milk = 38.7 lb. Bucket 1 weighed 23.2 lb w/ bucket = 21.7 lb w/o bucket. Bucket 2 weighed 18.5 lb w/bucket = 17 lb w/o bucket. Total =38.7lb/2 = 19.35 lb per treatment. The lighter weight bucket (which received the fortified treatment) was put back on scale and brought to 19.4 lbs using milk from the other bucket. However, we forgot to account for the 1.5 lb bucket weight, so 1.5 lbs less soymilk was transferred than should have been from bucket 1 to bucket 2. Bucket 2 received the vitamin and mineral fortification treatment; as a result, the fortified soymilk received the fortification treatment calculated for 19.4 lbs of soymilk in only 17.9 lbs of soymilk.

BATCH 2 (β):

Batch 2 Exit Temp = 84oC: Bucket 1 =78.6oC, Bucket 2 =81.3oC. Total weight of soy milk = 39 lb. Bucket 1 = 21.55lb w/ bucket = 20.5 lb w/o bucket. Bucket 2 weighed 20 lb w/ bucket = 18.5 lb w/o bucket. Total = 39 lb/2 = 19.5 lb/treatment. We factored in the bucket weight when dividing this time.

43 Appropriate amounts of fortified and unfortified premix were weighed out, pre-blended by shaking in a gallon Ziploc bag, and added. For fortified soy milk, premix was added at a rate of 39.687 g/lb of milk. Unfortified premix was added at 38.92g/lb milk.

Salt was used at 0.8g/kg of soymilk, white sugar at 80g/kg of soymilk, and vitamin-mineral blend (DSM) at 1.687g/kg of soymilk.

Fortified premix = sugar, salt, V/M. Batch 1 (α) = 768.3 g; Batch 2 (β) = 773.9 g Unfortified premix = sugar, salt. Batch 1 (α) = 753.02 g; Batch 2 (β) = 758.94 g

Premixes were stirred into the soymilk with a large wire whisk for 1 minute. Both fortified and unfortified portions of each batch received the premix addition and whisking at the same time. Premixes readily dissolved without issue.

The soymilk was held for 1 minute prior to filling the sample bottles as per the sampling schedules, giving the premixes 2 full minutes of contact time in the hot soymilk.

Filled and capped bottles were placed in water baths to cool for 30 min. Bottles for hot hold samples were 2 oz glass bottles with paper-lined plastic caps. Bottles for cooling and light treatments were 8 oz clear PET with the same plastic caps that are commonly used for soda pop.

Water baths for samples in 8 oz PET bottles were as follows:

Ambient water bath: 10 gallons water. Ambient air temperature was approximately 40 oC, though water bath was at about 80oF (26oC) when samples were first added for each batch.

Ice bath: 5 gallons water + 22 lb ice, prepared to below 4oC for each batch before the addition of samples. Plenty of ice was retained for both batches throughout the soymilk cooling process.

Water bath for 2 oz glass bottles was as follows:

Ice bath: 2 gallons water + 22 lb bag of ice

Sampling and study design:

Sampling Methods at NSRL Soymilk from each batch was all gathered in a large metal bucket with an oblong rim portion for easier pouring. From this container, the soymilk was poured into two relatively equal portions in tared 5 gallon plastic buckets (one at a time) on a large, non-digital top-loading scale and weighed. Having been pre-blended in more than adequate quantities, the premix weights were then adjusted to meet fortification targets for the actual quantity of soymilk in each of the treatment buckets and the premixes were added to 5 gallon buckets and mixed as quickly as possible. Sample bottles were filled from a hand-held pitcher that was filled from the large 5 gallon treatment buckets. The soymilk in the pitcher was whisked between filling every sample bottle, and every few minutes during the process of filling the sample bottles, the remaining soymilk in the pitcher was whisked back into the large quantity in the 5 gallon treatment bucket.

44 The pitcher was then re-filled and the filling of sample bottles continued. This was done in an attempt to have samples that were both equal and representative in terms of content and temperature.

BATCH 1 (α): Time from pressure cooker expulsion to fortification: 6 min Time spent vigorously whisking in the fortificant/premix: 1 min

“Hot Hold” Treatment Description: Triplicate samples were taken from un-cooled soymilk during the first (hottest) 15 minutes in 2 oz glass bottles. Samples were named using four digits representing: batch, fortification status, sampling time, and sample replicate (eg. αF1c or αU3a).

The following times are measured from the end of the 1 minute step of mixing in the fortificant/premixes, and temperatures were taken inserting a thermocouple probe into the last- filled sample bottle immediately prior to placing them in the ice-water bath:

Sample Names Time sample Temperature Temperature Time sample (samples taken bottle put in ice fortified (oC) unfortified (oC) removed from in triplicate) bath (min:s) ice bath (min:s) αF1/αU1 (a,b,c) 2:00 82 > ? > 71.5 80 > ? > 68.9 40:00 αF2/αU2 (a,b,c) 8:45 71.5 68.9 40:00 αF3/αU3 (a,b,c) 15:45 66.5 64.2 55:00

“Cooling Method Comparison” Treatment Description: Matched pairs of triplicate samples were taken in 16oz PET clear plastic bottles and cooled in either an ice-water or an ambient water bath. Samples were named using four digits representing: batch, fortification status, cooling method or water bath used, and sample replicate as follows: - αFC1-3, αFW1-3 - αUC1-3, αUW1-3

The following times are measured from the end of the 1 minute step of mixing in the fortificant/premixes, and temperatures were taken inserting a thermocouple probe into a sacrificial sample bottle (not one of the three replicates used for vitamin analysis):

Time (min:sec) Ice Bath Ambient Bath Temperature of Temperature of Temperature Temperature representative representative (oC) (oC) bottle (αFC4) in bottle (αUW4) Ice Bath (oC) in Ice Bath (oC) 8:45 ? ? 71.5 68.9 15:00 3.5 28.5 11.5 31.5 20:00 3.5 28.7 7.2 30.3 25:00 5.6 29.0 6.7 30.2 30:00 6.0 29.1 7.4 29.9 35:00 6.3 29.1 6.6 29.9 40:00 Bottles removed and refrigerated

“Light” Treatment Description: Ten samples from the fortified treatment and ten samples from the unfortified treatment were taken in 16oz PET clear plastic bottles and cooled in the ice-water bath. All samples were put in the large ice bath approximately 15 minutes and 45 seconds after the fortification step was complete (vigorous whisking for 1 minute) and removed to refrigeration 40 minutes later. At the time the samples were put in the ice bath to cool the soymilk of the fortified portion of the batch was at 66.5oC and the soymilk of the unfortified treatment was at 64.2oC.

After refrigerated shipment to BYU, triplicate samples were drawn in 2oz glass bottles from two of the fortified 16oz sample bottles and two of the unfortified 16oz sample bottles as baseline measures and frozen to await analysis at -80oC. Four of the remaining 16oz sample bottles from each of the fortified and unfortified treatments were then held in a walk-in refrigerator (set at 4oC) exposed to direct fluorescent lighting, and the other four 16oz sample bottles were held in the same refrigerator within a completely light-protected cardboard box. On day 5 of refrigeration at BYU, triplicate samples were then drawn in 2oz glass bottles from two of the fortified 16oz sample bottles exposed to light and two of the fortified 16oz sample bottles stored in the dark. The same was done for the unfortified samples. Finally, on day 12 of refrigeration at BYU, triplicate samples were drawn in 2oz glass bottles from final two of the fortified 16oz sample bottles exposed to light and two of the fortified 16oz sample bottles stored in the dark. The same was done for the unfortified samples. Triplicate 2oz quantities were taken only after thoroughly shaking the 16oz samples from which they were drawn to ensure representative samples were obtained. Samples were named using 7 digits or components representing: batch, fortification status, ice-bath cooling treatment, sampling time, light or dark refrigeration, duplicate 16oz bottle replicate, and triplicate 2oz sample replicate.

Primary treatments are shown below. Triplicate samples from all these were identified with a sub-sample number of 1,2, or 3, and frozen in 2oz glass bottles at -80oC until analysis.

Sampling Fortified Treatment Unfortified Treatment Time (days) Baseline αFC0L/DA αFC0L/DB αUC0L/DA αUC0L/DB Light Dark Light Dark 5 αFC5LA αFC5LB αFC5DA αFC5DB αUC5LA αUC5LB αUC5DA αUC5DB 12 αFC12LA αFC12LB αFC12DA αFC12DB αUC12LA αUC12LB αUC12DA αUC12DB

BATCH 2 (β): Time from pressure cooker expulsion to fortification: 8 min Time spent vigorously whisking in the fortificant/premix: 1 min Soymilk temperatures at the time of treatment (fortified or unfortified premix addition) were as follows: Fortified Treatment Portion: 81.3oC Unfortified Treatment Portion: 78.6oC

Sample Names Time Fortified Unfortified Ice bath Time (samples taken sample Soymilk Soymilk Temperature sample in triplicate) bottle put Temperature Temperature (oC) removed in ice bath (oC) (oC) from ice (min:sec) bath (min:s) βF1/ βU1 (a,b,c) 1:30 ? ? 4.0 40:00 βF2/βU2 (a,b,c) 9:30 ? ? 3.1 40:00 βF3/βU3 (a,b,c) 16:00 ? ? 2.4 51:00

The following times are measured from the end of the 1 minute step of mixing in the fortificant/premixes, and temperatures were taken inserting a thermocouple probe into a sacrificial sample bottle (i.e. not one of the three replicates used for vitamin analysis):

Time Ice Bath Ambient Bath Temperature of Temperature of (min:sec) Temperature Temperature representative representative (oC) (oC) bottle in Ice bottle in Ice Bath (oC) Bath (oC) 9:30 Bottles placed in ice bath 11:00 2.8 26.8 23.1 35.2 15:00 3.2 26.8 9.8 29.0 20:00 4.9 27.0 8.3 28.2 25:00 6.5 27.2 8.0 28.0 30:00 6.8 27.3 7.6 27.9 35:00 6.7 27.5 6.9 27.8 40:00 6.6 27.6 6.8 27.9 40:00 Bottles removed and refrigerated

“Light Exposure” Treatment

47 Description: Ten samples from the fortified treatment and ten samples from the unfortified treatment were taken in 16oz PET clear plastic bottles and cooled in the ice-water bath. All samples were put in the large ice bath approximately 16 minutes after the fortification step was complete (vigorous whisking for 1 minute) and removed to refrigeration 35 minutes later.

Primary treatments are shown below. Triplicate samples from all these were identified with a sub-sample number of 1,2, or 3, and frozen in 2oz glass bottles at -80oC until analysis.

Sampling Fortified Treatment Unfortified Treatment Time (days) Baseline βFC0L/DA βFC0L/DB βUC0L/DA βUC0L/DB Light Dark Light Dark 5 βFC5LA βFC5LB βFC5DA βFC5DB βUC5LA βUC5LB βUC5DA βUC5DB 12 βFC12LA βFC12LB βFC12DA βFC12DB βUC12LA βUC12LB βUC12DA βUC12DB

Sensory Testing Soymilk from each batch not assigned to a hot hold, cooling method, or light exposure treatment was poured into 16oz PET clear plastic bottles and cooled in the ice-water bath along with the cooling method comparison samples. Later all this soymilk was used for a sensory test by combining all of the fortified soymilk and unfortified soymilk together the palatability of the fortified soymilk versus unfortified soymilk was compared at BYU’s Sensory Analysis Laboratory.

Children ages 7 through 17 were recruited (with parental consent). Potential panelists were screened for their willingness to try soymilk and those who were willing to do so participated in a sensory panel on August 8, 2014. Panelists were rewarded with their choice from a selection of candy treats and their parents were remunerated at the price of $15 per child or youth under the age of 18 who participated in the panel. The demographic data gathered from the panelists consisted of age, gender, and whether they liked soymilk, disliked soymilk, or were not sure if

48 they liked or disliked soymilk. The ballot included 6 questions rated using a 7-point hedonic scale assessing overall impression, flavor, color, smell, mouthfeel, and aftertaste. Panelists were then asked to rank the samples in order of preference by selecting which of the two samples they liked best, and a final question asked how likely they would be to completely drink a full cup of each sample (fortified and unfortified soymilk) in a school cafeteria setting. Undergraduate BYU Sensory Lab workers were available to clarify the ballot questions to the participants.

Half an hour before the taste panel began the soymilk was prepared by pouring excess fortified soymilk from both batches (αFC & βFC) out of the original 16oz PET clear plastic bottles into 2 large pitchers in a walk-in refrigerator, and the same was done for the excess unfortified soymilk from both batches (αUC & βUC). Throughout the panel, the soymilk was stored on ice in a pitcher and so as not to give time for potential sedimentation to occur about 2-3 ounces of fortified and unfortified soymilk were poured into 4 oz clear plastic cups immediately before presenting to the panelists. The sample cups had been previously labeled with a three-digit numerical code: Fortified = 375, Unfortified = 861. The two samples given to each panelist were presented in random order from left to right and BYU Sensory Lab workers ensured that the amount of soymilk in each cup was visually similar. In order to avoid sample drift by potential sedimentation in the fortified samples, both the fortified and unfortified soymilk in the dispensing pitchers was mixed with a whisk immediately prior to pouring each set of samples.

Intermittently throughout the duration of the panel the dispensing pitchers had to be replenished from the reserve soymilk in the walk-in refrigerator. Always both the fortified and unfortified pitchers were refilled at the same time and the process was identical as follows: The remaining soymilk was dumped back into the reserve soymilk in the walk-in refrigerator and the reserve soymilk was mixed by pouring from pitcher to pitcher among 3 pitchers until any sediment was visibly reconstituted and a homogeneous sample was likely, whereupon the dispensing pitcher was again refilled and removed from the walk-in refrigerator to the panel servicing area where it was stored on ice and used as previously discussed.

Data Analysis Panelists ranged in age from 7 to 17, with all ages represented except age 16 years. A total of 57 panelists participated, consisting of 27 girls (47.4%) and 30 boys (52.6%). Only 23 panelists (40.4%) reported having had soymilk before. Most panelists (46, 80.7%) were not sure if they liked soymilk or not, and only 2 panelists (3.5%) claimed to dislike soymilk before tasting the samples. Data from the two panelists who disliked soymilk was included in the final analysis.

Data analysis of hedonic scale questions was carried out using one way ANOVA with post-hoc Tukey’s HSD (α = 0.05). The question asking panelists to rank the samples in order of preference was analyzed using Friedman analysis of rank.) Color was the only soymilk attribute evaluated for which panelists indicated a clear statistically significant preference (p = 0.007), which was in favor of the whiter unfortified soymilk. The fortified soymilk had a distinctly yellow hue, making the unfortified soymilk look lighter in the side by side panel comparison. The only other characteristic even remotely close to statistically differentiated between the two samples was mouthfeel, with a possible trend toward the fortified sample made slightly thicker by the corn starch base of the fortification premix (p = 0.185).

49 Appendix B: DSM Premix Information and Composition

The same vitamin and mineral premix from DSM Nutritional Products was used for every aspect of this thesis. This premix was the official fortification regimen for the Programa de Alimentacion Escolar (PAE) – a government-sponsored initiative for the introduction of nutritionally enhanced products into school lunch programs across Ecuador. The following figures show the premix composition and chemical nature of the ingredients:

This product description, received through email from DSM Colombia via Carlos Teran, was the source of some confusion at first, as the portion size should have been 350 ml instead of 35 ml. Therefore, the bottom five numbers in the right-hand column should be, from top to bottom: 590.46, 350.00, 1.68703, 0.1687, and 590. Also included in this communication was a tentative price in US dollars of $13.3522./kg.

This letter of certification and ingredient disclosure was shipped with the premix.

*Vitamin E 50% was interpreted to be a cold-water-dispersible d,l-alpha tocopherol acetate. **Vitamin B12 was interpreted to be a cyanocobalamin triturate.

51 Appendix C: Detailed Analysis Methods

Vitamin A Analysis

All Vitamin A samples were analyzed using AOAC Official Method 2012.10, Simultaneous determination of 13-cis and all-trans Vitamin A Palmitate (Retinyl Palmitate), Vitamin A Acetate (Retinyl Acetate), and total Vitamin E (dl-α-Tocopherol and dl-α-Tocopherol Acetate) in Infant Formula and Adult Nutritionals. The only variation to the AOAC method was the following flow rate alteration which provided better resolution of the retinyl palmitate isomer peaks:

AOAC Method Flow Rate: 1.5 ml/min Flow Rate Used for this Thesis: 1 ml/min

Soy Milk Vitamin A Extraction Method (from AOAC 2012.10)

Apparatus: HPLC Water Bath Auto sampler vial cap crimper Centrifuge capable of 50ml centrifuge tubes

Materials: 50 ml plastic centrifuge tube 0.45 µm PTFE or sodium acetate filter 5 ml syringe Amber auto sampler vial Crimp-top cap for amber auto sampler vial A 50 or 80 ml beaker, and a 250 ml beaker Glass stir rod

Reagents: • 2% Papain Solution (prepares enough for 9 samples): o Add 0.0500g of hydroquinone and 2.0 g sodium acetate anhydrous to a 50 or 80 ml beaker, then add about 40 ml distilled water. Adjust pH to 5.0 with dilute HCl, and then add 1.0g of papain (from carica papaya). Use a stir rod to mix until all the papain is dispersed in the solution. Quantitatively transfer the solution to a 50 ml volumetric flask and fill to volume with distilled water. Prepare fresh on day of use. • Acidified Methanol Solution (prepares enough for 9 samples): o Add 5 ml of glacial acetic acid to 200 ml of methanol in a 250 ml beaker. Mix thoroughly with a stir rod. Prepare fresh on day of use.

52 Extraction Method: (Lights off or filtered, yellow lighting): 1. Pipet 5 ml of thoroughly agitated, refrigerated sample into 50 ml centrifuge tube and record weight; nitrogen flush sample and return to refrigerator. 2. Add 5 ml of 2% papain solution. 3. Cap and swirl vigorously or shake lightly to mix, and place the tubes in a 37oC water bath for 20-25 min. 4. Cool in ice-water bath for approximately 5 minutes 5. Add 20 ml acidified methanol to each tube and swirl vigorously or cap tightly and shake lightly to mix. 6. Accurately pipet 10.0 ml of iso-octane into each sample. Cap tightly to avoid leakage and shake vigorously for 10 minutes. This is the volumetric step in the extraction. 7. Centrifuge for 10 minutes (centrifuge in Vitamin analysis lab is to be used at the full speed setting ~3500-4000 RPM). 8. Without disturbing the lower aqueous portion, draw iso-octane supernatent into a 3 or 5 ml syringe. Add 0.45 filter to syringe; discard the first few drops, and then slowly filter into an amber HPLC vial. Fill vial as much as possible to exclude O2 in headspace. Cap vial. 9. Refrigerate until same-day analysis using HPLC.

HPLC Parameters: Column: Zorbax NH2 4.6x150 mm, 5 μm Injection Volume: 20 μl Mobile Phase A: Hexane Mobile Phase B: Hexane/methyl-t-butyl ether (75/25 v/v), add 3 ml methanol/L Elution Gradient: As per AOAC method 2012.10 Flow Rate: 1 ml/min Duration: 20 min Detection: Diode Array, Signal λ = 325,10 nm, Reference λ = 400, 100 nm Vitamin A Peaks: Cis-13 peak fully resolved at ~4.5 min, Cis-11 peak fully resolved at ~4.9 min, All-trans peak fully resolved at ~5.4 min (consecutive samples may progressively shorter retention times with each successive injection)

References: 1. Simultaneous Determination of 13-cis and all-trans Vitamin A Palmitate (Retinyl Palmitate), Vitamin A Acetate (Retinyl Acetate), and Total Vitamin E (α-Tocopherol and DL- α-Tocopherol Acetate) in Infant Formula and Adult Nutritionals by Normal Phase HPLC: First Action 2012.10. McMahon et al., Journal of AOAC International, Vol. 96, No. 5, 2013.

Comments: While this method seemed much better for soy milk than the saponification method used previously in the Vitamin Analysis lab by Amalie Kurzer. Retinyl palmitate standards seemed to be well-recovered, but throughout the course of the study it became apparent that matrix effects were impeding recovery of the Vitamin A from the DSM Premix. The data is as follows: Soymilk samples held in refrigeration for 0, 5, and 12 days had higher and higher Vitamin A values the longer they were held in cold storage. My explanatory

53 hypothesis for this phenomenon is as follows: The Vitamin A in the DSM Premix was likely a “cold water soluble” ingredient which had been spray dried with a slurry of starches and gums that would envelop the lipophilic retinyl palmitate making it hydrophilic and dispersible in water. Slow hydration and disintegration over time of the gum “shell” around the retinyl palmitate would allow for better access to the Vitamin A by the iso-octane extracting solvent and explain why more and more Vitamin A was recovered over time in refrigerated storage. Attempts to alter the method were unsuccessful at improving Vitamin A recovery. More acidic methanol did not help. Longer incubation time in the water bath (up to 1.5 hours) did not improve Vitamin A extraction, and incubation for an hour at higher temperature (60oC) was also unsuccessful. Doubled papain concentrations did not improve Vitamin A extraction and furthermore they interfered with the phase separation by creating a thicker mass of proteinaceous emulsified matter between the aqueous layer and the iso-octane in the centrifuge tubes (there was usually a small film between phases in this method that was light in color and quite compact – less than 1/8 inch thick). Unfortunately, I did not measure the increase in microbial count, if any, over the course of the refrigerated storage, as this could have shed light on whether or not microbial digestion was freeing up more of the bound Vitamin A.

54 Vitamin C Analysis

Version Date: March 12, 2015

Soy Milk Vitamin C Extraction Method (Modified from MLD-007-102)

Apparatus: HPLC- warmed up Sonicator Auto sampler vial cap crimper Stir plate Centrifuge

Materials (for one sample): Appropriate volumetric flask for solvent preparation* 25 ml volumetric flask Small magnetic stir bars (approx. 12 mm long) (6) 50 ml plastic centrifuge tube 0.45 µm PTFE or sodium acetate filter 3 ml syringe Amber auto sampler vial Crimp-top cap for amber auto sampler vial Glass or PTFE stopper (for 50 ml volumetric flask)

Reagents: • 0.5% MPA Solution – Metaphosphoric acid (0.5%), Dithiothreitol (0.2%): o Accurately weigh 0.025g of metaphosphoric acid and 0.100g DTT into a 50ml volumetric flask. Dissolve completely. Fill to volume with distilled water. Prepare fresh daily. • TCA/DTT/MPA Solution – Trichloroacetic acid (2%) / Dithiothreitol (0.2%) / Metaphosphoric acid (0.05%): Samples Volumetric Trichloroacetic Dithiotreitol 0.5% MPA Solution flask* acid (g) (g) (ml) 9-10 250 ml 5 0.5 25 7-8 200 ml 4 0.4 20 3-4 100 ml 2 0.2 10 1-2 50 ml 1 0.1 5 o Half fill volumetric flask with water; add dithiothreitol (DTT); add trichloroacetic acid (TCA); add metaphosphoric acid (MPA) solution with a graduated cylinder; swirl until dissolved – sonicate if necessary; fill to volume with distilled water. Prepare fresh daily.

Extraction Method: (Lights off or filtered, yellow lighting): 1. Pipet 1 ml of sample into 25 ml volumetric flask and record weight; bring to volume with TCA/DTT/MPA solution.

55 2. Add six stir bars; insert stopper; mix on stir plate at highest speed for 10 minutes. Ensure flasks do not migrate while stirring. 3. Pour solution into a 50 ml plastic centrifuge tube (rotate flask while pouring so that the maximum amount of solution and sample will be poured into the centrifuge tube). 4. Centrifuge for 5 minutes at 1600 x G (with centrifuge in Dr. Pikes lab is to be used at the full speed setting ~4000 RPM). 5. Fill 3 ml syringe with centrifuged solution. Add 0.45 filter to syringe; discard the first few drops then slowly filter into an amber HPLC vial. Fill vial as much as possible to exclude O2 in headspace. Cap vial. 6. Analyze immediately using HPLC. Refrigerate if analysis is delayed, but samples must be injected on the HPLC column within 4 hours of extraction to stay within 95% recovery compared to a freshly run sample.

HPLC Parameters: Column: C-18 Synergi hydro-pro 4.6x250 mm, 5 μm Injection Volume: 10 μl Mobile Phase: Sodium Acetate 0.5 M (pH 4) Flow Rate: 1 ml/min Duration: 12 min Detection: Diode Array, Signal λ = 254 nm, Reference λ = 360, 100 nm Vitamin C Peak: Fully resolved at ~2.5 min

References: 1. Liquid Chromatographic Determination of L-Ascorbic 2-Polyphosphate in Fish Feeds by Enzymatic Release of L-Ascorbate. X. Y. Wang, M. L. Liao, T. H. Hung, and P. A. Seib. J. Assoc. Off. Anal. Chem. Vol. 71, No. 6: 1158-1161 2. Some effects of replacement of metaphosphoric acid/acetic acid solvent system with trichloroacetic acid in microfluorometric determination of vitamin C. F. R. Visser. J of AOAC. Vol. 67, No. 5: 1020-1022. 3. Method modification for Liquid Chromatographic Determination of Thiamine, Riboflavin and Pyridoxine in Medical Foods. Chase, et al. Journal of AOAC International. 1993. Vol 76 No. 6, pp 1276-1280).

Comments: I replaced MPA with TCA in this method because at the concentrations required for protein precipitation the MPA sample extract gave a huge peak at about 2 min that overwhelmed the Vitamin C peak (at 2.2 min). I believed the peak to be from a MPA-metal complex with spectral properties at 254 nm. Metaphosphoric acid is a known metal chelator, and the soymilk samples were heavily fortified with minerals. A similar problem occurred with the use of EDTA (i.e. a large peak appears that cannot be resolved from the Vitamin C peak). Impure TCA will also cause interfering peaks. Reagent or analytical grade TCA at least 99% pure should be used in this method. The reason there is still a small amount of metaphosphoric acid in the sample, even though it introduces a small leading bump, is because it seemed to improve stability in the HPLC vial and lengthen the time post-analysis that samples could be injected on the column to beyond 4 hours. Vitamin C stability in the HPLC injection vial was investigated with the final method in

56 an Illinois soymilk sample extract. Repeat injections of the same sample every hour for 4 hrs returned peak areas well within 5% of each other with 2% TCA and VERY little MPA (0.005%), but repeatability began to drop off after 3 hrs in the TCA only sample after repeated re- injections, presumably due to Vitamin C degradation. The centrifuged pellet is very soft, and a very fast centrifuge would be ideal. I tried higher concentrations of TCA, but while the precipitation situation was about the same as at the lower concentrations, the Vitamin C stability decreased in the HPLC vial. After multiple injections of the Ecuador samples, a leading peak or shoulder appeared and was difficult to get rid of. This issue was resolved by washing the column with acetonitrile for 10 min after analyzing each set of samples (up to 15 at a time, though the leading shoulder was best kept at bay with more frequent column washes). Still, some of the Ecuador data had a low bulge in the baseline starting ahead of the vitamin C peak and continuing through it – but it wasn’t sharp and tall like the MPA peak, it was low and broad. Additional ingredients in the Ecuador samples over the Illinois samples include carrageenan, crystalline vanilla, and liquid coconut flavoring. I don’t know what caused the problem on the column. The soybean source was also obviously different.

Vitamin C method development reducing metaphosphoric acid (MPA) in favor of trichloroacetic acid (TCA)

Injection #1 Injection #2 Injection #3 Injection #4 Trial TCA MPA EDTA Peak Peak Time % 1st Peak Time % 1st Peak Time % 1st Date # (%) (%) (%) Area Area (hrs) peak Area (hrs) Peak Area (hrs) Peak Notes Inherited lab method. Huge leading 2/27/2015 1 - ? - peak from MPA. Decided MPA was a

problem after repeated similar results. First run w/ TCA (no MPA). Huge leading solvent peak was gone, but 3/6/2015 2 ? - - adding a tiny bit of MPA increased recovery…possibly a balancing act? 3/10/2015 3 2 - - 153.95 153.73 1.25 99.9% 152.22 2.5 98.9%

At 2.5 min the little tail on the Vit C 3/10/2015 4 4 - - 152.43 148.2 1.25 97.2% 143.7 2.5 94.3% peak bulges and begins forming a a low,

unresolved peak. Problem! Terrible recovery. Leading peak not as 3/10/2015 5 8 - - 126.00 111.13 1.25 88.2% 97.028 2.5 77.0% well resolved and an unresolved lagging

peak got worse over time. MPA caused a massive leading peak 3/10/2015 6 8 0.5 - 174.93 161.33 1.25 92.2% 152.02 2.5 86.9% totally unresolved with the Vit C peak,

distorting peak areas. MPA caused an enormous leading peak which interfered with the Vit C peak, 3/10/2015 7 8 2 - 201.023 179.25 1.25 89.2% 169.05 2.5 84.1% distorting peak areas.This further confirmed that MPA was a problem. 2% TCA was the best combination of 3/11/2015 8 2 - - 212.97 211.53 1.5 99.3% 208.68 3.5 98.0% precipitation during extraction and

stability in HPLC vial. Stability was good until 3 hrs and then 3/11/2015 9 2 - - 146.01 146.54 1.5 100.4% 141.6 3.5 97.0% 140.6 4 96.3% it began to drop off. Stability and recovery were great up to 3/11/2015 10 2 0.005 - 157.22 155.17 1.5 98.7% 157.7 3.5 100.3% 154.5 4 98.3% 4 hours! The method was modified to replicate this result in study samples.

58 Thiamine & Riboflavin Analysis (Combined Method)

Version Date: May 12, 2015 Method for Thiamine and Riboflavin Determination in Soy Milk

PROCEDURE All steps performed under subdued and/or yellow light DAY 1 Materials: 1. 125ml Erlenmeyer Flask (1 per sample) 2. 50 or 100 ml beakers 3. Glass stir rods 4. 0.1 N HCl 5. 2.5 M Sodium acetate 6. Takadiastase Method: 1. Keep thawed soy milk sample bottle in the refrigerator in a lightless environment (wrap completely in tinfoil – refrigerator light is not yellow like the lab lights & it degrades vitamins). 2. Shake the sample for 15 seconds or until visually homogeneous right before pipetting the aliquots for analysis (especially ensure that any sediment that can be seen on the bottom of the sample bottle is re-suspended in the soy milk solution). Shake again before pipetting if the sample has been sitting for more than a minute or two. 3. Tare the 125ml Erlenmeyer flask. Pipet 5 ml of unfortified or 1 ml of fortified homogeneous soy milk sample into 150 ml Erlenmeyer flask. Record sample weight. 4. Add 40ml of 0.1 N HCl (washing any sample off the sides of the flask). Swirl for 1 minute. 5. Adjust pH to 4.5 +/– 0.05 with 2.5 M Sodium Acetate. This pH is crucial to stabilize the vitamins and prevent further loss. To ensure an accurate reading, mix solution gently with pH probe to ensure homogeneity and make sure base drips directly into the sample and not on the pH probe or on the sides of the Erlenmeyer flask. Wear gloves and goggles. 6. Add 500 mg Fluka Takadiastase (phosphatase to free thiamine & Riboflavin bound to phosphate), swirl, wash down sides with distilled water if any enzyme or soy milk is not in solution. 7. Cover in foil and incubate @ 37°C for 18 hours.

59 DAY 2 Materials: 1. Whatman #541 filter paper 2. Glass funnels (long or short stem) 3. 100 ml volumetric flasks 4. 20 ml syringe 5. 50 ml screw cap centrifuge tubes. 6. 3 ml syringes 7. 0.2 µm filters 8. NaCl 9. 1 % Ferricyanide solution 10. 15% NaOH 11. Isobutanol (Isobutyl alcohol) Method: 8. After incubation, filter the solution using Whatman #541 filter paper, and a funnel. Collect the filtrate in a 100 ml volumetric flask. Using a small amount of distilled water (~5-10 ml), triple rinse the Erlenmeyer flask, filtering each time. Using distilled water, triple rinse the filter paper, making sure to agitate the unfiltered soy milk residue each time. Use a spray bottle that can provide good agitation of soy milk residue without using large amounts of liquid and over-filling the volumetric flask. 9. Bring the flask to volume. Cap with parafilm and mix by inversion to homogeneity. 10. Filter approximately 1.5 ml of filtrate into an amber HPLC vial for riboflavin determination using a 2 ml syringe & a 0.2 μm membrane. Discard the first few drops. 11. While filtering, add 2.50 g NaCl to large centrifuge tube (50 ml) and prepare the oxidizing reagent. 12. Preparing the oxidizing reagent: wear nitrile gloves throughout all oxidizing steps - Prepare (fresh, on day of use) 3% Potassium Ferricyanide. - Prepare (does not need to be prepared fresh, on the day of use) 15% NaOH. - Finally, prepare the oxidizing reagent (fresh on day of use). Use within 4 hours. 13. Add 10 ml filtered solution to the centrifuge tube. 14. Gently swirl (don’t shake) each tube until most of the salt is dissolved. (It usually takes about 1.5-2 minutes for the salt to dissolve) 15. While gently swirling, add 3 ml oxidizing reagent. (Add reagent all at once using pipette set for 3 ml, making sure that the stream of oxidizing solution does not hit the sides of the tube). 16. Prepare isobutanol for pipetting. Gently swirl for about 5 seconds. Immediately add 15 ml isobutanol with 5 ml pipette. 17. Cap and shake tube vigorously for 15-20 seconds.

60 18. Move on to the next tube. Prepare the samples this way in sets of no more than 4. (If doing more than 4 samples, get 4 of them to the centrifuge step before taking the other samples through the oxidizing step). 19. After isobutanol has been added to the tubes, shake them all for 2 minutes. 20. Centrifuge tubes at full speed for 4 minutes, or until clear supernatant can be obtained from each. 21. Fill an amber HPLC vial with supernatant for thiamine determination, using a Pasteur pipet. (If the supernatant leaves residue on the Pasteur pipet or shows other signs of impurities, filter into the HPLC vial through a 0.2 µm filter, discarding the first few drops.) 22. Store refrigerated (put in chromatography lab refrigerator) for same-day HPLC analysis.

REAGENTS 0.1 N HCl: In a 1000 ml volumetric flask add 8.2 ml of concentrated HCl. Bring to volume with distilled water. Can also be done by substituting 16.4 ml 1:1 diluted HCl for the concentrated HCl 2.5 M Sodium Acetate: In a 200 ml volumetric flask dissolve 68.04 g Sodium Acetate Trihydrate. Bring to volume with distilled water. 15% NaOH: Add 15 g NaOH to a 100 ml volumetric flask. Fill to volume with distilled water. 3% Potassium Ferricyanide:

In a 10 ml volumetric flask dissolve 0.3 g K3Fe(CN)6. Bring to volume with distilled water. Alkaline Potassium Ferricyanide (Oxidizing Reagent): Add 1 ml of 3% Potassium Ferricyanide to a 25ml volumetric flask. Fill to volume with 15% NaOH.

REFERENCES: Arella F, Lahely S, Bourguignon JB, Hasselmann C (1996) Liquid chromatographic determination of vitamins B-1 and B-2 in foods. A collaborative study. Food Chemistry 56 (1):81-86.

AOAC 2006 Method 953.17.

El-Arab AE, Ali M, Hussein L (2004) Vitamin B1 profile of the Egyptian core foods and adequacy of intake. J Food Compos Anal 17 (1):81-97.

61 COMMENTS For most matrices this method starts off with an autoclave step, but autoclaving the soymilk samples seemed to cause no improvement in recovery perhaps because in its production the soymilk is already thoroughly pressure-cooked. The cause for between-run variation in riboflavin recovery is difficult to diagnose. Careful rinsing, filtering, and homogenization technique, along with accurate volumetric dilutions and carefully controlled exposure to light will help with riboflavin repeatability. By far the most critical and sensitive step in this method for Thiamine is the oxidation step. Critical variables include the concentration of K3Fe(CN)6, swirling time prior to isobutanol addition, and the timing & strength of agitation. Most importantly, the addition of isobutanol must be accurate and repeatable between samples because this is a volumetric step in the method and the low surface tension of isobutanol makes accurate pipetting difficult. Finally, how long the samples sit before being centrifuged and if they are re-shaken immediately prior to centrifugation may impact the final Thiamine value. The object of this step is to repeatedly obtain a complete reaction from Thiamine to thiochrome and quench the reaction to avoid over-oxidation. Each analyst should be trained to perform this step from reagent preparation to HPLC vial in exactly the same manner, so that the technique is consistent from the standard curve solution to every single sample.

62 Total Folate Analysis

Testing Total Folate in Cereal Products - Revision 14 Compiled by Jordan Chapman with much help from Kirk Oler, Nathan Tidwell, and Dr. Eitenmiller Revised 12-8-10 By Shintaro Pang Revised 6-6-11 by Jonathan Kershaw and Eric Engstrom Revised Dec 16, 2015 by Dallin Hardy

Notes about the entire analysis The entire analysis needs to be done under subdued and/or yellow light. Ideally, all sterile filtering, microbial transfers, and definitely plating should be done in a bio-hood (thoroughly cleaned with 70% ethanol or quaternary ammonia), but a benchtop or still-air hood sprayed down generously with 70% ethanol can be a satisfactory substitute if great care is taken to avoid contamination (i.e. long sleeves, gloves sprayed down with 70% ethanol, no breathing on the workspace, spraying down everything liberally with 70% ethanol when it is brought into the workspace, no turbulent air flow, etc.). All non-disposable glassware must be burned for 1 hour at 250ºC (475oF) prior to use. Use a clean scoop for measuring out all samples and the folic acid standard so that no contamination occurs. If there is an * after the material, it needs to be autoclaved prior to use. Gloves should be worn throughout the entire procedure. If sterile transfers and techniques aren’t used throughout this analysis, many weeks of work will be wasted with contamination so do a good job! Sterile Techniques Sterile transfer techniques include wearing a clean lab coat and new gloves. Liberally spray down gloves, workspace, and the transfer loop (handle included!) with 70% ethanol. Work in a dust-free area and avoid turbulent airflow as much as possible (exception is a certified sterile, functional bio-hood), including heavy or direct breathing on your work. Doing transfers on a benchtop & plating in a hood has worked well in the past. If a re-usable transfer loop is used, flame the metal portion until it is red hot, then dip it in 90% ethanol until cool (it will probably boil the ethanol until it cools down); then re-flame it. Flame alone will sterilize the metal if it gets hot enough, but a hot loop will kill the bacteria it is transferring. Since ethanol burns at a very low temperature, the second flaming can allow the loop to be both cool and sterile. The loop should be swirled in the media sufficiently to get a representative transfer (the bacteria often settle). Open vials and containers should be tilted away from the operator and nothing non-sterile should be passed over them so that no contaminants drop into the working solutions. Keep sterile and inoculated containers capped as much as possible, and replace the cap quickly and carefully after transfers. After using the transfer loop, flame it again until red-hot to burn off any residue of media and bacteria so the loop will be cleaner for the next time. A sterile pour is achieved (eg. Autoclaved water or media) by passing a flame over the rim of the autoclave bottle (if it is glass) after carefully removing the lid, and then carefully pouring. The rim can then be flamed a second time if desired before replacing the cap. It is critical to avoid having a liquid connection (a microbial ‘highway’) from the inside of the container to the outside, so keep the threads clean and dry. Autoclaving water or media: Put the bottle in the autoclave with the cap slightly loose to allow pressure release as it boils, and then tighten the cap as soon as it is taken from the autoclave to

63 avoid contamination. Always mark containers with fresh autoclave tape and leave the darkened tape on as long as the contents remain sterile. Sterilize water in the 1L bottles for 20 min, and in the 2L bottles for 30 min (wet goods setting). Sterile pipet tips should be prepared by autoclaving on the dry goods setting for 15 min at 120oC. The tip box should be labeled with autoclave tape, and the darkened tape should be left on the box only as long as the tips inside are sterile. The tips in the box may be considered sterile after the box is opened only if it is closed immediately after each use and the environment is not likely to be contaminating. However, it is a simple thing to refill the tip box and re- autoclave the pipet tips, so this is usually done for the pipet tips used in the plating step of this method. Sterile pipets are crucial for any pipetting done in the sterile portions of method. The literature has ample evidence for pipets being contaminated – including multi-channel pipets. Jordan Chapman was adamant that background noise in the plates (i.e. high blank rows) was often evidence of contaminating microbes delivered via the plunger systems of the pipets used. Preparation for the analysis – Inoculum Materials: • Lactobacilli broth AOAC (Difco) • .22 um Sterilizing 250 ml filter system (VWR 87006-062) • Sterile needle (18-20 gauge (wide opening to minimize shear for the microbes), 1” works well) • Sterile syringe (1cc) • Lyophilized L. casei (ATCC 7469) • Appropriate size beaker • 1L Autoclave Bottle • Sterile 2 ml cryo vials

Instructions & Discussion Preparation of Lactobacillus broth 1. Prepare Lactobacillus broth by dissolving 3.8 g in 100 ml water in a beaker or Erlenmeyer flask. Boil for 2-3 min, then sterile filter using the .22 µm sterile filter system. Alternatively, the media can be autoclaved for 15 min at 120oC and then sterile transferred from the autoclave bottle directly into the sterile holding container without passing through the filter (i.e. using a sterile pour, put ~100 ml of media into the sterile container. Cap both containers as quickly as possible and refrigerate the autoclave bottle if extra media was prepared for later use). a. The autoclaved media is darker than the sterile filtered media (maillard browning?), and when it is autoclaved for longer than 15 min it gets extremely dark and may lose significant nutritive value. For this reason, I recommend preparing no more than 1L at a time in a smaller autoclave bottle and only autoclave it for 15min. I prepared 500 ml quantities in this way and my culture did not seem to be contaminated from this source because I repeatedly got very good data with cryo-protected cultures that had been grown in the autoclaved media. Starting a new L. casei culture 2. If preparing the inoculum from lyophilized ATCC stock cultures: Using sterile techniques and the sterile needle and syringe, take 1 ml of media made in step 1 and inject it through the membrane into the L. casei vial. Mix to homogeneity by gently drawing up the re- suspended microbe solution into the syringe and expelling it repeatedly (eg. 5 times). Sterile transfer ~0.15 ml of the re-suspended solution into the sterile media made in step 1. Incubate at

64 37oC for 24-48 hours. If growth occurs (as evidenced by obvious cloudiness) this solution can now be used or refrigerated for up to a week. However, immediately growing up a second generation is ideal because the freeze-dried organisms may not all thrive in a growth media setting at first (see #3). 3. To grow up a second generation of microbes, transfer using a sterile loop from the previous bacteria solution, as described in step 4. An incubation time of 24 hours is sufficient. Transferring the bacteria 4. If transferring bacteria from a refrigerated or recently incubated lab culture, use sterile transfer techniques to move one loop of media from the previous culture to the new media made in step #1. Incubate for 24-48 hours at 37oC. Preferably, keep the incubation time close to 24 hours, so that the bacterial population does not have time to seriously deplete their resources or enter a lag or death phase. A second generation is not necessary if the bacteria are already used to growing in media. a. To keep an active working culture, re-transfer from the refrigerated culture using a sterile loop into new lactobacilli broth every 7 days to keep the inoculum fresh – incubating at 37oC for 24 hours and then refrigerating the culture for up to a week. (The microbes can actually live for longer than two weeks in the refrigerator, and after more than 1 week one can usually still transfer enough viable microbes to re-start the culture; but this is not ideal as the culture will not be as healthy.) A culture that is not cryoprotected or continually transferred to new media will eventually die completely – probably within 3 weeks. Cryoprotecting the bacteria 5. As an alternative to transferring bacteria weekly, the bacteria can be cryo-protected and stored in the negative 80oC freezer. After consulting with Dr. Steele and fellow grad student James Fudge, I achieved great results on soy milk folate analyses using bacteria cryo-protected in the following way: a. First the ATCC stock culture was re-suspended in ~100 ml of media and incubated at 37oC for about 42 hours hours as described in step 2. I used the autoclave method (15 min @ 120oC) to prepare 400 ml of Lactobacillus Broth, and I used the base container from a sterile filter system to hold the inoculum. All pours and transfers were sterile. b. I immediately grew a second generation of L. Casei in ~100 ml of the same media, by pipetting 10 ul of generation 1 L. casei into a second sterile container (a loop could also be used). The media for generation 2 was sterile poured from the same refrigerated autoclave bottle that was used for the first generation. This was incubated for about 28 hours at 37oC. c. Cryoprotective media contains glycerol to minimize the negative effects of freezing on the bacteria. To prepare cryoprotective media with 20% glycerol content, I first autoclaved Lactobacillus broth prepared as in step #1 but containing an additional 40% glycerol by volume and cooled it in a refrigerator. Then, I swirled my 2nd generation L. casei inoculum to homogeneity and sterile pipetted 1 ml into fresh cryo vials. Finally, I sterile pipetted 1 ml of the 40% glycerol media into the same cryo vials, resulting in a 20% glycerol inoculum. All these cryo vials were put into the -80oC freezer. Since I had more media, I grew a third generation of L. casei and cryo-protected some of it as well as a backup. d. Protocol for use of the cryoprotected bacteria (based on recommendations by Dr. Steele): Thaw inoculum immediately before use and refreeze it immediately afterward, using the same cryo vial for one month of analyses before discarding it (in the micro lab bio-trash) and using from the next vial. Always use sterile transfers to get the bacteria from the cryo vial to the

65 depletion tube. If contamination is suspected, the vial should be discarded and a new one used. In this way, each analysis is only two generations of bacteria from the purchased cell line. e. Theoretically, the BYU analysis lab should never need to buy the ATCC lyophilized L. casei again. As soon as there are insufficient cryo vials for the next project, a new stock culture can be inoculated from one of the cryoprotected vials and after one or two generations of growth, a new set of inoculum cryo vials can be frozen using the procedure described above.

*In all the above steps, remember to plan time for preparing the Lactobacillus broth before the transfer step. If the media is prepared by the autoclave method it will need more than an hour of refrigeration time to cool.

Discussion of bacterial cryoprotection Jordan Chapman advised against cryoprotection in previous versions of this method. He wrote: “Cryoprotection of the bacteria is certainly accepted in many articles in the literature. However, I have gotten better results using the weekly transferred bacteria. Eitenmiller’s lab never cryoprotects the bacteria.” That being said, there are some downsides to weekly bacteria transfer. First of all, the sterile filter system is a nuisance and it is expensive. If the boiled broth is not well dissolved or if the broth is too cool when it is put into the filtration system, the filter will sometimes clog before it all passes through (possibly due to the weak suction of our water- faucet vacuum systems). At this point one must either start again or transfer the remaining media to a new filter cap and filter the remaining media into the holding container without contaminating what has already been filtered. Sterilizing media using the autoclave method could circumvent the filtering problem, but weekly transfers also mean more opportunities for contamination and generational variation. The reality is that both cryoprotection and the live culture method can provide a good inoculum. Rat Plasma/Serum Materials: • Funnel • Filter paper of large pore size • 250 ml Erlenmeyer flask • 50 ml Erlenmeyer flask • .22 um acrodisc filter • Test tubes to fit centrifuge

Instructions 1. Add 5% acetic acid solution to flask containing activated charcoal until charcoal is just submersed. Cover flask and mix on mechanical shaker for 1 hour. Drain the charcoal and rinse with water. 2. Rat serum is mixed with one-tenth weight of acid treated charcoal in 50 ml Erlenmeyer flask. Mix flask on ice with a mechanical shaker for one hour. a. Thaw rat plasma/serum completely before use to get a representative sample. 3. Centrifuge at ~5500 rpm for sufficient time to retrieve the supernatant. 4. Filter the supernatant using the acrodisc filter. 5. Aliquots can be stored in cryo vials at -70°C or colder.

Sources:

66 Tamura T. 1998. Determination of food folate. Nutritional Biochemistry 9:285-93. Phillips KM, Wunderlich KM, Holden JM, Exler J, Gebhart SE, Haytowitz DB, Beecher GR, Doherty RF. 2005. Stability of 5-methyltetrahydrofolate in frozen fresh fruits and vegetables. Food chem. 92:587-95. Partridge SM. 1949. Displacement Chromatography on Synthetic Ion-exchange Resins. Biochem J. 44:521-7.

Discussion I didn’t have time to fully check this method of rat plasma preparation, but I think that some clarification may be in order. First of all, what we need from the rat plasma is deconjugase enzyme. Native folate binds to one or more glutamate molecules, and in order to make it bioavailable for the L. casei, deconjugase is used to trim that down to only one. There are other sources of deconjugase. This method has been run successfully using extract of chicken pancreas and porcine pancreas. I think the AOAC method uses chicken pancreas. However, the rat source is attractive because it avoids grinding up pancreases and extracting the enzyme. Serum or plasma? Serum is simply plasma with the clotting factors removed. This method was initially run using serum. Of course deconjugase is also present in the plasma. I believe that the clots in the plasma may clog the filters more readily in step 4 of this preparation, so all else being equal I recommend going back to rat serum. Filtration is done using a small disc filter of 0.22um pore size on the end of a syringe. The treated serum is pushed through the filter into a cryo vial and the charcoal, etc, is retained. The purpose of the activated charcoal step is not clear in any of the writings from previous BYU graduate students, and I have not taken the time to read the above sources. However, activated charcoal is very adsorbant for minerals, oxygen, and acetic acid, and its purpose may be to remove iron from the rat serum to keep it from oxidizing the folate throughout the analysis. I tried using 10% acetic acid and it seemed to work the same. Knowing that charcoal is highly oxygen sensitive, I tried to dry the acid-treated/rinsed charcoal with nitrogen or in a desiccator, but I found that this was not very feasible and gave it up. For step 2, the vitamin analysis lab mechanical shaker is usually set at 140. This is sufficient to swirl the contents of the flask somewhat, but I am not convinced that the charcoal gets good contact with all the serum. I tried using one-fifth weight of charcoal to get better contact time, and I got great data with it.

Sample analysis Day 1: Tri-enzyme treatment Materials: • Working standard solution (1ug/ml) • 125 ml Erlenmeyer flask (1 per sample + 1 for standard) • pH 7.8 phosphate buffer (30 ml per sample + 30 ml for standard + extra) • 1:1 HCl (1 part HCl to 1 part H20) • Graduated cylinder capable of 30 ml • Octanol • Aluminum foil • Protease solution (1 ml per sample + 1 for standard + extra) • α-amylase solution (1 ml per sample + 1 for standard + extra)

67 Preparing the Folic Acid Analytical Standard (1ng/ml) 1. Accurately pipette 1.0 ml working standard solution (1ug/ml) to 125 ml Erlenmeyer flask. Record weight. Add 20 ml pH 7.8 phosphate buffer, mix, and add 30 ml water. 2. Continue treatment starting with step 4 for samples.

Sample Preparation 1. Accurately weigh an appropriate amount of sample (about 1.0 g for samples containing 0.5-1.0 ug/g of folate) into a 125-erlenmeyer flask. Label the flasks clearly with permanent marker. a. I analyzed 1 ml of fortified soymilk, and 2 ml of unfortified soymilk. 2. Add 20 ml pH 7.8 phosphate buffer; mix thoroughly. 3. Add 30 ml water to each flask. Autoclave Step 4. Add 0.5 ml octanol (antifoam) to all flasks. This can be increased to 1ml for some sample matrices if the flasks seem to consistently boil over and sample is lost in the autoclave. 5. Cover flasks with aluminum foil and autoclave 15 min at 121°-123°C, then cool in an ice water bath (about 4 minutes or until reaching room temperature). a. Previous protocol was to cover the flasks with inverted 50ml beakers, but foil seems to work just as well. The foil should be labeled with the sample code, because overflowing octanol can sometimes erase the permanent marker from the glassware. If the flasks are well covered, ice can be carefully scooped into the autoclave tray to cool the flasks without having to transfer them to another water bath. Protease Step 6. Add an additional 10 ml pH 7.8 phosphate buffer to each flask. 7. Add 1 ml protease solution, wrap flasks in foil, and incubate 3 hours at 37°C. The idea is to avoid light-induced folate degradation. If the incubator is completely dark the flasks don’t need to be covered in foil, but flasks should still be capped and a little square of foil is perfect for this. Amylase Step 8. Inactivate protease in a boiling water bath for 3 min, then cool in an ice water bath 4 min. 9. Adjust samples and standard to pH 4.5, + or - 0.05 with HCl. About 13-16 drops of 1:1 HCl:H2O will be needed. Use diluted HCl to bring samples into correct range if needed. NaOH may be used if necessary. a. Use safety goggles when adjusting pH. Use correct technique (i.e. make sure pH probe does not touch the concentrated pH-altering solutions directly during addition, and ensure that the sample is swirled or mixed to homogeneity before trusting the reading). Some sample types (eg. flour) will clump, trapping fluid and keeping pH from equilibrating well. Break up any clumps before trusting the pH reading. 10. Add 1 ml α-amylase solution to each flask. Add 0.1 ml treated rat plasma/serum. Rewrap or cap and incubate 18 hours in the dark at 37°C.

Day two of the analysis - Plating Materials: • Depletion media in plastic-capped test tube or screw cap culture tube • 100 ml volumetric flasks (1 per sample + 1 for standard)* • Whatman 2V filter paper (1 for each sample + 1 for standard)

68 • Acrodisc .22 um filters (1 for each sample + 1 for standard + 1 for water + 1/plate for double strength basal medium) • Sterile screw top cryo vials (2 ml) (1 for each sample + 1 for standard) • 10 ml syringe (1 for each sample + 1 for standard) • 20 ml syringe (1 for water + 1 for double strength basal medium) • Double strength basal medium (15 ml per plate) • Sterile pipetting troughs • Sterile microtiter plates – Falcon 1172 96-well non tissue culture plate flat bottomed, sold by VWR • 150 ul pipette tips * • Glass Funnels (1 for each sample + 1 for standard)* • Thermometer clamps (1 for each sample + 1 for standard) • Burette stands • Appropriately sized beaker or Erlenmeyer flask for making double strength basal medium • 5 or 10 ml volumetric flasks (1 for each sample + 1 for standard)* • 18x150mm test tubes (or screw cap culture tubes*) (1 per sample) • Parafilm • Large OSI test tube*

Depletion Step 1. Six hours previous to planned plating time, take a loop of bacteria (inoculum) and transfer it to a depletion media tube. Ensure sterility of transfer, including spraying gloves and working area with 70% ethanol and sterilizing the inoculation loop with the Bunsen burner. Place in incubator at 37°C for 6 hours. a. If plating is delayed, the depletion tube can be placed in the refrigerator after incubation to slow the growth until the plating step is ready to go. 2. Inactivate amylase in a boiling water bath for 3 min; cool in an ice water bath 4 min. Volumetric Step 3. Quantitatively transfer to a 100 ml volumetric flask and fill to volume. Swirl Erlenmeyer flasks before pouring. To ensure quantitative transfer, rinse each of the sample flasks and the funnel 3 times with distilled water. The first of the 3 rinses needs to be the most complete. Be careful to not overfill the volumetric flask! Cap the vol. flask with parfilm. Filtration Step 4. Suspend glass funnels using thermometer clamps and burette stands. Arrange filter paper in funnels and filter approximately 20 ml through 2V filter paper into a test tube or screw cap tube. Invert the volumetric flask 5 times before filtering to ensure homogeneity of the dilution step and to be sure that the filtered portion accurately represents the sample! Dilution Step 5. Pipette the appropriate amount of each sample and standard from the filtered culture tubes into a 5 or 10 ml volumetric flask, fill to volume with deionized water, & cap with parafilm. For my fortified soymilk samples, I diluted 1.5 ml of the standard into a 10 ml volumetric flask and diluted the fortified . For my unfortified soymilk samples, I diluted 1.0 ml of the standard into a a. A 1:10 or 1.5:10 dilution at this step usually gives good linearity in the lower dilutions for the standard. (A 1:10 dilution would mean that 1 ml of filtered sample is diluted to volume in a 10 ml vol flask). Previous grad students have plated with 1:5 dilutions for the standard, but

69 they usually also used 200 ml volumetric flasks in the volumetric step instead of 100 ml. As far as I can tell, the purpose of working out the right dilution for the samples in this step is to get the bacterial growth of the samples to be similar to the growth of the standard on the plate. The growth curves are easily visualized by graphing the sample absorbance readings on the same y- axis as the standard curve, as I do in my version of the excel data analysis spreadsheet. For the linear portion of the data, this method is essentially a comparison of the slopes of the analytical samples to the standard. I suspect that the more similar the growth curves of the samples are to the standard, the better. For example, if one sample has significantly more growth than the standard and one sample has significantly less growth (on the same plate), then the concentration of the sample with more growth may be over-estimated and the low-growth sample may be underestimated – especially if the standard curve regression is not chosen very carefully. It is easy to see how erroneous concentration estimates are made when the growth in the sample wells is very different from the standard. In the calculations for the 3 most dilute wells on the plate, the dilution factors are 64, 32, and 16. These are huge multiplicative factors, and if the growth of the sample is very different from the standard, then when the sample absorbances are plugged into the standard curve regression equation any error has the potential to be hugely magnified. That being said, the growth curves can be different from the standard and the concentrations can still be correctly predicted. However, at some point as the sample concentrations in the plate differ more and more from the standard then the result will be increasingly inaccurate concentration values. b. After this dilution step is successfully completed, the filtered culture tubes from step #4 have traditionally been discarded. However, in the early stages of a project when running a new type of sample, the dilutions may not yet be optimized. In these instances especially (and perhaps always) it is wise practice to cap the culture tubes from step #4 and store them in dark refrigeration. Then the next day, after the plate has been read, step #5 can be repeated using different dilutions without having to repeat the tedious tri-enzyme extraction steps. c. There are three places to control folate concentration in this method. One is in how much sample enters into the analysis at the very beginning. A second possibility is the volumetric dilution step. (Previous graduate students have used 200 ml volumetric flasks. A plus to using 200 ml volumetric flasks is that there is less danger of overfilling them.) The final option is in this dilution step. I think that it is ideal to adjust the amount of sample going into the analysis instead of using hugely disparate dilutions at this step, because variations in the plate can be magnified differently for different samples if the back-calculations include very different multiplicative factors. However, this dilution step is a valuable tool because there are limitations to the amount of sample that can be put through the analysis. The window of flexibility in the amount of sample analyzed is basically from as low as is needed to get a representative sample to as high as won’t overwhelm the amounts of the three enzymes used in digestion. Of course, the amount of enzyme used can always be adjusted (refer to the lab reference textbook: Eitenmiller’s Vitamin Analysis ). There may be other practical limitations to how much sample can be put through this method, such as clogging up the filters, etc. Sterile Filter Step 6. Using the bio-hood, attach .22 um filter to a 5 or 10 ml sterile syringe. Invert the diluted samples 5 times to ensure homogeneity. Pour each sample from the 5 or 10 ml volumetric flask into its own syringe and filter system and expel into sterile screw cap cryo vials. Ensure sterility by generously spraying bio-hood surface and gloves with 70% ethanol prior to filtering. Do not over-fill the cryo vials or the threads will be wet when the lid is put on, allowing an unbroken

70 liquid ‘microbial highway’ from the outside of the vial to the supposedly sterile interior. If samples are not ready to be plated, store the screw-cap vials in dark refrigeration (wrap with foil if necessary) until ready to use. a. Ideally, the samples should be plated within six hours. However, the cryo vials should be kept and the samples can be re-plated up to a week later if the data was poor. This will avoid repeating the 2 days of tedious enzyme extraction! I have re-plated successfully from refrigerated cryo vials as long as a week after the sterile filter step without noticeable degradation of folate, as far as I can tell. Plating Step 7. Filter sterilize water into pipetting trough using .22 um filter and 20 ml sterile syringe. a. Alternatively, the water can be autoclaved and sterile poured into the trough (see Sterile Techniques section). I have achieved great data this way, and it is easier to prepare multiple plates when there is ample sterile water that doesn’t have to be pushed through a syringe. 8. Using multi-channel pipet, pipette 150 ul of water to wells A1-H12 of a sterile microtiter plate. a. It is important to transfer exactly the same amount into each well, so every time a new pipet tip is used during any part of the plating step, the solution being pipetted should be drawn up and expelled once to coat the tip. The pipet should then dispense very similar amounts for every subsequent delivery with that same tip. b. Bacteria and/or spores in the pipets used for plating seem to be a potential source of contamination in this method. Non-autoclaveable pipets have been used for this method since it was first run at BYU. However, in late 2014 this method began to randomly produce poor data with high absorbance values in the blank row (~0.4 – 0.7 absorbance units). This encouraged us to develop a more robust method of data analysis with excel, but since the linearity of optical density readings begins to drop off between 0.6 – 0.8 absorbance units, no amount of improvement in the data analysis can compensate for absorbance values taken outside of the linear range. At one point we opened up the pipets that we were using for plating and washed them with 70% ethanol (we had no autoclaveable pipets). We consistently got good data with low blank rows (~2.5 – 3 absorbance units) for a few weeks afterward. We also discovered that if the plating media is boiled for longer than 1 minute, background noise in the plate was high. Then, just when we thought we had it figured out, contamination seemed to sporadically creep back in. An almost paranoid emphasis on sterile techniques would sometimes produce low blank rows, but, frustratingly, sometimes it would not. We were using a still-air hood situated near the air vent for the room; could it be that when the forced air was on in the building during plating our data turned out poorly? We couldn’t figure it out. Spraying everything in the plating area liberally with 70% ethanol (including the pipettes) and allowing absolutely nothing to enter the plating area without first being sprayed with 70% ethanol seemed to work…most of the time. It is recommended that fully autoclaveable pipettes are used for plating in the future, allowing for the option of sterilizing the pipettes and ruling out the possibility of contamination from their plunger channels. Fully autoclaveable pipets were purchased in Dec of 2015, and we have yet to see whether or not this will result in consistently good data. 9. Pipette 150 ul standard extract to wells A1-A2. For each sample extract, pipette 150 ul in duplicate into wells in row A, i.e., 150 ul of sample 1 extract to each well A3 and A4, 150 ul of sample 2 extract to A5 and A6, etc., up to A12. Therefore you can run five sample extracts and one standard extract per microtiter plate. a. See appendix B for a discussion on alternate plating options.

71 10. Mix contents of row A to homogeneity by using a multi-channel pipette and repeating aspiration and delivery steps. Pick a number of repetitions (I use 15) and keep it consistent across the plate. 11. Make serial dilutions (x2) of standard and samples by transferring 150 ul from wells A1- A12 to B1-B12 and mixing, then from B1-B12 to C1-C12 and mixing, etc. For samples G1-G12 withdraw 150 ul after mixing and discard. Use the same tips for all the dilutions to ensure complete and equal delivery of the pipet contents for each well during each dilution. 12. Filter sterilize double strength basal media using .22 um filter and 20 ml sterile syringe into the large OSI test tube. 13. Add 20 ul of incubated depleted bacteria culture to the media in the large test tube. 14. Vortex mix the test tube for 30 seconds. Pour into pipette trough. Add 150 ul of inoculated depletion media to wells H1-A12 (row H becomes the inoculated blank row). As in plating step 8, completely fill and drain the pipet tips back into the media reservoir before putting any inoculum onto the plate. a. Start at the bottom of the plate with the blank row, and release the inoculated media into the wells without touching the pipet tips to the folate and water solution. By releasing the bacteria from an eighth of an inch above the surface of the well contents, the tips will not transfer folate to the rest of the rows and ruin the dilutions. Starting with the blank row ensures that even if the pipet tips touch the sample in the wells, insignificant amounts of folate will be transferred from more dilute wells to more concentrated wells, thus having less impact on the data outcome. The tiny volumes used in this method save a lot of money on media compared to the AOAC method, but they also magnify errors. A lot of care and precision is required in the plating portion of this method. 15. Put cover on microtiter plate. Put plate into a sterile bag and incubate 20-24 hours at 37°C (22 hours was usually the perfect incubation time for my soymilk samples).

*Record EVERYTHING accurately! Carefully fill out the data sheet (See appendix F) or the entire analysis may need to be redone!

Day three of the analysis – Reading the plate Materials • 150 ul pipette tips * Mixing the plate 1. Immediately prior to putting the plate into the plate reader, mix contents of each well using a multi-channel pipette with sterile tips. Discard tips after mixing each row. Ensure there are absolutely no air bubbles in any of the wells; this will distort the optical density (absorbance) reading. Mixing the plate is very important. If mixing is done incorrectly the data can be very adversely affected, so read the following explanations carefully. a. Bubbles can be hard to pop, so they are best avoided by introducing no air while mixing and by expelling the last bit of liquid while holding the pipet tip an eighth of an inch above the surface of the liquid in the well. b. After the wells are mixed the microbes immediately begin to settle to the bottom, giving falsely high absorbance readings. This is why the first and second scans of the plate are sometimes quite different, with the second reading giving the higher absorbance values. Therefore the plate should be mixed immediately before it is inserted into the plate reader, and it should be mixed as quickly as possible (without introducing bubbles). That being said, the

72 mixing needs to be sufficient to get all the microbes stirred up from where they have settled on the bottom and evenly distributed throughout the well. As with the plating dilution step, pick an adequate number of aspirations and keep it consistent across the plate (I use 15). Be sure that you move the plunger up and down enough to move a lot of liquid, but, again, avoid air bubbles! c. The linear relationship in this method between folate concentration and microbial growth begins to drop off at the higher absorbance values (See Appendix ?). Generally, the more concentrated wells are under-represented by the spectrophotometer, having a lower absorbance reading than they should. Therefore, the best readings are obtained when the plate is mixed from the most concentrated wells first (row A) to the least concentrated row (the blank row: H). This way the more concentrated rows are the ones that settle, resulting in a more linear data output and keeping accuracy highest in the most dilute wells which are generally the best ones to use for the standard curve regression. The way the plating step is done in this version of the method, this means that the plate is simply mixed from the top down – starting at row A and ending with the blank row – row H. Reading the plate 2. Read the plate with the microtiter plate reader in Dr. Davidson’s lab. “Folic Acid Reverse Plate” is the name of the Fluostar program used to read the plate in this method version. Previously, the most concentrated well was row G, right next to the blank row. The assay labeled simply “Folic Acid” is programmed for the old way of plating. Instructions for the machine are in appendix B. Label the file using a distinct code representative of the samples being run or as per current lab protocol. 3. If you are not sure of how long the plate should be incubated, such as in the case of a new set of samples, do not let it incubate too long and be sure to mix the plate with sterile tips the first time it is read. This way, if the values are low it can be returned to incubation for another hour or two. Just stay within the spectrophotometer’s range of 0.1 – 1.1. In my experience an additional hour of incubation after an original incubation time of 22 hours will raise the highest dilution 0.1-0.2 and the blank 0.005-0.015 absorbance units. 4. Verify that the standard curve has a good linear portion for the lower dilutions and that a four point regression line fits it well. Re-read the plate twice, unless the data looks inconsistent, in which case read the plate one more time or re-read it twice again after re-mixing the wells. 5. Analyze microplate data using the most current Excel template. a. The excel templates for data analysis have been built for a plating method with the standard in the leftmost two columns, and then the rest of the data in duplicate columns to the right. However, I have noticed that the standard (furthest to the left in the plate) and the sample in the rightmost columns of the plate often have higher absorbance readings than the others – especially in the low dilutions. Thus the standard curve tends to be built using a higher value than most of the samples. The result of this is often erroneous predictions of the sample concentrations in the most diluted wells, and these values can often be negative and must be excluded from the analysis. Jacob Foist adroitly observed that the problem is very apparent if the raw absorbance values are multiplied by their respective dilution factors. Technically this should return similar values for all the wells in each column, but it does not. More often than not, the absorbance values seem to be low in the middle of the plate and high on the edges. I think that this is a function of an uneven light source beneath the plate. More light passes through the plate without scattering in the middle than at the edges. Perhaps this discussion belongs in the plating section, but I think it’s a plate-reader problem. Some possible solutions would be to not put any sample or standard in the first and last columns, but then only duplicates

73 of 5 samples could be run instead of 6. A better solution might be to plate all six samples in single columns on one half of the plate and then repeat the order on the second half of the plate. This would put one of the standard columns near the edge and one right in the center. Appendix A – Solutions 4N KOH Materials needed: • 1 L volumetric flask • Plastic bottle for storage • Potassium hydroxide

Add 224 g potassium hydroxide in 1 L water.

4N NaOH Materials needed: • 1 L volumetric flask • Plastic bottle for storage • Sodium hydroxide

Add 160 g NaOH in 1 L water.

Folic Acid Stock Solution (100ug/ml) Materials needed: • 500 ml low actinic volumetric flask • USP folic acid

Accurately weigh 50 mg USP folic acid that has been dried to constant weight and dissolve in .1M, pH 7.0 phosphate buffer in 500 ml low-actinic volumetric flask. Dilute to volume with .1M phosphate buffer. Top with opaque glass stopper. Store in refrigerator. Check the concentration of the solution photometrically every two weeks. Instructions included in appendix C.

Discussion Degradation of the standard solution over time is a problem that has plagued past versions of this method. When the standard solution has degraded but the excel analysis still considers it full strength, the sample values calculated from the best fit regression of the standard growth curve are unintentionally over-estimated. I tried adding ascorbic acid to a new stock solution (ascorbic acid is used as a protectant in every other solution in the method) to slow the degradation and had Jiping check the degradation over several months using HPLC. (Ascorbic acid absorbs at the same wavelengths as folic acid, thus rendering the spectrophotometric method unusable. This is probably why the standard stock solution for this method has not included vitamin C in the past.) The folic acid seemed to continue to degrade, and while the HPLC method for checking the concentration worked well at first, the folic acid peak began to split after a month or so. Having no way to know if the folic acid was simply converting to another form which the microbes could still use, it became difficult to know how to adjust the concentration value of the standard when analyzing the data. Therefore, while the methods for verifying the concentration of the folic acid stock solution (Appendix C) may help adjust the standard concentration value or help determine if the standard needs to be replaced, neither method is ideal because ideally the

74 standard solution would just always be the same! Accurate data analysis on the back end of the method depends entirely on knowing the concentration of the standard solution entering the analysis on the front end. There are a couple of viable solutions to this problem:

1. Cryo-protect the working standard solution or the folic acid stock solution, depending on which freezes better. The stock solution is in a 0.1 M phosphate buffer, while the matrix of the working standard is primarily water. It could be that the buffered solution would be kinder to the folic acid in a freeze-thaw cycle because the ice crystals would be more frequently disrupted by the buffer salts. On the other hand, smaller ice crystals may cause more folic acid breakdown. The methods would simply need to be compared to know if one is better than the other. I am hopeful that the folic acid would weather the freeze thaw cycle well, because the fortified folic acid in my soy milk samples seemed to be unaffected by my sample storage at -80oC. The frozen folic acid standard could be stored in 2 ml cryo vials in the -80oC freezer, and a vial could be thawed and used for each analysis run. If only 1 ml is used for the analysis, this would leave 1 ml of the frozen standard that could be occasionally analyzed using HPLC to make sure that the standard is the same for each analysis run. 2. Another option would be to carefully make a new folic acid stock solution each week or maybe every other week. The error introduced by the scale, the volumetric dilution steps, and the analyst may be much less than the calculation errors caused by the standard degrading over time. I recommend trying the cryo-protection method first, as that would ensure that the standard concentration was always exactly the same, and would involve less work.

0.1M pH 7.0 phosphate buffer Materials needed: • 1 L volumetric flask • 4 N KOH • 600 ml beaker • Potassium phosphate monobasic

Dissolve 13.61 g potassium phosphate monobasic in 1 L volumetric flask and dilute to 1 L. Pour the amount needed (at least 500 ml) into the beaker and adjust the pH to 7.0 with 4N potassium hydroxide.

Working standard solution (1ug/ml) Materials needed: • 600 ml beaker • 500 ml low-actinic volumetric flask • (2) 50 ml beakers • HCl (1 part HCl to 1 part H20) • 4N KOH • Folic acid stock solution (100ug/ml)

Bring stock solution to room temperature (to accurately pipette). Fill large beaker with ~475 ml of water. If preparing a working standard solution for fortified samples or the Wyoming standard, add 5 ml of the stock solution. If preparing a working standard for unfortified samples add 1 ml. Adjust pH to 7.5 with dilute HCl or KOH depending on the starting pH of the solution.

75 Use safety goggles when working with strong acids and bases. Quantitatively transfer (triple rinse) the adjusted solution to the 500 ml flask and fill to volume with water. Prepare fresh on day of use. Alternatively, the stock solution can be pipetted by weight instead of with a volumetric pipet, so that the stock solution does not have to be brought to room temperature. For all of my soy milk samples, both fortified and unfortified, the working standard solution was prepared using a recorded weight close to 1g (1 ml) of stock solution. The data analysis then adjusts for the measured weight of the standard similarly to the samples. If pipetting is done by weight, pipet the stock solution first, and then add 475 ml of water to the beaker, and continue as above.

pH 7.8 phosphate buffer Materials needed: • Sodium phosphate dibasic • Ascorbic acid • 200 ml volumetric flask • 4N NaOH • 500 ml beaker

Fill flask about half full with water. Add 2.84 g sodium phosphate dibasic and 2.0 g ascorbic acid to flask and dilute to volume. Transfer to beaker. Adjust pH to 7.8 with 4N NaOH. Prepare fresh on day of use.

Protease solution Make just enough for the day’s analysis, 1 ml per sample, plus one extra milliliter. Materials needed: • Protease – from Streptomyces griseus, Pronase E from Sigma No P-5147 • Appropriate size volumetric flask or beaker (1 ml per sample + 1 ml extra)

Partially fill the volumetric flask with water. Add .002 g protease for every 1 ml water (ex. add .01 g protease to 5 ml) and fill to volume. Prepare fresh on day of use.

If the protease solution is prepared in a beaker, remember that too much water or too little protease will dilute the enzyme, so err on the side of ensuring that the solution is at least as concentrated as is necessary. This amount of protease is adequate for most matrices when 1g of dry sample is extracted, but this much enzyme can only process a certain amount of substrate during the method incubation time. If the protein burden of the sample is too high, more protease will be necessary. See the AOAC method or the lab’s vitamin analysis reference textbook for details on the trienzyme extraction.

α-amylase solution Make just enough for the day’s analysis, 1 ml per sample plus 1 ml extra Materials needed: • α-amylase – from Aspergillus oryzae, Sigma No A-6211 • Appropriate size volumetric flask or beaker (1 ml per sample + 1 ml extra)

76 Partially fill the volumetric flask with water. Add .02 g amylase for every 1 ml water (ex. add .1 g amylase to 5 ml). Prepare fresh on day of use.

If the amylase solution is prepared in a beaker, remember that too much water or too little amylase will dilute the enzyme, so err on the side of ensuring that the solution is at least as concentrated as described in the above paragraph. The amount of amylase in this method is adequate for most matrices when 1g of dry sample is extracted, but this much enzyme can only process a certain amount of substrate during the method incubation time. If the carbohydrate burden of the sample is too high, more amylase will be necessary. See the AOAC method or the lab’s vitamin analysis reference textbook for details on the trienzyme extraction.

Depletion media Materials needed: • Beaker capable of holding 100 ml • Lactobacillus broth • Folic acid casei media • Screw cap tubes (or culture tubes with autoclaveable caps) capable of holding 10 ml plus room to mix

Add 1.9 g Lactobacillus broth and 4.7 g folic acid casei media to 100 ml water in beaker and boil for 2-3 min. After cooling quickly, dispense 10 ml each into screw cap tubes. Autoclave loosely capped tubes at 121°C for 15 min and quickly cool to room temp. Tighten down the caps and store in refrigerator until use.

Double strength basal medium Materials needed: • Folic acid casei media • Ascorbic acid • Appropriate size beaker (15 ml per plate)

Suspend 2.35 g casei media and 12.5 mg ascorbic acid for every 25 ml water. Boil for exactly 1 minute and cool in ice water bath. When cooled cover flask with foil and leave in bath or refrigerate until ready to use. Do not boil the plating media longer than 1 minute even if some of the crystalline ingredients have not fully dissolved! We spent months trying to diagnose problems with background noise in this method, and boiling this media too long consistently correlated with a higher blank row and higher background absorbance values across the entire plate. We also tried autoclaving the media according to the instructions on the media bottle label, and it resulted in a more darkly colored media (visibly different in the plate) and a high blank row and background noise. The two variables that seemed to correlate most strongly with poor data were over-cooking this media and contamination from the pipettes.

Appendix B – Microplate reader

Kenealey Lab Plate Reader Setup Reader model: FLUOstar OPTIMA (made by BMG Laboratories)

77 Software version: 1.30 revision 3 Firmware version: 1.14-0 BMG Labtec technical support: 1-877-264-5227

1) Log onto “Lab User” – password: “davidson” 2) Double-click “FLUOstar OPTIMA” icon on desktop to launch program 3) Select “User” and press “Run” 4) “Setup” menu a. “Reader Configuration” – select “Absorbance” b. “Filters” – select A-620 in excitation column; emission column should read “empty” c. “Microplates” – select “Falcon 3070 Culture Plate” from list or add new plate (get specs from manufacturer) 5) “Test Setup” menu a. “Test Protocols” – select name of test or create new test name (i.e. name of the project you’re doing). If you have already created one and it doesn’t appear on the list, restart the program. b. “Basic Parameters” tab i. “Positioning Delay” – enter “0.5” seconds ii. “Flying Mode” leave unchecked iii. “No. of kinetic windows” – enter “1” – ignore any further kinetic settings iv. “Filter Setting” – enter “1” for “no. of chromatics” v. “Excitation filter” – select “A-620” (whenever doing absorbance readings you want to use only excitation) vi. “Emission filter” – select “empty” vii. Ignore “Gain” (this is for a manual gain adjustment, you want an automatic gain adjustment which will be described later) viii. “Well Scanning” – choose pattern of how to read each well. We used “none.” See Help File for more information. c. “Layout” tab i. Assign each well as “S” (standard), “X” (unknown), or “B” (blank) as illustrated below:

A S7 S7 X7 X7 X14 X14 B S6 S6 X6 X6 X13 X13 C S5 S5 X5 X5 X12 X12 D S4 S4 X4 X4 X11 X11 E S3 S3 X3 X3 X10 X10 F S2 S2 X2 X2 X9 X9 G S1 S1 X1 X1 X8 X8 H B B B B B B etc…

The ‘Folic Acid Reverse Plate’ version of this method was developed in part to make mixing the wells intuitive – from the top down. This test should be run when the samples are serially diluted on the plate from row A to row G, leaving the most concentrated wells at the top of the plate. This layout is programmed in the FLUOstar program as follows:

78 A S1 S1 X1 X1 X8 X8 B S2 S2 X2 X2 X9 X9 C S3 S3 X3 X3 X10 X10 D S4 S4 X4 X4 X11 X11 E S5 S5 X5 X5 X12 X12 F S6 S6 X6 X6 X13 X13 G S7 S7 X7 X7 X14 X14 H B B B B B B etc…

Careful observation of the raw absorbance data reveals that the plate reader may not function consistently across the plate. When analyzing “Folic Acid Reverse Plate” data in excel, if the raw data is multiplied by its corresponding dilution factor (1, 2, 4, 8, 16, 32, & 64) technically the values should all be exactly the same. However, the first and last columns on the plate tend to be suspiciously high and the columns in the center of the plate seem to read consistently low for the most dilute wells. The result is that the standard curve regression equation does not return an accurate concentration estimate for the most diluted wells of the samples, often returning negative concentration values for the samples in the middle of the plate. It is possible that since the plate reader has a single bulb light source it does not distribute light intensity evenly across all the columns, resulting in more light throughput in the center of the plate (interpreted as lower absorbance/reflection/scattering) and less light throughput on the outside edges of the plate (interpreted as more absorbance/reflection/scattering). This problem would be difficult to remedy. It is also possible that the multi-channel pipette is not mixing the outside columns well, and the problem might be remedied by plating such that for the standard and each sample on the plate the duplicate columns are split up such that one is near the edge of the plate and one is near the center. The “Layout” tab in the FLUOstar program would then look something like this (the standard need not be the one in the far left column, since the outside columns seem to be the worst):

A S1 X1 X8 X15 X22 X29 X36 S1 X1 X8 X15 X22 etc. B S2 X2 X9 X16 X23 X30 X37 S2 X2 X9 X16 X23 etc. C S3 X3 X10 X17 X24 X31 X38 S3 X3 X10 X17 X24 etc. D S4 X4 X11 X18 X25 X32 X39 S4 X4 X11 X18 X25 etc. E S5 X5 X12 X19 X26 X33 X40 S5 X5 X12 X19 X26 etc. F S6 X6 X13 X20 X27 X34 X41 S6 X6 X13 X20 X27 etc. G S7 X7 X14 X21 X28 X35 X42 S7 X7 X14 X21 X28 etc. H B B B B B B B B B B B B etc.

79 d. “Concentrations/ Volumes/ Shaking” tab i. “Concentration” – enter “0” for “start concentration” and select “factor” ii. “Volume” – enter “0” for “start volume” and select “factor” iii. “Shaking Options” – select “No Shaking” from drop down menu below “Additional Shaking” – ignore all other shaking options iv. Set dilutions of standard in the table to the right as follows:

S1 1 S2 0.5 S3 0.25 S4 0.125 S5 0.0625 S6 0.03125 S7 0.015625

At this step we ignored that the standard in row G gets diluted twice during plating (once during the serial dilution step when 150 μL of the analytical standard is added to 150 μL water, and again when 150 ul of inoculated double strength basal media is added to the plate). Since concentrations are calculated purely by comparison of the growth curves in the different columns and the dilutions that occur during plating are the same for all samples, it is simply easier to arbitrarily say that the growth of the standard was due to the folic acid concentration that was present at the sterile filter step. The data analysis will then estimate the concentrations of the other samples as they were at the sterile filter step and not the actual concentrations present in the wells of the plate. v. “Injection Timing” tab – everything should be grayed out

Kenealy Lab Plate Reading: First Time Computer Password: davidson 1) FLUOstar OPTIMA Main Program Screen a. Press “Plate Out” button (farthest left icon button) to open reader b. Place plate in reader. Well A1 should be in upper left corner when facing plate reader. c. Press “Plate In” button (next to “Plate Out”) d. Press “Measure” button (stop light icon button) e. Double-click appropriate test name (created in “Setup” step 2) a.) f. “Plate ID’s” tab – enter a name from this particular test run (e.g. June 5 hr21) g. “Gain Adjustment” tab vi. “Required value” – set to “80%” vii. “Filter setting” should read: # Excit. Emiss. Gain 1 A-620 empty (ignore what's here)

If “Filter setting” appears differently than above go back to steps 2) b. v., vi. and make the necessary changes. viii. Press “Gain Adjustment” button at bottom of screen to set the gain adjustment automatically. A number ~51,000 will appear below “Measurement value for 0% absorption

80 (100% transmission)”. This automatic gain setting will automatically be used instead of the manual gain setting. h. “Sample ID’s/Dilution Factors” tab ix. “Dilution” column of will be grayed out for cells corresponding to blanks and standards. Sort the table by content via clicking the “Content” button at the top of the table or by selecting it from the drop-down menu to the right. Leave “Sample ID” column blank. x. Set dilution factors for sample unknowns as follows: Well Content Sample ID Dilution G3 X1 1 G4 X1 1 F3 X2 2 F4 X2 2 E3 X3 4 E4 X3 4 D3 X4 8 D4 X4 8 C3 X5 16 C4 X5 16 B3 X6 32 B4 X6 32 A3 X7 64 A4 X7 64

Repeat four more times (once for each of the five non-standard samples) For the “Folic Acid Reverse Plate” test, the dilution factors should read 1 for the wells of row A, 2 for the wells of row B, etc. down to 64 for the wells of row G, consistent with the way the serial dilutions were made during plating. i. Press “Start Test Run” button at bottom of screen. Reading a plate with the same setting as described herein takes about two minutes. j. When reading is completed, eject plate with the “Plate Out” button. k. Exit program and log off computer.

Kenealy Lab Plate Reading: Second Time and Beyond After the initial setup, the following steps are all you should to read additional plates. 1) Place plate in reader (see “Plate Reading: First Time” steps 1) a-c) 2) Press “Measure” button 3) Double-click appropriate test name 4) “Gain Adjustment” tab – press “Gain Adjustment” button and wait for reading to appear. 5) Press “Start Test Run”

Kenealey Lab Data Analysis 1) FLUOstar OPTIMA Main Program Screen a. Press “Evaluation Part” button (spreadsheet icon button on main screen) b. Double-click test run of interest c. To Export raw data:

81 i. Press “Raw Data” tab (bottom of screen) ii. Press “Order by Columns” toggle button (top of screen) iii. Copy and paste data to the Excel folate data analysis template prepared for that particular project or set of samples, rename the template file as per lab protocol, and save file to a flash drive (this computer doesn’t have internet access). 2) “Standard Curve” tab a. Select “4 Parameter Fit” from the drop-down menu below the graph b. Press “Reset” button c. Press “Optimize Fit” button d. Press “Lin” (linear) toggles button for both x and y axis. The y axis is absorbance units. The x axis is concentration of folic acid.

Appendix C – Checking concentration of stock solution Prepare fifty ml 0.1M, pH7 phosphate buffer by adding 0.68 g potassium phosphate monobasic to a 50 ml volumetric flask and filling to volume. Transfer enough to fill a plastic cuvette. Transfer enough folic acid stock solution to fill another cuvette. Using the pH7 buffer as a blank, determine the absorbance of the folic acid solution at 282 and 346 nm. Determine the concentration of the solution using both values as follows: Folic acid conc, ug/ml = Absorbance(282nm) x 1000 x 441.4 / 27600 Folic acid conc, ug/ml = Absorbance(346nm) x 1000 x 441.4 / 7200 Use the average of the values obtained. The concentration of the solution should be approximately 100 ug/ml. If the solution is not between 90 and 110 ug/ml it should be remade. The experimental value for the concentration of the solution should be used for the standard curve on any plate until a new concentration is measured.

Discussion: There are some red flags in the above instructions. First of all, unless the “plastic cuvettes” referred to are the special new ones that are made for use in the UV range, measurements at these wavelengths should be taken using quartz cuvettes. Also, we have found that a ten-fold dilution of the stock solution with the buffer is required to get the readings at each wavelength within the ideal absorbance range of most spectrophotometers (0.1 – 1.0 absorbance units). Finally, it should be emphasized that the range of acceptable values mentioned here is far too great to be analyzed as 100 ug/ml. The recommendation is that any solution further than 10 ug/ml away from the target of 100 ug/ml should be remade, and always the experimentally verified concentration value should be used to analyze the data. The spectrophotometer concentration check can work, and it is better than nothing, but it is probably not ideal. The undeniably good advice here is to ensure that an accurate concentration for the standard is used for the analysis of the data. The concentration of every other sample on the plate is determined by comparison to this value, and it must be accurate.

In search of a better way to measure the concentration of the standard folic acid stock solution, we developed an HPLC method (from Lebiedzinska et al.) in which the folic acid stock solution can be injected directly onto an HPLC column. The HPLC method seemed to return the stock solution concentration at least as accurately as the spectrophotometer method for about a month and a half. However, the folic acid peak began to split as the standard apparently degraded in ways perhaps not captured spectrophotometrically and quantification became more difficult. The

82 HPLC method is easier to use, and could probably be used to test freshly made standard solutions, but I recommend that the folic acid stock solution be either remade more frequently or cryo-protected so that there isn’t so much variation between plates. This is discussed under the Folic Acid Stock Solution section in this version of the method.

Reference: Lebiedzinska A, Dabrowska M, Szefer P, Marszall M. High-Performance Liquid Chromatography Method for the Determination of Folic Acid in Fortified Food Products. Toxicology Mechanisms and Methods. 2008, 18:463-467.

Appendix D – Analysis of data Using Microsoft Excel makes this very easy. Average the absorbance readings from the duplicate columns. Then plot the first column of data (averaged from the duplicate columns on the plate) on the x-axis and plot the standard curves’ concentrations (2ng/ml, 1ng/ml, .5, etc) on the y axis. (It should be noted that the concentration value used for the standard is the sterile filtered concentration. The further dilutions with water and media in the plate are ignored because the same dilutions happen to all the samples during plating, and relative comparisons between the samples and the standard are more important than the absolute values obtained in this test.) Then, using only the points that are on the linear portion of the curve, determine the formula for the linear trendline of the standard curve. Plug the unknown absorbance values into the formula. Then multiply that number by the appropriate dilution factor (64, 32, 16, etc). Average the values obtained that seem to be predicted accurately by the standard curve (excluding those that are way too small or too big). Calculate the average, standard deviation, percent relative standard deviation (%RSD - StdDev*100/Average), and the Horrat value (%RSD/21.333333). If the Horrat value is between .3 and 1.3, the value can be trusted and used. If not, redo the sample on a different plate.

Discussion of Optical Density Readings Since this method relies on optical density (OD) as a measure of microbial growth, a short discussion of the limitations of this technique is in order. According to Brock Biology of Microorganisms, 13th Ed (see graph on pg 132), the ability of the spectrophotometer to accurately measure microbial growth is limited; the linearity (i.e. accuracy) of turbidity as a measure of concentration usually begins to drop off between absorbance readings of 0.6-0.8. On pages 96-97, Bacterial Physiology: Microbiology, 2nd Edition (Davis, et al.) also states that the accuracy of the prediction of microbial count based off of turbidity readings begins to be suspect above a certain concentration, but the limitation is listed in terms of microbial mass (~0.5 mg/ml) instead of absorbance units. Depending on the estimated weight of the microbes this could be anything above as low an OD reading as 0.4. I found multiple technical papers produced by Thermal Fischer discussing similar limitations in the context of specific applications for their particular instruments. In almost all of the data I have seen produced at BYU with this method, the textbook explanation seems to hold true. The most concentrated wells consistently have lower absorbance readings than would be expected, and the turbidity-concentration correlation begins to be non-linear somewhere between an absolute absorbance of 0.5-0.8. For instance, if the blank row is between 0.25 and 0.30 absorbance units, the linearity of the blank-adjusted data would begin to drop off after absorbance readings of about 0.35-0.45 (total absorbance of between 0.6 & 0.75). This occurs even if the plate is mixed correctly before it is read (See

83 Analysis Day 3, Mixing the Plate). Therefore the folate that goes onto the plate should be concentrated enough for the serial dilutions to spread across the entire accurate absorbance range, but dilute enough for the most concentrated wells to be within the accurate range or close to it; and the method for choosing the linear regression of the standard curve should exclude points outside the range of accurate, linear absorbance readings. Knowledge of the limitations of optical density measurements helps one to know how much background noise in the plate is acceptable. For example, if the blank row reads at 0.7 absorbance units, all of the blank-adjusted data will be outside of the linear range and the samples will need to be re-plated. However, if the blank row is 0.4 to 0.5 absorbance units, the data may be salvageable if at least the 4 most diluted wells fall within the approximately 0.2 absorbance units of linear range available. However, the lower the blank row the better the data will be, because the slope is a multiplicative factor which will magnify variance in the absorbance readings. In the above hypothetical scenario in which the blank row reads 0.4-0.5 absorbance units, the slope of the linear regression would be much steeper than if the blank row were lower. For example the slope might be 3.0 instead of 1.5, thus doubling the effect of variation in the raw microbial growth readings simply by plugging the raw data into a standard curve regression equation with a numerically higher slope value.

References Davis, Dulbecco, Eisen, Ginsberg. Bacterial Physiology: Microbiology, 2nd Edition. Maryland: Harper and Row, 1973, pages 96-97 Madigan, Michael T; Martinko, John M; Stahl, David A; Clark, David P. Brock Biology of Microorganisms 13th ed. San Francisco, CA: Benjamin Cummings, 2012, pages 132-133.

Discussion of the Horwitz Ratio (i.e. HorRat or Horrat) Since this method relies heavily on the Horrat value to determine the acceptability of the data analysis, an explanation of this metric is warranted. In 1980, while analyzing large quantities of data, Dr. William Horwitz discovered that regardless of the analyte or method, the observed relative standard deviation (RSD) increased predictably and exponentially as smaller and smaller quantities were analyzed. This general observation is not extraordinary or unexpected, but what was unique and important about Dr. Horwitz’s discovery was how universally he could predict the relative standard deviation for any anylate measured with any method simply using the predicted RSD taken from his large empirical data set. Thus, for any expected concentration of an analyte, when expressed as a unitless mass ratio (such as ug/g or ppm), there is a fairly predictable expected relative standard deviation. Dr. Horwitz outlined a mathematical equation to predict what the relative standard deviation should be for the measurement of any given concentration of an analyte. If the relative standard deviation (RSD) of the data from any given analysis were too much greater than the predicted relative standard deviation (PRSD) identified by the Horwitz equation, it would be assumed that the analysis was poorly run (i.e. variation is too high) and should be redone. If the RSD of the data from any given analysis were too much lower than the PRSD, it would be assumed that some form of data manipulation occurred (i.e. the data was ‘too good to be true’). Therefore, the Horwitz Ratio (a.k.a. HorRat or Horrat value) was born as a measure of the quality of the data for any given analysis. The Horrat value is calculated by dividing the actual relative standard deviation from any given analysis by the relative standard deviation predicted by the Horwitz equation (i.e. Horrat value = RSD/PRSD).

84 According to Rivera and Rodriguez, “acceptable values under repeatability conditions are between 0.3 and 1.3 [for single-laboratory studies]…AOAC considers that within-laboratory acceptance predicted target values for repeatability are given at 1⁄2 of [PRSD], which represents the best case.” In other words, an ideal HorRat value calculated on tests done within a single laboratory is about 0.5. According to Faigeli and Ambrus, “A Horrat value of about 0.5-0.7 is expected from within-laboratory studies (FDA-TD) and about 1 from among-laboratory studies (FAPAS & CA), although values in the interval from about 0.5 to 2 times the expected value may be acceptable.” Rivera and Rodriguez further comment on the interpretation of the HorRat: “AOAC’s supplement explains that values at extremes of the limits of the acceptable range must be interpreted with caution. With a series of low values of HorRat, check for unreported averaging or prior knowledge of the analyte content; with a series of high values of HorRat, check for method deficiencies such as unrestricted times, temperatures, masses, volumes and concentrations; unrecognized impurities (detergent residues on glassware, peroxides in ether); incomplete extractions and transfers and uncontrolled parameters in specific instrumental techniques.”

References Aleš Fajgelj and Árpád Ambrus. Principles and Practices of Method Validation. Royal Chemistry Society, 2000. Carlos Rivera and Rosario Rodriguez. Horwitz Equation as Quality Benchmark in ISO/IEC 17025 Testing Laboratory. Bufete de ingenieros industriales, S.C. Pimentel 4104 –B; Col. Las Granjas. Chihuahua Chihuahua Mexico. C.P. 31160. Aleš Fajgelj and Árpád Ambrus. Principles and Practices of Method Validation. Royal Chemistry Society, 2000. M Thompson. The Amazing Horwitz Function. AMC Technical Brief No.17, Royal Chemistry Society, July 2004. Horwitz, William. The Horwitz ratio (HorRat): A useful index of method performance with respect to precision. Journal of AOAC International. 2006, 89(4):1095-1099.

Discussion of the Horrat Value Calculation The concentration of folate is not what is being measured here. The real thing that is being measured is the optical density of microbes that are making the solution turbid, and for the Horrat value calculation we therefore need an estimate of microbial concentration as measured by a mass ratio. According to Brock Biology of Microorganisms 13th ed., an optical density reading (OD) at 600nm (what we often call ‘absorbance’ in the BYU method) of about 0.1 equates to about 10^6 microbes per ml, and an OD600 of 1.0 is equal to about 10^10 organisms per ml. This range represents what is generally considered the optimal range for most spectrophotometers (0.1 -1.0 absorbance units). However, what are the equivalent masses for these bacterial counts? Bacterial mass changes throughout its life cycle. Some estimates of the mass of a bacterium are found on this website: http://hypertextbook.com/facts/2003/LouisSiu.shtml. Estimates vary greatly. E. coli is a gram negative rod and has been estimated to be as heavy as 1*10^-10g/bacterium, but I think 1*10^- 12 to 1*10^-14 g per bacterium is a good estimate for L. casei, which is also gram negative rod. I believe that the predicted relative standard deviation (PRSD) value of 21.333333333 in the Horrat equation has been the PRSD value since this method was first used at BYU. I do not

85 know exactly how this value was chosen, but it must have taken some careful work to calculate it, because the PRSD varies according to the concentration of analyte that is being measured, and this method uses serial dilutions; therefore there is a different amount of microbial growth in each row, and data from the same dilutions is not always used to calculate the final concentration. In order to measure an accurate mass ratio to which the horrat value can be applied in this method, something similar to the following would need to be done: turbidity measurements would have to be taken, followed by collecting the dry weight of the bacteria from the same wells (ie. Stop the bacterial growth, dehydrate the substance of the wells, subtract the media weight, and calculate an analysis-specific weight for the L. casei organisms). Perhaps Dr. Eitenmiller or his graduate students did this in order to calculate the value of 21.333333333. However it was obtained, this value seems to work well. All my calculations suggest that 21.333333333 is probably a very good PRSD value to use in the Horrat equation.

In conclusion, there are some definite limitations to inferring microbial concentration from optical density readings and the relationships often break down at high turbidities. However, the Horrat value is a good metric for determining if the data is valid, and the PRSD value of 21.333333333 seems to be well chosen (i.e. this value makes pretty good sense with the BYU method microbe concentrations, depending on the estimated mass of the microbes). It seems to work so we’ll keep it, but Dr. Eitenmiller and/or Jordan Chapman should be asked about how this value was obtained.

Automated Data Analysis Choosing Data Points for the Concentration Calculation I have found that arbitrarily choosing which wells to include in the concentration calculation often feels like data snooping, and trying different combinations of wells in order to get the best Horrat value is tedious and time-intensive. Therefore, with the help of Jacob Foist, I programmed an excel template that analyzes the concentration estimates from all possible combinations of three or more wells in the dilution series for each sample on the micro-titer plate. Each of these concentration values is then included in an overall weighted average, and the weighting is determined by the Horrat value calculated from the same wells.

Choosing the Data Points to Include in the Standard Curve Regression Another cause of inconsistent data analysis is the choice of how to run the standard curve regression. I have found that linear regression equations from various combinations of the seven dilutions can all have satisfactory linearity while returning very different concentration values for the same sample. However, I have come up with certain logical guidelines to take the subjectivity out of choosing the standard curve regression. First and foremost, it is more important that the growth in the most dilute wells be represented accurately by the regression equation than the growth in the most concentrated wells. This is partly because variation will be magnified many more fold in the dilute wells (the dilution factors used in the calculations are on the order of 32 and 64 as opposed to 1 and 2 in the most concentrated wells) and partly because the more diluted wells are usually the only ones that are measured in the linear range of the spectrophotometer. Therefore, the most accurate data analysis is achieved when the standard curve regression is built off of the data from the most dilute wells. In order to speed up the choice of the standard curve regression, I programmed the excel template to recommend a standard curve, and Jacob Foist created auto-populating graphs to visualize the regression

86 equation in the context of the entire plate. The recommended standard curve is not always the best choice, but it is usually a good choice; and reviewing the reasoning behind the automated recommendation will help remove some of the subjectivity from choosing a standard curve regression. The automatically recommended standard curve regression meets the following criteria:

1. The pooled concentration estimate from all seven dilutions is one of the best at accurately predicting the concentration of the standard stock solution. 2. The standard deviation of all 7 concentration estimates for the standard stock solution is low – meaning that the standard curve regression has the ability to predict a similar concentration for all the dilutions. 3. The automated combinatorics analysis of the standard stock solution returns a concentration similar to the theoretical value. 4. The variance in the slope of the linear regression equation is small (after adjusting for the number of points used in the regression by dividing by degrees of freedom). This is a measure of goodness of fit.

(The first two criterion are measures of the ability of the regression to predict the concentration of all 7 dilutions accurately, while the last two criterion measure compatibility with the automated data analysis and the goodness of fit of the linear regression. As a single metric, I found that criterion #3 is the least reliable for choosing a good regression, while #1 and #2 often return one of the best regression equations. However, I believe that my combined metric is better than any single criterion alone. The difficulty with measures of goodness of fit is that they are overly dependent on the number of points included in the regression. I believe that the variance of the slope is a better measure of the goodness of fit of the linear regression than the r- squared value, because adjusting variance for the number of points used in the regression makes this measure comparable between the different possible regressions. Having no way of averaging dissimilar metrics, I used multiplicative factors to adjust the face value of each individual metric in order to balance the final combined metric. After some experimentation, I settled on the following adjustment factors for the four above criterion: #1. No adjustment. #2. Factor of 0.5. The standard deviation is usually larger in magnitude than the other criteria and therefore has undue influence in the final metric. #3. No adjustment. This alone is not a reliable metric and its influence in the final combined metric must be limited. #4. Factor of 2. This metric is very useful for weeding out those regressions that seem to predict the standard stock solution well but are taken from a non-linear portion of the data.)

Weighted Averaging Based on the Horrat Value Considering information from multiple sources, the ideal range of horrat values for within-laboratory data can be thought of as between 0.5 and 0.7, with an acceptable range of between 0.3 and 1.3. When considering values outside of the ideal range, one might consider the distribution of Horrat values to be skewed by a factor of 3 toward the higher numbers (i.e. the acceptable range allows for values 0.2 lower than the ideal range, but up to 0.6 higher). Therefore, comparable values outside of the ideal 0.5-0.7 range would be within increasing intervals of 0.1 below for every 0.3 above (i.e. Horrat values of 0.4 and 1.0 would be equivalent). It was with this knowledge in mind that I weighted the average concentration from the combinatorics matrix in my automated data analysis spreadsheet.

87 The weighted average excel formula is explained as follows: I weighted the concentration values based on a skewed normal distribution of the Horrat value. Horrat values below 0.3 and above 1.3 are completely excluded from the final average concentration. I considered 0.55 to be the “mean” of the distribution (the ideal value), and adjusted the weights as though the values of 0.3 and 1.3 were 4 standard deviations from the mean. (Obviously the term “standard deviation” is somewhat of a misnomer here because it is not measured but is arbitrarily inferred from the range of acceptable Horrat values.) The weight values were taken from the normal distribution; concentrations with Horrat values between 0.4875 and 0.7375 (1 “Std Dev” from 0.55) are weighted at 0.68; Horrat values outside of this range down to 0.425 and up to 0.925 (2 Std Dev) are weighted at 0.2718; Horrat values outside of this range down to 0.3625 and up to 1.1125 (3 Std Dev) are weighted at 0.0428; Horrat values outside of this range down to 0.3 and up to 1.3 (the 4th Std Dev from the mean) are weighted at 0.0027). All the weights add to a total of 1.0.

Appendix E – Additional Notes and Discussion about the analysis Included here are different things learned about the analysis from the experience of myself and of previous graduate students. Some of this information came through conversations and emails with Dr. Eitenmiller and his graduate student Sungeun Cho. • The amount of inoculated depletion media that is added to the folic acid free media for plating is very important. Too much and the bacteria will overgrow and it will make it hard to get good data. o It is possible that the high background noise that randomly occurred in our plates over the past two years have been due to this, but I don’t think so. It is something we never checked. It might be worth trying a transfer of 15 ul instead of 20 ul. It is also possible that contamination occurs in this transfer step, which is why it is wise to sterilize the pipettes on the outside as well as the inside. • Use of only the linear portion of the standard curve is important. Use the fourth degree polynomial in the program to verify that the microbes grew predictably (the more you do it, the more you’ll get a feel for it) but don’t use it to determine the unknowns’ concentrations. It will become less accurate as the curve plateaus. o I have not made a practice of checking the data in the FLUOStar program using a fourth order polynomial, but it might be a good practice to pick up again. I don’t think it has been done frequently since the days of Jordan Chapman. I do know that the best standard curve linear regressions are those that best predict the most dilute wells, which are also the most linear. • The AOAC method says to add a buffer to the wells whereas Jordan Chapman started to use water based on Eitenmillers’ suggestion. We have achieved great results using the water, and to my knowledge buffer has never been used in the BYU lab for this part of the method. • You should store the filtered extract in dark refrigeration for 1 week in case you have to re-run the samples. However, analysis should be attempted within 1 or 2 days of the extraction because folate loss is likely to occur during storage. This is based off of Sungeun Cho’s advice. I have rerun sterile filtered samples after a week of refrigeration and the data still seemed to be good. I think that since the analysis is comparative, as long as degradation occurs equally in all the samples the data may still be valid. • As long as the bacteria are transferred in a sterile manner every 5-7 days, you should be able to use the same bacteria for an unlimited period of time.

88 • Knowing that samples run in duplicate should be the same concentration, the data from duplicate samples should analyze to return similar values but beware of data snooping. • The automated spreadsheet returns two different weighted average concentrations, along with the weighted averages of their standard deviations and their Horrat values. One includes every combination down to “seven choose four” and the other includes every combination of three or more dilutions/wells. Looking at both 7C4 and 7C3 concentration estimates can be informative. The best data will return very similar concentration estimates from 7C4 and 7C3, so if the two values are very disparate, then the dilutions pattern and growth pattern must not match very well. When choosing which • Previous methods had this advice for the dilution step (If samples are fortified, pipette 1 ml of each sample plus standard into 5 ml volumetric flask and fill to volume. For the WY standard pipette 2 ml into a 5 ml vol. flask and fill to volume.

89 Appendix D: Standard Curves

Vitamin A

Reference Standard Preparation: A standard solution of 1ng/μl retinyl palmitate in iso-octane was produced with serial dilutions as follows:

Exactly 100 mg of neat retinyl palmitate was weighed directly into a 100 ml volumetric flask using a calibrated analytical balance and the flask was filled to volume with iso-octane, making a 1 mg/ml solution.

Exactly 0.5 ml of the 1 mg/ml solution was pipetted into a 500 ml volumetric flask and filled to volume with iso-octane, making a 1 μg/ml solution.

Exactly 200 μl of the 1 μg/ml solution was pipetted into a 200 ml volumetric flask and filled to volume with iso-octane, making a 1 ng/ml solution.

100 mg pipetted 500 μg pipetted 200 ng injected 100 ml 0.5 ml 500 ml 200μl 200 ml on column

A sample of the 1 ng/ml standard solution was crimped in a 1 ml amber class HPLC vial and immediately analyzed with the following consecutive injection volumes as per the method of McMahon et al., Journal of AOAC International, Vol. 96, No. 5, 2013 (except that the flow rate was increased to 1 ml/min):

1 μl, 8 μl, 20 μl, 35 μl, 50 μl

Results and Standard Curve Calculations: It was apparent that some 13-cis isomer was present in the retinyl palmitate standard or was produced in the preparation of the standard solution. As it was impossible to know how much of the 13-cis isomer had been isomerized in sample storage and preparation and the 13-cis isomer was to be measured in the study samples and partially included in total bioactive vitamin A calculations, linear regression of the combined peak areas was performed using Microsoft Excel to yield a standard curve. No 11-cis isomer was detected.

90 Standard Curve Run Mar 4, 2015 13-cis Retinyl Sum of RP as uL ng Peak Area Palmitate Peak Peak Areas % of injected injected (mAU) Area (mAU) (mAU) Total 1 1 0.352503 4.79116 5.143663 93.15 8 8 1.20637 37.0285 38.23487 96.84 20 20 3.1173 94.14625 97.26355 96.79 35 35 5.94409 165.36656 171.3107 96.53 50 50 8.81336 236.88263 245.696 96.41 average: 96.65

Vitamin A Standard Curve 300

y = 4.9174x - 0.5866 250

R² = 1

200

150

100 Total Peak Area (mAU) Area Peak Total 50

0 0 10 20 30 40 50 60 Retinyl Palmitate Injected (ng)

To calculate vitamin A concentrations in study samples from HPLC peak areas, the inverse of the standard curve formula was used: y = 0.2034x + 0.1201

91 Inverse of Vitamin A Standard Curve 60

y = 0.2034x + 0.1201 50 R² = 1 40

10 Retinyl (ng) Palmitate

0 0 50 100 150 200 250 300 Absorbance Units (mAU)

HPLC Parameters: Column: Zorbax NH2 4.6x150 mm, 5 μm Mobile Phase A: Hexane Mobile Phase B: Hexane/methyl-t-butyl ether (75/25 v/v), add 3 mL methanol/L Elution Gradient: As per AOAC method 2012.10 Flow Rate: 1 ml/min Detection: Diode Array, Signal λ = 325,10 nm, Reference λ = 400, 100 nm

Certified Reference Material: A pharmaceutical secondary standard of retinyl palmitate (neat), purchased from Sigma Aldrich on January 28, 2015 was used for the standard curve calculations – traceable to USP 1602502. (Sigma-Aldrich order 3013579745 for Michael L Dunn, Order Date: 01/28/2015)

92 Vitamin C

Reference Standard Preparation: A standard solution of 5 μg/ml ascorbic acid in trichloroacetic acid (2%), dithiothreitol (0.2%), metaphosphoric acid (0.05%) solution (TCA/DTT/MPA solution) was produced with serial dilutions as follows:

Exactly 0.5 ml of the 1 mg/ml solution was pipetted into a 500 ml volumetric flask and filled to volume with iso-octane, making a 1 μg/ml solution.

Exactly 200 μl of the 1 μg/ml solution was pipetted into a 200 ml volumetric flask and filled to volume with iso-octane, making a 1 ng/ml solution.

100 mg pipetted 500 μg pipetted 200 ng injected 100 ml 0.5 ml 500 ml 200μl 200 ml on column

1 μl, 5 μl, 10 μl, 18 μl, 25 μl

Results and Standard Curve Calculations: Linear regression of the peak area was performed using Microsoft Excel to yield a standard curve.

Vitamin C Standard Curve, Run March 26, 2015

Concentration of Injection Vitamin C injected Standard Solution Volume on column (ug/ml) (ul) (ng) Peak Area 5 1 5 15.651 5 5 25 77.407 5 10 50 154.876 5 18 90 281.695 5 25 125 394.807

Vitamin C Standard Curve 450 y = 3.1593x - 1.5138 400 R² = 0.9999

350 300 250 200 150 Peak Area (mAU) Area Peak 100 50 0 0 25 50 75 100 125 150 Vitamin C injected (ng)

To calculate vitamin C concentrations in study samples from HPLC peak areas, the inverse of the standard curve formula was used for simplicity: y = 3.1593x + 1.5138

Inverse of Vitamin C Standard Curve 140 y = 0.3165x + 0.4839 120 R² = 0.9999

100

Vitamin C (ng) 40

0 0 100 200 300 400 500 Peak Area (mAU)

Certified Reference Material: As the vitamin C analytical standard, L-Ascorbic acid was purchased from Sigma Aldrich on 3/18/2015, Catalog #: PHR1008-2G, Batch P500008 (Carton# 995485695, Delivery # 0850091770, Sold to Account 49435877 – Michael L. Dunn).

95 Thiamine

After confirming very accurate returns with a thiamine HCL standard solution just going through

thiochrome oxidation step, the thiamine std curve already developed by Virginia West was used

for thiamine analysis. Formula: y=[([x]-0.2231)/17.78]*100, where x=peak area from the

sample injection and y=thiamine concentration in the sample injection (ng/ml).

Reference:

West VA. 2015. Stability of Selected B-Vitamins in Thermally-Treated Pinto Beans [master’s thesis]. [Provo, (UT)]. Brigham Young University.

Riboflavin

After confirming satisfactory analytical results with a riboflavin standard solution, the riboflavin std curve already developed by Victoria Scott was used for riboflavin analysis. Formula: y=[((11.189*x)-4.1201)/1000]*100, where x=peak area from the sample injection and y=riboflavin concentration in the sample injection (ng/ml).

Reference:

Scott VA. 2015. Stability of Whole Wheat Flour, Rolled Oats, and Brown Rice During Long- Term Storage and Preparation [master’s thesis]. [Provo, (UT)]. Brigham Young University.

Folate

A folic acid stock solution was prepared as per the method description, and an internal standard was run in each 96-well microtiter plate against which the concentration of folate in the samples was calculated.

96 Appendix E: BYU Sensory Panel Ballot (NSRL Soymilk)

ILLINOIS SOY MILK CONSUMER TEST (Aug 6, 2014) 2401

Name

Welcome to the Food Science Sensory Laboratory. Please make sure your parent signed a consent form before participating. During this panel if you don’t want to participate anymore you can leave at any time.

Today you will taste 2 different samples of SOY MILK. Please read all instructions and questions carefully. Before you get the sample, please answer some questions by marking the correct circles.

* How old are you? ○ 7 years old ○ 8 years old ○ 9 years old ○ 10 years old ○ 11years old ○ 12 years old ○ 13 years old ○ 14 years old ○ 15 years old ○ 16 years old ○ 17 years old

* Are you a boy or a girl? ○ Boy ○ Girl

* Have you had SOY MILK before? ○ Yes ○ I’m not sure ○ No

* What do you think about SOY MILK? ○ I like soy milk ○ I’m not sure if I like it or not ○ I don’t like soy milk

...Please turn the page and continue...

97 SOY MILK CONSUMER TEST

Now look for the set of lights to the right of the computer screen and press the red button next to the green “READY” light to show that you are ready to get your samples. Please be patient; they will arrive shortly.

If you need help during this test press the red button next to the “help” light or ask a helper in the booth area.

You will answer 8 different questions about each SOY MILK sample, so try to keep enough SOY MILK to answer all of the questions. If you do run out of sample, push the “help” button or ask a helper.

TASTE EACH SAMPLE in the order they are given to you from left to right. Take a bite of cracker and a sip of water between samples to refresh your sense of taste.

* OVERALL, what do you think about each sample? Sample #'s ______

Really good O O Good O O Just a little good O O Maybe good or maybe bad O O Just a little bad O O Bad O O Really bad O O

* What do you think about the FLAVOR of each sample?

Really good O O Good O O Just a little good O O Maybe good or maybe bad O O Just a little bad O O Bad O O Really bad O O

...Please turn the page and continue…

98 SOY MILK CONSUMER TEST

* What do you think about the COLOR of each sample? Sample #'s ______Really good O O Good O O Just a little good O O Maybe good or maybe bad O O Just a little bad O O Bad O O Really bad O O

* What do you think about the way each sample SMELLS? Sample #'s ______Really good O O Good O O Just a little good O O Maybe good or maybe bad O O Just a little bad O O Bad O O Really bad O O

* What do you think about the way each sample FEELS IN YOUR MOUTH? Sample #'s ______Really good O O Good O O Just a little good O O Maybe good or maybe bad O O Just a little bad O O Bad O O Really bad O O

...Please turn the page and continue…

SOY MILK CONSUMER TEST

* What do you think about the TASTE IN YOUR MOUTH AFTER YOU’VE SWALLOWED each sample? Sample #'s ______Really good O O Good O O Just a little good O O Maybe good or maybe bad O O Just a little bad O O Bad O O Really bad O O

* Please click on the number of the sample you liked the BEST.

______Liked BEST

* If you were served a full cup of this sample in your school cafeteria, would you DRINK all of it or not? Sample #'s ______Definitely would drink all of it O O Probably would drink all of it O O Maybe drink/maybe not drink all of it O O Probably would not drink all of it O O Definitely would not drink all of it O O

You are finished with the questions! You may finish the samples or just place the rest on the tray. Push the tray into the pass-through compartment and PRESS THE BUTTON BY THE RED “FINISHED” LIGHT.

You are now done! You may go back out of the tasting room to the receptionist at the front desk.

THANK YOU!

100 Appendix F: NSRL Raw Data

Vitamin A

Total vitamin A is reported in mcg/g soymilk.

NSRL vitamin A "Hot Hold" raw data for statistical analysis.

Arbitrary Small Analysis Sample Sample Unique % % % All- % Total Run # Code Fort Lot Treat Bottle Bottle 13-cis 11 -cis trans Bioactive IU RAE Vit_A 1 αF1c F α 1 3 1 5.61 0.00 94.39 98.49 2.993 0.898 1.643 1 αF1c F α 1 3 1 5.68 0.00 94.32 98.47 2.984 0.895 1.638 1 αF1c F α 1 3 1 5.79 0.00 94.21 98.44 3.029 0.909 1.663 1 αF2c F α 2 3 2 5.65 0.00 94.35 98.47 4.438 1.332 2.437 1 αF2c F α 2 3 2 5.80 0.00 94.20 98.43 4.456 1.337 2.446 1 αF2c F α 2 3 2 5.60 0.00 94.40 98.49 4.432 1.330 2.433 1 αF3c F α 3 3 3 6.06 0.00 93.94 98.36 4.351 1.305 2.389 1 αF3c F α 3 3 3 6.06 0.00 93.94 98.36 4.373 1.312 2.401 1 αF3c F α 3 3 3 6.02 1.31 92.67 97.51 4.399 1.320 2.415 2 αF1b F α 1 2 4 5.29 0.00 94.71 98.57 4.502 1.351 2.472 2 αF1b F α 1 2 4 5.34 0.00 94.66 98.56 4.493 1.348 2.466 2 αF1b F α 1 2 4 5.26 0.00 94.74 98.58 4.295 1.288 2.358 2 αF2b F α 2 2 5 5.53 0.00 94.47 98.51 5.631 1.689 3.091 2 αF2b F α 2 2 5 5.55 0.00 94.45 98.50 5.588 1.676 3.068 2 αF2b F α 2 2 5 5.56 0.00 94.44 98.50 5.449 1.635 2.992 2 αF3b F α 3 2 6 5.72 1.15 93.13 97.70 6.047 1.814 3.320 2 αF3b F α 3 2 6 5.63 1.08 93.29 97.77 6.055 1.817 3.324 2 αF3b F α 3 2 6 5.60 1.16 93.24 97.72 6.096 1.829 3.347 3 βF1c F β 1 3 7 6.11 0.00 93.89 98.35 3.283 0.985 1.802 3 βF1c F β 1 3 7 5.80 0.00 94.20 98.43 3.253 0.976 1.786 3 βF1c F β 1 3 7 6.21 0.00 93.79 98.32 3.309 0.993 1.817 3 βF2c F β 2 3 8 5.87 1.30 92.84 97.56 4.616 1.385 2.534 3 βF2c F β 2 3 8 5.95 1.46 92.59 97.43 4.504 1.351 2.473 3 βF2c F β 2 3 8 5.83 1.33 92.84 97.55 4.631 1.389 2.543 3 βF3c F β 3 3 9 6.13 1.54 92.33 97.33 4.374 1.312 2.401 3 βF3c F β 3 3 9 6.15 1.56 92.29 97.31 4.447 1.334 2.442 3 βF3c F β 3 3 9 6.17 1.53 92.30 97.33 4.333 1.300 2.379

101 4 βF1b F β 1 2 10 5.46 0.00 94.54 98.53 4.449 1.335 2.442 4 βF1b F β 1 2 10 5.51 0.00 94.49 98.51 4.420 1.326 2.427 4 βF1b F β 1 2 10 4.88 0.00 95.12 98.68 4.278 1.283 2.349 4 βF2b F β 2 2 11 5.80 0.00 94.20 98.43 4.976 1.493 2.732 4 βF2b F β 2 2 11 5.71 1.22 93.07 97.66 5.010 1.503 2.751 4 βF2b F β 2 2 11 5.60 0.00 94.40 98.49 4.894 1.468 2.687 4 βF3b F β 3 2 12 6.04 1.38 92.59 97.46 4.928 1.478 2.705 4 βF3b F β 3 2 12 5.94 1.41 92.65 97.47 4.850 1.455 2.663 4 βF3b F β 3 2 12 5.86 1.34 92.80 97.53 4.842 1.453 2.658

102 NSRL vitamin A "Cooling Method" raw data for statistical analysis.

Small Arbitrary sub- Analysis Sample Large Sample % 13- % 11- % All- % Total Run # Code Fort Lot Treat Bottle Bottle cis cis trans Bioactive IU RAE Vit_A 5 αFCA3 F α C A 3 6.12 0.00 93.88 98.35 3.536 1.061 1.941 5 αFCA3 F α C A 3 4.97 0.00 95.03 98.66 3.419 1.026 1.877 5 αFCA3 F α C A 3 5.86 0.00 94.14 98.42 3.497 1.049 1.920 6 αFCB3 F α C B 3 5.90 0.00 94.10 98.41 3.677 1.103 2.019 6 αFCB3 F α C B 3 5.96 0.00 94.04 98.39 3.662 1.099 2.010 6 αFCB3 F α C B 3 4.50 3.31 92.20 96.60 3.498 1.049 1.920 5 αFWA3 F α W A 3 5.98 0.00 94.02 98.39 3.566 1.070 1.958 5 αFWA3 F α W A 3 5.62 0.00 94.38 98.48 3.550 1.065 1.949 5 αFWA3 F α W A 3 6.00 0.00 94.00 98.38 3.564 1.069 1.957 6 αFWB3 F α W B 3 5.59 0.00 94.41 98.49 3.560 1.068 1.954 6 αFWB3 F α W B 3 5.93 0.00 94.07 98.40 3.589 1.077 1.970 6 αFWB3 F α W B 3 5.85 0.00 94.15 98.42 3.577 1.073 1.964 7 αFCA2 F α C A 2 5.49 1.07 93.44 97.81 7.413 2.224 4.070 7 αFCA2 F α C A 2 5.52 1.06 93.42 97.81 6.598 1.979 3.622 7 αFCA2 F α C A 2 5.31 0.95 93.74 97.94 7.422 2.227 4.075 7 αFCB2 F α C B 2 5.52 1.00 93.48 97.85 6.761 2.028 3.712 7 αFCB2 F α C B 2 5.47 0.97 93.56 97.88 6.726 2.018 3.693 7 αFCB2 F α C B 2 5.40 0.00 94.60 98.54 6.531 1.959 3.586 7 αFWA2 F α W A 2 5.60 1.29 93.11 97.64 4.796 1.439 2.633 7 αFWA2 F α W A 2 5.65 1.31 93.04 97.61 5.415 1.624 2.973 7 αFWA2 F α W A 2 5.30 1.01 93.69 97.90 5.424 1.627 2.978 7 αFWB2 F α W B 2 5.57 1.22 93.22 97.69 5.641 1.692 3.097 7 αFWB2 F α W B 2 5.35 1.08 93.57 97.84 5.694 1.708 3.126 7 αFWB2 F α W B 2 4.91 1.03 94.05 97.99 5.443 1.633 2.988 8 βFCA3 F β C A 3 5.87 1.39 92.74 97.50 4.703 1.411 2.582 9 βFCA3 F β C A 3 5.80 1.49 92.71 97.45 4.740 1.422 2.602 10 βFCA3 F β C A 3 5.92 1.34 92.73 97.51 4.842 1.452 2.658 8 βFCB3 F β C B 3 5.83 1.31 92.86 97.56 4.800 1.440 2.635 9 βFCB3 F β C B 3 5.87 1.40 92.72 97.49 4.677 1.403 2.567 10 βFCB3 F β C B 3 5.83 1.33 92.84 97.55 4.807 1.442 2.639 8 βFWA3 F β W A 3 5.69 1.04 93.27 97.78 4.842 1.453 2.658 9 βFWA3 F β W A 3 4.78 1.17 94.05 97.94 6.073 1.822 3.334 10 βFWA3 F β W A 3 5.96 1.32 92.71 97.52 5.023 1.507 2.758 8 βFWB3 F β W B 3 5.75 1.33 92.92 97.57 5.116 1.535 2.809 9 βFWB3 F β W B 3 5.79 1.38 92.84 97.53 5.016 1.505 2.754 10 βFWB3 F β W B 3 5.96 1.37 92.67 97.49 5.146 1.544 2.825

103 Small Arbitrary sub- Analysis Sample Large Sample % 13- % 11- % All- % Total Run # Code Fort Lot Treat Bottle Bottle cis cis trans Bioactive IU RAE Vit_A 11 βFCA2 F β C A 2 5.51 1.21 93.28 97.71 6.149 1.845 3.376 11 βFCA2 F β C A 2 5.55 1.20 93.25 97.71 5.977 1.793 3.281 11 βFCA2 F β C A 2 5.57 1.32 93.12 97.63 5.957 1.787 3.271 11 βFCB2 F β C B 2 5.67 1.30 93.03 97.61 5.548 1.664 3.046 11 βFCB2 F β C B 2 5.42 1.04 93.54 97.85 5.398 1.619 2.963 11 βFCB2 F β C B 2 5.47 1.01 93.51 97.85 5.324 1.597 2.923 11 βFWA2 F β W A 2 5.61 1.20 93.19 97.69 6.222 1.867 3.416 11 βFWA2 F β W A 2 5.66 1.28 93.06 97.63 6.296 1.889 3.456 11 βFWA2 F β W A 2 5.62 1.27 93.11 97.64 6.331 1.899 3.476 11 βFWB2 F β W B 2 5.37 1.02 93.61 97.88 6.188 1.856 3.397 11 βFWB2 F β W B 2 5.46 1.05 93.49 97.84 6.220 1.866 3.415 11 βFWB2 F β W B 2 4.58 1.23 94.19 97.95 5.916 1.775 3.248

104 NSRL vitamin A "Light Exposure" raw data for statistical analysis.

Arbitrary Sam- Small Analysis Sample pling Large sample % 13- % 11- % All- % Bio- Total Run # Code Fort Lot Treat Day Bottle Bottle cis cis trans active IU RAE Vit_A 1 αFC0L/DA3 F α O 0 A 3 5.70 0.00 94.30 98.46 5.926 1.778 3.253 1 αFC0L/DA3 F α O 0 A 3 5.88 1.01 93.11 97.74 6.137 1.841 3.369 2 αFC0L/DA3 F α O 0 A 3 4.91 2.06 93.03 97.32 5.824 1.747 3.197 4 αFC0L/DB3 F α O 0 B 3 5.44 0.75 93.81 98.04 7.459 2.238 4.095 5 αFC0L/DB3 F α O 0 B 3 5.71 0.90 93.39 97.86 7.529 2.259 4.133 6 αFC0L/DB3 F α O 0 B 3 5.31 0.00 94.69 98.57 7.074 2.122 3.884 7 βFC0L/DA3 F β O 0 A 3 5.63 0.84 93.53 97.93 6.817 2.045 3.743 8 βFC0L/DA3 F β O 0 A 3 5.77 1.06 93.16 97.74 6.738 2.021 3.699 9 βFC0L/DA3 F β O 0 A 3 6.00 1.20 92.81 97.59 6.751 2.025 3.706 10 βFC0L/DB3 F β O 0 B 3 5.61 1.04 93.36 97.80 7.179 2.154 3.941 11 βFC0L/DB3 F β O 0 B 3 5.81 1.20 92.99 97.64 7.164 2.149 3.933 12 βFC0L/DB3 F β O 0 B 3 5.98 1.36 92.66 97.49 6.344 1.903 3.483 12 βFC0L/DB3 F β O 0 B 3 5.88 1.01 93.11 97.75 6.164 1.849 3.384 1 αFC5DA3 F α D 5 A 3 6.04 1.35 92.61 97.48 5.690 1.707 3.124 2 αFC5DA3 F α D 5 A 3 6.02 1.29 92.69 97.53 5.733 1.720 3.147 2 αFC5DA3 F α D 5 A 3 6.04 1.36 92.60 97.47 5.762 1.729 3.163 4 αFC5DB3 F α D 5 B 3 5.73 0.94 93.32 97.83 7.058 2.117 3.875 5 αFC5DB3 F α D 5 B 3 5.82 0.94 93.24 97.81 6.998 2.099 3.842 6 αFC5DB3 F α D 5 B 3 5.78 1.06 93.16 97.74 6.510 1.953 3.574 7 βFC5DA3 F β D 5 A 3 6.10 1.28 92.62 97.51 5.669 1.701 3.112 8 βFC5DA3 F β D 5 A 3 6.05 1.25 92.69 97.54 5.874 1.762 3.225 9 βFC5DA3 F β D 5 A 3 6.04 1.35 92.61 97.48 5.793 1.738 3.180 10 βFC5DB3 F β D 5 B 3 5.95 1.25 92.80 97.57 6.699 2.010 3.678 11 βFC5DB3 F β D 5 B 3 5.99 1.37 92.64 97.48 6.428 1.928 3.529 12 βFC5DB3 F β D 5 B 3 5.96 1.35 92.69 97.50 6.430 1.929 3.530 1 αFC12DA3 F α D 12 A 3 5.98 1.10 92.93 97.66 5.963 1.789 3.274 2 αFC12DA3 F α D 12 A 3 5.99 1.11 92.91 97.65 5.994 1.798 3.291 2 αFC12DA3 F α D 12 A 3 6.03 1.11 92.86 97.64 6.073 1.822 3.334 4 αFC12DB3 F α D 12 B 3 5.78 1.42 92.80 97.50 7.432 2.229 4.080 5 αFC12DB3 F α D 12 B 3 5.81 1.32 92.86 97.56 7.321 2.196 4.019 6 αFC12DB3 F α D 12 B 3 5.87 1.38 92.75 97.50 7.158 2.147 3.930 7 βFC12DA3 F β D 12 A 3 6.22 1.28 92.50 97.48 6.267 1.880 3.441 8 βFC12DA3 F β D 12 A 3 6.24 1.34 92.41 97.43 6.448 1.934 3.540 9 βFC12DA3 F β D 12 A 3 6.35 1.37 92.28 97.38 6.166 1.850 3.385 10 βFC12DB3 F β D 12 B 3 6.18 1.43 92.39 97.39 7.057 2.117 3.874 11 βFC12DB3 F β D 12 B 3 6.19 1.46 92.35 97.37 7.270 2.181 3.991

105 Arbitrary Sam- Small Analysis Sample pling Large sample % 13- % 11- % All- % Bio- Total Run # Code Fort Lot Treat Day Bottle Bottle cis cis trans active IU RAE Vit_A 12 βFC12DB3 F β D 12 B 3 6.30 1.44 92.26 97.35 6.522 1.956 3.580 1 αFC5LA3 F α L 5 A 3 5.69 5.05 89.26 95.13 5.906 1.772 3.242 1 αFC5LA3 F α L 5 A 3 5.84 5.26 88.90 94.95 5.985 1.796 3.286 4 αFC5LB3 F α L 5 B 3 5.50 5.04 89.47 95.19 6.954 2.086 3.818 5 αFC5LB3 F α L 5 B 3 5.63 5.18 89.19 95.06 6.935 2.080 3.807 6 αFC5LB3 F α L 5 B 3 5.53 5.09 89.37 95.14 6.584 1.975 3.615 7 βFC5LA3 F β L 5 A 3 5.94 5.28 88.78 94.91 6.107 1.832 3.353 8 βFC5LA3 F β L 5 A 3 5.74 5.08 89.18 95.10 5.918 1.775 3.249 9 βFC5LA3 F β L 5 A 3 5.91 5.40 88.69 94.84 5.938 1.781 3.260 10 βFC5LB3 F β L 5 B 3 5.88 5.46 88.66 94.81 5.966 1.790 3.275 11 βFC5LB3 F β L 5 B 3 5.86 5.54 88.60 94.76 6.216 1.865 3.412 12 βFC5LB3 F β L 5 B 3 6.08 5.79 88.13 94.53 5.567 1.670 3.056 1 αFC12LA3 F α L 12 A 3 5.84 8.92 85.24 92.54 5.425 1.628 2.979 2 αFC12LA3 F α L 12 A 3 6.00 9.03 84.97 92.42 5.496 1.649 3.017 3 αFC12LA3 F α L 12 A 3 5.96 8.87 85.16 92.53 5.359 1.608 2.942 3 αFC12LA3 F α L 12 A 3 5.99 8.91 85.10 92.50 5.364 1.609 2.945 3 αFC12LA3 F α L 12 A 3 6.00 8.95 85.05 92.47 5.399 1.620 2.964 3 αFC12LA3 F α L 12 A 3 6.00 9.01 84.99 92.44 5.343 1.603 2.934 4 αFC12LB3 F α L 12 B 3 5.70 8.38 85.92 92.93 6.069 1.821 3.332 5 αFC12LB3 F α L 12 B 3 5.72 8.56 85.72 92.81 6.164 1.849 3.384 6 αFC12LB3 F α L 12 B 3 5.74 8.56 85.70 92.80 6.031 1.809 3.311 7 βFC12LA3 F β L 12 A 3 5.96 8.90 85.15 92.52 5.465 1.640 3.000 8 βFC12LA3 F β L 12 A 3 6.03 9.02 84.95 92.42 5.324 1.597 2.923 9 βFC12LA3 F β L 12 A 3 6.07 9.18 84.74 92.30 5.481 1.644 3.009 10 βFC12LB3 F β L 12 B 3 5.90 8.25 85.85 92.96 6.054 1.816 3.324 11 βFC12LB3 F β L 12 B 3 5.94 8.32 85.74 92.91 6.050 1.815 3.321 12 βFC12LB3 F β L 12 B 3 5.90 8.35 85.75 92.90 5.522 1.657 3.031

106 Vitamin C

Total vitamin C is reported in mg/100g soymilk. NSRL vitamin C "Hot Hold" raw data for statistical analysis.

Arbitrary Small Analysis Sample Sample Unique Run # Code Fortification Lot Treatment Bottle Bottle Vit_C 1 αF1c F α 1 3 1 12.279 1 αF1c F α 1 3 1 12.699 1 αF1c F α 1 3 1 12.834 1 αF2c F α 2 3 2 12.306 1 αF2c F α 2 3 2 12.379 1 αF2c F α 2 3 2 12.396 1 αF3c F α 3 3 3 11.990 1 αF3c F α 3 3 3 12.015 1 αF3c F α 3 3 3 12.004 2 αF1b F α 1 2 4 12.816 2 αF1b F α 1 2 4 12.740 2 αF1b F α 1 2 4 12.764 2 αF2b F α 2 2 5 12.211 2 αF2b F α 2 2 5 12.341 2 αF2b F α 2 2 5 12.286 2 αF3b F α 3 2 6 12.155 2 αF3b F α 3 2 6 12.004 2 αF3b F α 3 2 6 12.050 3 βF1c F β 1 3 7 11.859 3 βF1c F β 1 3 7 11.730 3 βF1c F β 1 3 7 11.896 3 βF2c F β 2 3 8 11.634 3 βF2c F β 2 3 8 11.712 3 βF2c F β 2 3 8 11.747 3 βF3c F β 3 3 9 11.260 3 βF3c F β 3 3 9 11.262 3 βF3c F β 3 3 9 11.428 4 βF1b F β 1 2 10 12.585 4 βF1b F β 1 2 10 12.602 4 βF1b F β 1 2 10 12.594 4 βF2b F β 2 2 11 11.822 4 βF2b F β 2 2 11 11.909 4 βF2b F β 2 2 11 11.846 4 βF3b F β 3 2 12 11.631 4 βF3b F β 3 2 12 11.769 4 βF3b F β 3 2 12 11.726 5 βF1a F β 1 1 13 12.719 5 βF1a F β 1 1 13 12.726 5 βF1a F β 1 1 13 12.594 5 βF2a F β 2 1 14 12.131 5 βF2a F β 2 1 14 11.847 5 βF3a F β 3 1 15 11.935 5 βF3a F β 3 1 15 11.840

107 NSRL vitamin C "Cooling Method" raw data for statistical analysis.

Small Arbitrary Sub- Analysis Sample Large Sample Run # Code Treatment Lot Bottle Bottle Vit_C 6 αFCA3 C α 1 3 11.377 6 αFCA3 C α 1 3 11.412 6 αFCA3 C α 1 3 11.461 7 αFCB3 C α 2 3 11.729 7 αFCB3 C α 2 3 11.690 7 αFCB3 C α 2 3 11.638 6 αFWA3 W α 3 3 11.682 6 αFWA3 W α 3 3 11.723 6 αFWA3 W α 3 3 11.657 7 αFWB3 W α 4 3 11.888 7 αFWB3 W α 4 3 11.996 7 αFWB3 W α 4 3 11.905 8 αFCA2 C α 5 2 12.789 8 αFCA2 C α 5 2 12.919 8 αFCA2 C α 5 2 12.119 8 αFCB2 C α 6 2 12.983 8 αFCB2 C α 6 2 13.080 8 αFCB2 C α 6 2 13.067 8 αFWA2 W α 7 2 13.135 8 αFWA2 W α 7 2 13.031 8 αFWA2 W α 7 2 12.999 8 αFWB2 W α 8 2 13.043 8 αFWB2 W α 8 2 13.077 8 αFWB2 W α 8 2 13.032 9 βFCA3 C β 9 3 9.907 10 βFCA3 C β 9 3 10.154 11 βFCA3 C β 9 3 10.925 9 βFCB3 C β 10 3 9.982 10 βFCB3 C β 10 3 10.019 11 βFCB3 C β 10 3 10.802 9 βFWA3 W β 11 3 9.442 10 βFWA3 W β 11 3 9.796 11 βFWA3 W β 11 3 10.571 9 βFWB3 W β 12 3 9.743 10 βFWB3 W β 12 3 9.927 11 βFWB3 W β 12 3 10.785

108 Small Arbitrary Sub- Analysis Sample Large Sample Run # Code Treatment Lot Bottle Bottle Vit_C 12 βFCA2 C β 13 2 10.389 12 βFCA2 C β 13 2 12.313 12 βFCA2 C β 13 2 12.211 12 βFCB2 C β 14 2 11.166 12 βFCB2 C β 14 2 11.885 12 βFCB2 C β 14 2 13.772 12 βFWA2 W β 15 2 12.701 12 βFWA2 W β 15 2 10.780 12 βFWA2 W β 15 2 12.348 12 βFWB2 W β 16 2 13.848 12 βFWB2 W β 16 2 15.099 12 βFWB2 W β 16 2 14.474

109 NSRL vitamin C "Light Exposure" raw data for statistical analysis.

Small Arbitrary Sub- Analysis Sample Large Sample Unique Day # Code Lot Bottle Bottle Bottle Treatment Day Vit_C 1 αFC0L/DA2 α A 2 1 O 0 11.935 1 αFC0L/DA2 α A 2 1 O 0 11.689 1 αFC0L/DA2 α A 2 1 O 0 12.422 2 αFC0L/DB3 α B 3 2 O 0 11.311 3 αFC0L/DB3 α B 3 2 O 0 11.419 4 αFC0L/DB3 α B 3 2 O 0 11.282 5 βFC0L/DA3 β A 3 3 O 0 10.497 6 βFC0L/DA3 β A 3 3 O 0 10.832 7 βFC0L/DA3 β A 3 3 O 0 10.303 8 βFC0L/DB3 β B 2 4 O 0 9.700 9 βFC0L/DB3 β B 2 4 O 0 10.102 10 βFC0L/DB3 β B 2 4 O 0 9.818 11 αFC5DA2 α A 2 5 D 5 12.090 11 αFC5DA2 α A 2 5 D 5 11.953 11 αFC5DA2 α A 2 5 D 5 10.213 12 αFC5DA2 α A 2 5 D 5 11.488 12 αFC5DA2 α A 2 5 D 5 11.344 13 αFC5DB3 α B 3 6 D 5 10.723 14 αFC5DB3 α B 3 6 D 5 10.771 15 αFC5DB3 α B 3 6 D 5 10.526 17 βFC5DA3 β A 3 7 D 5 9.318 16 βFC5DA3 β A 3 7 D 5 9.860 18 βFC5DA3 β A 3 7 D 5 8.985 19 βFC5DB3 β B 3 8 D 5 9.720 19 βFC5DB3 β B 3 8 D 5 9.713 20 βFC5DB3 β B 3 8 D 5 9.293 21 αFC12DA2 α A 2 9 D 12 11.414 21 αFC12DA2 α A 2 9 D 12 11.307 21 αFC12DA2 α A 2 9 D 12 9.889 22 αFC12DA2 α A 2 9 D 12 10.642 22 αFC12DA2 α A 2 9 D 12 10.668 23 αFC12DB3 α B 3 10 D 12 11.217 24 αFC12DB3 α B 3 10 D 12 10.939 25 αFC12DB3 α B 3 10 D 12 10.810 27 βFC12DA3 β A 3 11 D 12 9.554 26 βFC12DA3 β A 3 11 D 12 10.201

110 Small Arbitrary Sub- Analysis Sample Large Sample Unique Day # Code Lot Bottle Bottle Bottle Treatment Day Vit_C 28 βFC12DA3 β A 3 11 D 12 9.636 29 βFC12DB3 β B 3 12 D 12 8.968 29 βFC12DB3 β B 3 12 D 12 9.031 29 βFC12DB3 β B 3 12 D 12 9.206 30 αFC5LA2 α A 2 13 L 5 10.975 30 αFC5LA2 α A 2 13 L 5 10.879 30 αFC5LA2 α A 2 13 L 5 10.081 31 αFC5LB3 α B 3 14 L 5 10.368 31 αFC5LB3 α B 3 14 L 5 10.203 32 αFC5LB3 α B 3 14 L 5 10.207 34 βFC5LA3 β A 3 15 L 5 8.549 33 βFC5LA3 β A 3 15 L 5 8.899 35 βFC5LA3 β A 3 15 L 5 8.072 36 βFC5LB2 β B 2 16 L 5 7.023 36 βFC5LB2 β B 2 16 L 5 6.921 36 βFC5LB2 β B 2 16 L 5 7.252 37 αFC12LA2 α A 2 17 L 12 9.224 37 αFC12LA2 α A 2 17 L 12 10.688 37 αFC12LA2 α A 2 17 L 12 10.168 38 αFC12LB3 α B 3 18 L 12 9.980 39 αFC12LB3 α B 3 18 L 12 9.898 39 αFC12LB3 α B 3 18 L 12 9.793 41 βFC12LA3 β A 3 19 L 12 8.684 40 βFC12LA3 β A 3 19 L 12 9.220 42 βFC12LA3 β A 3 19 L 12 7.982 43 βFC12LB3 β B 3 20 L 12 8.252 43 βFC12LB3 β B 3 20 L 12 8.235 44 βFC12LB3 β B 3 20 L 12 7.712 45 βFC12LB2 β B 2 21 L 12 7.930 45 βFC12LB2 β B 2 21 L 12 8.019 45 βFC12LB2 β B 2 21 L 12 8.205

111 Thiamine

Total thiamine is reported in mg/100g soymilk.

NSRL thaimine "Hot Hold" raw data for statistical analysis.

Arbitrary Small Analysis Sample Sample Unique Run # Code Fortification Lot Treatment Bottle Bottle Thiamine 1 αU1c U α 1 3 1 0.050 1 αU1c U α 1 3 1 0.046 1 αU1c U α 1 3 1 0.050 4 αU2c U α 2 3 2 0.046 4 αU2c U α 2 3 2 0.047 4 αU2c U α 2 3 2 0.052 5 αU3c U α 3 3 3 0.057 5 αU3c U α 3 3 3 0.062 5 αU3c U α 3 3 3 0.050 2 αU1b U α 1 2 4 0.046 2 αU1b U α 1 2 4 0.046 3 αU1b U α 1 2 4 0.046 2 αU2b U α 2 2 5 0.046 3 αU2b U α 2 2 5 0.045 3 αU2b U α 2 2 5 0.044 2 αU3b U α 3 2 6 0.042 3 αU3b U α 3 2 6 0.045 3 αU3b U α 3 2 6 0.047 6 βU1c U β 1 3 7 0.043 6 βU1c U β 1 3 7 0.044 6 βU1c U β 1 3 7 0.047 9 βU2c U β 2 3 8 0.039 9 βU2c U β 2 3 8 0.037 9 βU2c U β 2 3 8 0.048 10 βU3c U β 3 3 9 0.042 10 βU3c U β 3 3 9 0.045 10 βU3c U β 3 3 9 0.046 7 βU1b U β 1 2 10 0.040 7 βU1b U β 1 2 10 0.038 8 βU1b U β 1 2 10 0.045 7 βU2b U β 2 2 11 0.037 7 βU2b U β 2 2 11 0.035 8 βU2b U β 2 2 11 0.045

112 Arbitrary Small Analysis Sample Sample Unique Run # Code Fortification Lot Treatment Bottle Bottle Thiamine 7 βU3b U β 3 2 12 0.040 7 βU3b U β 3 2 12 0.035 8 βU3b U β 3 2 12 0.037 1 αF1c F α 1 3 13 0.439 1 αF1c F α 1 3 13 0.338 1 αF1c F α 1 3 13 0.353 4 αF2c F α 2 3 14 0.315 4 αF2c F α 2 3 14 0.359 5 αF3c F α 3 3 15 0.340 5 αF3c F α 3 3 15 0.374 5 αF3c F α 3 3 15 0.398 2 αF1b F α 1 2 16 0.310 2 αF1b F α 1 2 16 0.297 3 αF1b F α 1 2 16 0.309 2 αF2b F α 2 2 17 0.316 2 αF2b F α 2 2 17 0.305 3 αF2b F α 2 2 17 0.334 2 αF3b F α 3 2 18 0.336 3 αF3b F α 3 2 18 0.318 3 αF3b F α 3 2 18 0.315 6 βF1c F β 1 3 19 0.302 6 βF1c F β 1 3 19 0.348 9 βF2c F β 2 3 20 0.298 9 βF2c F β 2 3 20 0.292 9 βF2c F β 2 3 20 0.295 10 βF3c F β 3 3 21 0.303 10 βF3c F β 3 3 21 0.340 10 βF3c F β 3 3 21 0.293 7 βF1b F β 1 2 22 0.293 7 βF1b F β 1 2 22 0.287 8 βF1b F β 1 2 22 0.287 7 βF2b F β 2 2 23 0.311 7 βF2b F β 2 2 23 0.283 8 βF2b F β 2 2 23 0.304 7 βF3b F β 3 2 24 0.296 7 βF3b F β 3 2 24 0.288 8 βF3b F β 3 2 24 0.334

113

NSRL thiamine "Cooling Method" raw data for statistical analysis.

Arbitrary Small Analysis Sample Large Sample Run # Code Fortification Lot Treatment Bottle Bottle Thiamine 13 αUCA3 U α C 1 3 0.043 13 αUCA3 U α C 1 3 0.045 13 αUCA3 U α C 1 3 0.043 14 αUCB3 U α C 2 3 0.050 14 αUCB3 U α C 2 3 0.049 14 αUCB3 U α C 2 3 0.043 13 αUWA3 U α W 3 3 0.043 13 αUWA3 U α W 3 3 0.042 13 αUWA3 U α W 3 3 0.044 14 αUWB3 U α W 4 3 0.043 14 αUWB3 U α W 4 3 0.040 14 αUWB3 U α W 4 3 0.062 15 αFCA3 F α C 5 3 0.363 15 αFCA3 F α C 5 3 0.313 15 αFCA3 F α C 5 3 0.329 16 αFCB3 F α C 6 3 0.395 16 αFCB3 F α C 6 3 0.371 16 αFCB3 F α C 6 3 0.347 15 αFWA3 F α W 7 3 0.418 15 αFWA3 F α W 7 3 0.321 15 αFWA3 F α W 7 3 0.315 16 αFWB3 F α W 8 3 0.344 17 βUCA3 U β C 9 3 0.039 18 βUCA3 U β C 9 3 0.069 19 βUCA3 U β C 9 3 0.079 22 βUCA2 U β C 9 2 0.047 23 βUCA2 U β C 9 2 0.035 24 βUCA2 U β C 9 2 0.035 17 βUCB3 U β C 10 3 0.042 20 βUCB3 U β C 10 3 0.049 21 βUCB3 U β C 10 3 0.055 22 βUCB2 U β C 10 2 0.042 23 βUCB2 U β C 10 2 0.033 24 βUCB2 U β C 10 2 0.034 17 βUWA3 U β W 11 3 0.040 18 βUWA3 U β W 11 3 0.069 19 βUWA3 U β W 11 3 0.049 22 βUWA2 U β W 11 2 0.035 23 βUWA2 U β W 11 2 0.035

114 Arbitrary Small Analysis Sample Large Sample Run # Code Fortification Lot Treatment Bottle Bottle Thiamine 24 βUWA2 U β W 11 2 0.036 17 βUWB3 U β W 12 3 0.041 20 βUWB3 U β W 12 3 0.043 21 βUWB3 U β W 12 3 0.052 22 βUWB2 U β W 12 2 0.039 23 βUWB2 U β W 12 2 0.032 24 βUWB2 U β W 12 2 0.034 17 βFCA3 F β C 13 3 0.317 18 βFCA3 F β C 13 3 0.314 19 βFCA3 F β C 13 3 0.400 22 βFCA2 F β C 13 2 0.280 23 βFCA2 F β C 13 2 0.278 24 βFCA2 F β C 13 2 0.339 18 βFCB3 F β C 14 3 0.359 20 βFCB3 F β C 14 3 0.350 21 βFCB3 F β C 14 3 0.390 22 βFCB2 F β C 14 2 0.340 23 βFCB2 F β C 14 2 0.283 24 βFCB2 F β C 14 2 0.267 17 βFWA3 F β W 15 3 0.305 18 βFWA3 F β W 15 3 0.323 19 βFWA3 F β W 15 3 0.350 22 βFWA2 F β W 15 2 0.292 23 βFWA2 F β W 15 2 0.272 24 βFWA2 F β W 15 2 0.275 18 βFWB3 F β W 16 3 0.342 20 βFWB3 F β W 16 3 0.420 21 βFWB3 F β W 16 3 0.408 22 βFWB2 F β W 16 2 0.287 23 βFWB2 F β W 16 2 0.272 24 βFWB2 F β W 16 2 0.287

115 NSRL thiamine "Light Exposure" raw data for statistical analysis.

Small Sample Large Sample Unique Sampling Analyst Code Fort Lot Bottle Bottle Bottle Treat Day Thiamine TM αUC0L/DA3 U α A 3 1 O 0 0.051 FB αUC0L/DA3 U α A 3 1 O 0 0.045 AH αUC0L/DA3 U α A 3 1 O 0 0.059 MC αUC0L/DB3 U α B 3 2 O 0 0.055 GL/MC αUC0L/DB3 U α B 3 2 O 0 0.048 NC αUC0L/DB3 U α B 3 2 O 0 0.046 NC αUC0L/DB3 U α B 3 2 O 0 0.046 MC αUC0L/DB3 U α B 3 2 O 0 0.048 TM αFC0L/DA3 F α A 3 3 O 0 0.319 FB αFC0L/DA3 F α A 3 3 O 0 0.303 AH αFC0L/DA3 F α A 3 3 O 0 0.342 GL αFC0L/DA2 F α A 2 3 O 0 0.303 MC αFC0L/DA2 F α A 2 3 O 0 0.384 MC αFC0L/DB3 F α B 3 4 O 0 0.451 GL/MC αFC0L/DB3 F α B 3 4 O 0 0.310 NC αFC0L/DB3 F α B 3 4 O 0 0.310 NC αFC0L/DB3 F α B 3 4 O 0 0.328 MC αFC0L/DB3 F α B 3 4 O 0 0.323 NC βUC0L/DA3 U β A 3 5 O 0 0.038 MC βUC0L/DA3 U β A 3 5 O 0 0.034 MC βUC0L/DA3 U β A 3 5 O 0 0.036 MC βUC0L/DB3 U β B 3 6 O 0 0.041 GL/MC βUC0L/DB3 U β B 3 6 O 0 0.038 MC βUC0L/DB3 U β B 3 6 O 0 0.038 NC βFC0L/DA3 F β A 3 7 O 0 0.322 MC βFC0L/DA3 F β A 3 7 O 0 0.285 MC βFC0L/DA3 F β A 3 7 O 0 0.295 MC βFC0L/DB3 F β B 3 8 O 0 0.307 GL/MC βFC0L/DB3 F β B 3 8 O 0 0.289 βFC0L/DB3 F β B 3 8 O 0 0.310 FB αUC5DA3 U α A 3 9 D 5 0.047 EL αUC5DA3 U α A 3 9 D 5 0.048 FB αUC5DA3 U α A 3 9 D 5 0.043 MC αUC5DB3 U α B 3 10 D 5 0.053 NC αUC5DB3 U α B 3 10 D 5 0.046 MC αUC5DB4 U α B 3 10 D 5 0.049

116 Small Sample Large Sample Unique Sampling Analyst Code Fort Lot Bottle Bottle Bottle Treat Day Thiamine NC/MJ αUC5DB3 U α B 3 10 D 5 0.048 FB αFC5DA3 F α A 3 11 D 5 0.315 EL αFC5DA3 F α A 3 11 D 5 0.341 AH αFC5DA3 F α A 3 11 D 5 0.371 MC αFC5DA2 F α A 2 11 D 5 0.379 MC αFC5DA2 F α A 2 11 D 5 0.356 MC αFC5DB3 F α B 3 12 D 5 0.431 NC αFC5DB3 F α B 3 12 D 5 0.390 MC αFC5DB3 F α B 3 12 D 5 0.309 NC αFC5DB3 F α B 3 12 D 5 0.372 NC/MJ αFC5DB3 F α B 3 12 D 5 0.334 GL/MC βUC5DA3 U β A 3 13 D 5 0.036 MC βUC5DA3 U β A 3 13 D 5 0.035 NC βUC5DA3 U β A 3 13 D 5 0.035 MC βUC5DB3 U β B 3 14 D 5 0.046 MC βUC5DB3 U β B 3 14 D 5 0.038 GL βUC5DB3 U β B 3 14 D 5 0.038 GL/MC βFC5DA3 F β A 3 15 D 5 0.277 MC βFC5DA3 F β A 3 15 D 5 0.283 NC βFC5DA3 F β A 3 15 D 5 0.290 MC βFC5DB3 F β B 3 16 D 5 0.289 MC βFC5DB3 F β B 3 16 D 5 0.306 GL βFC5DB3 F β B 3 16 D 5 0.291 EL αUC12DA3 U α A 3 17 D 12 0.046 FB αUC12DA3 U α A 3 17 D 12 0.043 GL/MC αUC12DB3 U α B 3 18 D 12 0.046 NC αUC12DB3 U α B 3 18 D 12 0.046 NC αUC12DB3 U α B 3 18 D 12 0.046 MC αUC12DB3 U α B 3 18 D 12 0.046 FB αFC12DA3 F α A 3 19 D 12 0.303 EL αFC12DA3 F α A 3 19 D 12 0.321 AH αFC12DA3 F α A 3 19 D 12 0.353 MC αFC12DA2 F α A 2 19 D 12 0.283 GL αFC12DA2 F α A 2 19 D 12 0.271 MC αFC12DA2 F α A 2 19 D 12 0.321 GL/MC αFC12DB3 F α B 3 20 D 12 0.308 NC αFC12DB3 F α B 3 20 D 12 0.309 MC αFC12DB3 F α B 3 20 D 12 0.327

117 Small Sample Large Sample Unique Sampling Analyst Code Fort Lot Bottle Bottle Bottle Treat Day Thiamine NC αFC12DB3 F α B 3 20 D 12 0.346 NC βUC12DA3 U β A 3 21 D 12 0.040 MC βUC12DA3 U β A 3 21 D 12 0.037 MC βUC12DA3 U β A 3 21 D 12 0.035 GL/MC βUC12DB3 U β B 3 22 D 12 0.036 MC βUC12DB3 U β B 3 22 D 12 0.035 GL βUC12DB3 U β B 3 22 D 12 0.036 NC βFC12DA3 F β A 3 23 D 12 0.324 GL/MC βFC12DA3 F β A 3 23 D 12 0.277 NC βFC12DA3 F β A 3 23 D 12 0.280 GL/MC βFC12DB3 F β B 3 24 D 12 0.297 MC βFC12DB3 F β B 3 24 D 12 0.300 GL βFC12DB3 F β B 3 24 D 12 0.324 TM αUC5LA3 U α A 3 25 L 5 0.059 FB αUC5LA3 U α A 3 25 L 5 0.047 AH αUC5LA3 U α A 3 25 L 5 0.060 FB αUC5LA3 U α A 3 25 L 5 0.043 MC αUC5LB3 U α B 3 26 L 5 0.051 MC αUC5LB3 U α B 3 26 L 5 0.045 NC αUC5LB3 U α B 3 26 L 5 0.047 MC αUC5LB3 U α B 3 26 L 5 0.049 FB αFC5LA3 F α A 3 27 L 5 0.335 MC αFC5LA2 F α A 2 27 L 5 0.293 GL αFC5LA2 F α A 2 27 L 5 0.282 MC αFC5LA2 F α A 2 27 L 5 0.397 MC αFC5LB3 F α B 3 28 L 5 0.409 NC αFC5LB3 F α B 3 28 L 5 0.304 MC αFC5LB3 F α B 3 28 L 5 0.342 NC αFC5LB3 F α B 3 28 L 5 0.332 NC/MJ αFC5LB3 F α B 3 28 L 5 0.333 GL/MC βUC5LA3 U β A 3 29 L 5 0.034 MC βUC5LA3 U β A 3 29 L 5 0.036 NC βUC5LA3 U β A 3 29 L 5 0.036 MC βUC5LB3 U β B 3 30 L 5 0.044 MC βUC5LB3 U β B 3 30 L 5 0.037 GL βUC5LB3 U β B 3 30 L 5 0.038 GL/MC βFC5LA3 F β A 3 31 L 5 0.287

118 Small Sample Large Sample Unique Sampling Analyst Code Fort Lot Bottle Bottle Bottle Treat Day Thiamine MC βFC5LA3 F β A 3 31 L 5 0.277 NC βFC5LA3 F β A 3 31 L 5 0.306 MC βFC5LB3 F β B 3 32 L 5 0.285 MC βFC5LB3 F β B 3 32 L 5 0.303 GL βFC5LB3 F β B 3 32 L 5 0.291 FB αUC12LA3 U α A 3 33 L 12 0.047 AH αUC12LA3 U α A 3 33 L 12 0.060 FB αUC12LA3 U α A 3 33 L 12 0.044 GL/MC αUC12LB3 U α B 3 34 L 12 0.045 NC αUC12LB3 U α B 3 34 L 12 0.015 NC αUC12LB3 U α B 3 34 L 12 0.048 MC αUC12LB3 U α B 3 34 L 12 0.048 TM αFC12LA3 F α A 3 35 L 12 0.321 FB αFC12LA3 F α A 3 35 L 12 0.320 AH αFC12LA3 F α A 3 35 L 12 0.348 MC αFC12LA2 F α A 2 35 L 12 0.317 GL αFC12LA2 F α A 2 35 L 12 0.264 MC αFC12LA2 F α A 2 35 L 12 0.301 GL/MC αFC12LB3 F α B 3 36 L 12 0.351 NC αFC12LB3 F α B 3 36 L 12 0.328 MC αFC12LB3 F α B 3 36 L 12 0.317 NC βUC12LA3 U β A 3 37 L 12 0.040 MC βUC12LA3 U β A 3 37 L 12 0.036 MC βUC12LA3 U β A 3 37 L 12 0.036 GL/MC βUC12LB3 U β B 3 38 L 12 0.038 MC βUC12LB3 U β B 3 38 L 12 0.039 GL βUC12LB3 U β B 3 38 L 12 0.036 NC βFC12LA3 F β A 3 39 L 12 0.320 GL/MC βFC12LA3 F β A 3 39 L 12 0.279 NC βFC12LA3 F β A 3 39 L 12 0.283 GL/MC βFC12LB3 F β B 3 40 L 12 0.288 MC βFC12LB3 F β B 3 40 L 12 0.282 GL βFC12LB3 F β B 3 40 L 12 0.277

119 Riboflavin

Total riboflavin is reported in mg/100g soymilk. NSRL riboflavin "Hot Hold" raw data for statistical analysis. Arbitrary Small Analysis Sample Sample Unique Run # Code Fortification Lot Treatment Bottle Bottle Riboflavin 1 αU1c U α 1 3 1 0.022 1 αU1c U α 1 3 1 0.019 1 αU1c U α 1 3 1 0.017 4 αU2c U α 2 3 2 0.020 4 αU2c U α 2 3 2 0.024 4 αU2c U α 2 3 2 0.016 5 αU3c U α 3 3 3 0.020 5 αU3c U α 3 3 3 0.029 5 αU3c U α 3 3 3 0.020 2 αU1b U α 1 2 4 0.019 2 αU1b U α 1 2 4 0.018 3 αU1b U α 1 2 4 0.017 2 αU2b U α 2 2 5 0.019 3 αU2b U α 2 2 5 0.021 3 αU2b U α 2 2 5 0.020 2 αU3b U α 3 2 6 0.018 3 αU3b U α 3 2 6 0.013 3 αU3b U α 3 2 6 0.020 6 βU1c U β 1 3 7 0.022 6 βU1c U β 1 3 7 0.022 6 βU1c U β 1 3 7 0.023 9 βU2c U β 2 3 8 0.016 9 βU2c U β 2 3 8 0.017 9 βU2c U β 2 3 8 0.016 10 βU3c U β 3 3 9 0.022 10 βU3c U β 3 3 9 0.030 10 βU3c U β 3 3 9 0.023 7 βU1b U β 1 2 10 0.010 7 βU1b U β 1 2 10 0.008 8 βU1b U β 1 2 10 0.016 7 βU2b U β 2 2 11 0.014 7 βU2b U β 2 2 11 0.011 8 βU2b U β 2 2 11 0.014 7 βU3b U β 3 2 12 0.011

120 Arbitrary Small Analysis Sample Sample Unique Run # Code Fortification Lot Treatment Bottle Bottle Riboflavin 7 βU3b U β 3 2 12 0.011 8 βU3b U β 3 2 12 0.011 1 αF1c F α 1 3 13 0.223 1 αF1c F α 1 3 13 0.276 1 αF1c F α 1 3 13 0.243 4 αF2c F α 2 3 14 0.206 4 αF2c F α 2 3 14 0.243 4 αF2c F α 2 3 14 0.231 5 αF3c F α 3 3 15 0.244 5 αF3c F α 3 3 15 0.232 5 αF3c F α 3 3 15 0.247 2 αF1b F α 1 2 16 0.217 2 αF1b F α 1 2 16 0.196 3 αF1b F α 1 2 16 0.230 2 αF2b F α 2 2 17 0.219 2 αF2b F α 2 2 17 0.235 3 αF2b F α 2 2 17 0.220 2 αF3b F α 3 2 18 0.216 3 αF3b F α 3 2 18 0.218 3 αF3b F α 3 2 18 0.221 6 βF1c F β 1 3 19 0.236 6 βF1c F β 1 3 19 0.243 9 βF2c F β 2 3 20 0.225 9 βF2c F β 2 3 20 0.218 9 βF2c F β 2 3 20 0.201 9 βF2c F β 2 3 20 0.312 10 βF3c F β 3 3 21 0.231 10 βF3c F β 3 3 21 0.239 10 βF3c F β 3 3 21 0.206 7 βF1b F β 1 2 22 0.182 7 βF1b F β 1 2 22 0.188 8 βF1b F β 1 2 22 0.185 7 βF2b F β 2 2 23 0.196 7 βF2b F β 2 2 23 0.192 8 βF2b F β 2 2 23 0.179 7 βF3b F β 3 2 24 0.181 7 βF3b F β 3 2 24 0.182 8 βF3b F β 3 2 24 0.194

121 NSRL riboflavin "Cooling Method" raw data for statistical analysis.

Arbitrary Small Analysis Sample Large Sample Run # Code Fortification Lot Treatment Bottle Bottle Riboflavin 13 αUCA3 U α C 1 3 0.019 13 αUCA3 U α C 1 3 0.018 13 αUCA3 U α C 1 3 0.019 14 αUCB3 U α C 2 3 0.018 14 αUCB3 U α C 2 3 0.019 14 αUCB3 U α C 2 3 0.021 13 αUWA3 U α W 3 3 0.018 13 αUWA3 U α W 3 3 0.019 13 αUWA3 U α W 3 3 0.021 14 αUWB3 U α W 4 3 0.017 14 αUWB3 U α W 4 3 0.018 14 αUWB3 U α W 4 3 0.017 15 αFCA3 F α C 5 3 0.222 15 αFCA3 F α C 5 3 0.217 15 αFCA3 F α C 5 3 0.250 16 αFCB3 F α C 6 3 0.229 16 αFCB3 F α C 6 3 0.251 16 αFCB3 F α C 6 3 0.247 15 αFWA3 F α W 7 3 0.216 15 αFWA3 F α W 7 3 0.220 15 αFWA3 F α W 7 3 0.219 16 αFWB3 F α W 8 3 0.254 16 αFWB3 F α W 8 3 0.281 16 αFWB3 F α W 8 3 0.259 17 βUCA3 U β C 9 3 0.019 18 βUCA3 U β C 9 3 0.046 19 βUCA3 U β C 9 3 0.021 22 βUCA2 U β C 9 2 0.018 23 βUCA2 U β C 9 2 0.011 24 βUCA2 U β C 9 2 0.011 17 βUCB3 U β C 10 3 0.016 20 βUCB3 U β C 10 3 0.022 21 βUCB3 U β C 10 3 0.029 22 βUCB2 U β C 10 2 0.014 23 βUCB2 U β C 10 2 0.013 24 βUCB2 U β C 10 2 0.014

122 Arbitrary Small Analysis Sample Large Sample Run # Code Fortification Lot Treatment Bottle Bottle Riboflavin 17 βUWA3 U β W 11 3 0.047 18 βUWA3 U β W 11 3 0.020 19 βUWA3 U β W 11 3 0.019 22 βUWA2 U β W 11 2 0.012 23 βUWA2 U β W 11 2 0.011 24 βUWA2 U β W 11 2 0.014 17 βUWB3 U β W 12 3 0.018 20 βUWB3 U β W 12 3 0.018 21 βUWB3 U β W 12 3 0.020 22 βUWB2 U β W 12 2 0.012 23 βUWB2 U β W 12 2 0.011 24 βUWB2 U β W 12 2 0.013 17 βFCA3 F β C 13 3 0.203 18 βFCA3 F β C 13 3 0.237 19 βFCA3 F β C 13 3 0.302 22 βFCA2 F β C 13 2 0.180 23 βFCA2 F β C 13 2 0.181 24 βFCA2 F β C 13 2 0.181 18 βFCB3 F β C 14 3 0.228 20 βFCB3 F β C 14 3 0.231 21 βFCB3 F β C 14 3 0.280 22 βFCB2 F β C 14 2 0.209 23 βFCB2 F β C 14 2 0.184 24 βFCB2 F β C 14 2 0.181 17 βFWA3 F β W 15 3 0.302 18 βFWA3 F β W 15 3 0.224 19 βFWA3 F β W 15 3 0.249 22 βFWA2 F β W 15 2 0.184 23 βFWA2 F β W 15 2 0.178 24 βFWA2 F β W 15 2 0.171 18 βFWB3 F β W 16 3 0.230 20 βFWB3 F β W 16 3 0.211 21 βFWB3 F β W 16 3 0.314 22 βFWB2 F β W 16 2 0.181 23 βFWB2 F β W 16 2 0.176 24 βFWB2 F β W 16 2 0.191

123 NSRL riboflavin "Light Exposure" raw data for statistical analysis.

Sub- Sample Large Sample Unique Sampling Analyst Code Fortification Lot Bottle Bottle Bottle Treatment Day Riboflavin TM αUC0L/DA3 U α A 3 1 O 0 0.021 FB αUC0L/DA3 U α A 3 1 O 0 0.021 AH αUC0L/DA3 U α A 3 1 O 0 0.036 FB αUC0L/DA3 U α A 3 1 O 0 0.021 MC αUC0L/DB3 U α B 3 2 O 0 0.027 GL/MC αUC0L/DB3 U α B 3 2 O 0 0.017 NC αUC0L/DB3 U α B 3 2 O 0 0.018 NC αUC0L/DB3 U α B 3 2 O 0 0.020 MC αUC0L/DB3 U α B 3 2 O 0 0.018 TM αFC0L/DA3 F α A 3 3 O 0 0.238 FB αFC0L/DA3 F α A 3 3 O 0 0.220 AH αFC0L/DA3 F α A 3 3 O 0 0.238 MC αFC0L/DA2 F α A 2 3 O 0 0.342 GL αFC0L/DA2 F α A 2 3 O 0 0.182 MC αFC0L/DA2 F α A 2 3 O 0 0.214 MC αFC0L/DB3 F α B 3 4 O 0 0.322 GL/MC αFC0L/DB3 F α B 3 4 O 0 0.211 NC αFC0L/DB3 F α B 3 4 O 0 0.222 NC αFC0L/DB3 F α B 3 4 O 0 0.254 MC αFC0L/DB3 F α B 3 4 O 0 0.233 NC βUC0L/DA3 U β A 3 5 O 0 0.018 MC βUC0L/DA3 U β A 3 5 O 0 0.018 MC βUC0L/DA3 U β A 3 5 O 0 0.019 MC βUC0L/DB3 U β B 3 6 O 0 0.017 GL/MC βUC0L/DB3 U β B 3 6 O 0 0.017 MC βUC0L/DB3 U β B 3 6 O 0 0.016 NC βFC0L/DA3 F β A 3 7 O 0 0.205 MC βFC0L/DA3 F β A 3 7 O 0 0.204 MC βFC0L/DA3 F β A 3 7 O 0 0.228 MC βFC0L/DB3 F β B 3 8 O 0 0.207 GL/MC βFC0L/DB3 F β B 3 8 O 0 0.208 MC βFC0L/DB3 F β B 3 8 O 0 0.213 FB αUC5DA3 U α A 3 9 D 5 0.021 EL αUC5DA3 U α A 3 9 D 5 0.019 FB αUC5DA3 U α A 3 9 D 5 0.018 MC αUC5DB3 U α B 3 10 D 5 0.023

124 Sub- Sample Large Sample Unique Sampling Analyst Code Fortification Lot Bottle Bottle Bottle Treatment Day Riboflavin NC αUC5DB3 U α B 3 10 D 5 0.017 MC αUC5DB3 U α B 3 10 D 5 0.019 NC/MJ αUC5DB3 U α B 3 10 D 5 0.018 MJ αUC5DB3 U α B 3 10 D 5 0.018 FB αFC5DA3 F α A 3 11 D 5 0.242 EL αFC5DA3 F α A 3 11 D 5 0.237 AH αFC5DA3 F α A 3 11 D 5 0.277 MC αFC5DA2 F α A 2 11 D 5 0.198 MC αFC5DA2 F α A 2 11 D 5 0.213 MC αFC5DB3 F α B 3 12 D 5 0.303 NC αFC5DB3 F α B 3 12 D 5 0.218 MC αFC5DB3 F α B 3 12 D 5 0.221 NC αFC5DB3 F α B 3 12 D 5 0.229 αFC5DB3 F α B 3 12 D 5 0.219 NC/MJ αFC5DB3 F α B 3 12 D 5 0.206 GL/MC βUC5DA3 U β A 3 13 D 5 0.018 MC βUC5DA3 U β A 3 13 D 5 0.017 NC βUC5DA3 U β A 3 13 D 5 0.019 MC βUC5DB3 U β B 3 14 D 5 0.023 MC βUC5DB3 U β B 3 14 D 5 0.016 GL βUC5DB3 U β B 3 14 D 5 0.017 GL/MC βFC5DA3 F β A 3 15 D 5 0.212 MC βFC5DA3 F β A 3 15 D 5 0.209 NC βFC5DA3 F β A 3 15 D 5 0.220 MC βFC5DB3 F β B 3 16 D 5 0.216 MC βFC5DB3 F β B 3 16 D 5 0.216 GL βFC5DB3 F β B 3 16 D 5 0.215 EL αUC12DA3 U α A 3 17 D 12 0.018 FB αUC12DA3 U α A 3 17 D 12 0.022 FB αUC12DA3 U α A 3 17 D 12 0.020 GL/MC αUC12DB3 U α B 3 18 D 12 0.020 NC αUC12DB3 U α B 3 18 D 12 0.018 NC αUC12DB3 U α B 3 18 D 12 0.017 FB αFC12DA3 F α A 3 19 D 12 0.227 EL αFC12DA3 F α A 3 19 D 12 0.249 AH αFC12DA3 F α A 3 19 D 12 0.286 MC αFC12DA2 F α A 2 19 D 12 0.201 GL αFC12DA2 F α A 2 19 D 12 0.186

125 Sub- Sample Large Sample Unique Sampling Analyst Code Fortification Lot Bottle Bottle Bottle Treatment Day Riboflavin MC αFC12DA2 F α A 2 19 D 12 0.193 GL/MC αFC12DB3 F α B 3 20 D 12 0.241 NC αFC12DB3 F α B 3 20 D 12 0.203 MC αFC12DB3 F α B 3 20 D 12 0.220 NC αFC12DB3 F α B 3 20 D 12 0.226 NC βUC12DA3 U β A 3 21 D 12 0.016 MC βUC12DA3 U β A 3 21 D 12 0.018 MC βUC12DA3 U β A 3 21 D 12 0.017 GL/MC βUC12DB3 U β B 3 22 D 12 0.017 MC βUC12DB3 U β B 3 22 D 12 0.017 GL βUC12DB3 U β B 3 22 D 12 0.017 NC βFC12DA3 F β A 3 23 D 12 0.210 GL/MC βFC12DA3 F β A 3 23 D 12 0.214 NC βFC12DA3 F β A 3 23 D 12 0.208 GL/MC βFC12DB3 F β B 3 24 D 12 0.203 MC βFC12DB3 F β B 3 24 D 12 0.223 GL βFC12DB3 F β B 3 24 D 12 0.208 TM αUC5LA3 U α A 3 25 L 5 0.027 FB αUC5LA3 U α A 3 25 L 5 0.017 AH αUC5LA3 U α A 3 25 L 5 0.030 FB αUC5LA3 U α A 3 25 L 5 0.019 MC αUC5LB3 U α B 3 26 L 5 0.020 MC αUC5LB3 U α B 3 26 L 5 0.019 NC αUC5LB3 U α B 3 26 L 5 0.017 MC αUC5LB3 U α B 3 26 L 5 0.018 FB αFC5LA3 F α A 3 27 L 5 0.211 MC αFC5LA2 F α A 2 27 L 5 0.179 GL αFC5LA2 F α A 2 27 L 5 0.170 MC αFC5LA2 F α A 2 27 L 5 0.223 MC αFC5LB3 F α B 3 28 L 5 0.271 NC αFC5LB3 F α B 3 28 L 5 0.204 MC αFC5LB3 F α B 3 28 L 5 0.217 NC αFC5LB3 F α B 3 28 L 5 0.225 NC/MJ αFC5LB3 F α B 3 28 L 5 0.200 MJ αFC5LB3 F α B 3 28 L 5 0.203 GL/MC βUC5LA3 U β A 3 29 L 5 0.017 MC βUC5LA3 U β A 3 29 L 5 0.016 NC βUC5LA3 U β A 3 29 L 5 0.017

126 Sub- Sample Large Sample Unique Sampling Analyst Code Fortification Lot Bottle Bottle Bottle Treatment Day Riboflavin MC βUC5LB3 U β B 3 30 L 5 0.020 MC βUC5LB3 U β B 3 30 L 5 0.017 GL βUC5LB3 U β B 3 30 L 5 0.018 GL/MC βFC5LA3 F β A 3 31 L 5 0.205 MC βFC5LA3 F β A 3 31 L 5 0.190 NC βFC5LA3 F β A 3 31 L 5 0.211 MC βFC5LB3 F β B 3 32 L 5 0.196 MC βFC5LB3 F β B 3 32 L 5 0.205 GL βFC5LB3 F β B 3 32 L 5 0.210 FB αUC12LA3 U α A 3 33 L 12 0.017 AH αUC12LA3 U α A 3 33 L 12 0.027 FB αUC12LA3 U α A 3 33 L 12 0.018 GL/MC αUC12LB3 U α B 3 34 L 12 0.020 NC αUC12LB3 U α B 3 34 L 12 0.017 NC αUC12LB3 U α B 3 34 L 12 0.014 MC αUC12LB3 U α B 3 34 L 12 0.018 TM αFC12LA3 F α A 3 35 L 12 0.218 FB αFC12LA3 F α A 3 35 L 12 0.200 AH αFC12LA3 F α A 3 35 L 12 0.210 MC αFC12LA2 F α A 2 35 L 12 0.161 GL αFC12LA2 F α A 2 35 L 12 0.154 MC αFC12LA2 F α A 2 35 L 12 0.170 GL/MC αFC12LB3 F α B 3 36 L 12 0.214 NC αFC12LB3 F α B 3 36 L 12 0.198 MC αFC12LB3 F α B 3 36 L 12 0.194 NC βUC12LA3 U β A 3 37 L 12 0.016 MC βUC12LA3 U β A 3 37 L 12 0.015 MC βUC12LA3 U β A 3 37 L 12 0.015 GL/MC βUC12LB3 U β B 3 38 L 12 0.014 MC βUC12LB3 U β B 3 38 L 12 0.015 GL βUC12LB3 U β B 3 38 L 12 0.016 NC βFC12LA3 F β A 3 39 L 12 0.181 GL/MC βFC12LA3 F β A 3 39 L 12 0.193 NC βFC12LA3 F β A 3 39 L 12 0.191 GL/MC βFC12LB3 F β B 3 40 L 12 0.192 MC βFC12LB3 F β B 3 40 L 12 0.190 GL βFC12LB3 F β B 3 40 L 12 0.184

127 Folate

Total folate is measured in mcg/100g soymilk.

NSRL folate "Hot Hold" raw data for statistical analysis.

File Small (_auto or Sample Sample Unique _manual) Horrat Code Fortification Lot Treatment Bottle Bottle Folate 1HFA 1.6380 αU1c U α 1 3 1 0.1363 1HFA 0.7579 αU1c U α 1 3 1 0.1609 1HFA 0.7160 αU1a U α 1 1 2 0.1372 1HFA 0.9579 αU1a U α 1 1 2 0.1437 1HFA 0.7272 αU1a U α 1 1 2 0.1157 1HFA 0.8381 αU1a U α 1 1 2 0.1167 1HFB 0.8637 βU1a U β 1 1 3 0.1858 1HFB 1.1618 βU1a U β 1 1 3 0.1856 1HFB 0.5000 βU1a U β 1 1 3 0.1566 1HFB 0.7729 βU1a U β 1 1 3 0.1773 1HFA 1.0190 αF1c F α 1 3 4 0.6370 1HFA 0.9309 αF1c F α 1 3 4 0.6117 1HFA 0.6456 αF1a F α 1 1 5 0.6092 1HFA 0.6128 αF1a F α 1 1 5 0.5615 1HFA 0.6762 αF1a F α 1 1 5 0.5195 1HFA 0.8894 αF1a F α 1 1 5 0.6800 1HFB 0.7700 βF1a F β 1 1 6 0.5467 1HFB 1.0630 βF1a F β 1 1 6 0.6579 1HFB 0.3089 βF1a F β 1 1 6 0.5743 1HFB 0.6924 βF1a F β 1 1 6 0.6221 2HFA 0.5463 αU2c U α 2 3 7 0.1594 2HFA 0.5466 αU2c U α 2 3 7 0.1369 2HFA 0.7899 αU2a U α 2 1 8 0.1467 2HFA 0.6390 αU2a U α 2 1 8 0.1555 2HFA 0.8871 αU2a U α 2 1 8 0.1279 2HFA 0.7747 αU2a U α 2 1 8 0.1651 2HFB 0.7288 βU2a U β 2 1 9 0.1994 2HFB 0.7049 βU2a U β 2 1 9 0.1768 2HFB 0.7062 βU2a U β 2 1 9 0.1451 2HFB 0.6244 βU2a U β 2 1 9 0.1613 2HFA 1.1037 αF2c F α 2 3 10 0.6103 2HFA 0.3372 αF2c F α 2 3 10 0.5759

128 File Small (_auto or Sample Sample Unique _manual) Horrat Code Fortification Lot Treatment Bottle Bottle Folate 2HFA 0.7371 αF2a F α 2 1 11 0.6019 2HFA 0.8893 αF2a F α 2 1 11 0.5981 2HFA 1.0736 αF2a F α 2 1 11 0.5421 2HFA 0.7983 αF2a F α 2 1 11 0.6635 2HFB 0.7033 βF2a F β 2 1 12 0.5816 2HFB 0.7253 βF2a F β 2 1 12 0.6685 2HFB 0.6299 βF2a F β 2 1 12 0.6107 2HFB 0.5976 βF2a F β 2 1 12 0.5918 3HFA 1.0348 αU3c U α 3 3 13 0.1649 3HFA 0.9089 αU3c U α 3 3 13 0.1483 3HFA 0.8924 αU3a U α 3 1 14 0.1272 3HFA 1.0957 αU3a U α 3 1 14 0.1225 3HFA 0.7151 αU3a U α 3 1 14 0.1793 3HFA 0.7946 αU3a U α 3 1 14 0.1338 3HFA 0.7583 αU3a U α 3 1 14 0.1166 3HFA 1.1973 αU3a U α 3 1 14 0.1530 3HFB 0.7303 βU3a U β 3 1 15 0.1724 3HFB 0.6820 βU3a U β 3 1 15 0.1937 3HFA 0.6672 αU3c F α 3 3 16 0.6183 3HFA 0.8713 αU3c F α 3 3 16 0.6231 3HFA 0.7177 αF3a F α 3 1 17 0.5800 3HFA 0.7583 αF3a F α 3 1 17 0.6096 3HFA 0.7620 αF3a F α 3 1 17 0.6303 3HFA 0.8951 αF3a F α 3 1 17 0.5399 3HFA 1.0969 αF3a F α 3 1 17 0.6206 3HFA 0.9637 αF3a F α 3 1 17 0.6018 3HFB 0.7103 βF3a F β 3 1 18 0.5929 3HFB 0.6496 βF3a F β 3 1 18 0.6373

129 NSRL folate "Cooling Method" raw data for statistical analysis.

Sub- File Sample Large Sample (_auto) Horrat Code Fortification Lot Treatment Bottle Bottle Folate 1CUA 0.6646 αUCA2 U α C 1 2 0.1492 1CUA 0.7315 αUCA2 U α C 1 2 0.1461 2CUA 0.5909 αUCB2 U α C 2 2 0.1575 2CUA 0.6288 αUCB2 U α C 2 2 0.1419 1CUA 0.7690 αUWA2 U α W 3 2 0.1556 1CUA 0.7082 αUWA2 U α W 3 2 0.1505 2CUA 0.6888 αUWB2 U α W 4 2 0.1571 2CUA 0.8215 αUWB2 U α W 4 2 0.1412 1CFA 0.8756 αFCA3 F α C 5 3 0.6723 1CFA 0.9422 αFCA3 F α C 5 3 0.6506 2CFA 0.7165 αFCB3 F α C 6 3 0.5881 2CFA 0.7169 αFCB3 F α C 6 3 0.6171 1CFA 0.6334 αFWA3 F α W 7 3 0.6491 1CFA 0.5033 αFWA3 F α W 7 3 0.6071 2CFA 0.6327 αFWB3 F α W 8 3 0.5872 2CFA 0.7197 αFWB3 F α W 8 3 0.6172 1CFB 0.6777 βUCA3 U β C 9 3 0.1733 2CFB 0.7200 βUCA3 U β C 9 3 0.1757 1CFB 1.1651 βUCA2 U β C 9 2 0.1651 1CFB 1.2403 βUCA2 U β C 9 2 0.1619 1CFB 1.1549 βUCA2 U β C 9 2 0.1829 1CFB 1.3469 βUCA2 U β C 9 2 0.1883 2CFB 0.7339 βUCB3 U β C 10 3 0.1585 1CFB 0.7543 βUCB3 U β C 10 3 0.1822 2CFB 1.0870 βUCB2 U β C 10 2 0.1949 2CFB 1.8533 βUCB2 U β C 10 2 0.1576 1CFB 1.262 βUWA3 U β W 11 3 0.1461 1CFB 0.4454 βUWA3 U β W 11 3 0.1686 3CFB 1.2454 βUWA2 U β W 11 2 0.1667 3CFB 0.8743 βUWA2 U β W 11 2 0.1603 3CFB 0.8358 βUWA2 U β W 11 2 0.1637 3CFB 0.8166 βUWA2 U β W 11 2 0.1985 1CFB 1.1257 βUWB3 U β W 12 3 0.2219 1CFB 1.6643 βUWB3 U β W 12 3 0.1299 1CFB 0.4648 βFCA3 F β C 13 3 0.5992 1CFB 1.2017 βFCA3 F β C 13 3 0.6315

130 Sub- File Sample Large Sample (_auto) Horrat Code Fortification Lot Treatment Bottle Bottle Folate 1CFB 1.1479 βFCA2 F β C 13 2 0.6299 1CFB 1.0943 βFCA2 F β C 13 2 0.6317 1CFB 1.1035 βFCA2 F β C 13 2 0.5836 1CFB 1.1826 βFCA2 F β C 13 2 0.5911 2CFB 0.6400 βFCB3 F β C 14 3 0.5985 2CFB 0.6625 βFCB3 F β C 14 3 0.5931 2CFB 2.8977 βFCB2 F β C 14 2 0.5984 2CFB 1.0626 βFCB2 F β C 14 2 0.5713 1CFB 0.4567 βFWA3 F β W 15 3 0.5931 1CFB 0.9273 βFWA3 F β W 15 3 0.6263 3CFB 0.3743 βFWA2 F β W 15 2 0.5989 3CFB 1.0653 βFWA2 F β W 15 2 0.6051 3CFB 1.3233 βFWA2 F β W 15 2 0.5952 3CFB 0.7370 βFWA2 F β W 15 2 0.5653 2CFB 0.9206 βFWB3 F β W 16 3 0.6002 1CFB 1.1518 βFWB3 F β W 16 3 0.6193

131 NSRL folate "Light Exposure" raw data for statistical analysis.

Sub- File Sample Large Sample Unique (_auto) Horrat Code Fortification Lot Bottle Bottle Bottle Treatment Day Folate 1LFA 0.5155 αUC0L/DA1 U α A 1 1 O 0 0.1279 1LFA 0.6636 αUC0L/DA1 U α A 1 1 O 0 0.1450 2LFA 0.4683 αUC0L/DB3 U α B 3 2 O 0 0.1225 2LFA 0.5918 αUC0L/DB3 U α B 3 2 O 0 0.1234 2LFA 0.8353 αUC0L/DB3 U α B 3 2 O 0 0.1030 2LFA 1.1864 αUC0L/DB3 U α B 3 2 O 0 0.1244 2LFA 0.5614 αUC0L/DB3 U α B 3 2 O 0 0.1296 2LFA 0.7807 αUC0L/DB3 U α B 3 2 O 0 0.1311 1LFA 1.1139 αFC0L/DA1 F α A 1 3 O 0 0.5692 1LFA 1.1336 αFC0L/DA1 F α A 1 3 O 0 0.6254 2LFA 0.8311 αFC0L/DB3 F α B 3 4 O 0 0.5880 2LFA 0.8258 αFC0L/DB3 F α B 3 4 O 0 0.6226 2LFA 0.7859 αFC0L/DB3 F α B 3 4 O 0 0.6021 2LFA 0.9530 αFC0L/DB3 F α B 3 4 O 0 0.5744 2LFA 0.8357 αFC0L/DB3 F α B 3 4 O 0 0.5880 2LFA 0.8368 αFC0L/DB3 F α B 3 4 O 0 0.6081 1LFB 0.6707 βUC0L/DA2 U β A 2 5 O 0 0.1473 1LFB 0.7566 βUC0L/DA2 U β A 2 5 O 0 0.1525 2LFB 0.6162 βUC0L/DB3 U β B 3 6 O 0 0.1637 2LFB 1.1513 βUC0L/DB3 U β B 3 6 O 0 0.1649 2LFB 0.9339 βUC0L/DB3 U β B 3 6 O 0 0.1597 2LFB 0.5195 βUC0L/DB3 U β B 3 6 O 0 0.1413 1LFB 0.7529 βFC0L/DA2 F β A 2 7 O 0 0.5901 1LFB 0.7551 βFC0L/DA2 F β A 2 7 O 0 0.5910 2LFB 0.7921 βFC0L/DB3 F β B 3 8 O 0 0.5148 2LFB 0.8135 βFC0L/DB3 F β B 3 8 O 0 0.6586 2LFB 0.9050 βFC0L/DB3 F β B 3 8 O 0 0.6215 2LFB 0.6573 βFC0L/DB3 F β B 3 8 O 0 0.5769 3LUA 0.7503 αUC5DA1 U α A 1 9 D 5 0.1448 3LUA 0.7920 αUC5DA1 U α A 1 9 D 5 0.1372 4LFA 0.7299 αUC5DB3 U α B 3 10 D 5 0.1394 4LFA 0.7129 αUC5DB3 U α B 3 10 D 5 0.1297 4LFA 1.0459 αUC5DB2 U α B 2 10 D 5 0.1382 4LFA 1.5551 αUC5DB2 U α B 2 10 D 5 0.1033 4LFA 0.9046 αUC5DB2 U α B 2 10 D 5 0.1123 4LFA 0.7898 αUC5DB2 U α B 2 10 D 5 0.1116

132 Sub- File Sample Large Sample Unique (_auto) Horrat Code Fortification Lot Bottle Bottle Bottle Treatment Day Folate 3LFA 1.0362 αFC5DA2 F α A 2 11 D 5 0.5984 3LFA 1.3957 αFC5DA2 F α A 2 11 D 5 0.5799 4LFA 0.8711 αFC5DB3 F α B 3 12 D 5 0.5675 4LFA 0.7847 αFC5DB3 F α B 3 12 D 5 0.6339 4LFA 0.9501 αFC5DB2 F α B 2 12 D 5 0.6136 4LFA 1.2635 αFC5DB2 F α B 2 12 D 5 0.5890 4LFA 0.7763 αFC5DB2 F α B 2 12 D 5 0.6009 4LFA 0.6640 αFC5DB2 F α B 2 12 D 5 0.5805 3LFB 0.5198 βUC5DA2 U β A 2 13 D 5 0.1043 3LFB 0.7813 βUC5DA2 U β A 2 13 D 5 0.1664 4LFB 1.2982 βUC5DB3 U β B 3 14 D 5 0.1266 4LFB 0.9736 βUC5DB3 U β B 3 14 D 5 0.1015 4LFB 1.1150 βUC5DB3 U β B 3 14 D 5 0.1411 4LFB 0.8506 βUC5DB3 U β B 3 14 D 5 0.1118 7LFB 0.6344 βFC5DA3 F β A 3 15 D 5 0.5893 7LFB 1.0306 βFC5DA3 F β A 3 15 D 5 0.5730 3LFB 0.7316 βFC5DA2 F β A 2 15 D 5 0.6058 3LFB 0.7384 βFC5DA2 F β A 2 15 D 5 0.5785 4LFB 0.5694 βFC5DB3 F β B 3 16 D 5 0.5504 4LFB 1.3474 βFC5DB3 F β B 3 16 D 5 0.6043 4LFB 1.0791 βFC5DB3 F β B 3 16 D 5 0.5710 4LFB 1.0936 βFC5DB3 F β B 3 16 D 5 0.5621 3LUA 0.8679 αUC12DA1 U α A 1 17 D 12 0.1392 3LUA 0.6676 αUC12DA1 U α A 1 17 D 12 0.1628 6LFA 0.9326 αUC12DB3 U α B 3 18 D 12 0.1231 6LFA 1.2382 αUC12DB3 U α B 3 18 D 12 0.1326 6LFA 1.2747 αUC12DB3 U α B 3 18 D 12 0.1153 6LFA 1.2015 αUC12DB3 U α B 3 18 D 12 0.1497 5LFA 1.2476 αFC12DA2 F α A 2 19 D 12 0.6046 5LFA 1.2745 αFC12DA2 F α A 2 19 D 12 0.5535 6LFA 0.7459 αFC12DB3 F α B 3 20 D 12 0.5967 6LFA 1.5830 αFC12DB3 F α B 3 20 D 12 0.5650 6LFA 1.1890 αFC12DB3 F α B 3 20 D 12 0.5883 6LFA 1.2052 αFC12DB3 F α B 3 20 D 12 0.5954 5LFB 0.9064 βUC12DA2 U β A 2 21 D 12 0.1434 5LFB 0.5158 βUC12DA2 U β A 2 21 D 12 0.1244 5LFB 0.7090 βUC12DA2 U β A 2 21 D 12 0.1393 5LFB 0.6664 βUC12DA2 U β A 2 21 D 12 0.1335

133 Sub- File Sample Large Sample Unique (_auto) Horrat Code Fortification Lot Bottle Bottle Bottle Treatment Day Folate 6LFB 1.1858 βUC12DB3 U β B 3 22 D 12 0.1345 6LFB 1.2584 βUC12DB3 U β B 3 22 D 12 0.1384 6LFB 0.8613 βUC12DB3 U β B 3 22 D 12 0.1411 6LFB 0.5401 βUC12DB3 U β B 3 22 D 12 0.1512 6LFB 1.0931 βUC12DB3 U β B 3 22 D 12 0.1554 6LFB 1.4832 βUC12DB3 U β B 3 22 D 12 0.1765 5LFB 0.5939 βFC12DA2 F β A 2 23 D 12 0.5764 5LFB 0.7216 βFC12DA2 F β A 2 23 D 12 0.5643 5LFB 0.6813 βUC12DA2 F β A 2 23 D 12 0.5908 5LFB 0.7097 βUC12DA2 F β A 2 23 D 12 0.5938 6LFB 1.2343 βFC12DB3 F β B 3 24 D 12 0.5778 6LFB 1.1471 βFC12DB3 F β B 3 24 D 12 0.5697 6LFB 1.0804 βFC12DB3 F β B 3 24 D 12 0.6281 6LFB 0.9524 βFC12DB3 F β B 3 24 D 12 0.5420 6LFB 0.9607 βFC12DB3 F β B 3 24 D 12 0.5420 6LFB 1.2070 βFC12DB3 F β B 3 24 D 12 0.6403 7LUA 0.6251 αUC5LA1 U α A 1 25 L 5 0.1383 7LUA 0.6403 αUC5LA1 U α A 1 25 L 5 0.1285 8LFA 0.7964 αUC5LB3 U α B 3 26 L 5 0.1590 8LFA 0.7523 αUC5LB3 U α B 3 26 L 5 0.1212 8LFA 1.1404 αUC5LB3 U α B 3 26 L 5 0.1412 8LFA 1.1956 αUC5LB3 U α B 3 26 L 5 0.1586 3LFA 0.7081 αFC5LA2 F α A 3 27 L 5 0.6166 3LFA 0.7325 αFC5LA2 F α A 2 27 L 5 0.5757 8LFA 0.7015 αFC5LB3 F α B 3 28 L 5 0.6014 8LFA 0.6945 αFC5LB3 F α B 3 28 L 5 0.5829 8LFA 0.8709 αFC5LB3 F α B 3 28 L 5 0.5598 8LFA 0.8730 αFC5LB3 F α B 3 28 L 5 0.6090 7LUB 1.1309 βUC5LA3 U β A 3 29 L 5 0.1443 7LUB 0.8181 βUC5LA3 U β A 3 29 L 5 0.1385 7LUB 0.8855 βUC5LA3 U β A 3 29 L 5 0.1246 7LUB 1.0041 βUC5LA3 U β A 3 29 L 5 0.1264 8LFB 1.2059 βUC5LB3 U β B 3 30 L 5 0.1397 8LFB 1.1885 βUC5LB3 U β B 3 30 L 5 0.1373 8LFB 1.0502 βUC5LB3 U β B 3 30 L 5 0.1600 8LFB 0.9264 βUC5LB3 U β B 3 30 L 5 0.1487 7LFB 0.9740 βFC5LA3 F β A 3 31 L 5 0.5962 7LFB 0.6977 βFC5LA3 F β A 3 31 L 5 0.5845

134 Sub- File Sample Large Sample Unique (_auto) Horrat Code Fortification Lot Bottle Bottle Bottle Treatment Day Folate 7LFB 0.6839 βFC5LA3 F β A 3 31 L 5 0.5810 8LFB 0.7398 βFC5LB3 F β B 3 32 L 5 0.5084 8LFB 0.8867 βFC5LB3 F β B 3 32 L 5 0.5945 8LFB 0.7720 βFC5LB3 F β B 3 32 L 5 0.6024 8LFB 0.6921 βFC5LB3 F β B 3 32 L 5 0.5913 7LUA 0.8608 αUC12LA1 U α A 1 33 L 12 0.1210 7LUA 0.7907 αUC12LA1 U α A 1 33 L 12 0.1201 10LFA 1.0068 αUC12LB3 U α B 3 34 L 12 0.1243 10LFA 0.8616 αUC12LB3 U α B 3 34 L 12 0.1527 10LFA 0.8211 αUC12LB2 U α B 2 34 L 12 0.1462 10LFA 0.8465 αUC12LB2 U α B 2 34 L 12 0.1543 5LFA 1.2476 αFC12LA2 F α A 2 35 L 12 0.5509 5LFA 1.3282 αFC12LA2 F α A 2 35 L 12 0.5753 10LFA 0.8967 αFC12LB3 F α B 3 36 L 12 0.5905 10LFA 0.9346 αFC12LB3 F α B 3 36 L 12 0.5572 10LFA 1.0530 αFC12LB2 F α B 2 36 L 12 0.5885 10LFA 0.9759 αFC12LB2 F α B 2 36 L 12 0.6059 9LFB 0.7541 βUC12LA2 U β A 2 37 L 12 0.1256 9LFB 1.1642 βUC12LA2 U β A 2 37 L 12 0.1463 9LFB 0.7493 βUC12LA2 U β A 2 37 L 12 0.1261 9LFB 0.7953 βUC12LA2 U β A 2 37 L 12 0.1625 10LFB 1.2078 βUC12LB3 U β B 3 38 L 12 0.1232 10LFB 0.7555 βUC12LB3 U β B 3 38 L 12 0.1076 10LFB 0.9924 βUC12LB3 U β B 3 38 L 12 0.1217 10LFB 1.1227 βUC12LB3 U β B 3 38 L 12 0.1275 9LFB 0.7261 βFC12LA2 F β A 2 39 L 12 0.5426 9LFB 0.8010 βFC12LA2 F β A 2 39 L 12 0.5681 9LFB 0.7147 βFC12LA2 F β A 2 39 L 12 0.5713 9LFB 0.7809 βFC12LA2 F β A 2 39 L 12 0.5589 10LFB 1.1097 βFC12LB3 F β B 3 40 L 12 0.5552 10LFB 0.8071 βFC12LB3 F β B 3 40 L 12 0.5251 10LFB 1.1578 βFC12LB3 F β B 3 40 L 12 0.5641 10LFB 1.3115 βFC12LB3 F β B 3 40 L 12 0.5809

135 pH pH measurements were taken on analytical sample replicates from Cooling Study samples and stored at -80 oC. Samples were thawed in a 37oC water bath with occasional shaking to facilitate heat transfer immediately prior to measurement (Orion pH meter calibrated with pH 4 & 7 standards).

pH pH Sample Fortification Reading Reading Code Batch Status #1 #2 αUWB1 1 U 6.59 6.55 αUWA1 1 U 6.62 6.58 αUCB1 1 U 6.61 6.56 αUCA1 1 U 6.62 6.58 αFWA1 1 F 6.48 6.43 αFWB1 1 F 6.48 6.44 αFCA1 1 F 6.49 6.44 αFCB1 1 F 6.5 6.47 βUWA1 2 U 6.67 6.63 βUWB1 2 U 6.67 6.63 βUCA1 2 U 6.67 6.63 βUCB1 2 U 6.67 6.63 βFWB1 2 F 6.54 6.52 βFWA1 2 F 6.58 6.53 βFCB1 2 F 6.54 6.51 βFCA1 2 F 6.55 6.51

136 Appendix G: NSRL Raw Statistical Analysis Output

Vitamin A

The SAS System, 12:18 Wednesday, January 30, 2019 Cooling Data

Analysis for: cis_13 cis_11 All_trans

The Mixed Procedure

Model Information Data Set WORK.GOOD WORK.GOOD WORK.GOOD Dependent Variable cis_13 cis_11 All_trans Covariance Structure Variance Components Variance Components Variance Components Estimation Method REML REML REML Residual Variance Method Profile Profile Profile Fixed Effects SE Method Model-Based Model-Based Model-Based Degrees of Freedom Method Containment Containment Containment

Class Level Information Class Levels (Values) Levels (Values) Levels (Values) Lot 2 (a ß) 2 (a ß) 2 (a ß) Large_Bottle 2 (A B) 2 (A B) 2 (A B) Treatment 2 (C W) 2 (C W) 2 (C W)

Dimensions Covariance Parameters 2 2 2 Columns in X 3 3 3 Columns in Z 8 8 8 Subjects 1 1 1 Max Obs per Subject 48 48 48

Number (No.) of Observations No. of Observations Read 48 48 48 No. of Observations Used 48 48 48 No. of Observations Not Used 0 0 0

Iteration History Iteration (Evaluations) -2 Res Log Like -2 Res Log Like -2 Res Log Like (Criterion) (Criterion) (Criterion) 0 (1) 43.04900652 96.81582487 91.34159935

1 (1) 43.04900652 94.59634746 86.02485335 (0.00000000) (0.00000000) (0.00000000)

Convergence Convergence Convergence criteria met. criteria met. criteria met.

137 The SAS System 12:18 Wednesday, January 30, 2019 Cooling Data

Analysis for: cis_13 cis_11 All-trans

The Mixed Procedure

Covariance Parameter Estimates Cov Parm Estimate Estimate Estimate Lot*Large_B(Treatme) 0 0.07775 0.1148 Residual 0.1300 0.3575 0.2816

Fit Statistics -2 Res Log Likelihood 43.0 94.6 86.0 AIC (Smaller is Better) 45.0 98.6 90.0 AICC (Smaller is Better) 45.1 98.9 90.3 BIC (Smaller is Better) 45.1 98.8 90.2

Type 3 Tests of Fixed Effects Effect Treatment Treatment Treatment Num DF, Den DF 1, 6 1, 6 1, 6 F Value 0.10 0.11 0.18 Pr > F 0.7665 0.7516 0.6897

Least Squares Means Effect Treatment Treatment Treatment

Treatment C C C Estimate 5.5976 0.9874 93.4150 Std Err 0.07360 0.1853 0.2011 DF 6 6 6 t Value 76.06 5.33 464.53 Pr > |t| <.0001 0.0018 <.0001

Treatment W W W Estimate 5.5653 0.9005 93.5342 Std Err 0.07360 0.1853 0.2011 DF 6 6 6 t Value 75.62 4.86 465.12 Pr > |t| <.0001 0.0028 <.0001

Differences of Least Squares Means Effect Treatment Treatment Treatment Treatments C, W C, W C, W Estimate 0.03234 0.08685 -0.1192 Std Error 0.1041 0.2620 0.2844 DF 6 6 6 t Value 0.31 0.33 -0.42 Pr > |t| 0.7665 0.7516 0.6897 Adjustment Tukey Tukey Tukey Adj P 0.7665 0.7516 0.6897

138 The SAS System, 12:18 Wednesday, January 30, 2019 Cooling Data

Analysis for: Bioactive RAE

The Mixed Procedure

Model Information Data Set WORK.GOOD WORK.GOOD Dependent Variable Bioactive RAE Covariance Structure Variance Components Variance Components Estimation Method REML REML Residual Variance Method Profile Profile Fixed Effects SE Method Model-Based Model-Based Degrees of Freedom Method Containment Containment

Class Level Information

Class Levels (Values) Levels (Values) Lot 2 (a ß) 2 (a ß) Large_Bottle 2 (A B) 2 (A B) Treatment 2 (C W) 2 (C W)

Dimensions

Covariance Parameters 2 2 Columns in X 3 3 Columns in Z 8 8 Subjects 1 1 Max Obs per Subject 48 48

Number of Observations

Number of Observations Read 48 48 Number of Observations Used 48 48 Number of Observations Not Used 0 0

Iteration History

Iteration (Evaluations) -2 Res Log Like -2 Res Log Like (Criterion) (Criterion)

0 (1) 52.65513331 39.58837580

1 (1) 48.67198111 39.49268956 (0.00000000) (0.00000000)

Convergence Convergence criteria met. criteria met.

139 The SAS System 12:18 Wednesday, January 30, 2019 Cooling Data

Analysis for: Bioactive RAE

The Mixed Procedure

Covariance Parameter Estimates Cov Parm Estimate Estimate Lot*Large_B(Treatme) 0.04181 0.004028 Residual 0.1275 0.1174

Fit Statistics -2 Res Log Likelihood 48.7 39.5 AIC (Smaller is Better) 52.7 43.5 AICC (Smaller is Better) 53.0 43.8 BIC (Smaller is Better) 52.8 43.7

Type 3 Tests of Fixed Effects Effect Treatment Treatment Num DF, Den DF 1, 6 1, 6 F Value 0.14 0.16 Pr > F 0.7227 0.7049

Least Squares Means Effect Treatment Treatment

Treatment C C Estimate 97.8370 1.5708 Standard Error 0.1256 0.07681 DF 6 6 t Value 779.23 20.45 Pr > |t| <.0001 <.0001

Treatment W W Estimate 97.9031 1.5276 Standard Error 0.1256 0.07681 DF 6 6 t Value 779.76 19.89 Pr > |t| <.0001 <.0001

Differences of Least Squares Means Effect Treatment Treatment Treatments C, W C, W Estimate -0.06606 0.04316 Standard Error 0.1776 0.1086 DF 6 6 t Value -0.37 0.40 Pr > |t| 0.7227 0.7049 Adjustment Tukey Tukey Adj P 0.7227 0.7049

140 The SAS System, 12:18 Wednesday, January 30, 2019 Hot Hold

Analysis for: cis_13 cis_11 All_trans

The Mixed Procedure

Class Level Information

Class Levels (Values) Levels (Values) Levels (Values) Lot 2 (a ß) 2 (a ß) 2 (a ß) Unique_Bottle 12 (1 2 3 4 5 6 12 (1 2 3 4 5 6 12 (1 2 3 4 5 6 7 8 9 10 11 12) 7 8 9 10 11 12) 7 8 9 10 11 12) Treatment 3 (1 2 3) 3 (1 2 3) 3 (1 2 3)

Dimensions Covariance Parameters 2 2 2 Columns in X 4 4 4 Columns in Z 12 12 12 Subjects 1 1 1 Max Obs per Subject 36 36 36

Number (No.) of Observations No. of Observations Read 36 36 36 No. of Observations Used 36 36 36 No. of Observations Not Used 0 0 0

Iteration History

Iteration (Evaluations) -2 Res Log Like -2 Res Log Like -2 Res Log Like (Criterion) (Criterion) (Criterion) 0 (1) 11.68036005 54.36020492 66.94011384

1 (1) -10.11234865 39.44368495 47.60380944 (0.00000000) (0.00000000) (0.00000000)

Convergence Convergence Convergence criteria met. criteria met. criteria met.

141 The SAS System 12:18 Wednesday, January 30, 2019 Hot Hold

Analysis for: cis_13 cis_11 All-trans

The Mixed Procedure

Covariance Parameter Estimates Cov Parm Estimate Estimate Estimate Lot*Unique_(Treatme) 0.05947 0.1864 0.3033 Residual 0.01789 0.09007 0.1069

Fit Statistics -2 Res Log Likelihood -10.1 39.4 47.6 AIC (Smaller is Better) -6.1 43.4 51.6 AICC (Smaller is Better) -5.7 43.8 52.0 BIC (Smaller is Better) -5.1 44.4 52.6

Type 3 Tests of Fixed Effects Effect Treatment Treatment Treatment Num DF, Den DF 2, 9 2, 9 2, 9 F Value 2.15 5.91 6.69 Pr > F 0.1722 0.0229 0.0166

Least Squares Means cis_13 Standard Effect Treatment Estimate Error DF t Value Pr > |t|

Treatment 1 5.5776 0.1279 9 43.61 <.0001 Treatment 2 5.7053 0.1279 9 44.61 <.0001 Treatment 3 5.9471 0.1279 9 46.50 <.0001

cis_11 Standard Effect Treatment Estimate Error DF t Value Pr > |t|

Treatment 1 0 0.2326 9 0.00 1.0000 Treatment 2 0.4416 0.2326 9 1.90 0.0901 Treatment 3 1.1229 0.2326 9 4.83 0.0009

All_trans Standard Effect Treatment Estimate Error DF t Value Pr > |t|

Treatment 1 94.4224 0.2911 9 324.35 <.0001 Treatment 2 93.8531 0.2911 9 322.40 <.0001 Treatment 3 92.9301 0.2911 9 319.22 <.0001

142 The SAS System 12:18 Wednesday, January 30, 2019 Hot Hold

Differences of Least Squares Means cis_13 Standard Effect Treatment Treatment Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatment 1 2 -0.1277 0.1809 9 -0.71 0.4981 Tukey 0.7662 Treatment 1 3 -0.3695 0.1809 9 -2.04 0.0715 Tukey 0.1576 Treatment 2 3 -0.2418 0.1809 9 -1.34 0.2141 Tukey 0.4115 cis_11 Standard Effect Treatment Treatment Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatment 1 2 -0.4416 0.3289 9 -1.34 0.2123 Tukey 0.4087 Treatment 1 3 -1.1229 0.3289 9 -3.41 0.0077 Tukey 0.0190 Treatment 2 3 -0.6813 0.3289 9 -2.07 0.0682 Tukey 0.1511

All_trans Standard Effect Treatment Treatment Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatment 1 2 0.5693 0.4117 9 1.38 0.2001 Tukey 0.3891 Treatment 1 3 1.4923 0.4117 9 3.62 0.0055 Tukey 0.0138 Treatment 2 3 0.9231 0.4117 9 2.24 0.0517 Tukey 0.1169

143 The SAS System, 12:18 Wednesday, January 30, 2019 Hot Hold

Analysis for: Bioactive RAE

The Mixed Procedure

Class Level Information

Class Levels (Values) Levels (Values) Lot 2 (a ß) 2 (a ß) Unique_Bottle 12 (1 2 3 4 5 6 12 (1 2 3 4 5 6 7 8 9 10 11 12) 7 8 9 10 11 12) Treatment 3 (1 2 3) 3 (1 2 3)

Dimensions Covariance Parameters 2 2 Columns in X 4 4 Columns in Z 12 12 Subjects 1 1 Max Obs per Subject 36 36

Number of Observations Number of Observations Read 36 36 Number of Observations Used 36 36 Number of Observations Not Used 0 0

Iteration History

Iteration (Evaluations) -2 Res Log Like -2 Res Log Like (Criterion) (Criterion)

0 (1) 30.77954589 -9.24558563

1 (1) 14.02196938 -107.77855062 (0.00000000) (0.00000000)

Convergence Convergence criteria met. criteria met.

144 The SAS System 12:18 Wednesday, January 30, 2019 Hot Hold

Analysis for: Bioactive RAE

The Mixed Procedure

Covariance Parameter Estimates Cov Parm Estimate Estimate Lot*Unique_(Treatme) 0.09577 0.04270 Residual 0.04036 0.000358

Fit Statistics -2 Res Log Likelihood 14.0 -107.8 AIC (Smaller is Better) 18.0 -103.8 AICC (Smaller is Better) 18.4 -103.4 BIC (Smaller is Better) 19.0 -102.8

Type 3 Tests of Fixed Effects Effect Treatment Treatment Num DF, Den DF 2, 9 2, 9 F Value 6.58 3.59 Pr > F 0.0173 0.0715

Least Squares Means Bioactive Effect Treatment Estimate Std Error DF t Value Pr > |t| Treatment 1 98.4941 0.1652 9 596.06 <.0001 Treatment 2 98.1681 0.1652 9 594.09 <.0001 Treatment 3 97.6532 0.1652 9 590.97 <.0001

RAE Effect Treatment Estimate Std Error DF t Value Pr > |t| Treatment 1 1.1322 0.1035 9 10.94 <.0001 Treatment 2 1.4656 0.1035 9 14.16 <.0001 Treatment 3 1.4774 0.1035 9 14.28 <.0001

Differences of Least Squares Means Bioactive Standard Effect Treatment Treatment Estimate Error DF t Value Pr > |t| Adjustment Adj P Treatment 1 2 0.3259 0.2337 9 1.39 0.1966 Tukey 0.3835 Treatment 1 3 0.8408 0.2337 9 3.60 0.0058 Tukey 0.0144 Treatment 2 3 0.5149 0.2337 9 2.20 0.0550 Tukey 0.1239

RAE Standard Effect Treatment Treatment Estimate Error DF t Value Pr > |t| Adjustment Adj P Treatment 1 2 -0.3334 0.1463 9 -2.28 0.0487 Tukey 0.1106 Treatment 1 3 -0.3452 0.1463 9 -2.36 0.0427 Tukey 0.0978 Treatment 2 3 -0.01179 0.1463 9 -0.08 0.9375 Tukey 0.9964

145 The SAS System, 12:18 Wednesday, January 30, 2019 Light Exposure

Analysis for: cis_13 cis_11 All_trans

The Mixed Procedure Model Information Data Set WORK.GOOD WORK.GOOD WORK.GOOD Dependent Variable cis_13 cis_11 All_trans Covariance Structure Variance Components Variance Components Variance Components Estimation Method REML REML REML Residual Variance Method Profile Profile Profile Fixed Effects SE Method Model-Based Model-Based Model-Based Degrees of Freedom Method Containment Containment Containment

Class Level Information Class Levels (Values) Levels (Values) Levels (Values) Lot 2 (a ß) 2 (a ß) 2 (a ß) Large_Bottle 2 (A B) 2 (A B) 2 (A B) Treatday 5 (D 5, D 12, 5 (D 5, D 12, 5 (D 5, D 12, L 5, L 12, O 0) L 5, L 12, O 0) L 5, L 12, O 0)

Dimensions Covariance Parameters 2 2 2 Columns in X 6 6 6 Columns in Z 20 20 20 Subjects 1 1 1 Max Obs per Subject 63 63 63

Number (No.) of Observations No. of Observations Read 63 63 63 No. of Observations Used 63 63 63 No. of Observations Not Used 0 0 0

Iteration History Iteration (Evaluations) -2 Res Log Like -2 Res Log Like -2 Res Log Like (Criterion) (Criterion) (Criterion) 0 (1) -13.32735300 43.79840792 72.28073510

1 (cis_13 & All_trans: 2, -26.68193205 37.18025598 50.78240753 cis_11: 3) (0.00000287) (0.00003773) (0.00001781)

2 (1) -26.68212441 37.17892357 50.78190508 (0.00000000) (0.00000003) (0.00000000)

3 (1) - 37.17892264 - - (0.00000000) -

Convergence Convergence Convergence criteria met. criteria met. criteria met.

146 The SAS System 12:18 Wednesday, January 30, 2019 Light Exposure

Analysis for: cis_13 cis_11 All-trans

The Mixed Procedure

Covariance Parameter Estimates Cov Parm Estimate Estimate Estimate Lot*Large_B(Treatda) 0.02116 0.03734 0.1124 Residual 0.02054 0.06940 0.07162

Fit Statistics -2 Res Log Likelihood -26.7 37.2 50.8 AIC (Smaller is Better) -22.7 41.2 54.8 AICC (Smaller is Better) -22.5 41.4 55.0 BIC (Smaller is Better) -20.7 43.2 56.8

Type 3 Tests of Fixed Effects Effect Treatday Treatday Treatday Num DF, Den DF 4, 15 4, 15 4, 15 F Value 3.87 800.68 347.28 Pr > F 0.0236 <.0001 <.0001

Least Squares Means cis_13 Standard Effect Treatday Estimate Error DF t Value Pr > |t| Treatday D 5 5.9602 0.08367 15 71.23 <.0001 Treatday D 12 6.0783 0.08367 15 72.64 <.0001 Treatday L 5 5.7805 0.08484 15 68.14 <.0001 Treatday L 12 5.9063 0.08225 15 71.81 <.0001 Treatday O 0 5.6530 0.08300 15 68.11 <.0001 cis_11 Standard Effect Treatday Estimate Error DF t Value Pr > |t| Treatday D 5 1.2331 0.1230 15 10.03 <.0001 Treatday D 12 1.3135 0.1230 15 10.68 <.0001 Treatday L 5 5.2833 0.1255 15 42.10 <.0001 Treatday L 12 8.7099 0.1195 15 72.90 <.0001 Treatday O 0 0.9447 0.1214 15 7.78 <.0001

All_trans Standard Effect Treatday Estimate Error DF t Value Pr > |t| Treatday D 5 92.8067 0.1846 15 502.75 <.0001 Treatday D 12 92.6082 0.1846 15 501.67 <.0001 Treatday L 5 88.9389 0.1865 15 476.92 <.0001 Treatday L 12 85.3925 0.1824 15 468.10 <.0001 Treatday O 0 93.4061 0.1836 15 508.88 <.0001

147 The SAS System 12:18 Wednesday, January 30, 2019 Light Exposure

Differences of Least Squares Means cis_13 Standard Effect Treatday _Treatday Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatday D 5 D 12 -0.1181 0.1183 15 -1.00 0.3340 Tukey-Kramer 0.8521 Treatday D 5 L 5 0.1797 0.1192 15 1.51 0.1523 Tukey-Kramer 0.5730 Treatday D 5 L 12 0.05385 0.1173 15 0.46 0.6528 Tukey-Kramer 0.9899 Treatday D 5 O 0 0.3071 0.1179 15 2.61 0.0199 Tukey-Kramer 0.1192 Treatday D 12 L 5 0.2978 0.1192 15 2.50 0.0245 Tukey-Kramer 0.1429 Treatday D 12 L 12 0.1720 0.1173 15 1.47 0.1634 Tukey-Kramer 0.5980 Treatday D 12 O 0 0.4253 0.1179 15 3.61 0.0026 Tukey-Kramer 0.0186 Treatday L 5 L 12 -0.1259 0.1182 15 -1.07 0.3037 Tukey-Kramer 0.8211 Treatday L 5 O 0 0.1274 0.1187 15 1.07 0.3000 Tukey-Kramer 0.8170 Treatday L 12 O 0 0.2533 0.1169 15 2.17 0.0467 Tukey-Kramer 0.2440 cis_11 Standard Effect Treatday _Treatday Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatday D 5 D 12 -0.08045 0.1739 15 -0.46 0.6502 Tukey-Kramer 0.9896 Treatday D 5 L 5 -4.0503 0.1757 15 -23.05 <.0001 Tukey-Kramer <.0001 Treatday D 5 L 12 -7.4768 0.1714 15 -43.61 <.0001 Tukey-Kramer <.0001 Treatday D 5 O 0 0.2884 0.1728 15 1.67 0.1158 Tukey-Kramer 0.4798 Treatday D 12 L 5 -3.9698 0.1757 15 -22.60 <.0001 Tukey-Kramer <.0001 Treatday D 12 L 12 -7.3964 0.1714 15 -43.14 <.0001 Tukey-Kramer <.0001 Treatday D 12 O 0 0.3689 0.1728 15 2.14 0.0497 Tukey-Kramer 0.2562 Treatday L 5 L 12 -3.4266 0.1733 15 -19.78 <.0001 Tukey-Kramer <.0001 Treatday L 5 O 0 4.3387 0.1746 15 24.85 <.0001 Tukey-Kramer <.0001 Treatday L 12 O 0 7.7653 0.1703 15 45.60 <.0001 Tukey-Kramer <.0001

All_trans Standard Effect Treatday _Treatday Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatday D 5 D 12 0.1986 0.2611 15 0.76 0.4586 Tukey-Kramer 0.9380 Treatday D 5 L 5 3.8679 0.2624 15 14.74 <.0001 Tukey-Kramer <.0001 Treatday D 5 L 12 7.4142 0.2595 15 28.57 <.0001 Tukey-Kramer <.0001 Treatday D 5 O 0 -0.5994 0.2603 15 -2.30 0.0361 Tukey-Kramer 0.1976 Treatday D 12 L 5 3.6693 0.2624 15 13.98 <.0001 Tukey-Kramer <.0001 Treatday D 12 L 12 7.2156 0.2595 15 27.80 <.0001 Tukey-Kramer <.0001 Treatday D 12 O 0 -0.7980 0.2603 15 -3.07 0.0079 Tukey-Kramer 0.0522 Treatday L 5 L 12 3.5464 0.2609 15 13.59 <.0001 Tukey-Kramer <.0001 Treatday L 5 O 0 -4.4672 0.2617 15 -17.07 <.0001 Tukey-Kramer <.0001 Treatday L 12 O 0 -8.0136 0.2588 15 -30.97 <.0001 Tukey-Kramer <.0001

148

The SAS System, 12:18 Wednesday, January 30, 2019 Light Exposure

Analysis for: Bioactive RAE

The Mixed Procedure

Class Level Information Class Levels (Values) Levels (Values) Lot 2 (a ß) 2 (a ß) Large_Bottle 2 (A B) 2 (A B) Treatday 5 (D 5, D 12, 5 (D 5, D 12, L 5, L 12, O 0) L 5, L 12, O 0)

Dimensions Covariance Parameters 2 2 Columns in X 6 6 Columns in Z 20 20 Subjects 1 1 Max Obs per Subject 63 63

Number of Observations Number of Observations Read 63 63 Number of Observations Used 63 63 Number of Observations Not Used 0 0

Iteration History Iteration (Evaluations) -2 Res Log Like -2 Res Log Like (Criterion) (Criterion) 0 (1) 4.66082928 -44.71501538

1 (Bioactive: 3, RAE: 2) -8.49386970 -93.23543645 (0.00002442) (0.00023774)

2 (1) -8.49530025 -93.26075574 (0.00000002) (0.00000275)

3 (1) -8.49530131 -93.26103245 (0.00000000) (0.00000000)

Convergence Convergence criteria met. criteria met.

149 The SAS System 12:18 Wednesday, January 30, 2019 Light Exposure

Analysis for: Bioactive RAE

The Mixed Procedure

Covariance Parameter Estimates Cov Parm Estimate Estimate Lot*Large_B(Treatda) 0.02737 0.02175 Residual 0.02849 0.004651

Fit Statistics -2 Res Log Likelihood -8.5 -93.3 AIC (Smaller is Better) -4.5 -89.3 AICC (Smaller is Better) -4.3 -89.0 BIC (Smaller is Better) -2.5 -87.3

Type 3 Tests of Fixed Effects Effect Treatday Treatday Num DF, Den DF 4, 15 4, 15 F Value 563.78 2.58 Pr > F <.0001 0.0800

Least Squares Means Bioactive Standard Effect Treatday Estimate Error DF t Value Pr > |t| Treatday D 5 97.5769 0.09601 15 1016.34 <.0001 Treatday D 12 97.4919 0.09601 15 1015.45 <.0001 Treatday L 5 94.9533 0.09741 15 974.81 <.0001 Treatday L 12 92.6601 0.09428 15 982.79 <.0001 Treatday O 0 97.8514 0.09519 15 1027.93 <.0001

RAE Standard Effect Treatday Estimate Error DF t Value Pr > |t| Treatday D 5 1.8661 0.07633 15 24.45 <.0001 Treatday D 12 1.9918 0.07633 15 26.09 <.0001 Treatday L 5 1.8511 0.07664 15 24.15 <.0001 Treatday L 12 1.7081 0.07600 15 22.47 <.0001 Treatday O 0 2.0099 0.07617 15 26.39 <.0001

150 The SAS System 12:18 Wednesday, January 30, 2019 Light Exposure

Differences of Least Squares Means Bioactive Standard Effect Treatday _Treatday Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatday D 5 D 12 0.08499 0.1358 15 0.63 0.5407 Tukey-Kramer 0.9685 Treatday D 5 L 5 2.6237 0.1368 15 19.18 <.0001 Tukey-Kramer <.0001 Treatday D 5 L 12 4.9169 0.1346 15 36.54 <.0001 Tukey-Kramer <.0001 Treatday D 5 O 0 -0.2745 0.1352 15 -2.03 0.0605 Tukey-Kramer 0.2990 Treatday D 12 L 5 2.5387 0.1368 15 18.56 <.0001 Tukey-Kramer <.0001 Treatday D 12 L 12 4.8319 0.1346 15 35.91 <.0001 Tukey-Kramer <.0001 Treatday D 12 O 0 -0.3595 0.1352 15 -2.66 0.0179 Tukey-Kramer 0.1087 Treatday L 5 L 12 2.2932 0.1356 15 16.92 <.0001 Tukey-Kramer <.0001 Treatday L 5 O 0 -2.8981 0.1362 15 -21.28 <.0001 Tukey-Kramer <.0001 Treatday L 12 O 0 -5.1913 0.1340 15 -38.75 <.0001 Tukey-Kramer <.0001

RAE Standard Effect Treatday _Treatday Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatday D 5 D 12 -0.1257 0.1079 15 -1.16 0.2625 Tukey-Kramer 0.7708 Treatday D 5 L 5 0.01503 0.1082 15 0.14 0.8913 Tukey-Kramer 0.9999 Treatday D 5 L 12 0.1580 0.1077 15 1.47 0.1630 Tukey-Kramer 0.5972 Treatday D 5 O 0 -0.1438 0.1078 15 -1.33 0.2023 Tukey-Kramer 0.6760 Treatday D 12 L 5 0.1407 0.1082 15 1.30 0.2130 Tukey-Kramer 0.6948 Treatday D 12 L 12 0.2837 0.1077 15 2.63 0.0188 Tukey-Kramer 0.1136 Treatday D 12 O 0 -0.01811 0.1078 15 -0.17 0.8689 Tukey-Kramer 0.9998 Treatday L 5 L 12 0.1430 0.1079 15 1.32 0.2051 Tukey-Kramer 0.6809 Treatday L 5 O 0 -0.1588 0.1081 15 -1.47 0.1623 Tukey-Kramer 0.5956 Treatday L 12 O 0 -0.3018 0.1076 15 -2.80 0.0133 Tukey-Kramer 0.0840

151 Vitamin C

The SAS System 15:12 Monday, June 20, 2016 44 Analysis for 'Vitamin C Raw Data$' Cooling Data

The Mixed Procedure

Model Information

Data Set WORK.GOOD Dependent Variable Vit_C Covariance Structure Variance Components Estimation Method REML Residual Variance Method Profile Fixed Effects SE Method Model-Based Degrees of Freedom Method Containment

Class Level Information

Class Levels Values Lot 2 a ß Large_Bottle 16 1 10 11 12 13 14 15 16 2 3 4 5 6 7 8 9 Treatment 2 C W

Dimensions

Covariance Parameters 2 Columns in X 3 Columns in Z 16 Subjects 1 Max Obs per Subject 48

Number of Observations

Number of Observations Read 48 Number of Observations Used 48 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 162.77767972 1 1 126.58383668 0.00000000

Convergence criteria met.

152 The SAS System 15:12 Monday, June 20, 2016 45 Analysis for 'Vitamin C Raw Data$' Cooling Data

The Mixed Procedure

Covariance Parameter Estimates

Cov Parm Estimate

Lot*Large_B(Treatme) 1.5281 Residual 0.3600

Fit Statistics

-2 Res Log Likelihood 126.6 AIC (Smaller is Better) 130.6 AICC (Smaller is Better) 130.9 BIC (Smaller is Better) 132.1

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F Treatment 1 14 0.33 0.5730

Least Squares Means

Standard Effect Treatment Estimate Error DF t Value Pr > |t|

Treatment C 11.6578 0.4539 14 25.68 <.0001 Treatment W 12.0282 0.4539 14 26.50 <.0001

Differences of Least Squares Means

Standard Effect Treatment Treatment Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatment C W -0.3705 0.6419 14 -0.58 0.5730 Tukey 0.5730

153 The SAS System 15:12 Monday, June 20, 2016 46 Analysis for 'Vitamin C Raw Data$' Hot Hold

The Mixed Procedure

Model Information

Class Level Information

Class Levels Values Lot 2 a ß Unique_Bottle 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Treatment 3 1 2 3

Dimensions

Covariance Parameters 2 Columns in X 4 Columns in Z 15 Subjects 1 Max Obs per Subject 43

Number of Observations

Number of Observations Read 43 Number of Observations Used 43 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 29.94895671 1 3 -19.47956756 0.00012708 2 1 -19.48570832 0.00000049 3 1 -19.48573095 0.00000000

154 The SAS System 15:12 Monday, June 20, 2016 47 Analysis for 'Vitamin C Raw Data$' Hot Hold

The Mixed Procedure

Convergence criteria met.

Covariance Parameter Estimates

Cov Parm Estimate Lot*Unique_(Treatme) 0.1010 Residual 0.01094

Fit Statistics

-2 Res Log Likelihood -19.5 AIC (Smaller is Better) -15.5 AICC (Smaller is Better) -15.2 BIC (Smaller is Better) -14.1

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F Treatment 2 12 6.02 0.0155

Least Squares Means

Standard Effect Treatment Estimate Error DF t Value Pr > |t|

Treatment 1 12.4958 0.1447 12 86.37 <.0001 Treatment 2 12.0372 0.1449 12 83.06 <.0001 Treatment 3 11.7966 0.1449 12 81.40 <.0001

Differences of Least Squares Means

Standard Effect Treatment Treatment Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatment 1 2 0.4586 0.2048 12 2.24 0.0448 Tukey-Kramer 0.1045 Treatment 1 3 0.6992 0.2048 12 3.41 0.0051 Tukey-Kramer 0.0132 Treatment 2 3 0.2406 0.2050 12 1.17 0.2632 Tukey-Kramer 0.4900

155 The SAS System 15:12 Monday, June 20, 2016 48 Analysis for 'Vitamin C Raw Data$' Light Exposure

The Mixed Procedure

Model Information

Class Level Information

Class Levels Values Lot 2 a ß Unique_Bottle 21 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 Treatday 5 D 5 D 12 L 5 L 12 O 0

Dimensions

Covariance Parameters 2 Columns in X 6 Columns in Z 21 Subjects 1 Max Obs per Subject 67

Number of Observations

Number of Observations Read 67 Number of Observations Used 67 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 196.62055479 1 2 130.92871624 0.00000212 2 1 130.92869820 0.00000000

156 The SAS System 15:12 Monday, June 20, 2016 49 Analysis for 'Vitamin C Raw Data$' Light Exposure

The Mixed Procedure

Convergence criteria met.

Covariance Parameter Estimates Cov Parm Estimate Lot*Unique_(Treatda) 1.1744 Residual 0.1772

Fit Statistics -2 Res Log Likelihood 130.9 AIC (Smaller is Better) 134.9 AICC (Smaller is Better) 135.1 BIC (Smaller is Better) 137.0

Type 3 Tests of Fixed Effects Num Den Effect DF DF F Value Pr > F Treatday 4 16 2.48 0.0860

Least Squares Means Standard Effect Treatday Estimate Error DF t Value Pr > |t| Treatday D 5 10.2690 0.5540 16 18.54 <.0001 Treatday D 12 10.1624 0.5540 16 18.35 <.0001 Treatday L 5 9.1191 0.5553 16 16.42 <.0001 Treatday L 12 8.9326 0.4967 16 17.98 <.0001 Treatday O 0 10.9426 0.5553 16 19.71 <.0001

Differences of Least Squares Means Standard Effect Treatday _Treatday Estimate Error DF t Value Pr > |t| Adjustment Adj P Treatday D 5 D 12 0.1065 0.7834 16 0.14 0.8935 Tukey-Kramer 0.9999 Treatday D 5 L 5 1.1499 0.7844 16 1.47 0.1620 Tukey-Kramer 0.5971 Treatday D 5 L 12 1.3363 0.7440 16 1.80 0.0914 Tukey-Kramer 0.4090 Treatday D 5 O 0 -0.6736 0.7844 16 -0.86 0.4032 Tukey-Kramer 0.9077 Treatday D 12 L 5 1.0433 0.7844 16 1.33 0.2021 Tukey-Kramer 0.6775 Treatday D 12 L 12 1.2298 0.7440 16 1.65 0.1178 Tukey-Kramer 0.4876 Treatday D 12 O 0 -0.7801 0.7844 16 -0.99 0.3347 Tukey-Kramer 0.8539 Treatday L 5 L 12 0.1865 0.7450 16 0.25 0.8056 Tukey-Kramer 0.9990 Treatday L 5 O 0 -1.8235 0.7853 16 -2.32 0.0338 Tukey-Kramer 0.1885 Treatday L 12 O 0 -2.0099 0.7450 16 -2.70 0.0158 Tukey-Kramer 0.0986

157 Thiamine

The SAS System 09:22 Monday, July 11, 2016 369 Analysis for Thiamine Cooling Data

The Mixed Procedure

Model Information

Data Set WORK.GOOD Dependent Variable Thiamine Covariance Structure Variance Components Estimation Method REML Residual Variance Method Profile Fixed Effects SE Method Model-Based Degrees of Freedom Method Containment

Class Level Information

Class Levels Values Lot 2 a ß Large_Bottle 16 1 10 11 12 13 14 15 16 2 3 4 5 6 7 8 9 Treatment 2 C W Fortification 2 F U

Dimensions Covariance Parameters 2 Columns in X 9 Columns in Z 16 Subjects 1 Max Obs per Subject 70

Number of Observations

Number of Observations Read 70 Number of Observations Used 70 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -252.10556297 1 1 -252.10556297 0.00000000

Convergence criteria met.

158 The SAS System 09:22 Monday, July 11, 2016 370 Analysis for Thiamine Cooling Data

The Mixed Procedure

Covariance Parameter Estimates

Cov Parm Estimate Lot*Larg(Trea*Forti) 0 Residual 0.001080

Fit Statistics

-2 Res Log Likelihood -252.1 AIC (Smaller is Better) -250.1 AICC (Smaller is Better) -250.0 BIC (Smaller is Better) -249.3

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F Treatment 1 12 0.50 0.4935 Fortification 1 12 1323.50 <.0001 Treatment*Fortificat 1 12 0.11 0.7452

Least Squares Means

Standard Effect Treatment Fortification Estimate Error DF t Value Pr > |t| Treatment*Fortificat C F 0.3351 0.007745 12 43.27 <.0001 Treatment*Fortificat C U 0.04636 0.007745 12 5.99 <.0001 Treatment*Fortificat W F 0.3269 0.008215 12 39.80 <.0001 Treatment*Fortificat W U 0.04342 0.007745 12 5.61 0.0001

Differences of Least Squares Means

Standard Effect Treat. Fort. Treat. Fort. Estimate Error DF t Value Pr > |t| Adjustment Adj P Treat*Fort C F C U 0.2887 0.01095 12 26.36 <.0001 Tukey-Kramer <.0001 Treat*Fort C F W F 0.008171 0.01129 12 0.72 0.4831 Tukey-Kramer 0.8858 Treat*Fort C F W U 0.2917 0.01095 12 26.63 <.0001 Tukey-Kramer <.0001 Treat*Fort C U W F -0.2806 0.01129 12 -24.85 <.0001 Tukey-Kramer <.0001 Treat*Fort C U W U 0.002940 0.01095 12 0.27 0.7930 Tukey-Kramer 0.9929 Treat*Fort W F W U 0.2835 0.01129 12 25.11 <.0001 Tukey-Kramer <.0001

159 The SAS System 09:22 Monday, July 11, 2016 372 Analysis for Thiamine Cooling Data

------Fortification=F ------

The Mixed Procedure

Model Information

Class Level Information

Class Levels Values Lot 2 a ß Large_Bottle 8 13 14 15 16 5 6 7 8 Treatment 2 C W

Dimensions

Covariance Parameters 2 Columns in X 3 Columns in Z 8 Subjects 1 Max Obs per Subject 34

Number of Observations

Number of Observations Read 34 Number of Observations Used 34 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -100.79729426 1 1 -100.79729426 0.00000000

Convergence criteria met.

160 The SAS System 09:22 Monday, July 11, 2016 373 Analysis for Thiamine Cooling Data

------Fortification=F ------

The Mixed Procedure

Covariance Parameter Estimates

Cov Parm Estimate

Lot*Large_B(Treatme) 0 Residual 0.002102

Fit Statistics

-2 Res Log Likelihood -100.8 AIC (Smaller is Better) -98.8 AICC (Smaller is Better) -98.7 BIC (Smaller is Better) -98.7

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F

Treatment 1 6 0.27 0.6226

Least Squares Means

Standard Effect Treatment Estimate Error DF t Value Pr > |t|

Treatment C 0.3351 0.01081 6 31.01 <.0001 Treatment W 0.3269 0.01146 6 28.52 <.0001

Differences of Least Squares Means

Standard Effect Treatment Treatment Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatment C W 0.008171 0.01575 6 0.52 0.6226 Tukey-Kramer 0.6226

161 The SAS System 09:22 Monday, July 11, 2016 374 Analysis for Thiamine Cooling Data

------Fortification=U ------

The Mixed Procedure

Model Information

Class Level Information

Class Levels Values Lot 2 a ß Large_Bottle 8 1 10 11 12 2 3 4 9 Treatment 2 C W

Dimensions

Covariance Parameters 2 Columns in X 3 Columns in Z 8 Subjects 1 Max Obs per Subject 36

Number of Observations

Number of Observations Read 36 Number of Observations Used 36 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -205.43564704 1 1 -205.43564704 0.00000000

Convergence criteria met.

162 The SAS System 09:22 Monday, July 11, 2016 375 Analysis for Thiamine Cooling Data

------Fortification=U ------

The Mixed Procedure

Covariance Parameter Estimates

Cov Parm Estimate Lot*Large_B(Treatme) 0 Residual 0.000117

Fit Statistics

-2 Res Log Likelihood -205.4 AIC (Smaller is Better) -203.4 AICC (Smaller is Better) -203.3 BIC (Smaller is Better) -203.4

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F Treatment 1 6 0.66 0.4467

Least Squares Means

Standard Effect Treatment Estimate Error DF t Value Pr > |t|

Treatment C 0.04636 0.002554 6 18.15 <.0001 Treatment W 0.04342 0.002554 6 17.00 <.0001

Differences of Least Squares Means

Standard Effect Treatment Treatment Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatment C W 0.002940 0.003611 6 0.81 0.4467 Tukey 0.4467

163 The SAS System 09:22 Monday, July 11, 2016 376 Analysis for Thiamine Hot Hold

The Mixed Procedure

Model Information

Class Level Information

Class Levels Values

Lot 2 a ß Unique_Bottle 24 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Treatment 3 1 2 3 Fortification 2 F U

Dimensions

Covariance Parameters 2 Columns in X 12 Columns in Z 24 Subjects 1 Max Obs per Subject 70

Number of Observations

Number of Observations Read 70 Number of Observations Used 70 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion

0 1 -279.27523174 1 2 -296.40632736 0.00000000

164 The SAS System 09:22 Monday, July 11, 2016 377 Analysis for Thiamine Hot Hold

The Mixed Procedure Convergence criteria met.

Covariance Parameter Estimates Cov Parm Estimate Lot*Uniq(Trea*Forti) 0.000360 Residual 0.000296

Fit Statistics -2 Res Log Likelihood -296.4 AIC (Smaller is Better) -292.4 AICC (Smaller is Better) -292.2 BIC (Smaller is Better) -290.1

Type 3 Tests of Fixed Effects Num Den Effect DF DF F Value Pr > F Treatment 2 18 0.41 0.6728 Fortification 1 18 991.70 <.0001 Treatment*Fortificat 2 18 0.22 0.8071

Least Squares Means Effect Fortification Treatment Estimate Std Err DF t Value Pr > |t| Treatment*Fortificat F 1 0.3239 0.01084 18 29.88 <.0001 Treatment*Fortificat U 1 0.04523 0.01071 18 4.22 0.0005 Treatment*Fortificat F 2 0.3117 0.01084 18 28.76 <.0001 Treatment*Fortificat U 2 0.04319 0.01071 18 4.03 0.0008 Treatment*Fortificat F 3 0.3278 0.01071 18 30.61 <.0001 Treatment*Fortificat U 3 0.04562 0.01071 18 4.26 0.0005

Differences of Least Squares Means Effect Fort. Treat. Fort. Treat. Estimate Std Err DF t Value Pr > |t| Adjustment Adj P Treat*Fort F 1 U 1 0.2787 0.01524 18 18.29 <.0001 Tukey-Kramer <.0001 Treat*Fort F 1 F 2 0.01223 0.01533 18 0.80 0.4356 Tukey-Kramer 0.9643 Treat*Fort F 1 U 2 0.2808 0.01524 18 18.43 <.0001 Tukey-Kramer <.0001 Treat*Fort F 1 F 3 -0.00382 0.01524 18 -0.25 0.8050 Tukey-Kramer 0.9998 Treat*Fort F 1 U 3 0.2783 0.01524 18 18.27 <.0001 Tukey-Kramer <.0001 Treat*Fort U 1 F 2 -0.2665 0.01524 18 -17.49 <.0001 Tukey-Kramer <.0001 Treat*Fort U 1 U 2 0.002039 0.01514 18 0.13 0.8944 Tukey-Kramer 1.0000 Treat*Fort U 1 F 3 -0.2825 0.01514 18 -18.66 <.0001 Tukey-Kramer <.0001 Treat*Fort U 1 U 3 -0.00039 0.01514 18 -0.03 0.9799 Tukey-Kramer 1.0000 Treat*Fort F 2 U 2 0.2685 0.01524 18 17.62 <.0001 Tukey-Kramer <.0001 Treat*Fort F 2 F 3 -0.01604 0.01524 18 -1.05 0.3063 Tukey-Kramer 0.8933 Treat*Fort F 2 U 3 0.2661 0.01524 18 17.46 <.0001 Tukey-Kramer <.0001 Treat*Fort U 2 F 3 -0.2846 0.01514 18 -18.79 <.0001 Tukey-Kramer <.0001 Treat*Fort U 2 U 3 -0.00243 0.01514 18 -0.16 0.8745 Tukey-Kramer 1.0000 Treat*Fort F 3 U 3 0.2821 0.01514 18 18.63 <.0001 Tukey-Kramer <.0001

165 The SAS System 09:22 Monday, July 11, 2016 380 Analysis for Thiamine Hot Hold

------Fortification=F ------

The Mixed Procedure

Model Information

Class Level Information

Class Levels Values Lot 2 a ß Unique_Bottle 12 13 14 15 16 17 18 19 20 21 22 23 24 Treatment 3 1 2 3

Dimensions

Covariance Parameters 2 Columns in X 4 Columns in Z 12 Subjects 1 Max Obs per Subject 34

Number of Observations

Number of Observations Read 34 Number of Observations Used 34 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -113.56887126 1 3 -121.55719241 0.00001602 2 1 -121.55865549 0.00000002 3 1 -121.55865730 0.00000000

166 The SAS System 09:22 Monday, July 11, 2016 381 Analysis for Thiamine Hot Hold

------Fortification=F ------

The Mixed Procedure

Convergence criteria met.

Covariance Parameter Estimates

Cov Parm Estimate Lot*Unique_(Treatme) 0.000694 Residual 0.000604

Fit Statistics

-2 Res Log Likelihood -121.6 AIC (Smaller is Better) -117.6 AICC (Smaller is Better) -117.1 BIC (Smaller is Better) -116.6

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F Treatment 2 9 0.31 0.7414

Least Squares Means

Standard Effect Treatment Estimate Error DF t Value Pr > |t|

Treatment 1 0.3239 0.01515 9 21.38 <.0001 Treatment 2 0.3117 0.01515 9 20.57 <.0001 Treatment 3 0.3278 0.01496 9 21.91 <.0001

Differences of Least Squares Means

Standard Effect Treatment Treatment Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatment 1 2 0.01225 0.02143 9 0.57 0.5816 Tukey-Kramer 0.8382 Treatment 1 3 -0.00382 0.02129 9 -0.18 0.8617 Tukey-Kramer 0.9825 Treatment 2 3 -0.01607 0.02129 9 -0.75 0.4698 Tukey-Kramer 0.7386

167 The SAS System 09:22 Monday, July 11, 2016 382 Analysis for Thiamine Hot Hold

------Fortification=U ------

The Mixed Procedure

Model Information

Class Level Information

Class Levels Values Lot 2 a ß Unique_Bottle 12 1 2 3 4 5 6 7 8 9 10 11 12 Treatment 3 1 2 3

Dimensions

Covariance Parameters 2 Columns in X 4 Columns in Z 12 Subjects 1 Max Obs per Subject 36

Number of Observations

Number of Observations Read 36 Number of Observations Used 36 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -239.20553570 1 1 -253.88398191 0.00000000

Convergence criteria met.

168 The SAS System 09:22 Monday, July 11, 2016 383 Analysis for Thiamine Hot Hold

------Fortification=U ------

The Mixed Procedure

Covariance Parameter Estimates

Cov Parm Estimate Lot*Unique_(Treatme) 0.000025 Residual 0.000012

Fit Statistics

-2 Res Log Likelihood -253.9 AIC (Smaller is Better) -249.9 AICC (Smaller is Better) -249.5 BIC (Smaller is Better) -248.9

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F

Treatment 2 9 0.23 0.7987

Least Squares Means

Standard Effect Treatment Estimate Error DF t Value Pr > |t|

Treatment 1 0.04523 0.002716 9 16.65 <.0001 Treatment 2 0.04319 0.002716 9 15.90 <.0001 Treatment 3 0.04562 0.002716 9 16.80 <.0001

Differences of Least Squares Means

Standard Effect Treatment Treatment Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatment 1 2 0.002039 0.003841 9 0.53 0.6083 Tukey 0.8584 Treatment 1 3 -0.00039 0.003841 9 -0.10 0.9219 Tukey 0.9944 Treatment 2 3 -0.00243 0.003841 9 -0.63 0.5432 Tukey 0.8068

169 The SAS System 09:22 Monday, July 11, 2016 384 Analysis for Thiamine Light Exposure

The Mixed Procedure

Model Information Data Set WORK.GOOD Dependent Variable Thiamine Covariance Structure Variance Components Estimation Method REML Residual Variance Method Profile Fixed Effects SE Method Model-Based Degrees of Freedom Method Containment

Class Level Information

Class Levels Values Lot 2 a ß Unique_Bottle 40 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 Fortification 2 F U Treatday 5 D 5 D 12 L 5 L 12 O 0

Dimensions Covariance Parameters 2 Columns in X 18 Columns in Z 40 Subjects 1 Max Obs per Subject 144

Number of Observations

Number of Observations Read 144 Number of Observations Used 144 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -555.79964722 1 2 -566.08286366 0.00000009 2 1 -566.08290208 0.00000000

170 The SAS System 09:22 Monday, July 11, 2016 385 Analysis for Thiamine Light Exposure

The Mixed Procedure

Convergence criteria met.

Covariance Parameter Estimates

Cov Parm Estimate

Lot*Uniq(Fort*Treat) 0.000220 Residual 0.000582

Fit Statistics

-2 Res Log Likelihood -566.1 AIC (Smaller is Better) -562.1 AICC (Smaller is Better) -562.0 BIC (Smaller is Better) -558.7

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F

Treatday 4 30 0.56 0.6925 Fortification 1 30 1923.90 <.0001 Fortificati*Treatday 4 30 0.35 0.8404

Least Squares Means

Standard Effect Fortification Treatday Estimate Error DF t Value Pr > |t|

Fort*Treatday F D 5 0.3283 0.009629 30 34.09 <.0001 Fort*Treatday F D 12 0.3085 0.009653 30 31.95 <.0001 Fort*Treatday F L 5 0.3154 0.009736 30 32.39 <.0001 Fort*Treatday F L 12 0.3055 0.009801 30 31.17 <.0001 Fort*Treatday F O 0 0.3212 0.009629 30 33.35 <.0001 Fort*Treatday U D 5 0.04293 0.01000 30 4.29 0.0002 Fort*Treatday U D 12 0.04075 0.01024 30 3.98 0.0004 Fort*Treatday U L 5 0.04422 0.009848 30 4.49 <.0001 Fort*Treatday U L 12 0.04109 0.01000 30 4.11 0.0003 Fort*Treatday U O 0 0.04414 0.009888 30 4.46 0.0001

171 The SAS System 09:22 Monday, July 11, 2016 386 Analysis for Thiamine Light Exposure The Mixed Procedure Differences of Least Squares Means Effect Fort Treatday Fort Treatday Estimate Std Err DF t Val Pr > |t| Adjustment Adj P Fort*Treatday F D 5 F D 12 0.01981 0.01363 30 1.45 0.1567 Tukey-Kramer 0.8997 Fort*Treatday F D 5 F L 5 0.01288 0.01369 30 0.94 0.3543 Tukey-Kramer 0.9935 Fort*Treatday F D 5 F L 12 0.02278 0.01374 30 1.66 0.1078 Tukey-Kramer 0.8093 Fort*Treatday F D 5 F O 0 0.007112 0.01362 30 0.52 0.6053 Tukey-Kramer 0.9999 Fort*Treatday F D 5 U D 5 0.2853 0.01389 30 20.55 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 5 U D12 0.2875 0.01406 30 20.45 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 5 U L 5 0.2840 0.01377 30 20.62 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 5 U L 12 0.2872 0.01389 30 20.68 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 5 U O 0 0.2841 0.01380 30 20.59 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 12 F L 5 -0.00693 0.01371 30 -0.51 0.6172 Tukey-Kramer 1.0000 Fort*Treatday F D 12 F L 12 0.002968 0.01376 30 0.22 0.8307 Tukey-Kramer 1.0000 Fort*Treatday F D 12 F O 0 -0.01270 0.01363 30 -0.93 0.3592 Tukey-Kramer 0.9940 Fort*Treatday F D 12 U D 5 0.2655 0.01390 30 19.10 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 12 U D 12 0.2677 0.01407 30 19.02 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 12 U L 5 0.2642 0.01379 30 19.16 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 12 U L 12 0.2674 0.01390 30 19.23 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 12 U O 0 0.2643 0.01382 30 19.13 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 5 F L 12 0.009894 0.01382 30 0.72 0.4794 Tukey-Kramer 0.9992 Fort*Treatday F L 5 F O 0 -0.00577 0.01369 30 -0.42 0.6765 Tukey-Kramer 1.0000 Fort*Treatday F L 5 U D 5 0.2725 0.01396 30 19.52 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 5 U D 12 0.2746 0.01413 30 19.43 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 5 U L 5 0.2712 0.01385 30 19.58 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 5 U L 12 0.2743 0.01396 30 19.65 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 5 U O 0 0.2712 0.01388 30 19.55 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 12 F O 0 -0.01566 0.01374 30 -1.14 0.2633 Tukey-Kramer 0.9759 Fort*Treatday F L 12 U D 5 0.2626 0.01401 30 18.75 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 12 U D 12 0.2647 0.01418 30 18.67 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 12 U L 5 0.2613 0.01389 30 18.80 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 12 U L 12 0.2644 0.01401 30 18.88 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 12 U O 0 0.2613 0.01392 30 18.77 <.0001 Tukey-Kramer <.0001 Fort*Treatday F O 0 U D 5 0.2782 0.01389 30 20.04 <.0001 Tukey-Kramer <.0001 Fort*Treatday F O 0 U D 12 0.2804 0.01406 30 19.95 <.0001 Tukey-Kramer <.0001 Fort*Treatday F O 0 U L 5 0.2769 0.01377 30 20.11 <.0001 Tukey-Kramer <.0001 Fort*Treatday F O 0 U L 12 0.2801 0.01389 30 20.17 <.0001 Tukey-Kramer <.0001 Fort*Treatday F O 0 U O 0 0.2770 0.01380 30 20.07 <.0001 Tukey-Kramer <.0001 Fort*Treatday U D 5 U D 12 0.002181 0.01432 30 0.15 0.8799 Tukey-Kramer 1.0000 Fort*Treatday U D 5 U L 5 -0.00129 0.01404 30 -0.09 0.9276 Tukey-Kramer 1.0000 Fort*Treatday U D 5 U L 12 0.001841 0.01415 30 0.13 0.8973 Tukey-Kramer 1.0000 Fort*Treatday U D 5 U O 0 -0.00121 0.01407 30 -0.09 0.9320 Tukey-Kramer 1.0000 Fort*Treatday U D 12 U L 5 -0.00347 0.01421 30 -0.24 0.8089 Tukey-Kramer 1.0000 Fort*Treatday U D 12 U L 12 -0.00034 0.01432 30 -0.02 0.9812 Tukey-Kramer 1.0000 Fort*Treatday U D 12 U O 0 -0.00339 0.01424 30 -0.24 0.8133 Tukey-Kramer 1.0000 Fort*Treatday U L 5 U L 12 0.003127 0.01404 30 0.22 0.8252 Tukey-Kramer 1.0000 Fort*Treatday U L 5 U O 0 0.000075 0.01396 30 0.01 0.9958 Tukey-Kramer 1.0000

172 The SAS System 09:22 Monday, July 11, 2016 392 Analysis for Thiamine Light Exposure

The Mixed Procedure

Differences of Least Squares Means

Effect Fortification Treatday Fortification _Treatday Estimate

Fortificati*Treatday U L 12 U O 0 -0.00305

Differences of Least Squares Means

Standard Effect Fortification Treatday Fortification _Treatday Error DF

Fortificati*Treatday U L 12 U O 0 0.01407 30

Differences of Least Squares Means

Effect Fortification Treatday Fortification _Treatday t Value

Fortificati*Treatday U L 12 U O 0 -0.22

Differences of Least Squares Means

Effect Fortification Treatday Fortification _Treatday Pr > |t|

Fortificati*Treatday U L 12 U O 0 0.8297

Differences of Least Squares Means

Effect Fortification Treatday Fortification _Treatday Adjustment

Fortificati*Treatday U L 12 U O 0 Tukey-Kramer

Differences of Least Squares Means

Effect Fortification Treatday Fortification _Treatday Adj P

Fortificati*Treatday U L 12 U O 0 1.0000

173 The SAS System 09:22 Monday, July 11, 2016 393 Analysis for Thiamine Light Exposure

------Fortification=F ------

The Mixed Procedure

Class Level Information

Class Levels Values Lot 2 a ß Unique_Bottle 20 3 4 7 8 11 12 15 16 19 20 23 24 27 28 31 32 35 36 39 40 Treatday 5 D 5 D 12 L 5 L 12 O 0

Dimensions

Covariance Parameters 2 Columns in X 6 Columns in Z 20 Subjects 1 Max Obs per Subject 78

Number of Observations

Number of Observations Read 78 Number of Observations Used 78 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -261.84176687 1 2 -267.63576629 0.00000401 2 1 -267.63658361 0.00000000

174 The SAS System 09:22 Monday, July 11, 2016 394 Analysis for Thiamine Light Exposure

------Fortification=F ------

The Mixed Procedure Convergence criteria met.

Covariance Parameter Estimates Cov Parm Estimate Lot*Unique_(Treatda) 0.000416 Residual 0.001026

Fit Statistics -2 Res Log Likelihood -267.6 AIC (Smaller is Better) -263.6 AICC (Smaller is Better) -263.5 BIC (Smaller is Better) -261.6

Type 3 Tests of Fixed Effects Num Den Effect DF DF F Value Pr > F Treatday 4 15 0.49 0.7410

Least Squares Means

Effect Treatday Estimate Std Err DF t Value Pr > |t| Treatday D 5 0.3281 0.01307 15 25.11 <.0001 Treatday D 12 0.3084 0.01310 15 23.55 <.0001 Treatday L 5 0.3153 0.01321 15 23.88 <.0001 Treatday L 12 0.3055 0.01329 15 22.98 <.0001 Treatday O 0 0.3211 0.01307 15 24.57 <.0001

Differences of Least Squares Means

Effect Treatday _Treatday Estimate Std Err DF t Value Pr > |t| Adjustment Adj P

Treatday D 5 D 12 0.01968 0.01850 15 1.06 0.3044 Tukey-Kramer 0.8219 Treatday D 5 L 5 0.01282 0.01858 15 0.69 0.5008 Tukey-Kramer 0.9556 Treatday D 5 L 12 0.02265 0.01864 15 1.22 0.2431 Tukey-Kramer 0.7432 Treatday D 5 O 0 0.007038 0.01848 15 0.38 0.7086 Tukey-Kramer 0.9950 Treatday D 12 L 5 -0.00686 0.01860 15 -0.37 0.7174 Tukey-Kramer 0.9956 Treatday D 12 L 12 0.002971 0.01866 15 0.16 0.8756 Tukey-Kramer 0.9998 Treatday D 12 O 0 -0.01264 0.01850 15 -0.68 0.5050 Tukey-Kramer 0.9572 Treatday L 5 L 12 0.009831 0.01874 15 0.52 0.6075 Tukey-Kramer 0.9834 Treatday L 5 O 0 -0.00578 0.01858 15 -0.31 0.7601 Tukey-Kramer 0.9977 Treatday L 12 O 0 -0.01561 0.01864 15 -0.84 0.4155 Tukey-Kramer 0.9147

175 The SAS System 09:22 Monday, July 11, 2016 396 Analysis for Thiamine Light Exposure

------Fortification=U ------

The Mixed Procedure

Model Information

Class Level Information

Class Levels Values Lot 2 a ß Unique_Bottle 20 1 2 5 6 9 10 13 14 17 18 21 22 25 26 29 30 33 34 37 38 Treatday 5 D 5 D 12 L 5 L 12 O 0

Dimensions

Covariance Parameters 2 Columns in X 6 Columns in Z 20 Subjects 1 Max Obs per Subject 66

Number of Observations

Number of Observations Read 66 Number of Observations Used 66 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -408.23779265 1 2 -423.28221684 0.00000000

176 The SAS System 09:22 Monday, July 11, 2016 397 Analysis for Thiamine Light Exposure

------Fortification=U ------

The Mixed Procedure Convergence criteria met.

Covariance Parameter Estimates Cov Parm Estimate Lot*Unique_(Treatda) 0.000033 Residual 0.000032

Fit Statistics -2 Res Log Likelihood -423.3 AIC (Smaller is Better) -419.3 AICC (Smaller is Better) -419.1 BIC (Smaller is Better) -417.3

Type 3 Tests of Fixed Effects Num Den Effect DF DF F Value Pr > F Treatday 4 15 0.22 0.9212

Least Squares Means

Effect Treatday Estimate Std Error DF t Value Pr > |t| Treatday D 5 0.04283 0.003290 15 13.02 <.0001 Treatday D 12 0.04073 0.003335 15 12.21 <.0001 Treatday L 5 0.04402 0.003264 15 13.48 <.0001 Treatday L 12 0.04113 0.003290 15 12.50 <.0001 Treatday O 0 0.04401 0.003273 15 13.45 <.0001

Differences of Least Squares Means

Effect Treatday _Treatday Estimate Std Err DF t Value Pr > |t| Adjustment Adj P

Treatday D 5 D 12 0.002098 0.004685 15 0.45 0.6607 Tukey 0.9908 Treatday D 5 L 5 -0.00119 0.004635 15 -0.26 0.8006 Tukey 0.9989 Treatday D 5 L 12 0.001701 0.004653 15 0.37 0.7198 Tukey 0.9958 Treatday D 5 O 0 -0.00118 0.004641 15 -0.25 0.8028 Tukey 0.9990 Treatday D 12 L 5 -0.00329 0.004667 15 -0.70 0.4917 Tukey 0.9522 Treatday D 12 L 12 -0.00040 0.004685 15 -0.08 0.9336 Tukey 1.0000 Treatday D 12 O 0 -0.00328 0.004673 15 -0.70 0.4938 Tukey 0.9530 Treatday L 5 L 12 0.002893 0.004635 15 0.62 0.5420 Tukey 0.9688 Treatday L 5 O 0 0.000012 0.004623 15 0.00 0.9980 Tukey 1.0000 Treatday L 12 O 0 -0.00288 0.004641 15 -0.62 0.5441 Tukey 0.9694

177 Riboflavin

The SAS System 09:22 Monday, July 11, 2016 339 Analysis for Riboflavin Cooling Data

The Mixed Procedure

Model Information Data Set WORK.GOOD Dependent Variable Riboflavin Covariance Structure Variance Components Estimation Method REML Residual Variance Method Profile Fixed Effects SE Method Model-Based Degrees of Freedom Method Containment

Class Level Information

Class Levels Values Lot 2 a ß Large_Bottle 16 1 10 11 12 13 14 15 16 2 3 4 5 6 7 8 9 Treatment 2 C W Fortification 2 F U

Dimensions Covariance Parameters 2 Columns in X 9 Columns in Z 16 Subjects 1 Max Obs per Subject 72

Number of Observations

Number of Observations Read 72 Number of Observations Used 72 Number of Observations Not Used 0

Iteration History Iteration Evaluations -2 Res Log Like Criterion 0 1 -278.72341645 1 1 -278.72341645 0.00000000

Convergence criteria met.

178 The SAS System 09:22 Monday, July 11, 2016 340 Analysis for Riboflavin Cooling Data

The Mixed Procedure

Covariance Parameter Estimates Cov Parm Estimate Lot*Larg(Trea*Forti) 0 Residual 0.000820

Fit Statistics -2 Res Log Likelihood -278.7 AIC (Smaller is Better) -276.7 AICC (Smaller is Better) -276.7 BIC (Smaller is Better) -276.0

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F Treatment 1 12 0.01 0.9094 Fortification 1 12 927.05 <.0001 Treatment*Fortificat 1 12 0.08 0.7810

Least Squares Means

Effect Treatment Fortification Estimate Std Error DF t Value Pr > |t| Treatment*Fortificat C F 0.2229 0.006748 12 33.03 <.0001 Treatment*Fortificat C U 0.01934 0.006748 12 2.87 0.0142 Treatment*Fortificat W F 0.2256 0.006748 12 33.43 <.0001 Treatment*Fortificat W U 0.01821 0.006748 12 2.70 0.0194

Differences of Least Squares Means

Effect Treat Fort Treat Fort Estimate Std Error DF t Value Pr > |t| Adjustment Adj P Treat*Fort C F C U 0.2035 0.009543 12 21.33 <.0001 Tukey <.0001 Treat*Fort C F W F -0.00270 0.009543 12 -0.28 0.7818 Tukey 0.9917 Treat*Fort C F W U 0.2047 0.009543 12 21.45 <.0001 Tukey <.0001 Treat*Fort C U W F -0.2062 0.009543 12 -21.61 <.0001 Tukey <.0001 Treat*Fort C U W U 0.001134 0.009543 12 0.12 0.9074 Tukey 0.9994 Treat*Fort W F W U 0.2074 0.009543 12 21.73 <.0001 Tukey <.0001

179 The SAS System 09:22 Monday, July 11, 2016 342 Analysis for Riboflavin Cooling Data

------Fortification=F ------

The Mixed Procedure

Class Level Information

Class Levels Values Lot 2 a ß Large_Bottle 8 13 14 15 16 5 6 7 8 Treatment 2 C W

Dimensions

Covariance Parameters 2 Columns in X 3 Columns in Z 8 Subjects 1 Max Obs per Subject 36

Number of Observations

Number of Observations Read 36 Number of Observations Used 36 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -117.12808231 1 1 -117.12808231 0.00000000

Convergence criteria met.

180 The SAS System 09:22 Monday, July 11, 2016 343 Analysis for Riboflavin Cooling Data

------Fortification=F ------

The Mixed Procedure

Covariance Parameter Estimates

Cov Parm Estimate Lot*Large_B(Treatme) 0 Residual 0.001576

Fit Statistics

-2 Res Log Likelihood -117.1 AIC (Smaller is Better) -115.1 AICC (Smaller is Better) -115.0 BIC (Smaller is Better) -115.0

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F Treatment 1 6 0.04 0.8449

Least Squares Means

Standard Effect Treatment Estimate Error DF t Value Pr > |t|

Treatment C 0.2229 0.009357 6 23.82 <.0001 Treatment W 0.2256 0.009357 6 24.11 <.0001

Differences of Least Squares Means

Standard Effect Treatment Treatment Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatment C W -0.00270 0.01323 6 -0.20 0.8449 Tukey 0.8449

181 The SAS System 09:22 Monday, July 11, 2016 344 Analysis for Riboflavin Cooling Data

------Fortification=U ------

The Mixed Procedure

Class Level Information

Class Levels Values Lot 2 a ß Large_Bottle 8 1 10 11 12 2 3 4 9 Treatment 2 C W

Dimensions

Covariance Parameters 2 Columns in X 3 Columns in Z 8 Subjects 1 Max Obs per Subject 36

Number of Observations

Number of Observations Read 36 Number of Observations Used 36 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion

0 1 -226.57313721 1 1 -226.57313721 0.00000000

Convergence criteria met.

182 The SAS System 09:22 Monday, July 11, 2016 345 Analysis for Riboflavin Cooling Data

------Fortification=U ------

The Mixed Procedure

Covariance Parameter Estimates

Cov Parm Estimate Lot*Large_B(Treatme) 0 Residual 0.000063

Fit Statistics

-2 Res Log Likelihood -226.6 AIC (Smaller is Better) -224.6 AICC (Smaller is Better) -224.4 BIC (Smaller is Better) -224.5

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F Treatment 1 6 0.18 0.6833

Least Squares Means

Standard Effect Treatment Estimate Error DF t Value Pr > |t| Treatment C 0.01934 0.001871 6 10.34 <.0001 Treatment W 0.01821 0.001871 6 9.73 <.0001

Differences of Least Squares Means

Standard Effect Treatment Treatment Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatment C W 0.001134 0.002646 6 0.43 0.6833 Tukey 0.6833

183 The SAS System 09:22 Monday, July 11, 2016 346 Analysis for Riboflavin Hot Hold

The Mixed Procedure

Class Level Information

Class Levels Values Lot 2 a ß Unique_Bottle 24 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Treatment 3 1 2 3 Fortification 2 F U

Dimensions

Covariance Parameters 2 Columns in X 12 Columns in Z 24 Subjects 1 Max Obs per Subject 72

Number of Observations

Number of Observations Read 72 Number of Observations Used 72 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion

0 1 -309.47115533 1 2 -323.50447750 0.00000001 2 1 -323.50448053 0.00000000

184 The SAS System 09:22 Monday, July 11, 2016 347 Analysis for Riboflavin Hot Hold

The Mixed Procedure Convergence criteria met.

Covariance Parameter Estimates Cov Parm Estimate Lot*Uniq(Trea*Forti) 0.000231 Residual 0.000240

Fit Statistics -2 Res Log Likelihood -323.5 AIC (Smaller is Better) -319.5 AICC (Smaller is Better) -319.3 BIC (Smaller is Better) -317.1

Type 3 Tests of Fixed Effects Effect Num DF Den DF F Value Pr > F Treatment 2 18 0.01 0.9942 Fortification 1 18 780.95 <.0001 Treatment*Fortificat 2 18 0.04 0.9627

Least Squares Means Effect Fortification Treatment Estimate Std Error DF t Value Pr > |t| Treatment*Fortificat F 1 0.2208 0.008950 18 24.67 <.0001 Treatment*Fortificat U 1 0.01780 0.008822 18 2.02 0.0587 Treatment*Fortificat F 2 0.2201 0.008747 18 25.17 <.0001 Treatment*Fortificat U 2 0.01735 0.008822 18 1.97 0.0649 Treatment*Fortificat F 3 0.2177 0.008822 18 24.68 <.0001 Treatment*Fortificat U 3 0.01902 0.008822 18 2.16 0.0449

Differences of Least Squares Means Effect Fort Treat Fort Treat Estimate Std Error DF t Value Pr > |t| Adjustment Adj P Treat*Fort F 1 U 1 0.2030 0.01257 18 16.15 <.0001 Tukey-Kramer <.0001 Treat*Fort F 1 F 2 0.000687 0.01251 18 0.05 0.9568 Tukey-Kramer 1.0000 Treat*Fort F 1 U 2 0.2035 0.01257 18 16.19 <.0001 Tukey-Kramer <.0001 Treat*Fort F 1 F 3 0.003115 0.01257 18 0.25 0.8070 Tukey-Kramer 0.9998 Treat*Fort F 1 U 3 0.2018 0.01257 18 16.06 <.0001 Tukey-Kramer <.0001 Treat*Fort U 1 F 2 -0.2023 0.01242 18 -16.29 <.0001 Tukey-Kramer <.0001 Treat*Fort U 1 U 2 0.000457 0.01248 18 0.04 0.9712 Tukey-Kramer 1.0000 Treat*Fort U 1 F 3 -0.1999 0.01248 18 -16.02 <.0001 Tukey-Kramer <.0001 Treat*Fort U 1 U 3 -0.00121 0.01248 18 -0.10 0.9236 Tukey-Kramer 1.0000 Treat*Fort F 2 U 2 0.2028 0.01242 18 16.32 <.0001 Tukey-Kramer <.0001 Treat*Fort F 2 F 3 0.002428 0.01242 18 0.20 0.8472 Tukey-Kramer 1.0000 Treat*Fort F 2 U 3 0.2011 0.01242 18 16.19 <.0001 Tukey-Kramer <.0001 Treat*Fort U 2 F 3 -0.2004 0.01248 18 -16.06 <.0001 Tukey-Kramer <.0001 Treat*Fort U 2 U 3 -0.00167 0.01248 18 -0.13 0.8950 Tukey-Kramer 1.0000 Treat*Fort F 3 U 3 0.1987 0.01248 18 15.93 <.0001 Tukey-Kramer <.0001

185 The SAS System 09:22 Monday, July 11, 2016 350 Analysis for Riboflavin Hot Hold

------Fortification=F ------

The Mixed Procedure

Class Level Information

Class Levels Values Lot 2 a ß Unique_Bottle 12 13 14 15 16 17 18 19 20 21 22 23 24 Treatment 3 1 2 3

Dimensions

Covariance Parameters 2 Columns in X 4 Columns in Z 12 Subjects 1 Max Obs per Subject 36

Number of Observations

Number of Observations Read 36 Number of Observations Used 36 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -132.86845552 1 2 -139.68840567 0.00000000

186 The SAS System 09:22 Monday, July 11, 2016 351 Analysis for Riboflavin Hot Hold

------Fortification=F ------

The Mixed Procedure

Convergence criteria met.

Covariance Parameter Estimates

Cov Parm Estimate Lot*Unique_(Treatme) 0.000443 Residual 0.000471

Fit Statistics

-2 Res Log Likelihood -139.7 AIC (Smaller is Better) -135.7 AICC (Smaller is Better) -135.3 BIC (Smaller is Better) -134.7

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F Treatment 2 9 0.02 0.9825

Least Squares Means

Standard Effect Treatment Estimate Error DF t Value Pr > |t| Treatment 1 0.2208 0.01243 9 17.77 <.0001 Treatment 2 0.2201 0.01214 9 18.14 <.0001 Treatment 3 0.2177 0.01224 9 17.78 <.0001

Differences of Least Squares Means

Standard Effect Treatment Treatment Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatment 1 2 0.000671 0.01737 9 0.04 0.9700 Tukey-Kramer 0.9992 Treatment 1 3 0.003106 0.01744 9 0.18 0.8626 Tukey-Kramer 0.9827 Treatment 2 3 0.002435 0.01724 9 0.14 0.8908 Tukey-Kramer 0.9891

187 The SAS System 09:22 Monday, July 11, 2016 352 Analysis for Riboflavin Hot Hold

------Fortification=U ------

The Mixed Procedure

Class Level Information

Class Levels Values Lot 2 a ß Unique_Bottle 12 1 2 3 4 5 6 7 8 9 10 11 12 Treatment 3 1 2 3

Dimensions

Covariance Parameters 2 Columns in X 4 Columns in Z 12 Subjects 1 Max Obs per Subject 36

Number of Observations

Number of Observations Read 36 Number of Observations Used 36 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -247.63624852 1 1 -265.54066807 0.00000000

Convergence criteria met.

188 The SAS System 09:22 Monday, July 11, 2016 353 Analysis for Riboflavin Hot Hold

------Fortification=U ------

The Mixed Procedure

Covariance Parameter Estimates

Cov Parm Estimate Lot*Unique_(Treatme) 0.000021 Residual 8.285E-6

Fit Statistics

-2 Res Log Likelihood -265.5 AIC (Smaller is Better) -261.5 AICC (Smaller is Better) -261.1 BIC (Smaller is Better) -260.6

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F Treatment 2 9 0.12 0.8850

Least Squares Means

Standard Effect Treatment Estimate Error DF t Value Pr > |t| Treatment 1 0.01780 0.002454 9 7.26 <.0001 Treatment 2 0.01735 0.002454 9 7.07 <.0001 Treatment 3 0.01902 0.002454 9 7.75 <.0001

Differences of Least Squares Means

Standard Effect Treatment Treatment Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatment 1 2 0.000457 0.003470 9 0.13 0.8980 Tukey 0.9905 Treatment 1 3 -0.00121 0.003470 9 -0.35 0.7346 Tukey 0.9353 Treatment 2 3 -0.00167 0.003470 9 -0.48 0.6417 Tukey 0.8816

189 The SAS System 09:22 Monday, July 11, 2016 354 Analysis for Riboflavin Light Exposure

The Mixed Procedure

Class Level Information

Dimensions

Covariance Parameters 2 Columns in X 18 Columns in Z 40 Subjects 1 Max Obs per Subject 149

Number of Observations

Number of Observations Read 149 Number of Observations Used 149 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -654.42569058 1 1 -654.42569058 0.00000000

190 The SAS System 09:22 Monday, July 11, 2016 355 Analysis for Riboflavin Light Exposure

The Mixed Procedure

Convergence criteria met.

Covariance Parameter Estimates

Cov Parm Estimate Lot*Uniq(Fort*Treat) 0 Residual 0.000435

Fit Statistics

-2 Res Log Likelihood -654.4 AIC (Smaller is Better) -652.4 AICC (Smaller is Better) -652.4 BIC (Smaller is Better) -650.7

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F Treatday 4 30 5.16 0.0028 Fortification 1 30 3253.27 <.0001 Fortificati*Treatday 4 30 4.17 0.0084

Least Squares Means

Standard Effect Fortification Treatday Estimate Error DF t Value Pr > |t|

Fortificati*Treatday F D 5 0.2265 0.005059 30 44.78 <.0001 Fortificati*Treatday F D 12 0.2185 0.005215 30 41.90 <.0001 Fortificati*Treatday F L 5 0.2076 0.005215 30 39.81 <.0001 Fortificati*Treatday F L 12 0.1901 0.005386 30 35.30 <.0001 Fortificati*Treatday F O 0 0.2318 0.005059 30 45.82 <.0001 Fortificati*Treatday U D 5 0.01887 0.005575 30 3.39 0.0020 Fortificati*Treatday U D 12 0.01805 0.006022 30 3.00 0.0054 Fortificati*Treatday U L 5 0.01958 0.005575 30 3.51 0.0014 Fortificati*Treatday U L 12 0.01724 0.005785 30 2.98 0.0057 Fortificati*Treatday U O 0 0.02041 0.005386 30 3.79 0.0007

191 The SAS System 09:22 Monday, July 11, 2016 356 Analysis for Riboflavin Light Exposure The Mixed Procedure Differences of Least Squares Means Effect Fort Treatday Fort Treatday Estimate Std Error DF t Val Pr > |t| Adjustment Adj P Fort*Treatday F D 5 F D 12 0.008020 0.007266 30 1.10 0.2784 Tukey-Kramer 0.9805 Fort*Treatday F D 5 F L 5 0.01894 0.007266 30 2.61 0.0141 Tukey-Kramer 0.2580 Fort*Treatday F D 5 F L 12 0.03643 0.007389 30 4.93 <.0001 Tukey-Kramer 0.0010 Fort*Treatday F D 5 F O 0 -0.00528 0.007155 30 -0.74 0.4659 Tukey-Kramer 0.9989 Fort*Treatday F D 5 U D 5 0.2077 0.007528 30 27.58 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 5 U D 12 0.2085 0.007865 30 26.51 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 5 U L 5 0.2070 0.007528 30 27.49 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 5 U L 12 0.2093 0.007685 30 27.23 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 5 U O 0 0.2061 0.007389 30 27.90 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 12 F L 5 0.01092 0.007375 30 1.48 0.1492 Tukey-Kramer 0.8895 Fort*Treatday F D 12 F L 12 0.02841 0.007497 30 3.79 0.0007 Tukey-Kramer 0.0203 Fort*Treatday F D 12 F O 0 -0.01330 0.007266 30 -1.83 0.0770 Tukey-Kramer 0.7115 Fort*Treatday F D 12 U D 5 0.1996 0.007634 30 26.15 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 12 U D 12 0.2005 0.007966 30 25.17 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 12 U L 5 0.1989 0.007634 30 26.06 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 12 U L 12 0.2013 0.007789 30 25.84 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 12 U O 0 0.1981 0.007497 30 26.43 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 5 F L 12 0.01749 0.007497 30 2.33 0.0265 Tukey-Kramer 0.3984 Fort*Treatday F L 5 F O 0 -0.02422 0.007266 30 -3.33 0.0023 Tukey-Kramer 0.0597 Fort*Treatday F L 5 U D 5 0.1887 0.007634 30 24.72 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 5 U D 12 0.1896 0.007966 30 23.80 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 5 U L 5 0.1880 0.007634 30 24.63 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 5 U L 12 0.1904 0.007789 30 24.44 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 5 U O 0 0.1872 0.007497 30 24.97 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 12 F O 0 -0.04171 0.007389 30 -5.64 <.0001 Tukey-Kramer 0.0001 Fort*Treatday F L 12 U D 5 0.1712 0.007752 30 22.09 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 12 U D 12 0.1721 0.008079 30 21.30 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 12 U L 5 0.1705 0.007752 30 22.00 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 12 U L 12 0.1729 0.007904 30 21.87 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 12 U O 0 0.1697 0.007617 30 22.28 <.0001 Tukey-Kramer <.0001 Fort*Treatday F O 0 U D 5 0.2129 0.007528 30 28.29 <.0001 Tukey-Kramer <.0001 Fort*Treatday F O 0 U D 12 0.2138 0.007865 30 27.18 <.0001 Tukey-Kramer <.0001 Fort*Treatday F O 0 U L 5 0.2122 0.007528 30 28.19 <.0001 Tukey-Kramer <.0001 Fort*Treatday F O 0 U L 12 0.2146 0.007685 30 27.92 <.0001 Tukey-Kramer <.0001 Fort*Treatday F O 0 U O 0 0.2114 0.007389 30 28.61 <.0001 Tukey-Kramer <.0001 Fort*Treatday U D 5 U D 12 0.000826 0.008206 30 0.10 0.9205 Tukey-Kramer 1.0000 Fort*Treatday U D 5 U L 5 -0.00070 0.007884 30 -0.09 0.9295 Tukey-Kramer 1.0000 Fort*Treatday U D 5 U L 12 0.001639 0.008034 30 0.20 0.8397 Tukey-Kramer 1.0000 Fort*Treatday U D 5 U O 0 -0.00154 0.007752 30 -0.20 0.8442 Tukey-Kramer 1.0000 Fort*Treatday U D 12 U L 5 -0.00153 0.008206 30 -0.19 0.8533 Tukey-Kramer 1.0000 Fort*Treatday U D 12 U L 12 0.000813 0.008350 30 0.10 0.9231 Tukey-Kramer 1.0000 Fort*Treatday U D 12 U O 0 -0.00236 0.008079 30 -0.29 0.7719 Tukey-Kramer 1.0000 Fort*Treatday U L 5 U L 12 0.002343 0.008034 30 0.29 0.7726 Tukey-Kramer 1.0000 Fort*Treatday U L 5 U O 0 -0.00083 0.007752 30 -0.11 0.9151 Tukey-Kramer 1.0000

192 The SAS System 09:22 Monday, July 11, 2016 362 Analysis for Riboflavin Light Exposure

The Mixed Procedure

Differences of Least Squares Means

Effect Fortification Treatday Fortification _Treatday Estimate Fortificati*Treatday U L 12 U O 0 -0.00318

Differences of Least Squares Means

Standard Effect Fortification Treatday Fortification _Treatday Error DF

Fortificati*Treatday U L 12 U O 0 0.007904 30

Differences of Least Squares Means

Effect Fortification Treatday Fortification _Treatday t Value Fortificati*Treatday U L 12 U O 0 -0.40

Differences of Least Squares Means

Effect Fortification Treatday Fortification _Treatday Pr > |t| Fortificati*Treatday U L 12 U O 0 0.6907

Differences of Least Squares Means

Effect Fortification Treatday Fortification _Treatday Adjustment Fortificati*Treatday U L 12 U O 0 Tukey-Kramer

Differences of Least Squares Means

Effect Fortification Treatday Fortification _Treatday Adj P Fortificati*Treatday U L 12 U O 0 1.0000

193 The SAS System 09:22 Monday, July 11, 2016 363 Analysis for Riboflavin Light Exposure

------Fortification=F ------

The Mixed Procedure

Class Level Information

Class Levels Values Lot 2 a ß Unique_Bottle 20 3 4 7 8 11 12 15 16 19 20 23 24 27 28 31 32 35 36 39 40 Treatday 5 D 5 D 12 L 5 L 12 O 0

Dimensions

Covariance Parameters 2 Columns in X 6 Columns in Z 20 Subjects 1 Max Obs per Subject 81

Number of Observations

Number of Observations Read 81 Number of Observations Used 81 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -313.74360880 1 1 -313.74360880 0.00000000

194 The SAS System 09:22 Monday, July 11, 2016 364 Analysis for Riboflavin Light Exposure

------Fortification=F ------

The Mixed Procedure

Convergence criteria met.

Covariance Parameter Estimates

Cov Parm Estimate Lot*Unique_(Treatda) 0 Residual 0.000785

Fit Statistics -2 Res Log Likelihood -313.7 AIC (Smaller is Better) -311.7 AICC (Smaller is Better) -311.7 BIC (Smaller is Better) -310.7

Type 3 Tests of Fixed Effects Effect Num DF Den DF F Value Pr > F Treatday 4 15 5.54 0.0061

Least Squares Means

Effect Treatday Estimate Std Error DF t Value Pr > |t|

Treatday D 5 0.2265 0.006797 15 33.33 <.0001 Treatday D 12 0.2185 0.007007 15 31.19 <.0001 Treatday L 5 0.2076 0.007007 15 29.63 <.0001 Treatday L 12 0.1901 0.007236 15 26.27 <.0001 Treatday O 0 0.2318 0.006797 15 34.11 <.0001

Differences of Least Squares Means

Effect Treatday Treatday Estimate Std Error DF t Value Pr > |t| Adjustment Adj P

Treatday D 5 D 12 0.008020 0.009762 15 0.82 0.4242 Tukey-Kramer 0.9198 Treatday D 5 L 5 0.01894 0.009762 15 1.94 0.0714 Tukey-Kramer 0.3396 Treatday D 5 L 12 0.03643 0.009928 15 3.67 0.0023 Tukey-Kramer 0.0165 Treatday D 5 O 0 -0.00528 0.009613 15 -0.55 0.5906 Tukey-Kramer 0.9803 Treatday D 12 L 5 0.01092 0.009909 15 1.10 0.2880 Tukey-Kramer 0.8031 Treatday D 12 L 12 0.02841 0.01007 15 2.82 0.0129 Tukey-Kramer 0.0817 Treatday D 12 O 0 -0.01330 0.009762 15 -1.36 0.1930 Tukey-Kramer 0.6587 Treatday L 5 L 12 0.01749 0.01007 15 1.74 0.1030 Tukey-Kramer 0.4428 Treatday L 5 O 0 -0.02422 0.009762 15 -2.48 0.0254 Tukey-Kramer 0.1474 Treatday L 12 O 0 -0.04171 0.009928 15 -4.20 0.0008 Tukey-Kramer 0.0059

195 The SAS System 09:22 Monday, July 11, 2016 366 Analysis for Riboflavin Light Exposure

------Fortification=U ------

The Mixed Procedure

Class Level Information

Class Levels Values Lot 2 a ß Unique_Bottle 20 1 2 5 6 9 10 13 14 17 18 21 22 25 26 29 30 33 34 37 38 Treatday 5 D 5 D 12 L 5 L 12 O 0

Dimensions

Covariance Parameters 2 Columns in X 6 Columns in Z 20 Subjects 1 Max Obs per Subject 68

Number of Observations

Number of Observations Read 68 Number of Observations Used 68 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -519.58240631 1 2 -522.15789215 0.00000005 2 1 -522.15790910 0.00000000

196 The SAS System 09:22 Monday, July 11, 2016 367 Analysis for Riboflavin Light Exposure

------Fortification=U ------

The Mixed Procedure

Convergence criteria met.

Covariance Parameter Estimates Cov Parm Estimate Lot*Unique_(Treatda) 2.633E-6 Residual 0.000010

Fit Statistics -2 Res Log Likelihood -522.2 AIC (Smaller is Better) -518.2 AICC (Smaller is Better) -518.0 BIC (Smaller is Better) -516.2

Type 3 Tests of Fixed Effects

Effect Num DF Den DF F Value Pr > F Treatday 4 15 0.99 0.4436

Least Squares Means

Standard Effect Treatday Estimate Std Error DF t Value Pr > |t| Treatday D 5 0.01885 0.001190 15 15.84 <.0001 Treatday D 12 0.01805 0.001233 15 14.64 <.0001 Treatday L 5 0.01947 0.001185 15 16.44 <.0001 Treatday L 12 0.01723 0.001208 15 14.27 <.0001 Treatday O 0 0.02027 0.001168 15 17.36 <.0001

Differences of Least Squares Means

Effect Treatday Treatday Estimate Std Error DF t Value Pr > |t| Adjustment Adj P

Treatday D 5 D 12 0.000805 0.001713 15 0.47 0.6454 Tukey 0.9890 Treatday D 5 L 5 -0.00062 0.001679 15 -0.37 0.7168 Tukey 0.9956 Treatday D 5 L 12 0.001619 0.001696 15 0.96 0.3547 Tukey 0.8706 Treatday D 5 O 0 -0.00141 0.001667 15 -0.85 0.4101 Tukey 0.9113 Treatday D 12 L 5 -0.00143 0.001710 15 -0.83 0.4176 Tukey 0.9160 Treatday D 12 L 12 0.000815 0.001726 15 0.47 0.6436 Tukey 0.9888 Treatday D 12 O 0 -0.00222 0.001698 15 -1.31 0.2113 Tukey 0.6919 Treatday L 5 L 12 0.002240 0.001692 15 1.32 0.2054 Tukey 0.6814 Treatday L 5 O 0 -0.00079 0.001663 15 -0.48 0.6409 Tukey 0.9884 Treatday L 12 O 0 -0.00303 0.001680 15 -1.80 0.0912 Tukey 0.4065

197 Folate

The SAS System 09:22 Monday, July 11, 2016 399 Analysis for Folate Cooling Data

The Mixed Procedure

Model Information Data Set WORK.GOOD Dependent Variable Folate Covariance Structure Variance Components Estimation Method REML Residual Variance Method Profile Fixed Effects SE Method Model-Based Degrees of Freedom Method Containment

Class Level Information

Class Levels Values Lot 2 a ß Large_Bottle 16 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Treatment 2 C W Fortification 2 F U

Dimensions

Covariance Parameters 2 Columns in X 9 Columns in Z 16 Subjects 1 Max Obs per Subject 52

Number of Observations

Number of Observations Read 52 Number of Observations Used 52 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -214.68283011 1 2 -217.46305681 0.00022043 2 1 -217.49963168 0.00000453 3 1 -217.50033360 0.00000000

198 The SAS System 09:22 Monday, July 11, 2016 400 Analysis for Folate Cooling Data

The Mixed Procedure

Convergence criteria met.

Covariance Parameter Estimates

Cov Parm Estimate Lot*Larg(Trea*Forti) 0.000191 Residual 0.000413

Fit Statistics

-2 Res Log Likelihood -217.5 AIC (Smaller is Better) -213.5 AICC (Smaller is Better) -213.2 BIC (Smaller is Better) -212.0

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F Treatment 1 12 0.17 0.6895 Fortification 1 12 2376.34 <.0001 Treatment*Fortificat 1 12 0.06 0.8129

Least Squares Means

Standard Effect Treatment Fortification Estimate Error DF t Value Pr > |t|

Treatment*Fortificat C F 0.6139 0.009019 12 68.07 <.0001 Treatment*Fortificat C U 0.1636 0.009019 12 18.14 <.0001 Treatment*Fortificat W F 0.6079 0.009363 12 64.93 <.0001 Treatment*Fortificat W U 0.1621 0.009363 12 17.31 <.0001

Differences of Least Squares Means

Standard Effect Treat Fort Treat Fort Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treat*Fort C F C U 0.4503 0.01276 12 35.31 <.0001 Tukey-Kramer <.0001 Treat*Fort C F W F 0.005987 0.01300 12 0.46 0.6534 Tukey-Kramer 0.9662 Treat*Fort C F W U 0.4519 0.01300 12 34.76 <.0001 Tukey-Kramer <.0001 Treat*Fort C U W F -0.4444 0.01300 12 -34.18 <.0001 Tukey-Kramer <.0001 Treat*Fort C U W U 0.001539 0.01300 12 0.12 0.9077 Tukey-Kramer 0.9994 Treat*Fort W F W U 0.4459 0.01324 12 33.68 <.0001 Tukey-Kramer <.0001

199 The SAS System 09:22 Monday, July 11, 2016 402 Analysis for Folate Cooling Data

------Fortification=F ------

The Mixed Procedure

Class Level Information

Class Levels Values Lot 2 a ß Large_Bottle 8 5 6 7 8 13 14 15 16 Treatment 2 C W

Dimensions

Covariance Parameters 2 Columns in X 3 Columns in Z 8 Subjects 1 Max Obs per Subject 26

Number of Observations

Number of Observations Read 26 Number of Observations Used 26 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -102.98881429 1 2 -106.58765443 0.00096997 2 1 -106.67152833 0.00007068 3 1 -106.67713107 0.00000048 4 1 -106.67716727 0.00000000

200 The SAS System 09:22 Monday, July 11, 2016 403 Analysis for Folate Cooling Data

------Fortification=F ------

The Mixed Procedure

Convergence criteria met.

Covariance Parameter Estimates

Cov Parm Estimate Lot*Large_B(Treatme) 0.000388 Residual 0.000400

Fit Statistics -2 Res Log Likelihood -106.7 AIC (Smaller is Better) -102.7 AICC (Smaller is Better) -102.1 BIC (Smaller is Better) -102.5

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F Treatment 1 6 0.15 0.7127

Least Squares Means

Standard Effect Treatment Estimate Error DF t Value Pr > |t|

Treatment C 0.6149 0.01143 6 53.80 <.0001 Treatment W 0.6085 0.01170 6 52.02 <.0001

Differences of Least Squares Means

Standard Effect Treatment Treatment Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatment C W 0.006315 0.01635 6 0.39 0.7127 Tukey-Kramer 0.7127

201 The SAS System 09:22 Monday, July 11, 2016 404 Analysis for Folate Cooling Data

------Fortification=U ------

The Mixed Procedure

Model Information

Data Set WORK.GOOD Dependent Variable Folate Covariance Structure Variance Components Estimation Method REML Residual Variance Method Profile Fixed Effects SE Method Model-Based Degrees of Freedom Method Containment

Class Level Information

Class Levels Values Lot 2 a ß Large_Bottle 8 1 2 3 4 9 10 11 12 Treatment 2 C W

Dimensions

Covariance Parameters 2 Columns in X 3 Columns in Z 8 Subjects 1 Max Obs per Subject 26

Number of Observations

Number of Observations Read 26 Number of Observations Used 26 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -112.66222708 1 2 -112.66693207 0.00000001

Convergence criteria met.

202 The SAS System 09:22 Monday, July 11, 2016 405 Analysis for Folate Cooling Data

------Fortification=U ------

The Mixed Procedure

Covariance Parameter Estimates

Cov Parm Estimate Lot*Large_B(Treatme) 7.512E-6 Residual 0.000427

Fit Statistics

-2 Res Log Likelihood -112.7 AIC (Smaller is Better) -108.7 AICC (Smaller is Better) -108.1 BIC (Smaller is Better) -108.5

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F

Treatment 1 6 0.15 0.7094

Least Squares Means

Standard Effect Treatment Estimate Error DF t Value Pr > |t|

Treatment C 0.1665 0.005726 6 29.08 <.0001 Treatment W 0.1632 0.006169 6 26.46 <.0001

Differences of Least Squares Means

Standard Effect Treatment Treatment Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatment C W 0.003290 0.008417 6 0.39 0.7094 Tukey-Kramer 0.7094

203 The SAS System 11:23 Friday, February 8, 2019 10 Analysis for Folate Hot Hold

The Mixed Procedure

Model Information

Class Level Information

Class Levels Values Lot 2 a ß Unique_Bottle 18 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Treatment 3 1 2 3 Fortification 2 F U

Dimensions

Covariance Parameters 2 Columns in X 12 Columns in Z 18 Subjects 1 Max Obs per Subject 60

Number of Observations

Number of Observations Read 60 Number of Observations Used 60 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -200.39202495 1 3 -200.39483212 0.00000000

Convergence criteria met.

204 The SAS System 11:23 Friday, February 8, 2019 11 Analysis for Folate Hot Hold

The Mixed Procedure

Covariance Parameter Estimates Cov Parm Estimate Lot*Uniq(Trea*Forti) 7.525E-6 Residual 0.001103

Fit Statistics -2 Res Log Likelihood -200.4 AIC (Smaller is Better) -196.4 AICC (Smaller is Better) -196.2 BIC (Smaller is Better) -194.6

Type 3 Tests of Fixed Effects Effect Num DF Den DF F Value Pr > F Treatment 2 12 0.07 0.9286 Fortification 1 12 2688.88 <.0001 Treatment*Fortificat 2 12 0.06 0.9460

Least Squares Means

Effect Fortification Treatment Estimate Std Error DF t Value Pr > |t| Treatment*Fortificat F 1 0.6021 0.01063 12 56.63 <.0001 Treatment*Fortificat U 1 0.1516 0.01063 12 14.26 <.0001 Treatment*Fortificat F 2 0.6044 0.01063 12 56.85 <.0001 Treatment*Fortificat U 2 0.1574 0.01063 12 14.80 <.0001 Treatment*Fortificat F 3 0.6055 0.01066 12 56.80 <.0001 Treatment*Fortificat U 3 0.1514 0.01066 12 14.20 <.0001

Differences of Least Squares Means

Effect Fort Treat Fort Treat Estimate Std Error DF t Value Pr > |t| Adjustment Adj P Treat*Fort F 1 U 1 0.4505 0.01504 12 29.96 <.0001 Tukey <.0001 Treat*Fort F 1 F 2 -0.00235 0.01504 12 -0.16 0.8786 Tukey 1.0000 Treat*Fort F 1 U 2 0.4447 0.01504 12 29.57 <.0001 Tukey <.0001 Treat*Fort F 1 F 3 -0.00344 0.01506 12 -0.23 0.8231 Tukey 0.9999 Treat*Fort F 1 U 3 0.4507 0.01506 12 29.94 <.0001 Tukey <.0001 Treat*Fort U 1 F 2 -0.4528 0.01504 12 -30.12 <.0001 Tukey <.0001 Treat*Fort U 1 U 2 -0.00581 0.01504 12 -0.39 0.7061 Tukey 0.9986 Treat*Fort U 1 F 3 -0.4539 0.01506 12 -30.15 <.0001 Tukey <.0001 Treat*Fort U 1 U 3 0.000211 0.01506 12 0.01 0.9890 Tukey 1.0000 Treat*Fort F 2 U 2 0.4470 0.01504 12 29.73 <.0001 Tukey <.0001 Treat*Fort F 2 F 3 -0.00109 0.01506 12 -0.07 0.9432 Tukey 1.0000 Treat*Fort F 2 U 3 0.4530 0.01506 12 30.09 <.0001 Tukey <.0001 Treat*Fort U 2 F 3 -0.4481 0.01506 12 -29.76 <.0001 Tukey <.0001 Treat*Fort U 2 U 3 0.006018 0.01506 12 0.40 0.6964 Tukey 0.9983 Treat*Fort F 3 U 3 0.4541 0.01508 12 30.12 <.0001 Tukey <.0001

205 The SAS System 11:23 Friday, February 8, 2019 14 Analysis for Folate Hot Hold

------Fortification=F ------

The Mixed Procedure

Class Level Information

Class Levels Values Lot 2 a ß Unique_Bottle 9 4 5 6 10 11 12 16 17 18 Treatment 3 1 2 3

Dimensions

Covariance Parameters 2 Columns in X 4 Columns in Z 9 Subjects 1 Max Obs per Subject 30

Number of Observations

Number of Observations Read 30 Number of Observations Used 30 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -89.89874180 1 1 -89.89874180 0.00000000

Convergence criteria met.

206 The SAS System 11:23 Friday, February 8, 2019 15 Analysis for Folate Hot Hold

------Fortification=F ------

The Mixed Procedure

Covariance Parameter Estimates

Cov Parm Estimate Lot*Unique_(Treatme) 0 Residual 0.001623

Fit Statistics

-2 Res Log Likelihood -89.9 AIC (Smaller is Better) -87.9 AICC (Smaller is Better) -87.7 BIC (Smaller is Better) -87.7

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F Treatment 2 6 0.02 0.9816

Least Squares Means

Standard Effect Treatment Estimate Error DF t Value Pr > |t|

Treatment 1 0.6020 0.01274 6 47.25 <.0001 Treatment 2 0.6044 0.01274 6 47.44 <.0001 Treatment 3 0.6054 0.01274 6 47.51 <.0001

Differences of Least Squares Means

Standard Effect Treatment Treatment Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatment 1 2 -0.00244 0.01802 6 -0.14 0.8968 Tukey 0.9900 Treatment 1 3 -0.00337 0.01802 6 -0.19 0.8579 Tukey 0.9810 Treatment 2 3 -0.00093 0.01802 6 -0.05 0.9605 Tukey 0.9985

207 The SAS System 11:23 Friday, February 8, 2019 16 Analysis for Folate Hot Hold

------Fortification=U ------

The Mixed Procedure

Class Level Information

Class Levels Values Lot 2 a ß Unique_Bottle 9 1 2 3 7 8 9 13 14 15 Treatment 3 1 2 3

Dimensions

Covariance Parameters 2 Columns in X 4 Columns in Z 9 Subjects 1 Max Obs per Subject 30

Number of Observations

Number of Observations Read 30 Number of Observations Used 30 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -117.04912634 1 3 -123.87311819 0.00063545 2 1 -123.93557547 0.00003607 3 1 -123.93883151 0.00000014 4 1 -123.93884410 0.00000000

208 The SAS System 11:23 Friday, February 8, 2019 17 Analysis for Folate Hot Hold

------Fortification=U ------

The Mixed Procedure

Convergence criteria met.

Covariance Parameter Estimates

Cov Parm Estimate Lot*Unique_(Treatme) 0.000332 Residual 0.000340

Fit Statistics -2 Res Log Likelihood -123.9 AIC (Smaller is Better) -119.9 AICC (Smaller is Better) -119.4 BIC (Smaller is Better) -119.5

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F Treatment 2 6 0.08 0.9272

Least Squares Means

Standard Effect Treatment Estimate Error DF t Value Pr > |t|

Treatment 1 0.1512 0.01214 6 12.46 <.0001 Treatment 2 0.1563 0.01214 6 12.88 <.0001 Treatment 3 0.1576 0.01235 6 12.76 <.0001

Differences of Least Squares Means

Standard Effect Treatment Treatment Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatment 1 2 -0.00509 0.01717 6 -0.30 0.7771 Tukey-Kramer 0.9532 Treatment 1 3 -0.00639 0.01732 6 -0.37 0.7250 Tukey-Kramer 0.9287 Treatment 2 3 -0.00130 0.01732 6 -0.08 0.9426 Tukey-Kramer 0.996

209 The SAS System 09:22 Monday, July 11, 2016 415 Analysis for Folate Light Exposure

The Mixed Procedure

Class Level Information

Dimensions

Covariance Parameters 2 Columns in X 18 Columns in Z 40 Subjects 1 Max Obs per Subject 145

Number of Observations

Number of Observations Read 145 Number of Observations Used 145 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -619.24894381 1 1 -619.24894381 0.00000000

210 The SAS System 09:22 Monday, July 11, 2016 416 Analysis for Folate Light Exposure

The Mixed Procedure

Convergence criteria met.

Covariance Parameter Estimates

Cov Parm Estimate Lot*Uniq(Fort*Treat) 0 Residual 0.000489

Fit Statistics

-2 Res Log Likelihood -619.2 AIC (Smaller is Better) -617.2 AICC (Smaller is Better) -617.2 BIC (Smaller is Better) -615.6

Type 3 Tests of Fixed Effects

Num Den Effect DF DF F Value Pr > F Treatday 4 30 2.45 0.0677 Fortification 1 30 14775.2 <.0001 Fortificati*Treatday 4 30 1.83 0.1485

Least Squares Means

Standard Effect Fortification Treatday Estimate Error DF t Value Pr > |t|

Fortificati*Treatday F D 5 0.5874 0.005529 30 106.23 <.0001 Fortificati*Treatday F D 12 0.5830 0.005529 30 105.45 <.0001 Fortificati*Treatday F L 5 0.5849 0.006134 30 95.35 <.0001 Fortificati*Treatday F L 12 0.5667 0.005911 30 95.88 <.0001 Fortificati*Treatday F O 0 0.5950 0.005911 30 100.67 <.0001 Fortificati*Treatday U D 5 0.1263 0.005911 30 21.37 <.0001 Fortificati*Treatday U D 12 0.1413 0.005529 30 25.55 <.0001 Fortificati*Treatday U L 5 0.1404 0.005911 30 23.76 <.0001 Fortificati*Treatday U L 12 0.1328 0.005911 30 22.47 <.0001 Fortificati*Treatday U O 0 0.1383 0.005911 30 23.40 <.0001

211 The SAS System 09:22 Monday, July 11, 2016 417 Analysis for Folate Light Exposure The Mixed Procedure Differences of Least Squares Means Effect Fort Treatday Fort Treatday Estimate Std Error DF t Val Pr > |t| Adjustment Adj P Fort*Treatday F D 5 F D 12 0.004354 0.007820 30 0.56 0.5818 Tukey-Kramer 0.9999 Fort*Treatday F D 5 F L 5 0.002499 0.008258 30 0.30 0.7643 Tukey-Kramer 1.0000 Fort*Treatday F D 5 F L 12 0.02064 0.008094 30 2.55 0.0161 Tukey-Kramer 0.2836 Fort*Treatday F D 5 F O 0 -0.00765 0.008094 30 -0.95 0.3520 Tukey-Kramer 0.9933 Fort*Treatday F D 5 U D 5 0.4611 0.008094 30 56.97 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 5 U D 12 0.4461 0.007820 30 57.05 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 5 U L 5 0.4469 0.008094 30 55.22 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 5 U L 12 0.4546 0.008094 30 56.16 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 5 U O 0 0.4491 0.008094 30 55.48 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 12 F L 5 -0.00186 0.008258 30 -0.22 0.8237 Tukey-Kramer 1.0000 Fort*Treatday F D 12 F L 12 0.01629 0.008094 30 2.01 0.0532 Tukey-Kramer 0.5974 Fort*Treatday F D 12 F O 0 -0.01201 0.008094 30 -1.48 0.1484 Tukey-Kramer 0.8882 Fort*Treatday F D 12 U D 5 0.4567 0.008094 30 56.43 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 12 U D 12 0.4418 0.007820 30 56.50 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 12 U L 5 0.4426 0.008094 30 54.68 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 12 U L 12 0.4502 0.008094 30 55.63 <.0001 Tukey-Kramer <.0001 Fort*Treatday F D 12 U O 0 0.4447 0.008094 30 54.95 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 5 F L 12 0.01815 0.008519 30 2.13 0.0415 Tukey-Kramer 0.5221 Fort*Treatday F L 5 F O 0 -0.01015 0.008519 30 -1.19 0.2427 Tukey-Kramer 0.9681 Fort*Treatday F L 5 U D 5 0.4586 0.008519 30 53.83 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 5 U D 12 0.4436 0.008258 30 53.72 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 5 U L 5 0.4444 0.008519 30 52.17 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 5 U L 12 0.4521 0.008519 30 53.07 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 5 U O 0 0.4466 0.008519 30 52.42 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 12 F O 0 -0.02830 0.008359 30 -3.39 0.0020 Tukey-Kramer 0.0531 Fort*Treatday F L 12 U D 5 0.4404 0.008359 30 52.69 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 12 U D 12 0.4255 0.008094 30 52.57 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 12 U L 5 0.4263 0.008359 30 51.00 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 12 U L 12 0.4339 0.008359 30 51.91 <.0001 Tukey-Kramer <.0001 Fort*Treatday F L 12 U O 0 0.4284 0.008359 30 51.25 <.0001 Tukey-Kramer <.0001 Fort*Treatday F O 0 U D 5 0.4687 0.008359 30 56.07 <.0001 Tukey-Kramer <.0001 Fort*Treatday F O 0 U D 12 0.4538 0.008094 30 56.06 <.0001 Tukey-Kramer <.0001 Fort*Treatday F O 0 U L 5 0.4546 0.008359 30 54.38 <.0001 Tukey-Kramer <.0001 Fort*Treatday F O 0 U L 12 0.4622 0.008359 30 55.30 <.0001 Tukey-Kramer <.0001 Fort*Treatday F O 0 U O 0 0.4567 0.008359 30 54.64 <.0001 Tukey-Kramer <.0001 Fort*Treatday U D 5 U D 12 -0.01496 0.008094 30 -1.85 0.0745 Tukey-Kramer 0.7011 Fort*Treatday U D 5 U L 5 -0.01415 0.008359 30 -1.69 0.1010 Tukey-Kramer 0.7911 Fort*Treatday U D 5 U L 12 -0.00650 0.008359 30 -0.78 0.4431 Tukey-Kramer 0.9984 Fort*Treatday U D 5 U O 0 -0.01200 0.008359 30 -1.44 0.1615 Tukey-Kramer 0.9059 Fort*Treatday U D 12 U L 5 0.000813 0.008094 30 0.10 0.9206 Tukey-Kramer 1.0000 Fort*Treatday U D 12 U L 12 0.008461 0.008094 30 1.05 0.3042 Tukey-Kramer 0.9865 Fort*Treatday U D 12 U O 0 0.002958 0.008094 30 0.37 0.7173 Tukey-Kramer 1.0000 Fort*Treatday U L 5 U L 12 0.007648 0.008359 30 0.91 0.3676 Tukey-Kramer 0.9947 Fort*Treatday U L 5 U O 0 0.002145 0.008359 30 0.26 0.7993 Tukey-Kramer 1.0000

212 The SAS System 09:22 Monday, July 11, 2016 423 Analysis for Folate Light Exposure

The Mixed Procedure

Differences of Least Squares Means

Effect Fortification Treatday Fortification _Treatday Estimate

Fortificati*Treatday U L 12 U O 0 -0.00550

Differences of Least Squares Means

Standard Effect Fortification Treatday Fortification _Treatday Error DF

Fortificati*Treatday U L 12 U O 0 0.008359 30

Differences of Least Squares Means

Effect Fortification Treatday Fortification _Treatday t Value

Fortificati*Treatday U L 12 U O 0 -0.66

Differences of Least Squares Means

Effect Fortification Treatday Fortification _Treatday Pr > |t|

Fortificati*Treatday U L 12 U O 0 0.5154

Differences of Least Squares Means

Effect Fortification Treatday Fortification _Treatday Adjustment

Fortificati*Treatday U L 12 U O 0 Tukey-Kramer

Differences of Least Squares Means

Effect Fortification Treatday Fortification _Treatday Adj P

Fortificati*Treatday U L 12 U O 0 0.9996

213 The SAS System 09:22 Monday, July 11, 2016 424 Analysis for Folate Light Exposure

------Fortification=F ------

The Mixed Procedure

Model Information

Class Level Information

Class Levels Values Lot 2 a ß Unique_Bottle 20 3 4 7 8 11 12 15 16 19 20 23 24 27 28 31 32 35 36 39 40 Treatday 5 D 5 D 12 L 5 L 12 O 0

Dimensions

Covariance Parameters 2 Columns in X 6 Columns in Z 20 Subjects 1 Max Obs per Subject 73

Number of Observations

Number of Observations Read 73 Number of Observations Used 73 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -287.41793385 1 1 -287.41793385 0.00000000

214 The SAS System 09:22 Monday, July 11, 2016 425 Analysis for Folate Light Exposure

------Fortification=F ------

The Mixed Procedure

Convergence criteria met.

Covariance Parameter Estimates Cov Parm Estimate Lot*Unique_(Treatda) 0 Residual 0.000702

Fit Statistics -2 Res Log Likelihood -287.4 AIC (Smaller is Better) -285.4 AICC (Smaller is Better) -285.4 BIC (Smaller is Better) -284.4

Type 3 Tests of Fixed Effects Effect Num DF Den DF F Value Pr > F Treatday 4 15 2.16 0.1234

Least Squares Means

Effect Treatday Estimate Std Error DF t Value Pr > |t|

Treatday D 5 0.5874 0.006624 15 88.68 <.0001 Treatday D 12 0.5830 0.006624 15 88.02 <.0001 Treatday L 5 0.5849 0.007349 15 79.59 <.0001 Treatday L 12 0.5667 0.007081 15 80.03 <.0001 Treatday O 0 0.5950 0.007081 15 84.03 <.0001

Differences of Least Squares Means

Effect Treatday Treatday Estimate Std Error DF t Value Pr > |t| Adjustment Adj P

Treatday D 5 D 12 0.004354 0.009368 15 0.46 0.6487 Tukey-Kramer 0.9894 Treatday D 5 L 5 0.002499 0.009894 15 0.25 0.8041 Tukey-Kramer 0.9990 Treatday D 5 L 12 0.02064 0.009697 15 2.13 0.0502 Tukey-Kramer 0.2585 Treatday D 5 O 0 -0.00765 0.009697 15 -0.79 0.4423 Tukey-Kramer 0.9298 Treatday D 12 L 5 -0.00186 0.009894 15 -0.19 0.8537 Tukey-Kramer 0.9997 Treatday D 12 L 12 0.01629 0.009697 15 1.68 0.1137 Tukey-Kramer 0.4739 Treatday D 12 O 0 -0.01201 0.009697 15 -1.24 0.2346 Tukey-Kramer 0.7303 Treatday L 5 L 12 0.01815 0.01021 15 1.78 0.0957 Tukey-Kramer 0.4204 Treatday L 5 O 0 -0.01015 0.01021 15 -0.99 0.3357 Tukey-Kramer 0.8537 Treatday L 12 O 0 -0.02830 0.01001 15 -2.83 0.0128 Tukey-Kramer 0.0809

215 The SAS System 09:22 Monday, July 11, 2016 427 Analysis for Folate Light Exposure

------Fortification=U ------

The Mixed Procedure

Class Level Information

Class Levels Values Lot 2 a ß Unique_Bottle 20 1 2 5 6 9 10 13 14 17 18 21 22 25 26 29 30 33 34 37 38 Treatday 5 D 5 D 12 L 5 L 12 O 0

Dimensions Covariance Parameters 2 Columns in X 6 Columns in Z 20 Subjects 1 Max Obs per Subject 72

Number of Observations

Number of Observations Read 72 Number of Observations Used 72 Number of Observations Not Used 0

Iteration History

Iteration Evaluations -2 Res Log Like Criterion 0 1 -346.31428031 1 3 -352.98409727 0.00076688 2 1 -353.19396862 0.00005766 3 1 -353.20842736 0.00000041 4 1 -353.20852643 0.00000000

216 The SAS System 09:22 Monday, July 11, 2016 428 Analysis for Folate Light Exposure

------Fortification=U ------

The Mixed Procedure Convergence criteria met.

Covariance Parameter Estimates Cov Parm Estimate Lot*Unique_(Treatda) 0.000084 Residual 0.000203

Fit Statistics -2 Res Log Likelihood -353.2 AIC (Smaller is Better) -349.2 AICC (Smaller is Better) -349.0 BIC (Smaller is Better) -347.2

Type 3 Tests of Fixed Effects Effect Num DF Den DF F Value Pr > F Treatday 4 15 0.94 0.4680

Least Squares Means

Effect Treatday Estimate Std Error DF t Value Pr > |t|

Treatday D 5 0.1282 0.006124 15 20.93 <.0001 Treatday D 12 0.1411 0.005904 15 23.89 <.0001 Treatday L 5 0.1400 0.006017 15 23.27 <.0001 Treatday L 12 0.1321 0.006017 15 21.95 <.0001 Treatday O 0 0.1405 0.006124 15 22.94 <.0001

Differences of Least Squares Means

Standard Effect Treatday _Treatday Estimate Error DF t Value Pr > |t| Adjustment Adj P

Treatday D 5 D 12 -0.01289 0.008507 15 -1.51 0.1506 Tukey-Kramer 0.5690 Treatday D 5 L 5 -0.01183 0.008585 15 -1.38 0.1884 Tukey-Kramer 0.6497 Treatday D 5 L 12 -0.00387 0.008585 15 -0.45 0.6587 Tukey-Kramer 0.9906 Treatday D 5 O 0 -0.01232 0.008661 15 -1.42 0.1754 Tukey-Kramer 0.6237 Treatday D 12 L 5 0.001055 0.008430 15 0.13 0.9020 Tukey-Kramer 0.9999 Treatday D 12 L 12 0.009019 0.008430 15 1.07 0.3016 Tukey-Kramer 0.8188 Treatday D 12 O 0 0.000568 0.008507 15 0.07 0.9476 Tukey-Kramer 1.0000 Treatday L 5 L 12 0.007963 0.008509 15 0.94 0.3642 Tukey-Kramer 0.8785 Treatday L 5 O 0 -0.00049 0.008585 15 -0.06 0.9555 Tukey-Kramer 1.0000 Treatday L 12 O 0 -0.00845 0.008585 15 -0.98 0.3406 Tukey-Kramer 0.8582

217 Sensory Panel Demographic Question Results

Project: 2401.2405 DEM

Question Number: 2 Question Type: Multiple Choice (Demographic) Question Title: How old are you?

C hoic es Number Value Choices 1 [6] Less than 7 years old 2 [5] 7 years old 3 [4] 8 years old 4 [3] 9 years old 5 [2] 10 years old 6 [1] 11 years old 7 [7] 12 years old 8 [8] 13 years old 9 [9] 14 years old 10 [10] 15 years old 11 [11] 16 years old 12 [12] 17 years old 13 [13] 18 years old or older

C r os s tabulation 1 2 3 4 5 6 7 8 9 10 Sample [6] [5] [4] [3] [2] [1] [7] [8] [9] [10] n/a 6 6 6 6 7 4 11 4 5 TOTALS 6 6 6 6 7 4 11 4 5

11 12 13 Sample [11] [12] [13] Total n/a 2 57 TOTALS 2 57

P e r c e n t a g e C r o s s t a b u l a t i o n 1 2 3 4 5 6 7 8 9 10 Sample [6] [5] [4] [3] [2] [1] [7] [8] [9] [10] n/a 10.5 10.5 10.5 10.5 12.3 7.0 19.3 7.0 8.8

11 12 13 Sample [11] [12] [13] Total n/a 3.5 100

218 Question Number: 3 Question Type: Multiple Choice (Demographic) Question Title: Are you a boy or a girl?

Choices Number Value Choices 1 [2] Girl 2 [1] Boy

Crosstabulation

1 2 Sample [2] [1] Total n/a 27 30 57 TOTALS 27 30 57

Percentage Crosstabulation

1 2 Sample [2] [1] Total n/a 47.4 52.6 100

Question Number: 4 Question Type: Multiple Choice (Demographic) Question Title: Have you had SOY MILK before?

Choices

Number Value Choices 1 [1] Yes 2 [2] No

Crosstabulation

1 2 Sample [1] [2] Total n/a 23 34 57 TOTALS 23 34 57

Percentage Crosstabulation

1 2 Sample [1] [2] Total n/a 40.4 59.7 100

219 Question Number: 5 Question Type: Multiple Choice (Demographic) Question Title: What do you think about SOY MILK?

Choices

Number Value Choices 1 [1] I like soy milk 2 [2] I'm not sure if I like soy milk or not 3 [3] I do not like soy milk

Crosstabulation

1 2 3 Sample [1] [2] [3] Total n/a 9 46 2 57 TOTALS 9 46 2 57

Percentage Crosstabulation

1 2 3 Sample [1] [2] [3] Total n/a 15.8 80.7 3.5 100

220 Project: 2401.2405 DEM

Question Number: 6 Question Type: Multiple Choice (Demographic) Question Title: What do you think about CHOCOLATE PUDDING?

Choices

Number Value Choices 1 [3] I like chocolate pudding 2 [2] I am not sure if I like it or not 3 [1] I do not like chocolate pudding

Crosstabulation

1 2 3 Sample [3] [2] [1] Total n/a 53 3 1 57 TOTALS 53 3 1 57

Percentage Crosstabulation

1 2 3 Sample [3] [2] [1] Total n/a 93.0 5.3 1.8 100

221

Project: 2401.2405 DEM

Question Number: 7 Question Type: Multiple Choice (Demographic) Question Title: How often do you eat CHOCOLATE PUDDING?

Choices

Number Value Choices 1 [1] More than once a week 2 [2] Once a week to every two weeks 3 [3] Once every two weeks to once a month 4 [4] Once a month to once every three months 5 [5] Less than once every three months

Crosstabulation

1 2 3 4 5 Sample [1] [2] [3] [4] [5] Total n/a 5 11 20 21 57 TOTALS 5 11 20 21 57

Percentage Crosstabulation

1 2 3 4 5 Sample [1] [2] [3] [4] [5] Total n/a 8.8 19.3 35.1 36.8 100

Sensory Panel Hedonic Question Results

Data analyzed using Compusense Five (5.6) on 2013/08/16, (c) Compusense Inc. 1986-2013.

Project: 2401 SM

Product Questionnaire Question Number: 1 Question Type: Category / Hedonics Question Title: OVERALL, what do you think about each sample? Attribute Number: 1 Attribute Title: Q#1.1 Design: T=2, K=2, B=72

P r oduc ts Product Code Name 1 - 375 375 Fortified 2 - 861 861 Unfortified

S c al e P a r a m e t e r s Value Descriptor 7 Really good 6 Good 5 Just a little good 4 Maybe good or maybe bad 3 Just a little bad 2 Bad 1 Really bad

Note: Numbers shown in brackets are the 'values' associated with the category selected.

C r o s s t a b u l a t i o n 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 375 10 21 14 3 5 1 3 57 2 - 861 6 23 16 5 4 1 2 57 TOTALS 16 44 30 8 9 2 5 114

P e r c e n t a g e C r o s s t a b u l a t i o n 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 375 17.5 36.8 24.6 5.3 8.8 1.8 5.3 100 2 - 861 10.5 40.4 28.1 8.8 7.0 1.8 3.5 100

C o u n t s , M e d i a n s , M e a n s a n d S D ' s Sample Standard Number Count Total Median Mean Deviation 1 - 375 57 298.00 6.00 5.23 1.570

223

Sample Standard Number Count Total Median Mean Deviation 2 - 861 57 296.00 6.00 5.19 1.381

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares F Value p-value Samples 1 0.035 0.035 0.05 0.8295 Judges 56 202.947 3.624 4.84 0.0000 Error 56 41.965 0.749 Total 113 244.947 2.168 Standard 0.114 Error (SEM) =

Multiple comparison tests may appear below. Tukey's HSD controls for maximum experimentwise error rate and can be used without F protection. Standard practice recommends that LSD and Duncan's be considered only if the ANOVA p-value is deemed acceptable to control for experimentwise error rates (under the complete null hypothesis). If automatic significance is selected, an available significance level is chosen for the multiple comparison test based on the observed p-value.

Tukey's HSD = 0.325 (5% Significance Level)

Sample Mean Significantly Different Than Sample 1 - 375 5.23 a 2 - 861 5.19 a

224

Project: 2401 SM

Question Number: 2 Question Type: Category / Hedonics Question Title: What do you think about the the FLAVOR of each sample? Attribute Number: 1 Attribute Title: Q#2.1 Design: T=2, K=2, B=72

Products Product Code Name 1 - 375 375 Fortified 2 - 861 861 Unfortified

Scale Parameters Value Descriptor 7 Really good 6 Good 5 Just a little good 4 Maybe good or maybe bad 3 Just a little bad 2 Bad 1 Really bad

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 375 11 17 13 5 7 3 1 57 2 - 861 6 27 7 5 5 6 1 57 TOTALS 17 44 20 10 12 9 2 114

Percentage Crosstabulation 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 375 19.3 29.8 22.8 8.8 12.3 5.3 1.8 100 2 - 861 10.5 47.4 12.3 8.8 8.8 10.5 1.8 100

Counts, Medians, Means and SD's Sample Standard Number Count Total Median Mean Deviation 1 - 375 57 292.00 5.00 5.12 1.559

225

Sample Standard Number Count Total Median Mean Deviation 2 - 861 57 287.00 6.00 5.04 1.614

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares F Value p-value Samples 1 0.219 0.219 0.34 0.5630 Judges 56 245.789 4.389 6.77 0.0000 Error 56 36.281 0.648 Total 113 282.289 2.498 Standard 0.106 Error (SEM) =

Tukey's HSD = 0.302 (5% Significance Level)

Sample Mean Significantly Different Than Sample 1 - 375 5.12 a 2 - 861 5.04 a

226

Project: 2401 SM

Question Number: 3 Question Type: Category / Hedonics Question Title: What do you think about the COLOR of each sample? Attribute Number: 1 Attribute Title: Q#3.1 Design: T=2, K=2, B=72

Products Product Code Name 1 - 375 375 Fortified 2 - 861 861 Unfortified

Scale Parameters Value Descriptor 7 Really good 6 Good 5 Just a little good 4 Maybe good or maybe bad 3 Just a little bad 2 Bad 1 Really bad

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 375 13 24 8 7 5 57 2 - 861 19 25 6 6 1 57 TOTALS 32 49 14 13 6 114

Percentage Crosstabulation 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 375 22.8 42.1 14.0 12.3 8.8 100 2 - 861 33.3 43.9 10.5 10.5 1.8 100

Counts, Medians, Means and SD's Sample Standard Number Count Total Median Mean Deviation 1 - 375 57 318.00 6.00 5.58 1.224

227

Sample Standard Number Count Total Median Mean Deviation 2 - 861 57 340.00 6.00 5.96 1.017

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares F Value p-value Samples 1 4.246 4.246 7.99 0.0065 Judges 56 112.070 2.001 3.77 0.0000 Error 56 29.754 0.531 Total 113 146.070 1.293 Standard 0.096 Error (SEM) =

Tukey's HSD = 0.274 (5% Significance Level)

Sample Mean Significantly Different Than Sample 2 - 861 5.96 a 1 1 - 375 5.58 b

228

Project: 2401 SM

Question Number: 4 Question Type: Category / Hedonics Question Title: What do you think about the way each sample SMELLS? Attribute Number: 1 Attribute Title: Q#4.1 Design: T=2, K=2, B=72

Products Product Code Name 1 - 375 375 Fortified 2 - 861 861 Unfortified

Scale Parameters Value Descriptor 7 Really good 6 Good 5 Just a little good 4 Maybe good or maybe bad 3 Just a little bad 2 Bad 1 Really bad

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 375 7 23 5 11 9 2 57 2 - 861 4 27 5 16 4 1 57 TOTALS 11 50 10 27 13 3 114

Percentage Crosstabulation 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 375 12.3 40.4 8.8 19.3 15.8 3.5 100 2 - 861 7.0 47.4 8.8 28.1 7.0 1.8 100

Counts, Medians, Means and SD's Sample Standard Number Count Total Median Mean Deviation 1 - 375 57 287.00 6.00 5.04 1.439

229

Sample Standard Number Count Total Median Mean Deviation 2 - 861 57 293.00 6.00 5.14 1.217

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares F Value p-value Samples 1 0.316 0.316 0.72 0.4009 Judges 56 174.123 3.109 7.05 0.0000 Error 56 24.684 0.441 Total 113 199.123 1.762 Standard 0.087 Error (SEM) =

Tukey's HSD = 0.249 (5% Significance Level)

Sample Mean Significantly Different Than Sample 2 - 861 5.14 a 1 - 375 5.04 a

230

Project: 2401 SM

Question Number: 5 Question Type: Category / Hedonics Question Title: What do you think about the way each sample FEELS IN YOUR MOUTH? Attribute Number: 1 Attribute Title: Q#5.1 Design: T=2, K=2, B=72

Products Product Code Name 1 - 375 375 Fortified 2 - 861 861 Unfortified

Scale Parameters Value Descriptor 7 Really good 6 Good 5 Just a little good 4 Maybe good or maybe bad 3 Just a little bad 2 Bad 1 Really bad

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 375 23 18 6 3 5 1 1 57 2 - 861 20 15 11 5 5 1 57 TOTALS 43 33 17 8 10 1 2 114

Percentage Crosstabulation 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 375 40.4 31.6 10.5 5.3 8.8 1.8 1.8 100 2 - 861 35.1 26.3 19.3 8.8 8.8 1.8 100

Counts, Medians, Means and SD's Sample Standard Number Count Total Median Mean Deviation 1 - 375 57 329.00 6.00 5.77 1.488

231

Sample Standard Number Count Total Median Mean Deviation 2 - 861 57 321.00 6.00 5.63 1.422

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares F Value p-value Samples 1 0.561 0.561 1.80 0.1848 Judges 56 219.860 3.926 12.61 0.0000 Error 56 17.439 0.311 Total 113 237.860 2.105 Standard 0.073 Error (SEM) =

Tukey's HSD = 0.21 (5% Significance Level)

Sample Mean Significantly Different Than Sample 1 - 375 5.77 a 2 - 861 5.63 a

232

Project: 2401 SM

Question Number: 6 Question Type: Category / Hedonics Question Title: What do you think about the TASTE IN YOUR MOUTH AFTER YOU HAVE SWALLOWED each sample? Attribute Number: 1 Attribute Title: Q#6.1 Design: T=2, K=2, B=72

Products Product Code Name 1 - 375 375 Fortified 2 - 861 861 Unfortified

Scale Parameters Value Descriptor 7 Really good 6 Good 5 Just a little good 4 Maybe good or maybe bad 3 Just a little bad 2 Bad 1 Really bad

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 375 8 15 8 11 4 9 2 57 2 - 861 8 12 12 7 7 8 3 57 TOTALS 16 27 20 18 11 17 5 114

Percentage Crosstabulation 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 375 14.0 26.3 14.0 19.3 7.0 15.8 3.5 100 2 - 861 14.0 21.1 21.1 12.3 12.3 14.0 5.3 100

Counts, Medians, Means and SD's Sample Standard Number Count Total Median Mean Deviation 1 - 375 57 262.00 5.00 4.60 1.781 2 - 861 57 256.00 5.00 4.49 1.814

233

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares F Value p-value Samples 1 0.316 0.316 0.58 0.4509 Judges 56 331.281 5.916 10.80 0.0000 Error 56 30.684 0.548 Total 113 362.281 3.206 Standard 0.098 Error (SEM) =

Tukey's HSD = 0.278 (5% Significance Level)

Sample Mean Significantly Different Than Sample 1 - 375 4.60 a 2 - 861 4.49 a

234

Project: 2401 SM Question Number: 7 Question Type: Ranking Question Title: RANK the samples in order of preference. Design: T=2, K=2, B=72 Products

Product Code Name 1 - 375 375 Fortified 2 - 861 861 Unfortified

Crosstabulation

Sample 1 2 Total 1 - 375 32 25 57 2 - 861 25 32 57 TOTALS 57 57 114

Percentage Crosstabulation

Sample 1 2 Total 1 - 375 56.1 43.9 100 2 - 861 43.9 56.1 100

Friedman Analysis of Rank This procedure is valid for Complete Block Experimental Designs with no missing data only. This is a Complete Block Design.

Calculated Degrees Friedman Statistic of Freedom p-value 0.85 1 0.354

Critical values corresponding to specific levels of significance: 10%=2.71 5%=3.84 1%=6.63 No difference between the samples at the 10% level. (0.85 < 2.71) No difference between the samples at the 5% level. (0.85 < 3.84) No difference between the samples at the 1% level. (0.85 < 6.63)

Tukey's HSD = 14.788 (5% Significance Level)

Rank Sample Total Significantly Different Than Sample 2 - 861 89.00 A 1 - 375 82.00 a

235

Project: 2401 SM

Question Number: 8 Question Type: Category / Hedonics Question Title: If you were served a full cup of this sample in your school cafeteria, would you DRINK all of it or not? Attribute Number: 1 Attribute Title: Q#8.1 Design: T=2, K=2, B=72

Products Product Code Name 1 - 375 375 Fortified 2 - 861 861 Unfortified

Scale Parameters Value Descriptor 5 Definitely would drink all of it 4 Probably would drink all of it 3 Maybe drink/maybe not drink all of it 2 Probably would not drink all of it 1 Definitely would not drink all of it

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation 1 2 3 4 5 Sample [5] [4] [3] [2] [1] Total 1 - 375 6 18 9 16 8 57 2 - 861 5 16 16 14 6 57 TOTALS 11 34 25 30 14 114

Percentage Crosstabulation 1 2 3 4 5 Sample [5] [4] [3] [2] [1] Total 1 - 375 10.5 31.6 15.8 28.1 14.0 100 2 - 861 8.8 28.1 28.1 24.6 10.5 100

Counts, Medians, Means and SD's

Sample Standard Number Count Total Median Mean Deviation 1 - 375 57 169.00 3.00 2.96 1.267 2 - 861 57 171.00 3.00 3.00 1.150

236

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares F Value p-value Samples 1 0.035 0.035 0.08 0.7843 Judges 56 137.965 2.464 5.31 0.0000 Error 56 25.965 0.464 Total 113 163.965 1.451 Standard 0.090 Error (SEM) =

Tukey's HSD = 0.256 (5% Significance Level)

Sample Mean Significantly Different Than Sample 2 - 861 3.00 a 1 - 375 2.96 a

237

Appendix H: Sedimentation Study

Introduction

I noticed that in the 473 mL PETE clear plastic bottles of NSRL soymilk, a whitish sediment approximately an eighth of an inch thick formed on the bottom of the treatment bottles in the storage study within the first few days of refrigeration. The sedimentation seemed to occurr equally in both light-exposed and dark-stored samples. Prior to taking NSRL analytical samples, the sediment was re-distributed throughout the soymilk via repeated shaking, inversions, and pouring back and forth between the sample bottle and an autoclaved beaker. The purpose of this study was to investigate the effectiveness of two different hydrocolloids, carrageenan and xanthan gum, in suspending the DSM fortification premix in soymilk of varying soybean to water ratios.

Experimental Design

Soybeans from the same source were made into soymilk of varying dry soybean to water ratios

(1:6, 1:7.5, and 1:9). One liter of unfortified salt and sugar control was produced from each of the soymilk concentrations, while the remainder was divided into six fortified treatments (1 liter each), three containing carrageenan at, respectively, 0.25, 0.4, and 0.55 grams per liter, and three containing xanthan gum at the same concentrations. The 1 liter quantities of soymilk from the unfortified control and each of the six treatments of fortified soymilk were divided into two 600 mL plastic beakers and placed in refrigerated storage for 7 days. Colorimeter readings were taken on the day the soymilk was made and again on days, 1, 3, 5, and 7 of storage to document sedimentation-related color changes.

238

Materials & Methods

Six hundred millilitre, round, clear plastic tricorner beakers were selected for their flat, uniform

bottoms containing no writing or etching (only a single blow-mold mark in the center) and were purchased from BYU Chem Stores (BYU Campus, Ezra Taft Bension Building). Muslin cloth bags were purchased through amazon.com from Celestial Gifts LLC (Cynthiana, KY) on

February 7, 2015. Organic yellow soybeans were purchased through the Amazon.com True Leaf

Market storefront from Gold Mine Natural Food Co. (Poway, CA) on March 17, 2015. A 500g sample of carrageenan (LF-180) was obtained from CPKelco (Atlanta, GA, USA) and a 500g sample of xanthan gum (Ticaxan®) was obtained from TIC Gums (White Marsh, MD, USA).

239

The vitamin-mineral premix was obtained from DSM Nutritional Products, Colombia,

S.A. The premix was formulated to meet governmental nutrient content guidelines for the

national school feeding program (“Programa de Alimentación Escolar” PAE)) of Ecuador. The

premix was intended to deliver 60% of Ecuador’s daily recommended values (“ingesta diaria

recomendada” (IDR)) for Vitamin A, Vitamin D, Vitamin E, Vitamin C, Thiamine, Riboflavin,

Vitamin B6, Niacin, Folate, Vitamin B12, Iron, Zinc, Copper, and Selenium, and 20% of the

IDR for calcium per 350 ml portion of fortified beverage. Corn maltodextrin served as the

diluent and carrier in the premix. The chemical forms of the fortificants added to the premix

were retinol palmitate (250,000 IU/g), Vitamin D3 (cholecalciferol), Vitamin E (dl-alpha tocopherol), thiamine mononitrate, riboflavin, pyridoxine hydrochloride, sodium ascorbate, folic acid, niacinamide, vitamin B12 (cyanocobalamin), tricalcium phosphate, copper gluconate, ferric pyrophosphate, sodium selenite, and zinc sulphate.

Method for Protein Evaluation: Analysis was performed in triplicate for samples from the most concentrated (1:6 dry soybeans:water) and the least concentrated (1:9 soybeans:water) soymilk.

Aluminum canisters were weighed, then 1 ml unfortified soymilk was added. Each sample was

then weighed again (canisters + wet soymilk) and placed in a 1600 Hafo Series drying oven for 8

hours at 100oC (Sheldon Mfg Inc, Model #1670, Serial #0400201, 12.2 Amps, 230V, 60 Hz).

Another 1 ml of soymilk was added to each canister, and the samples were dried for another 12

hours. The finished, dried samples were then weighed for total weight (dry milk + canister

weight). The weight of the remaining dried soymilk was calculated by subtracting the weight of

the canisters from the total weight.

The dry soymilk samples were then analyzed at the BYU Environmental Analytical

Laboratory for percent nitrogen using the Dumas Method as per established methods for the

240

equipment. The resulting % nitrogen values were then multiplied by the protein factor of 5.52 for

soymilk (Food Analysis fourth edition, Suzanne Nielsen, ISBN 978-1-4419-1477-4). The full analysis methods & machine specifications (including model) for nitrogen, carbon, & minerals, including sample preparation can be obtained by emailing Rachel (lab manager at the BYU

Environmental Analytical Laboratory) at [email protected].

Method for % Solids and % Fat: Percent solids and percent fat measurements were taken using

a combined two stage method. Soymilk samples in 2 oz (60 ml) clear glass bottles from each

batch were stored at -80oC until analyzed. After thawing, samples were transferred on ice from

refrigerated storage in the Eyering Science Center to the BYU Culinary Support Center for

analysis. Each sample was shaken for 15 seconds immediately prior to testing. A disposable

plastic pipette was then used to transfer between 4g and 5g of soymilk to glass fiber pads for

analysis using a CEM Corporation Smart Trac, with Smart System 5 microwave digestion.

Settings were set to those used for 2% dairy milk, and all methods used were according to

manufacturer instructions. The digested sample and glass fiber pad was then analyzed using

NMR rapid fat analysis according to manufacturer instructions. Two readings were averaged for

each treatment excepting the 1:6 dry soybean:water soymilk, for which only one analysis was

done.

Method for Preparation of Soymilk: The appropriate weight of soybeans as measured out for

each batch, to create 1:6, 1:7.5, and 1:9 dry bean-to-water weight ratios with 9kg of water. The soybeans were soaked overnight (~18hrs), drained, and the soaked weight measured. These were added to a steel kettle and heated until boiling. The boiled beans were blended to a homogeneous slurry using a Proctor Silex hand blender (120V, 60 Hz, 120W, Type HB07, Series

241

A98804J, Model 59735); the resulting bean/water blend was then poured through a muslin cloth into another steel pot to strain out the okara from the milk. The beans were pressed inside the cloth with a ceramic plate using human force to release any residual milk from the cloth.

Once the soymilk was strained and pressed, the soymilk from each batch was quickly divided into portions of 1000 ml of soymilk, and temperature in degrees Farenheit was measured immediately before fortification. The sample was then prepared according to treatment targets:

0.8 g salt, 80 g sugar, and 1.687g fortificant (fortified only) per each 1000 ml portion, with gums as per the treatment design. The treatment additives were blended in advance for each of the 1 L treatments by weighing out the respective amounts of salt, sugar, micronutrient premix (fortified only), and gums needed to fortify 1 L of soymilk, combining all components in a one-gallon

Ziploc plastic storage bag (with ample air), and shaken in a vigorous circular motion for sixty seconds to allow ample mixing.

Treatments Treatment Salt (g/L) Sugar (g/L) Fortification Carrageenan Xanthan Premix LF-180 Ticaxan (g/L) (g/L) (g/L) 1 0.8 80 - - - 2 0.8 80 1.687 - - 3 0.8 80 1.687 0.25 - 4 0.8 80 1.687 0.40 - 5 0.8 80 1.687 0.55 - 6 0.8 80 1.687 - 0.25 7 0.8 80 1.687 - 0.40 8 0.8 80 1.687 - 0.55 9 - - - - -

Fortified samples were mixed with a steel whisk for 1 minute. The 1000 ml sample was then separated into two 600 ml plastic containers, resulting in two 500 ml duplicate samples of each treatment. This process was repeated for each treatment.

242

Methods for Color and Viscosity Readings: After preparation, the soymilk samples were placed on a wheeled cart, where they would stay for the duration of the study, and transferred to the pilot plant walk-in refrigerator. After 24 hours of refrigeration a viscosity measurement was taken on each beaker (RV-02 thin spindle – SSP RV2; WTI 00:00:10; SSN 25:00; OSP; LEC;

LSC-S; DAI-S), and the values from the two beakers from each treatment were averaged to give a single treatment value. The temperature at the time of each measurement was recorded. The samples were then stirred with a plastic spoon for 15 seconds before measuring color, to roughly replicate the state of equal dispersion immediately after each sample was prepared. These baseline L*, a*, and b* color measurements were taken in triplicate readings immediately after stirring using the ColorFlux colorimeter in the BYU Quality Assurance Laboratory. The samples were measured with triplicate readings as per instrument instructions in the same 600 ml plastic containers in which they were prepared, and the containers were rotated approximately 120 degrees before each subsequent measurement. Triplicate color readings were averaged to give a single color value for each of the two beakers of every treatment and every day that color measurements were taken.

Samples were then transferred back to the refrigerator on the wheeled cart with as little physical agitation as possible. Subsequent color measurements were taken on the next day (day

2) and every other day thereafter for a week without stirring the soymilk in order to measure any color changes due to sedimentation. In summary, color measurements were taken on day 1 (after stirring the samples), and then again on day 2, day 4, day 6, and day 8 (all without stirring the samples). However, after discovering that color measurements stabilized by day 6 for the highest concentration of soymilk (1:6 soybean:water ratio) color measurements were not taken on day 8 for the other soymilk samples.

243

Results

Viscosity Raw Data (Soymilk temperatures ranged from 6.8oC to 15.3oC) 1:6 (soybean:water ratio) Salt & LF-180 LF-180 LF-180 Ticaxan Ticaxan Ticaxan Sugar Premix 0.25g 0.4g 0.55g 0.25g 0.45g 0.55g Viscosity (Centipoise) 17.8 16.8 32.0 82.6 174.6 47.5 84.5 112.8

1:7.5 (soybean:water ratio) Salt & LF-180 LF-180 LF-180 Ticaxan Ticaxan Ticaxan Sugar Premix 0.25g 0.4g 0.55g 0.25g 0.45g 0.55g Viscosity (Centipoise) 16.6 15.5 15.5 45.9 129.1 25.1 74.4 112.9

1:9 (soybean:water ratio) Salt & LF-180 LF-180 LF-180 Ticaxan Ticaxan Ticaxan Sugar Premix 0.25g 0.4g 0.55g 0.25g 0.45g 0.55g Viscosity (Centipoise) 8.8 8.0 25.0 52.8 103.7 13.6 96.5 177.0

Colorimeter raw data for high concentration soymilk (1:6 dry soybean:water ratio): Batch (soybean:water Treatment Color # Day ratio) Color Label Value 1 1:6 L* 1 79.07 1 1:6 L* 2 79.43 1 1:6 L* 3 78.96 1 1:6 L* 4 78.71 1 1:6 L* 5 79.00 1 1:6 L* 6 79.44 1 1:6 L* 7 79.08 1 1:6 L* 8 79.47 1 1:6 a* 1 -0.43 1 1:6 a* 2 -0.84 1 1:6 a* 3 -0.73 1 1:6 a* 4 -0.60 1 1:6 a* 5 -0.98 1 1:6 a* 6 -0.91 1 1:6 a* 7 -0.81 1 1:6 a* 8 -0.98

244

Batch (soybean:water Treatment Color # Day ratio) Color Label Value 1 1:6 b* 1 18.80 1 1:6 b* 2 21.11 1 1:6 b* 3 21.13 1 1:6 b* 4 21.32 1 1:6 b* 5 21.66 1 1:6 b* 6 21.26 1 1:6 b* 7 21.35 1 1:6 b* 8 21.46 2 1:6 L* 1 79.26 2 1:6 L* 2 80.85 2 1:6 L* 3 79.87 2 1:6 L* 4 79.15 2 1:6 L* 5 79.26 2 1:6 L* 6 79.54 2 1:6 L* 7 79.98 2 1:6 L* 8 80.74 2 1:6 a* 1 0.56 2 1:6 a* 2 0.36 2 1:6 a* 3 0.04 2 1:6 a* 4 0.00 2 1:6 a* 5 -0.43 2 1:6 a* 6 0.04 2 1:6 a* 7 0.00 2 1:6 a* 8 -0.17 2 1:6 b* 1 18.43 2 1:6 b* 2 15.37 2 1:6 b* 3 18.42 2 1:6 b* 4 20.44 2 1:6 b* 5 21.15 2 1:6 b* 6 20.60 2 1:6 b* 7 20.42 2 1:6 b* 8 20.31 4 1:6 L* 1 79.05 4 1:6 L* 2 80.72 4 1:6 L* 3 79.33 4 1:6 L* 4 79.03 4 1:6 L* 5 79.18 4 1:6 L* 6 79.14 4 1:6 L* 7 80.04 4 1:6 L* 8 80.51 4 1:6 a* 1 0.35

245

Batch (soybean:water Treatment Color # Day ratio) Color Label Value 4 1:6 a* 2 0.16 4 1:6 a* 3 -0.19 4 1:6 a* 4 -0.16 4 1:6 a* 5 -0.62 4 1:6 a* 6 -0.21 4 1:6 a* 7 -0.21 4 1:6 a* 8 -0.33 4 1:6 b* 1 18.10 4 1:6 b* 2 14.19 4 1:6 b* 3 18.24 4 1:6 b* 4 20.26 4 1:6 b* 5 20.70 4 1:6 b* 6 20.53 4 1:6 b* 7 20.18 4 1:6 b* 8 20.20 6 1:6 L* 1 79.06 6 1:6 L* 2 80.83 6 1:6 L* 3 79.82 6 1:6 L* 4 79.14 6 1:6 L* 5 79.18 6 1:6 L* 6 79.23 6 1:6 L* 7 80.12 6 1:6 L* 8 80.62 6 1:6 a* 1 0.57 6 1:6 a* 2 0.35 6 1:6 a* 3 0.02 6 1:6 a* 4 0.02 6 1:6 a* 5 -0.41 6 1:6 a* 6 -0.06 6 1:6 a* 7 -0.06 6 1:6 a* 8 -0.16 6 1:6 b* 1 18.21 6 1:6 b* 2 14.13 6 1:6 b* 3 18.38 6 1:6 b* 4 20.24 6 1:6 b* 5 21.01 6 1:6 b* 6 20.86 6 1:6 b* 7 20.26 6 1:6 b* 8 20.39

246

Colorimeter raw data for medium concentration soymilk (1:7.5 dry soybean:water ratio): Batch (soybean:water Treatment Color # Day ratio) Color Label Value 1 1:7.5 L* 1 78.71 1 1:7.5 L* 2 79.24 1 1:7.5 L* 3 78.58 1 1:7.5 L* 4 78.26 1 1:7.5 L* 5 78.14 1 1:7.5 L* 6 78.83 1 1:7.5 L* 7 78.49 1 1:7.5 L* 8 78.69 1 1:7.5 a* 1 -0.71 1 1:7.5 a* 2 -1.29 1 1:7.5 a* 3 -1.34 1 1:7.5 a* 4 -1.22 1 1:7.5 a* 5 -1.29 1 1:7.5 a* 6 -1.09 1 1:7.5 a* 7 -1.13 1 1:7.5 a* 8 -1.02 1 1:7.5 b* 1 17.88 1 1:7.5 b* 2 21.18 1 1:7.5 b* 3 21.58 1 1:7.5 b* 4 21.50 1 1:7.5 b* 5 21.54 1 1:7.5 b* 6 20.94 1 1:7.5 b* 7 21.28 1 1:7.5 b* 8 21.22 2 1:7.5 L* 1 78.53 2 1:7.5 L* 2 80.87 2 1:7.5 L* 3 79.38 2 1:7.5 L* 4 79.13 2 1:7.5 L* 5 78.55 2 1:7.5 L* 6 77.91 2 1:7.5 L* 7 79.66 2 1:7.5 L* 8 79.89 2 1:7.5 a* 1 -0.08 2 1:7.5 a* 2 -0.21 2 1:7.5 a* 3 -0.74 2 1:7.5 a* 4 -0.25 2 1:7.5 a* 5 -0.43

247

Batch (soybean:water Treatment Color # Day ratio) Color Label Value 2 1:7.5 a* 6 -0.74 2 1:7.5 a* 7 -0.18 2 1:7.5 a* 8 -0.11 2 1:7.5 b* 1 18.13 2 1:7.5 b* 2 14.24 2 1:7.5 b* 3 19.71 2 1:7.5 b* 4 20.34 2 1:7.5 b* 5 20.42 2 1:7.5 b* 6 20.05 2 1:7.5 b* 7 19.79 2 1:7.5 b* 8 19.93 4 1:7.5 L* 1 78.21 4 1:7.5 L* 2 80.89 4 1:7.5 L* 3 79.54 4 1:7.5 L* 4 79.54 4 1:7.5 L* 5 78.62 4 1:7.5 L* 6 76.52 4 1:7.5 L* 7 79.25 4 1:7.5 L* 8 79.68 4 1:7.5 a* 1 0.14 4 1:7.5 a* 2 -0.08 4 1:7.5 a* 3 -0.61 4 1:7.5 a* 4 -0.20 4 1:7.5 a* 5 -0.28 4 1:7.5 a* 6 -1.01 4 1:7.5 a* 7 -0.02 4 1:7.5 a* 8 -0.01 4 1:7.5 b* 1 18.31 4 1:7.5 b* 2 13.58 4 1:7.5 b* 3 19.50 4 1:7.5 b* 4 20.15 4 1:7.5 b* 5 20.23 4 1:7.5 b* 6 20.41 4 1:7.5 b* 7 19.57 4 1:7.5 b* 8 19.91 6 1:7.5 L* 1 78.14 6 1:7.5 L* 2 81.05 6 1:7.5 L* 3 79.77

248

Batch (soybean:water Treatment Color # Day ratio) Color Label Value 6 1:7.5 L* 4 79.77 6 1:7.5 L* 5 78.79 6 1:7.5 L* 6 76.23 6 1:7.5 L* 7 79.16 6 1:7.5 L* 8 79.88 6 1:7.5 a* 1 0.19 6 1:7.5 a* 2 -0.12 6 1:7.5 a* 3 -0.61 6 1:7.5 a* 4 -0.20 6 1:7.5 a* 5 -0.29 6 1:7.5 a* 6 -1.11 6 1:7.5 a* 7 0.06 6 1:7.5 a* 8 0.01 6 1:7.5 b* 1 18.49 6 1:7.5 b* 2 13.53 6 1:7.5 b* 3 19.57 6 1:7.5 b* 4 20.24 6 1:7.5 b* 5 20.29 6 1:7.5 b* 6 20.77 6 1:7.5 b* 7 19.70 6 1:7.5 b* 8 19.88

249

Colorimeter raw data for low concentration soymilk (1:9 dry soybean:water ratio): Batch (soybean:water Treatment Color # Day ratio) Color Label Value 1 1:9 L* 1 77.02 1 1:9 L* 2 77.66 1 1:9 L* 3 76.85 1 1:9 L* 4 76.44 1 1:9 L* 5 75.88 1 1:9 L* 6 77.92 1 1:9 L* 7 77.29 1 1:9 L* 8 77.10 1 1:9 a* 1 -1.35 1 1:9 a* 2 -1.91 1 1:9 a* 3 -1.92 1 1:9 a* 4 -1.98 1 1:9 a* 5 -2.05 1 1:9 a* 6 -1.92 1 1:9 a* 7 -1.71 1 1:9 a* 8 -1.60 1 1:9 b* 1 17.13 1 1:9 b* 2 21.52 1 1:9 b* 3 21.56 1 1:9 b* 4 21.72 1 1:9 b* 5 21.83 1 1:9 b* 6 21.49 1 1:9 b* 7 21.17 1 1:9 b* 8 21.19 2 1:9 L* 1 77.33 2 1:9 L* 2 81.65 2 1:9 L* 3 79.20 2 1:9 L* 4 78.63 2 1:9 L* 5 78.54 2 1:9 L* 6 76.61 2 1:9 L* 7 78.66 2 1:9 L* 8 78.98 2 1:9 a* 1 -0.78 2 1:9 a* 2 -0.55 2 1:9 a* 3 -0.85 2 1:9 a* 4 -0.88 2 1:9 a* 5 -0.69

250

Batch (soybean:water Treatment Color # Day ratio) Color Label Value 2 1:9 a* 6 -1.70 2 1:9 a* 7 -1.06 2 1:9 a* 8 -0.87 2 1:9 b* 1 17.31 2 1:9 b* 2 10.66 2 1:9 b* 3 17.18 2 1:9 b* 4 16.66 2 1:9 b* 5 17.27 2 1:9 b* 6 20.02 2 1:9 b* 7 19.42 2 1:9 b* 8 19.70 4 1:9 L* 1 76.78 4 1:9 L* 2 81.87 4 1:9 L* 3 79.45 4 1:9 L* 4 78.57 4 1:9 L* 5 78.49 4 1:9 L* 6 74.62 4 1:9 L* 7 77.84 4 1:9 L* 8 78.76 4 1:9 a* 1 -0.59 4 1:9 a* 2 -0.49 4 1:9 a* 3 -0.59 4 1:9 a* 4 -0.85 4 1:9 a* 5 -0.66 4 1:9 a* 6 -1.97 4 1:9 a* 7 -1.10 4 1:9 a* 8 -0.56 4 1:9 b* 1 17.43 4 1:9 b* 2 10.05 4 1:9 b* 3 17.47 4 1:9 b* 4 16.70 4 1:9 b* 5 17.30 4 1:9 b* 6 20.58 4 1:9 b* 7 19.45 4 1:9 b* 8 19.96 6 1:9 L* 1 77.02 6 1:9 L* 2 82.11 6 1:9 L* 3 80.39

251

Batch (soybean:water Treatment Color # Day ratio) Color Label Value 6 1:9 L* 4 78.88 6 1:9 L* 5 78.82 6 1:9 L* 6 74.47 6 1:9 L* 7 77.59 6 1:9 L* 8 78.86 6 1:9 a* 1 -0.51 6 1:9 a* 2 -0.50 6 1:9 a* 3 -0.61 6 1:9 a* 4 -0.85 6 1:9 a* 5 -0.67 6 1:9 a* 6 -2.06 6 1:9 a* 7 -1.16 6 1:9 a* 8 -0.63 6 1:9 b* 1 17.63 6 1:9 b* 2 10.06 6 1:9 b* 3 17.42 6 1:9 b* 4 16.76 6 1:9 b* 5 17.38 6 1:9 b* 6 20.58 6 1:9 b* 7 20.02 6 1:9 b* 8 20.23

252

Colorimeter Data as Figures

Color changes due to sedimentation in the most concentrated soymilk (1:6 soybeans:water).

Carrageenan Xanthan (1:6 soybeans:water) (1:6 soybeans:water) 83.00 83.00 Unfortified 81.50 81.50 Control

Fortified 80.00 80.00

Control

78.50 78.50 0.25 g/L L* Value L* Value 77.00 77.00 0.40 g/L

75.50 75.50 0.55 g/L 74.00 74.00 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 Day Day Carrageenan Xanthan (1:6 soybeans:water) (1:6 soybeans:water)

1.00 1.00 Unfortified 0.50 0.50 Control 0.00 0.00 Fortified Control -0.50 -0.50 -1.00 -1.00 0.25 g/L a* Value a* Value-1.50 -1.50 0.40 g/L -2.00 -2.00 -2.50 -2.50 0.55 g/L 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 Day Day

253

Carrageenan Xanthan (1:6 soybeans:water) (1:6 soybeans:water) 23.00 23.00 21.00 21.00 Unfortified Control 19.00 19.00

Fortified 17.00 17.00 Control

15.00 15.00 0.25 g/L b* Value b* Value 13.00 13.00 0.40 g/L 11.00 11.00 9.00 9.00 0.55 g/L 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 Day Day

Color changes due to sedimentation in medium concentration soymilk (1:7.5 soybeans:water).

Carrageenan Xanthan (1:7.5 soybeans:water) (1:7.5 soybeans:water) 83.00 83.00 Unfortified 81.50 81.50 Control

80.00 Fortified 80.00 Control 78.50 78.50 0.25 g/L L* Value L* Value 77.00 77.00 0.40 g/L 75.50 75.50 0.55 g/L 74.00 74.00 1 2 3 4 5 6 1 2 3 4 5 6 Day Day

254

Carrageenan Xanthan (1:7.5 soybeans:water) (1:7.5 soybeans:water) 1.00 1.00

0.50 0.50 Unfortified Control 0.00 0.00

Fortified -0.50 -0.50 Control

-1.00 -1.00 0.25 g/L a* Value a* Value -1.50 -1.50 0.40 g/L -2.00 -2.00

-2.50 -2.50 0.55 g/L 1 2 3 4 5 6 1 2 3 4 5 6 Day Day

Carrageenan Xanthan (1:7.5 soybeans:water) (1:7.5 soybeans:water) 23.00 23.00 21.00 21.00 Unfortified Control 19.00 19.00

Fortified 17.00 17.00 Control

15.00 15.00 0.25 g/L b* Value b* Value 13.00 13.00 0.40 g/L 11.00 11.00 9.00 9.00 0.55 g/L 1 2 3 4 5 6 1 2 3 4 5 6 Day Day

255

Color changes due to sedimentation in the least concentrated soymilk (1:9 soybeans:water).

Carrageenan Xanthan (1:9 soybeans:water) (1:9 soybeans:water) 83.00 83.00 Unfortified 81.50 81.50 Control

80.00 80.00 Fortified Control

78.50 78.50 0.25 g/L L* Value L* Value 77.00 77.00 0.40 g/L 75.50 75.50 0.55 g/L 74.00 74.00 1 2 3 4 5 6 1 2 3 4 5 6 Day Day

Carrageenan Xanthan (1:9 soybeans:water) (1:9 soybeans:water) 1.00 1.00 Unfortified 0.50 0.50 Control 0.00 0.00

Fortified -0.50 -0.50 Control

-1.00 -1.00 0.25 g/L a* Value a* Value -1.50 -1.50 0.40 g/L -2.00 -2.00 -2.50 -2.50 0.55 g/L 1 2 3 4 5 6 1 2 3 4 5 6 Day Day

256

Carrageenan Xanthan (1:9 soybeans:water) (1:9 soybeans:water) 23.00 23.00 21.00 21.00 Unfortified Control 19.00 19.00

Fortified 17.00 17.00 Control 15.00 15.00 0.25 g/L b* Value b* Value 13.00 13.00 0.40 g/L 11.00 11.00 9.00 9.00 0.55 g/L 1 2 3 4 5 6 1 2 3 4 5 6 Day Day

Results from Dumas method analysis for the sugar and salt only treatments of the most concentrated and least concentrated soymilks – and the calculated % protein (% N * 5.52). Nitrogen Carbon Name Mass Method % % 1:6 A 0.1539 High Nitrogen 7.8466 52.2 1:6 B 0.1556 High Nitrogen 7.9404 51.8 1:6 C 0.1549 High Nitrogen 8.0948 53.2 1:9 A 0.1850 High Nitrogen 4.5199 47.3 1:9 B 0.1870 High Nitrogen 4.3102 44.8 1:9 C 0.1839 High Nitrogen 4.5288 47.3

Element Average Std. Deviation RSD Count 1:6 Mass 0.15 0.0009 0.6% 3 1:6 Nitrogen % 7.96 0.1253 1.6% 3 1:6 Carbon % 52.40 0.7211 1.4% 3 1:9 Mass 0.19 0.0016 0.8% 3 1:9 Nitrogen % 4.45 0.1237 2.8% 3 1:9 Carbon % 46.47 1.4434 3.1% 3

1:6 % Protein: 43.9 1:9 % Protein: 24.6

257

Percent solids and percent fat for all soymilk treatments and concentrations: 1:6 1:7.5 1:9 Treatment % Solids % Fat % Solids % Fat % Solids % Fat Sugar/Salt Control 14.93 1.44 13.49 1.04 12.24 0.65 Premix Control 14.63 1.36 13.59 1.08 12.31 0.70 LF-180 0.25 g/L 14.72 1.33 13.41 1.06 12.31 0.77 LF-180 0.40 g/L 14.64 1.34 13.73 1.00 12.47 0.77 LF-180 0.55 g/L 14.52 1.33 13.56 1.05 12.57 0.78 Ticaxan 0.25 g/L 14.73 1.32 13.69 0.89 12.47 0.77 Ticaxan 0.40 g/L 13.84 1.34 13.60 0.89 12.72 0.79 Ticaxan 0.55 g/L 14.57 0.85 13.85 1.09 12.67 0.80 Raw Soymilk 7.91 1.81 - - 5.00 1.10

Discussion

The NSRL soymilk was produced with a dry soybean:water ratio of 1:8, plus any condensation from the steam injection, and the Ecuador soymilk was produced similarly. Therefore, the 1:7.5 and 1:9 soymilk concentrations are most pertinent.

The apparent reason for the atypical results of the xanthan gum samples was the fact that the whisk-in method was incapable of keeping the xanthan gum from clumping and it formed

“fish eyes.” This ultimately meant that the xanthan gum was unfit for the Ecuador soymilk application and the carrageenan was used instead.

Conclusions

Carrageenan can satisfactorily reduce the sedimentation of fortified micronutrients and is preferable to xanthan gum due to excellent dispersion and integration into the soymilk with low shear (i.e. hand whisking).

258

Appendix I: Hogar de Cristo soymilk production process summary November 2013 – Exploratory trip to document Hogar de Cristo soy cow process

This picture is looking into the top of the pressure cooker showing the colander with the grinding blender in the middle. Soaked soybean and water are steam heated to a gauge temperature of 80 degrees Celsius and gauge pressure of 25 psi (about 15 min of steam heating) and then ground while still hot and under pressure (approximately another 15 min).

This picture shows the entire soy cow with the pressure-cooking pot on the bench, the filtering pot on legs, and the soy milk and additives pot on the floor. The hot ground soybeans leave the pressure cooker at between 92 and 95 degrees Celsius. Most of the milk filters through the muslin cloth filter in the filtering pot on its own, and the rest is pressed from the okra by the screw press lid as shown on the right. The filtering process takes only

259

a few minutes and the soy milk is still about 80 degrees Celsius when the sugar & flavorings are all mixed in and the milk is put through the next filter into the bottling canister.

The flavored milk is then poured through a second muslin cloth bag into a bottling canister, and then quickly put into chlorine-washed and rinsed bottles while it is still above 70 degrees Celsius.

The filled and capped bottles are cooled in a water bath for about half an hour and then they are refrigerated. They cool to about 30-40 degrees Celsius in the water bath before entering refrigeration. It likely takes hours before the milk is at its storage temperature of about 4-5 degrees Celsius in refrigeration.

260

HDC soymilk processing flow diagram November, 2013

HDC soymilk processing flow diagram April, 2015

261

Appendix J: Ecuador Sensory Ballot

¡BIENVENIDO A UNA PRUEBA SENSORIAL DE LA LECHE DE SOYA!

Por favor, levanta la mano en cualquier momento si no entiendes las instrucciones o si tienes preguntas. Durante esta prueba, si no deseas participar más puedes salir en cualquier momento.

Si eres alérgico a la soya por favor díselo a un adulto y NO SIGAS CON ESTA PRUEBA.

Hoy vas probar 2 muestras diferentes de LECHE DE SOYA. Por favor, lee todas las instrucciones y preguntas cuidadosamente. Antes de obtener la juestra, por favor responde a estas preguntas, marcando con círculos.

* Cuántos años tienes? ______años

* ¿Eres varón o mujer? . Varón

. Mujer

* ¿Has probado la leche de soya antes?

. Sí

. No

. No sé

* ¿Qué opines de la leche de soya?

. Me gusta

. No sé

. No me gusta

Si no te gusta la leche de soya, hazlo saber a un ayudante y no tienes que beberlo ni coninuar con esta prueba.

… Por favor, pasa a la siguiente página …

262

PRUEBA DE CONSUMO DE LECHE DE SOYA

Ahora levanta la mano para mostrar que estás listo para obtener tus muestras. Por favor, sé paciente; llegarán en breve.

Si necesitas ayuda durante la prueba, levanta la mano o pide a un ayudante.

Responderás a 7 preguntas diferentes sobre cada muestra de leche de soya, así que guarda leche suficiente para evaluarla con todas las preguntas. Si se te acaba la muestra, levanta la mano o pídele a un ayudante.

Prueba las muestras en el orden en que se te fueron dadas de izquierda a derecha. Toma un poco de agua entre las muestras para limpiar la boca.

Contesta las preguntas marcando con círculo la respuesta más apropiada.

* EN GENERAL, ¿qué opines de las muestras?

* ¿Qué opines del SABOR de las muestras?

… Por favor, pasa a la siguiente página …

263

* ¿Qué opines del COLOR de las muestras?

* ¿Qué opines del AROMA de las muestras?

* ¿Qué opines de la SENSACIÓN de la muestra EN TU BOCA?

… Por favor, pasa a la siguiente página …

264

* ¿Qué opines del SABOR EN LA BOCA DESPUÉS DE TOMAR LA LECHE?

* Por favor, escribe el número de la muestra que te gustó más.

______

* Si recibieras una botella de esta leche de soya ¿te la beberías toda?

* ¡Has acabado las preguntas! Puedes beber las muestras o simplemente dejarlas. Favor de levanter la mano para entregar tu cuestionario.

¡GRACIAS!

265

Appendix K: Ecuador Implementation Study

Introduction

Nutritional status evaluations have documented a history of malnutrition amongst school-age children and youth in Ecuador. In the mid 1990’s pronounced sub-clinical Vitamin A deficiency was found to be widespread (48% of subjects) in 5 Ecuadorian provinces – more especially among children from poor families with uneducated mothers (Rodriguez et al. 1996). Nutrient-

fortified intervention programs have seen some success. Evaluation of an ongoing Ecuador

national program designed to alleviate malnutrition in infants was successful in arresting the

trends of both stunting and anemia as compared to a control group, and the positive results were

achieved with only 76% compliance (Lutter et al. 2006).

However, addressing the burden of early childhood malnutrition is a difficult problem.

Government programs do not always reach the marginalized urban poor – often squatting in

shanty towns just beyond city limits. In the mid-2000’s a national survey of health among mothers and children in Ecuador confirmed that chronic malnutrition persists as a signficant problem, leading to stunting (deficient height-for-age) and associated maladies (World Bank

2007). More recent studies have identified that as much as 23% of children under the age of 5

may be stunted and underweight – even more so among the urban poor than their rural

counterparts (Katuli et al. 2012; 2013). Thankfully, some of the developmental setbacks from

malnutrition in the crucial first years of life can be significantly corrected by growth recovery later on. For instance, significant benefits to academic achievement and cognitive ability are

possible through youth nutrition interventions (Crookston et al. 2013; Fink and Rockers 2014).

In Peru not-for-profit, charity, religious, and non-government organizations (NGO) have

seen collaborative success in the fight against the malnutrition which persists beyond the reach of

266

government programs (Acosta and Haddad 2013). In Ecuador, one such NGO on the outskirts of

Guayaquil is a catholic charity by the name of Hogar de Cristo (HDC). With two Rotary Club- donated “mechanical cows,” HDC has supplied unfortified soy milk to approximately 1,000 children in 5 different Guayaquil schools for multiple years. During the same time an extension program of the School of Nursing at Brigham Young University (Provo, Utah, USA) monitored the health of the children. The most prominent deficiency disease they identified was macrocytic anemia, with over 50% of the children affected in 5 different schools tested in 2012 (2013 data- sharing meeting with Sondra Heaston, Michael Dunn, and Dallin Hardy).

The ongoing presence of BYU nursing students has encouraged experimentation with a wide variety of intervention programs in the form of health education, anti-parasite medication, nutritional supplements, and school lunch programs, but the current project represents the first time micronutrient enriched food has been used in this setting. The addition of a sufficiently comprehensive nutrient fortification regimen to the soy milk being produced at Hogar de Cristo has the potential to reduce malnutrition, enhance physical development, and improve cognitive abilities of hundreds of young children. If successful, the Hogar de Cristo soy milk fortification project could potentially be a model for other charitable organizations with soy milk production facilities, including at least 11 more in Ecuador (Rotary [date unkown]).

We have shown previously that soymilk produced by similar methods as those used at

Hogar de Cristo is an effective delivery matrix for many of the nutrients needed by the

Ecuadorian children (Hardy, 2019). Therefore, the purpose of this project was to take advantage of the existing HDC soy milk processing plant and test the feasibility of fortifying the soy milk distributed by Hogar de Cristo. Fortification was implemented at the HDC production facility and sensory acceptability and nutrient stability were evaluated in the fortified soymilk.

267

Experimental Design

A total of eight batches of soymilk, produced over two days, were included in the study. Four

dedicated study batches were produced at the end of each production day, one of which was

unfortified and three of which were fortified. Aside from the vitamin-mineral fortification premix and the carrageenan to prevent settling, the additives for unfortified and fortified soymilks were identical. Samples for vitamin and mineral analysis were taken from the bulk soymilk immediately after premix addition at a second filter step while filling the bottling dispenser, and again after bottling and complete cooling immediately prior to refrigeration in order to evaluate vitamin stability during processing. Soymilk bottles from the unfortified batch and one of the fortified batches produced on day 1 were used for sensory analysis.

Materials & Methods

Soymilk Processing

On Monday a total of 14 batches and on Tuesday 16 batches of soy milk were prepared at Hogar de Cristo (Av. Casuarina. Coop. Sergio Toral, Mz. 101/Bloque 1 (Perimetral Noroeste)

Guayaquil – Ecuador). Soybeans were soaked overnight (over the weekend for soymilk produced on Monday) under refrigeration in large tubs prior to processing. For each batch of soymilk 4 kg of soaked soybeans were drained, rinsed, and added to the pressure cooker along with 14 L of water. The “soya cow” was then closed and steam was injected (with venting until the air was fully displaced with steam) until a gauge temperature of 80oC was reached (15-18.5

min). The steam valve to the pressure cooker was then closed to keep soymilk from flowing

back into the boiler unit and the beans were ground for 15 minutes. After grinding, the soymilk

was discharged through a muslin cloth into a large steel pot and the okara was pressed with a

268

manually operated screw press to extract as much milk as possible. The final four batches of

each day were dedicated to the study – the first of the study batches was unfortified and the last

three were micronutrient-fortified each day.

Premix Incorporation

For the two study batches of unfortified soymilk the following additives were placed in the

bottom of the steel pot catching the filtered soymilk: 3 lb (1.36 kg) sugar, 15 g salt, 1 g

crystallized vanilla (Aromcolor, Guayaquil, EC), 16 g liquid coconut flavour & colourant

(Aromcolor, Guayaquil, EC). For the six study batches of micronutrient-fortified soymilk, the same ingredients were placed in the bottom of the steel pot catching the hot soymilk except that only half of the sugar was added. The remaining 1.5 lb (0.68 kg) of sugar was reserved in a 1 gallon sealable bag. The soymilk was weighed immediately after filtering and the appropriate amounts of carrageenan and micronutrient premix were weighed out and added to the reserved

0.68 kg of sugar to achieve the target soymilk concentrations of 0.5 g/kg carrageenan and 1.687 g/kg micronutrient premix. The bag containing the sugar, fortificants, and carrageenan was sealed with a significant quantity of air and shaken vigorously for 1 min. The fortifying and thickening blend was then whisked into the soymilk for 1 minute.

Quantities of soymilk, carrageenan, and micronutrient premix, & addition temperatures

Soymilk Fortification Fortification Weight Carrageenan Premix Temperature Day (kg) Weight (g) Weight (g) (oC) 1 14.61 - - approx. 76 1 12.86 6.4 21.7 75 1 14.42 7.2 24.3 77 1 12.97 6.49 21.88 72 2 15.24 - - 76.5 2 13.97 6.985 23.26 69 2 13.24 6.62 22.34 75.1 2 13.88 6.94 23.4 73.5

269

Cooling, Packaging, Transport, & Sampling

Analytical samples were drawn in duplicate from the bulk soymilk immediately after

fortification in 60 mL glass bottles, cooled in an ambient temperature water bath (~30 oC) and

refrigerated. If bottling was delayed for any reason after the additives had been mixed into the raw soymilk, the steel pot containing bulk soymilk was set in an ambient temperature circulating water bath (overflowing with continual water inflow) which kept the bath temperature at approximately 30oC. As soon as possible, the soymilk was poured through a muslin cloth filter

into a steel tub with a dispensing spout at the bottom. The dispensing tub was placed on a table,

and previously clorox-washed 240 ml clear plastic bottles were filled by hand, capped, and

placed in a 28-30oC circulating water bath for 25-30 minutes prior to refrigeration. Analytical

samples and soymilk for the sensory panel were then stored in a walk-in refrigerator with near

constant light exposure throughout the workday. The temperature of the refrigerator was set at

4oC, but the temperature reached up to 11oC during the workday due to the continual addition of

warm (~30 oC) soymilk bottles and the constant opening and closing of the door. As each batch

was entering refrigeration, two more analytical samples were drawn in 60 ml glass bottles from

the 240 ml bottles. Analytical samples were refrigerated at HDC until production and sensory

analysis was complete (Wednesday, day 3) and then shipped on ice to Brigham Young

University (a process taking several days) and stored at -80oC until analyzed.

Nutritional Analysis

Vitamins: Vitamins were analyzed as previously reported by Hardy et al. (i.e. NSRL soymilk

analysis), except that for vitamin C analysis intermittent acetonitrile column washes between sets

of every several injections were necessary to eliminate a bulge in the baseline that obscured the

vitamin C peak.

270

Minerals: Zinc and iron were measured by ICP at BYU’s Environmental Analytical Laboratory

as a measure of batch to batch consistency.

Sensory Evaluation

With approval of the university’s Institutional Review board, and with parental consent, children

ages 5 through 17 were recruited at an Hogar de Cristo health clinic near the soymilk production

facility on Wednesday, April 19, 2015. Parents and guardians accompanying the children to the

clinic were also invited to taste the samples and fill out a paper ballot by circling their answer for

each question. Potential panelists were screened for soy allergies and their willingness to try

soymilk. Panelists were not compensated for their participation in the panel.

Throughout the panel, soymilk was stored in the clear plastic production bottles in a

refrigerator, and about 75 ml of fortified and unfortified soymilk were poured into 4 fl. oz. (119

ml) clear plastic cups which had been previously labelled with a three-digit blinding code (138

for unfortified soymilk and 643 for fortified soymilk). Each panelist was presented with two

soymilk samples (one fortified and one unfortified), side-by-side on the serving tray. Sample presentation from left to right was randomized, and BYU students ensured that the amount of soymilk in each cup was visually similar. The order of response on the ballot (i.e. which sample was evaluated first for each question) was the same for each complete ballot but randomized between ballots/panelists. All BYU students present spoke Spanish and were available to answer any questions of the child panelists or their parents. Due to the limited number of panelists obtained at the clinic during the morning, recruitment of children at the health clinic was discontinued at noon and samples were distributed instead to adult Hogar de Cristo staff during their lunchbreak in the organization’s large cafeteria. The same ballots were used for adults and children.

271

The ballots used pictorial scales as well as accompanying textual descriptors. Data was later manually entered into CompuSense 5.6 software for analysis. Incomplete ballots and ballots with ambiguous or uninterpretable responses were excluded. The demographic information gathered consisted of age, gender, and whether they liked soymilk, disliked soymilk, or were not sure if they liked or disliked soymilk. Participants then answered six questions assessing overall impression, flavour, colour, smell, mouthfeel, and aftertaste. Panelists were also asked to rank the samples in order of preference by selecting which of the two samples they liked best, and a final question asked how likely they would be to completely drink a full cup of each sample (fortified and unfortified soymilk) in a school cafeteria setting. All organoleptic properties except for mouthfeel were rated for each sample using a 7-point hedonic scale with the following descriptors: really bad, bad, just a little bad, maybe good or maybe bad, just a little good, good, really good. The mouthfeel of the soymilk was evaluated using a 5-point “just- about-right” (JAR) question in order to help optimize the level of carrageenan used for future production batches. The final ‘willingness to consume’ question was scored for each sample using a 5 point hedonic rating scale with the following descriptors: definitely would not drink all of it, probably would not drink all of it, maybe drink – maybe not drink all of it, probably would drink all of it, definitely would drink all of it. Undergraduate BYU Sensory Lab workers and parents were available to clarify the ballot questions to the participants.

Statistical Analysis

Vitamins: Fortified and unfortified samples taken from hot soymilk immediately after premix addition and again after bottling and cooling to refrigeration temperatures were analyzed for thiamine, riboflavin, folate, vitamin A, and vitamin C content using mixed models ANOVA

(α=0.05), blocking by analytical sample bottle and sampling time. An F test (α=0.05) was used

272

to evaluate the effect of sampling time. In order to account for time and temperature dependent matrix effects which interfered with vitamin A extraction, isomer data were analyzed as a percent of total retinyl palmitate vitamers.

Sensory Evaluation: Data from participants who reported a dislike of soymilk were included in the final analysis, because the sole exclusion criteria were soy allergy and an unwillingness to try soymilk. Statistical analysis of hedonic scale questions was carried out using one way ANOVA with post-hoc Tukey’s HSD (α=0.05). The question asking panelists to rank the samples in order of preference was analyzed using Friedman analysis of rank. The panel was administered using physical paper ballots, but the answers were later entered into Compusense 5.6 software

(Compusense Inc., Guelph ON, Canada) for statistical analysis.

Results

Vitamin Stability

There was no difference in vitamin levels between samples taken immediately after fortification and samples taken after bottling and fully cooling the soymilk as determined by the one-way

ANOVA F-test. However, as we have previously shown in a similar small batch setting of fortified soymilk production, incomplete recovery of retinyl palmitate immediately following fortification may cloud the interpretation of total vitamin A data. Changes in the vitamin A isomerization profile, however, which are inherently recovery-adjusted were also not significant.

273

Vitamin levels in fortified soymilk samples at the time of premix addition and after bottling.

Fortification Refrigeration Units p-value mcg Vitamin A 99.5 108.9 RAE/100g 0.5860 % all-trans 93.9 93.9 % 0.9947 % 13-cis 5.75 5.85 % 0.8451 % 11-cis 0.38 0.29 % 0.8568 % bioactive 98.2 98.2 % 0.8777 Vitamin C 6.52 6.35 mg/100g 0.5959 Thiamine 0.311 0.340 mg/100g 0.1497 Riboflavin 0.168 0.165 mg/100g 0.7291 Folate 51.5 49.9 mcg/100g 0.7080

Vitamin levels in unfortified soymilk samples at premix addition and after bottling.

Premix Addition Refrigeration Units p-value Thiamine 0.039 0.044 mg/100g 0.2386 Riboflavin 0.011 0.008 mg/100g 0.2120 Folate 96.5 79.9 mcg/100g 0.0505

Minerals

Average mineral content of the 2 unfortified batches and all 6 fortified batches

Fe 95% Zn 95% (ppm) CI (ppm) CI Unfortified 3.39 0.28 1.70 0.16 Fortified 19.83 1.36 14.39 0.46

274

Iron and Zinc concentrations in the 6 fortified batches (batches 1-3 were produced on day 1, and batches 4-6 were produced on day 2)

Mineral content by batch 35 30 25

ppm 15 Fe 10 Zn 5 0 0 1 2 3 4 5 6 7

Batch

Nutrient Delivery

Nutrient delivery of native and fortified nutrients by a 240 ml serving of soymilk as a percent of Ecuador’s daily recommended intakes [“ingesta diaria recomendada” (IDR)]

Contributions of Native & Fortified Nutrients toward Recommended Intakes

100% 90% 80% 70% 60% 50% 40% Fortified 30% Unfortified 20% 10% % IDR in240 ml% IDR serving 0%

275

Nutrient content in unfortified and fortified soymilk compared to the theoretical quantities present in the fortificant premix formula.

Vitamin A Vitamin C

500 40

450 35 400 30 350 300 25 250 20 200 15 150 10 100 50 mg per 240 mL serving 5 mcg RAE per 240 mL serving 0 0 Unfortified Fortified Fortificant Unfortified Fortified Fortificant Actual Actual Theoretical Actual Actual Theoretical

Folate 140

120 100 80 60 40

mcg per 240 serving ml 20 0 Unfortified Fortified Fortificant Actual Actual Theoretical

276

Thiamine Riboflavin 1.0 0.8

0.9 0.7 0.8 0.6 0.7 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 mg per 240 serving ml mg per 240 serving ml 0.1 0.1 0.0 0.0 Unfortified Fortified Fortificant Unfortified Fortified Fortificant Actual Actual Theoretical Actual Actual Theoretical *Vitamin levels of bottled and fully cooled samples only were reported for all "actual" results

277

Iron Zinc 6.0 4.5

4.0 5.0 3.5 4.0 3.0 2.5 3.0 2.0 2.0 1.5 1.0 mg per 240 serving ml

1.0 mg per 240 serving ml 0.5 0.0 0.0 Unfortified Fortified Fortificant Unfortified Fortified Fortificant Actual Actual Theoretical Actual Actual Theoretical

**Percent theoretical recoveries are as follows (measured nutrient values of unfortified soymilk were subtracted from the fortified soymilk values and this number was divided by the theoretical value – total quantity present in fortificant premix formula): Vitamin A: 59%; Vitamin C 44%; Thiamine: 89%; Riboflavin: 50%; Folate: 99%; Iron: 82%; Zinc: 76%.

pH Readings

pH of select fortified and unfortified Ecuador soymilk samples from several batches

Fortified pH pH Sample or reading reading Code Day Unfortified #1 #2 LSB-1 1 U 6.21 6.24 LSB-2 1 U 6.05 6.07 LF1a-2 1 F 6.20 6.23 LF1a-1 1 F 6.21 6.23 LF1b 1 F 6.16 6.20 LF1b 1 F 6.21 6.22 LF2b -1 1 F 6.22 6.22 LF2b -2 1 F 6.21 6.22 MSb-1 2 U 6.21 6.21 MSb-2 2 U 6.18 6.19 MF2b-1 2 F 6.27 6.27 MF2b-2 2 F 6.25 6.27

278

Sensory Panel

Demographic results from all complete ballots:

# of Panelists (%) Age 5-17 23 36.5 18+ 40 63.5 Gender M 25 39.7 F 38 60.3 Previously had soymilk? Y 53 84.1 N 10 15.9 Not Sure 0 0 What do you think about soymilk? Like it 47 74.6 Don't know 15 23.8 Don't like it 1 1.6

Panelist rankings of soymilk samples in order of preference (all complete ballots):

Ranked as 1st Ranked as 2nd Preference Preference # of # of Significantly Different Sample (%) (%) panelists panelists than Sample Unfortified 20 31.8 43 68.3 a

Fortified 43 68.3 20 31.8 b

Significant difference between the samples at the 5% level, p<0.001.

279

Panelist hedonic ratings of fortified (F) and unfortified (U) soymilk for sensory properties, likeability, and JAR questions – results from all complete ballots:

Attribute Sample Mean SD p-value U 6.0 a 1.09 Overall Likability 0.0189 F 6.4 b 0.98 U 5.9 a 1.32 Flavor 0.0296 F 6.3 b 1.11 U 5.8 a 1.16 Color 0.0292 F 6.1 b 0.92 U 5.8 a 1.21 Smell 0.0632 F 6.1 a 0.98 U 2.5 a 0.90 Mouthfeel* 0.0000 F 3.3 b 0.61 U 5.7 a 1.41 Aftertaste 0.0140 F 6.1 b 1.16 U 4.3 a 0.94 Willingness to Consume** 0.0248 F 4.6 b 0.77 Organoleptic properties were rated using the following 7-point hedonic scale with accompanying pictorial representations: 1-very bad, 2-I don't like it, 3-just a little bad, 4-not sure, 5-just a little good, 6-I like it 7-very good. *The question pertaining to texture was rated using the following 5-point Just-About-Right (JAR) scale with accompanying pictorial representations: 1-way too thick, 2-a little too thick, 3-perfect, 4-a little too thin, 5-way too thin. **The 'willingness to consume' question asked if the panelist would drink the entire bottle of soymilk if one was given to them. The question was rated using the following 5-point hedonic scale with accompanying pictorial representations: 1-definitely yes, 2-probably yes, 3-not sure, 4-probably not, 5- definitely not.

280

Demographic results from all complete child ballots (ages 5-17):

# of Panelists (%) Gender M 9 39.1 F 14 60.9 Previously had soymilk? Y 16 69.6 N 7 30.4 Not Sure 0 0 What do you think about soymilk? Like it 16 69.6 Don't know 7 30.4 Don't like it 0 0

Panelist hedonic ratings of fortified (F) and unfortified (U) soymilk for sensory properties, likeability, and JAR questions – results from all complete child ballots:

Attribute Sample Mean SD p-value U 6.4 a 1.03 Overall Likability 0.2130 F 6.7 a 0.64 U 6.3 a 1.21 Flavor 0.1424 F 6.7 a 0.71 U 5.9 a 1.28 Color 0.0364 F 6.4 b 0.73 U 5.8 a 1.28 Smell 0.0393 F 6.4 b 0.83 U 2.5 a 0.95 Mouthfeel* 0.0232 F 3.1 b 0.52 U 6.1 a 1.29 Aftertaste 0.0426 F 6.7 b 0.57 U 4.4 a 0.98 Willingness to Consume** 0.0426 F 4.9 b 0.46 Organoleptic properties were rated using the following 7-point hedonic scale with accompanying pictorial representations: 1-very bad, 2-I don't like it, 3-just a little bad, 4-not sure, 5-just a little good, 6-I like it 7-very good. *The question pertaining to texture was rated using the following 5-point Just-About-Right (JAR) scale with accompanying pictorial representations: 1-way too thick, 2-a little too thick, 3-perfect, 4-a little too thin, 5-way too thin. **The 'willingness to consume' question asked if the panelist would drink the entire bottle of soymilk if one was given to them. The question was rated using the following 5-point hedonic scale with accompanying pictorial representations: 1-definitely yes, 2-probably yes, 3-not sure, 4-probably not, 5- definitely not.

281

Panelist rankings of soymilk samples in order of preference (all complete child ballots):

Ranked as 1st Ranked as 2nd Preference Preference # of # of Significantly Different Sample (%) panelists panelists (%) than Sample Unfortified 8 34.8 15 65.2 a

Fortified 15 65.2 8 34.8 a

No difference between the samples at the 5% level, p=0.144.

Discussion

Vitamin stability was different in surprising ways from what we measured in the soymilk made previously at the National Soybean Research Laboratory (NSRL) using a very similar process.

Most notably, vitamin C levels did not significantly differ between samples taken immediately after fortification and samples taken after the soymilk was bottled and fully cooled immediately prior to being placed in the walk-in refrigerator. In addition, the vitamin A isomer profile was not altered during processing. However, in spite of the apparent stability during processing the total levels measured for these vitamins appeared to be lower (recoveries: vitamin C, 44% of theoretical; vitamin A 59% of theoretical). Vitamin C in particular was analyzed at only roughly half the level we recovered in our previous study, in spite of our efforts to deliver the same fortificant input targets. This degradation may be at least partially explained by the fact that the samples spent more than three full days in transit from Ecuador to BYU. Although the samples were kept on ice as much as possible, intermittent exposure to light and warm temperatures was unavoidable (e.g. long bus ride, long plane ride, long customs inspection).

The results for the other vitamins were largely as expected. On average, riboflavin was measured in fortified soymilk at only 50% of theoretical input, but nearly all of the thiamine and folate were recovered (89% and 99% of theoretical input, respectively). Interestingly, however,

282

and in stark contrast to the apparent stability of fortified folate, the loss of native folate in unfortified soymilk during processing was very nearly significant (p=0.0505). While great effort was taken to standardize folate analysis, the method is subject to some subjectivity due to the choice of which microtiter rows from the internal standard will be used to build the standard curve. At times multiple selections can satisfy pre-determined horrat value requirements. Thus, it cannot be ruled out that data for fortified samples has bias towards the theoretical values while the analysis of unfortified is unbiased.

Batch to batch variation over six fortified batches across two different production days, as measured by zinc levels, seemed to be within a reasonable range. However, statistical analysis confirming the same has yet to be completed. Iron levels were variable but still reasonably consistent considering the number of samples analyzed. It seems as though fortification can be performed sufficiently repeatably in this process.

Supplier-recommended label claims for the fortification premix were 41% of Ecuador’s daily recommended intakes (“ingesta diaria recomendada” – IDR) for all of the nutrients measured. As seen in the figure comparing the vitamin content of unfortified and fortified soymilks as percentages of the IDR, the quantities of thiamine and folate in the premix were more than adequate to meed this claim. With the help of native minerals iron comfortably met this label claim, zinc marginally exceeded it, and vitamin A was close to qualifying for it.

However, vitamin C and riboflavin did not even come close to qualifying for the proposed label claim. Still, the levels of all the analyzed nutrients, including vitamin C and riboflavin, were significantly increased over unfortified soymilk.

The sensory panel revealed dramatic differences between the fortified and unfortified soymilk. Both samples were well liked by the panelists (Hogar de Cristo staff informed us that

283

coconut was there most liked flavour), but the fortified soymilk with carrageenan was decisively

better liked for all properties except for smell in the analysis of the answers of both adult and

child panelists. The chosen concentration of carrageenan may have been slightly high, because

panelists rated the fortified soymilk as slightly thicker than optimal, but they also rated it as being closer to a perfect soymilk texture than the unfortified soymilk, which panelists

considered, on average, to be a little too thin. Impressively, most of the statistically significant

preferences of the full panel of adults and children emerged in from the analysis of just 23 child

responses when considered alone. While child ratings revealed no difference in overall

likeability, flavour, and the Friedman analysis of rank, all of the other properties evaluated were

significantly in favour of the fortified soymilk – including smell, which was not significantly

different when adult responses were also considered.

Conclusions

We discovered that micronutrient fortification of soymilk can be successfully implemented in a

small batch process and a humanitarian setting. The addition of micronutrients and carrageenan

significantly increased consumer willingness to consume the soymilk and improved many of the

organoleptic properties. Nutrient degradation was significant for some of the fortified vitamins,

but the loss did not seem to be associated with any one particular processing step. Significantly

increased overages of vitamin C and riboflavin in the fortification premix are recommended.

284

References

Acosta AM, Haddad L. 2014. The politics of success in the fight against malnutrition in Peru. Food Policy. 44:26-35. Rotary background of mechanical cow [Internet]. [cited 2019 March 5]. Available from: http://www.clubrunner.ca/Data/5360/966/HTML/125178/ROTARYBACKGROUNDOF MECHANICALCOW4.pdf Crookston BT, Schott W, Cueto S, Dearden KA, Engle P, Georgiadis A, Lundeen EA, Penny ME, Stein AD, Behrman JR. 2013. Postinfancy growth, schooling, and cognitive achievement: Young Lives. Am J Clin Nutr. 98(6):1555-63. Fink G, Rockers PC. 2014. Childhood growth, schooling, and cognitive development: further evidence from the Young Lives study. Am J Clin Nutr. 100(1):182-188. Katuli SD, Natto Z, Beeson LW, Cordero-MacIntyre ZR. 2012. Nutrition status of children in rural and urban areas in Ecuador. FASEB J. 26:1 Suppl. 814.5. Katuli S, Natto ZS, Beeson L, Cordero-MacIntyre ZR. 2013. Nutritional status of highland and lowland children in Ecuador. J Trop Pediatr. 59(1):3-9. Lutter C, Rodriguez A, Fuenmayor G, Sempertegui F. 2006. Evaluation of a national food and nutrition program in Ecuador on child growth and micronutrient status. FASEB J. 20(4):A557. Rodriguez A, Guaman G, Nelson DP. 1996. Vitamin A status of children in five Ecuadorian provinces. Bull Pan Am Health Organ. 30(3):234-41. World Bank. 2007. Nutritional failure in Ecuador: causes, consequences, and solutions. Washington (DC). The World Bank. Publication 38689.

285

Appendix L: Ecuador Raw Data

Vitamin A

Total vitamin A is reported in mg/100g soymilk.

Sample Treat- Sampling Sample % 13- % 11- % All- % Bio- Vitamin Run Code ment Day Batch Time Bottle cis cis trans active A 3 LF3a-1 F M 14 Fort 1 5.59 0.00 94.41 98.49 0.126 3 LF3a-1 F M 14 Fort 1 5.74 0.00 94.26 98.45 0.123 3 LF3a-1 F M 14 Fort 1 5.58 0.00 94.42 98.49 0.127 3 LF3a-2 F M 14 Fort 2 4.82 1.47 93.70 97.73 0.275 3 LF3a-2 F M 14 Fort 2 4.86 1.52 93.62 97.68 0.262 3 LF3a-2 F M 14 Fort 2 4.87 1.56 93.57 97.65 0.256 4 LF3c-1 F M 14 Bottling 1 5.55 0.00 94.45 98.50 0.142 4 LF3c-1 F M 14 Bottling 1 5.75 0.00 94.25 98.45 0.139 4 LF3c-1 F M 14 Bottling 1 5.69 0.00 94.31 98.46 0.151 4 LF3c-2 F M 14 Bottling 2 5.05 1.15 93.79 97.87 0.290 4 LF3c-2 F M 14 Bottling 2 5.08 1.17 93.76 97.86 0.300 4 MF2a-1 F T 15 Fort 1 6.41 0.00 93.59 98.27 0.162 4 MF2a-1 F T 15 Fort 1 6.12 0.00 93.88 98.35 0.150 4 MF2a-1 F T 15 Fort 1 6.50 0.00 93.50 98.25 0.146 4 MF2a-2 F T 15 Fort 2 6.21 0.00 93.79 98.32 0.184 4 MF2a-2 F T 15 Fort 2 6.13 0.00 93.87 98.34 0.183 4 MF2a-2 F T 15 Fort 2 6.22 0.00 93.78 98.32 0.193 4 MF2c-1 F T 15 Bottling 1 6.30 0.00 93.70 98.30 0.192 4 MF2c-1 F T 15 Bottling 1 6.28 0.00 93.72 98.31 0.194 4 MF2c-1 F T 15 Bottling 1 6.39 0.00 93.61 98.27 0.191 4 MF2c-2 F T 15 Bottling 2 6.33 0.00 93.67 98.29 0.196 4 MF2c-2 F T 15 Bottling 2 6.28 0.00 93.72 98.30 0.196 4 MF2c-2 F T 15 Bottling 2 6.39 0.00 93.61 98.27 0.200

286

Vitamin C

Total Vitamin C is reported in mg/100g soymilk.

Sample Sampling Sample Run Code Treatment Day Batch Time Bottle Vitamin C 12 LF1a-1 F M 12 Fort 1 6.080 13 LF1a-1 F M 12 Fort 1 6.434 13 LF1a-1 F M 12 Fort 1 4.864 13 LF1a-1 F M 12 Fort 1 5.204 14 LF1a-2 F M 12 Fort 2 7.389 15 LF1b-1 F M 12 Filter 1 6.460 16 LF1b-1 F M 12 Filter 1 5.510 17 LF1b-1 F M 12 Filter 1 5.488 17 LF1b-1 F M 12 Filter 1 4.266 18 LF1c-1 F M 12 Bottling 1 6.710 18 LF1c-1 F M 12 Bottling 1 7.028 18 LF1c-1 F M 12 Bottling 1 6.845 19 LF1c-2 F M 12 Bottling 2 5.577 20 LF1c-2 F M 12 Bottling 2 4.321 20 LF1c-2 F M 12 Bottling 2 4.437 21 LF2b-1 F M 13 Filter 1 6.642 22 LF2b-1 F M 13 Filter 1 5.827 22 LF2b-1 F M 13 Filter 1 6.031 23 LF2b-2 F M 13 Filter 2 6.112 24 LF2b-2 F M 13 Filter 2 4.124 24 LF2b-2 F M 13 Filter 2 4.916 24 LF2b-2 F M 13 Filter 2 4.992 25 LF2c-1 F M 13 Bottling 1 6.988 26 LF2c-1 F M 13 Bottling 1 6.768 26 LF2c-1 F M 13 Bottling 1 6.620 26 LF2c-1 F M 13 Bottling 1 6.551 27 LF2c-2 F M 13 Bottling 2 8.049 28 LF2c-2 F M 13 Bottling 2 7.996 28 LF2c-2 F M 13 Bottling 2 7.941 28 LF2c-2 F M 13 Bottling 2 7.766 29 LF3a-1 F M 14 Fort 1 6.536 30 LF3a-1 F M 14 Fort 1 4.876 30 LF3a-1 F M 14 Fort 1 6.219 31 LF3a-2 F M 14 Fort 2 7.046 32 LF3a-2 F M 14 Fort 2 5.175 32 LF3a-2 F M 14 Fort 2 5.986 32 LF3a-2 F M 14 Fort 2 5.990 33 LF3c-1 F M 14 Bottling 1 6.924 34 LF3c-1 F M 14 Bottling 1 6.286 34 LF3c-1 F M 14 Bottling 1 6.337 35 LF3c-2 F M 14 Bottling 2 6.320

287

Sample Sampling Sample Run Code Treatment Day Batch Time Bottle Vitamin C 36 LF3c-2 F M 14 Bottling 2 5.643 36 LF3c-2 F M 14 Bottling 2 5.947 36 LF3c-2 F M 14 Bottling 2 5.938 14 MF1a-1 F T 14 Fort 1 6.306 15 MF1a-1 F T 14 Fort 1 6.793 15 MF1a-1 F T 14 Fort 1 6.644 14 MF1a-2 F T 14 Fort 2 7.268 17 MF1a-2 F T 14 Fort 2 6.742 17 MF1a-2 F T 14 Fort 2 6.244 19 MF1b-1 F T 14 Filter 1 5.728 20 MF1b-1 F T 14 Filter 1 5.297 21 MF1b-1 F T 14 Filter 1 5.773 21 MF1b-1 F T 14 Filter 1 5.510 21 MF1b-1 F T 14 Filter 1 5.607 22 MF1b-2 F T 14 Filter 2 6.392 22 MF1b-2 F T 14 Filter 2 7.041 23 MF1c-1 F T 14 Bottling 1 5.449 24 MF1c-1 F T 14 Bottling 1 5.192 24 MF1c-1 F T 14 Bottling 1 5.244 24 MF1c-1 F T 14 Bottling 1 5.551 25 MF1c-2 F T 14 Bottling 2 6.395 26 MF1c-2 F T 14 Bottling 2 6.231 26 MF1c-2 F T 14 Bottling 2 6.265 26 MF1c-2 F T 14 Bottling 2 6.265 27 MF2a-1 F T 14 Fort 1 6.355 28 MF2a-1 F T 14 Fort 1 6.896 28 MF2a-1 F T 14 Fort 1 7.020 28 MF2a-1 F T 14 Fort 1 6.868 27 MF2a-2 F T 14 Fort 2 7.899 29 MF2a-2 F T 14 Fort 2 7.874 29 MF2a-2 F T 14 Fort 2 7.688 29 MF2a-2 F T 14 Fort 2 7.625 30 MF2b-1 F T 14 Filter 1 6.780 31 MF2b-1 F T 14 Filter 1 6.085 31 MF2b-1 F T 14 Filter 1 6.183 32 MF2b-2 F T 14 Filter 2 7.302 33 MF2b-2 F T 14 Filter 2 6.972 33 MF2b-2 F T 14 Filter 2 6.901 33 MF2b-2 F T 14 Filter 2 6.852 34 MF2c-1 F T 14 Bottling 1 6.230 35 MF2c-1 F T 14 Bottling 1 6.049 34 MF2c-2 F T 14 Bottling 2 6.468 35 MF2c-2 F T 14 Bottling 2 6.948 36 MF3a-1 F T 14 Fort 1 8.686 37 MF3a-1 F T 14 Fort 1 6.694

288

Sample Sampling Sample Run Code Treatment Day Batch Time Bottle Vitamin C 37 MF3a-1 F T 14 Fort 1 6.679 37 MF3a-1 F T 14 Fort 1 6.918 36 MF3a-2 F T 14 Fort 2 8.269 38 MF3a-2 F T 14 Fort 2 6.885 38 MF3a-2 F T 14 Fort 2 7.854 38 MF3a-2 F T 14 Fort 2 6.498 39 MF3b-1 F T 14 Filter 1 7.649 40 MF3b-1 F T 14 Filter 1 5.591 40 MF3b-1 F T 14 Filter 1 5.465 39 MF3b-2 F T 14 Filter 2 8.079 41 MF3b-2 F T 14 Filter 2 7.586 41 MF3b-2 F T 14 Filter 2 7.601 41 MF3b-2 F T 14 Filter 2 7.505 42 MF3c-1 F T 14 Bottling 1 5.007 43 MF3c-1 F T 14 Bottling 1 5.739 43 MF3c-1 F T 14 Bottling 1 4.461 44 MF3c-2 F T 14 Bottling 2 7.020 45 MF3c-2 F T 14 Bottling 2 6.590 45 MF3c-2 F T 14 Bottling 2 6.753 45 MF3c-2 F T 14 Bottling 2 6.709

289

Thiamine

Total Thiamine is reported in mg/100g soymilk.

Sample Sampling Sample Analyst Run Code Treatment Day Batch Time Bottle Thiamine M 1 Lsa-1 U M 11 Fort 1 0.035 M 1 Lsa-1 U M 11 Fort 1 0.039 F 3 Lsa-1 U M 11 Fort 1 0.044 M 1 Lsa-2 U M 11 Fort 2 0.038 M 1 Lsa-2 U M 11 Fort 2 0.036 F 4 Lsa-2 U M 11 Fort 2 0.046 M 5 Lsc-1 U M 11 Bottling 1 0.053 M 5 Lsc-1 U M 11 Bottling 1 0.051 F 6 Lsc-1 U M 11 Bottling 1 0.041 M 5 Lsc-2 U M 11 Bottling 2 0.056 M 5 Lsc-2 U M 11 Bottling 2 0.053 F 6 Lsc-2 U M 11 Bottling 2 0.045 M 7 LF3a-1 F M 14 Fort 1 0.289 M 7 LF3a-1 F M 14 Fort 1 0.316 F 8 LF3a-1 F M 14 Fort 1 0.344 M 7 LF3a-2 F M 14 Fort 2 0.321 M 7 LF3a-2 F M 14 Fort 2 0.304 F 8 LF3a-2 F M 14 Fort 2 0.329 M 9 LF3c-1 F M 14 Bottling 1 0.383 M 9 LF3c-1 F M 14 Bottling 1 0.390 F 11 LF3c-1 F M 14 Bottling 1 0.319 M 9 LF3c-2 F M 14 Bottling 2 0.395 M 9 LF3c-2 F M 14 Bottling 2 0.415 F 11 LF3c-2 F M 14 Bottling 2 0.307 M 1 Msa-1 U T 13 Fort 1 0.041 M 4 Msc-2 U T 13 Bottling 2 0.040 M 5 Msc-2 U T 13 Bottling 2 0.043 F 6 Msc-2 U T 13 Bottling 2 0.033 M 7 MF2a-1 F T 15 Fort 1 0.314 M 8 MF2a-1 F T 15 Fort 1 0.335 F 9 MF2a-1 F T 15 Fort 1 0.264 M 7 MF2a-2 F T 15 Fort 2 0.328 M 8 MF2a-2 F T 15 Fort 2 0.331 F 9 MF2a-2 F T 15 Fort 2 0.262 M 10 MF2c-1 F T 15 Bottling 1 0.287 M 11 MF2c-1 F T 15 Bottling 1 0.358 F 12 MF2c-1 F T 15 Bottling 1 0.273 M 10 MF2c-2 F T 15 Bottling 2 0.332 M 11 MF2c-2 F T 15 Bottling 2 0.354 F 12 MF2c-2 F T 15 Bottling 2 0.262

290

Riboflavin

Total riboflavin is reported in mg/100g soymilk.

Sample Sampling Sample Analyst Run Code Treatment Day Batch Time Bottle Riboflavin M 1 Lsa-1 U M 11 Fort 1 0.009 M 1 Lsa-1 U M 11 Fort 1 0.011 F 2 Lsa-1 U M 11 Fort 1 0.008 M 1 Lsa-2 U M 11 Fort 2 0.010 M 1 Lsa-2 U M 11 Fort 2 0.013 F 2 Lsa-2 U M 11 Fort 2 0.010 M 5 Lsc-1 U M 11 Bottling 1 0.006 M 5 Lsc-1 U M 11 Bottling 1 0.005 F 6 Lsc-1 U M 11 Bottling 1 0.005 M 5 Lsc-2 U M 11 Bottling 2 0.007 M 5 Lsc-2 U M 11 Bottling 2 0.005 F 6 Lsc-2 U M 11 Bottling 2 0.006 M 7 LF3a-1 F M 14 Fort 1 0.155 M 7 LF3a-1 F M 14 Fort 1 0.168 F 8 LF3a-1 F M 14 Fort 1 0.160 M 7 LF3a-2 F M 14 Fort 2 0.171 M 7 LF3a-2 F M 14 Fort 2 0.150 F 8 LF3a-2 F M 14 Fort 2 0.154 M 9 LF3c-1 F M 14 Bottling 1 0.161 M 9 LF3c-1 F M 14 Bottling 1 0.147 F 10 LF3c-1 F M 14 Bottling 1 0.143 M 9 LF3c-2 F M 14 Bottling 2 0.154 M 9 LF3c-2 F M 14 Bottling 2 0.170 F 10 LF3c-2 F M 14 Bottling 2 0.145 M 1 Msa-1 U T 13 Fort 1 0.012 M 2 Msa-1 U T 13 Fort 1 0.007 F 3 Msa-1 U T 13 Fort 1 0.014 M 1 Msa-2 U T 13 Fort 2 0.014 M 2 Msa-2 U T 13 Fort 2 0.012 F 3 Msa-2 U T 13 Fort 2 0.009 M 4 Msc-1 U T 13 Bottling 1 0.011 M 5 Msc-1 U T 13 Bottling 1 0.011 F 6 Msc-1 U T 13 Bottling 1 0.011 M 4 Msc-2 U T 13 Bottling 2 0.012 M 5 Msc-2 U T 13 Bottling 2 0.011 F 6 Msc-2 U T 13 Bottling 2 0.011 M 7 MF2a-1 F T 15 Fort 1 0.182 M 8 MF2a-1 F T 15 Fort 1 0.187 F 9 MF2a-1 F T 15 Fort 1 0.164 M 7 MF2a-2 F T 15 Fort 2 0.183 M 8 MF2a-2 F T 15 Fort 2 0.185

291

Sample Sampling Sample Analyst Run Code Treatment Day Batch Time Bottle Riboflavin F 9 MF2a-2 F T 15 Fort 2 0.159 M 10 MF2c-1 F T 15 Bottling 1 0.167 M 11 MF2c-1 F T 15 Bottling 1 0.201 F 12 MF2c-1 F T 15 Bottling 1 0.160 M 10 MF2c-2 F T 15 Bottling 2 0.169 M 11 MF2c-2 F T 15 Bottling 2 0.204 F 12 MF2c-2 F T 15 Bottling 2 0.160

292

Folate

Arbitrary Sampling Sample [FA] ug/g Plate # Sample Code Treatment Day Batch Time Bottle (WWB) 1 Lsa-1 U M 11 Fort 1 0.0789 1 Lsa-1 U M 11 Fort 1 0.0788 1 Lsa-2 U M 11 Fort 2 0.1051 1 Lsa-2 U M 11 Fort 2 0.1090 2 LSc-1 U M 11 Bottling 1 0.0847 2 LSc-1 U M 11 Bottling 1 0.0779 2 LSc-2 U M 11 Bottling 2 0.0763 2 LSc-2 U M 11 Bottling 2 0.0881 3 LF3a-1 F M 14 Fort 1 0.4106 3 LF3a-1 F M 14 Fort 1 0.4008 3 LF3a-2 F M 14 Fort 2 0.5787 3 LF3a-2 F M 14 Fort 2 0.6334 4 LF3c-1 F M 14 Bottling 1 0.4738 4 LF3c-1 F M 14 Bottling 1 0.4788 4 LF3c-2 F M 14 Bottling 2 0.4933 4 LF3c-2 F M 14 Bottling 2 0.4989 5 Msa-1 U T 13 Fort 1 0.1107 5 Msa-1 U T 13 Fort 1 0.1005 5 Msa-2 U T 13 Fort 2 0.0964 5 Msa-2 U T 13 Fort 2 0.0925 6 MSc-1 U T 13 Bottling 1 0.0735 6 MSc-1 U T 13 Bottling 1 0.0745 6 MSc-2 U T 13 Bottling 2 0.0799 6 MSc-2 U T 13 Bottling 2 0.0842 1 MF2a-1 F T 15 Fort 1 0.5596 7 MF2a-1 F T 15 Fort 1 0.5316 7 MF2a-2 F T 15 Fort 2 0.5195 7 MF2a-2 F T 15 Fort 2 0.4887 8 MF2c-1 F T 15 Bottling 1 0.5094 8 MF2c-1 F T 15 Bottling 1 0.5156 8 MF2c-2 F T 15 Bottling 2 0.5134 8 MF2c-2 F T 15 Bottling 2 0.5049

293

Minerals (Iron & Zinc)

Iron and zinc are reported in ppm.

Sampling Sample Date Run Sample ID Treatment Day Batch Time Bottle Fe Zn 9/22/2015 1 Msc-1 U T 13 Bottling 1 3.64 1.74 9/22/2015 1 Msc-2 U T 13 Bottling 2 3.60 1.79 9/22/2015 1 Msc-2 U T 13 Bottling 2 3.20 1.83 9/22/2015 1 MF1c-1 F T 14 Bottling 1 22.13 14.84 9/22/2015 1 MF1c-2 F T 14 Bottling 2 22.52 14.63 9/22/2015 1 MF1c-2 F T 14 Bottling 2 22.56 14.69 9/22/2015 1 MF2c-1 F T 15 Bottling 1 18.97 13.99 9/22/2015 1 MF2c-2 F T 15 Bottling 2 19.28 14.27 9/22/2015 1 MF3c-1 F T 16 Bottling 1 19.90 15.01 9/22/2015 1 MF3c-2 F T 16 Bottling 2 21.97 15.02 10/16/2015 2 LSC-1 U M 11 Bottling 1 3.33 1.55 10/16/2015 2 LSC-2 U M 11 Bottling 2 3.16 1.58 10/16/2015 2 LF1C-1 F M 12 Bottling 1 21.37 15.47 10/16/2015 2 LF1C-1 F M 12 Bottling 1 22.77 15.97 10/16/2015 2 LF1C-2 F M 12 Bottling 2 16.11 13.91 10/16/2015 2 LF2C-1 F M 13 Bottling 1 15.95 13.02 10/16/2015 2 LF2C-1 F M 13 Bottling 1 16.30 13.18 10/16/2015 2 LF2C-2 F M 13 Bottling 2 17.47 13.62 10/16/2015 2 LF3C-1 F M 14 Bottling 1 20.48 14.50 10/16/2015 2 LF3C-2 F M 14 Bottling 2 19.63 13.69

294

Appendix M: Ecuador Sensory Panel Statistical Analysis Output

Ecuador Sensory Panel – All complete ballots

Demographics

Project: 2518 REVISED SOY MILK PROJECT

Question Number: 2 Question Type: Multiple Choice (Demographic) Question Title: How old are you?

Choices

Number Value Choices 1 [1] 5-17 years old 2 [2] 18+ years old

Crosstabulation

1 2 Sample [1] [2] Total n/a 23 40 63 TOTALS 23 40 63

Percentage Crosstabulation

1 2 Sample [1] [2] Total n/a 36.5 63.5 100

295

Project: 2518 REVISED SOY MILK PROJECT

Question Number: 3 Question Type: Multiple Choice (Demographic) Question Title: Are you male or female?

Choices

Number Value Choices 1 [2] Male 2 [1] Female

Crosstabulation

1 2 Sample [2] [1] Total n/a 25 38 63 TOTALS 25 38 63

Percentage Crosstabulation

1 2 Sample [2] [1] Total n/a 39.7 60.3 100

296

Project: 2518 REVISED SOY MILK PROJECT

Question Number: 4 Question Type: Multiple Choice (Demographic) Question Title: Have you had soy milk before?

Choices

Number Value Choices 1 [3] Yes 2 [2] No 3 [1] I don't know

Crosstabulation

1 2 3 Sample [3] [2] [1] Total n/a 53 10 63 TOTALS 53 10 63

Percentage Crosstabulation

1 2 3 Sample [3] [2] [1] Total n/a 84.1 15.9 100

297

Project: 2518 REVISED SOY MILK PROJECT

Question Number: 5 Question Type: Multiple Choice (Demographic) Question Title: What do you think about soy milk?

Choices

Number Value Choices 1 [1] I like it 2 [2] I don't know 3 [3] I don't like it

Crosstabulation

1 2 3 Sample [1] [2] [3] Total n/a 47 15 1 63 TOTALS 47 15 1 63

Percentage Crosstabulation

1 2 3 Sample [1] [2] [3] Total n/a 74.6 23.8 1.6 100

298

Hedonics Project: 2518 REVISED SOY MILK PROJECT

Question Number: 6 Question Type: Category / Hedonics Question Title: IN GENERAL, how do you feel about the samples? Attribute Number: 1 Attribute Title: Q#6.1 Design: T=2, K=2, B=100

Products

Product Code Name 1 - 138 138 control 2 - 643 643 fortified carrageenan

Scale Parameters

Value Descriptor 7 like very much 6 like moderately 5 like slightly 4 neither like nor dislike 3 dislike slightly 2 dislike moderately 1 dislike very much

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation

1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 25 20 13 2 3 63 2 - 643 41 13 4 4 1 63 TOTALS 66 33 17 6 4 126

Percentage Crosstabulation

1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 39.7 31.8 20.6 3.2 4.8 100 2 - 643 65.1 20.6 6.4 6.4 1.6 100

Counts, Medians, Means and SD's

Sample Standard Number Count Total Median Mean Deviation 1 - 138 63 377.00 6.00 5.98 1.085 2 - 643 63 404.00 7.00 6.41 0.978

299

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares Squares F Value p-value Samples 1 5.786 5.786 5.81 0.0189 Judges 62 70.540 1.138 1.14 0.3002 Error 62 61.714 0.995 Total 125 138.040 1.104 Standard Error 0.125 (SEM) =

Tukey's HSD = 0.356 (5% Significance Level)

Sample Mean Significantly Different Than Sample 2 - 643 6.41 a 1 1 - 138 5.98 b

300

Project: 2518 REVISED SOY MILK PROJECT

Question Number: 7 Question Type: Category / Hedonics Question Title: What do you think about the TASTE of the samples? Attribute Number: 1 Attribute Title: Q#7.1 Design: T=2, K=2, B=100

Products

Product Code Name 1 - 138 138 control 2 - 643 643 fortified carrageenan

Scale Parameters

Value Descriptor 7 like very much 6 like moderately 5 like slightly 4 neither like nor dislike 3 dislike slightly 2 dislike moderately 1 dislike very much

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation

1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 25 21 10 2 2 3 63 2 - 643 40 12 6 3 1 1 63 TOTALS 65 33 16 5 3 4 126

Percentage Crosstabulation

1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 39.7 33.3 15.9 3.2 3.2 4.8 100 2 - 643 63.5 19.1 9.5 4.8 1.6 1.6 100

Counts, Medians, Means and SD's

Sample Standard Number Count Total Median Mean Deviation 1 - 138 63 371.00 6.00 5.89 1.321 2 - 643 63 399.00 7.00 6.33 1.107

301

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares Squares F Value p-value Samples 1 6.222 6.222 4.96 0.0296 Judges 62 106.444 1.717 1.37 0.1098 Error 62 77.778 1.254 Total 125 190.444 1.524 Standard Error 0.141 (SEM) =

Tukey's HSD = 0.399 (5% Significance Level)

Sample Mean Significantly Different Than Sample 2 - 643 6.33 a 1 1 - 138 5.89 b

302

Project: 2518 REVISED SOY MILK PROJECT

Question Number: 8 Question Type: Category / Hedonics Question Title: What do you think about the COLOR of the samples? Attribute Number: 1 Attribute Title: Q#8.1 Design: T=2, K=2, B=100

Products

Product Code Name 1 - 138 138 control 2 - 643 643 fortified carrageenan

Scale Parameters

Value Descriptor 7 like very much 6 like moderately 5 like slightly 4 neither like nor dislike 3 dislike slightly 2 dislike moderately 1 dislike very much

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation

1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 19 27 8 6 2 1 63 2 - 643 25 27 5 6 63 TOTALS 44 54 13 12 2 1 126

Percentage Crosstabulation

1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 30.2 42.9 12.7 9.5 3.2 1.6 100 2 - 643 39.7 42.9 7.9 9.5 100

Counts, Medians, Means and SD's

Sample Standard Number Count Total Median Mean Deviation 1 - 138 63 367.00 6.00 5.83 1.158 2 - 643 63 386.00 6.00 6.13 0.924

303

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares Squares F Value p-value Samples 1 2.865 2.865 4.98 0.0292 Judges 62 100.429 1.620 2.82 0.0000 Error 62 35.635 0.575 Total 125 138.929 1.111 Standard Error 0.095 (SEM) =

Tukey's HSD = 0.27 (5% Significance Level)

Sample Mean Significantly Different Than Sample 2 - 643 6.13 a 1 1 - 138 5.83 b

304

Project: 2518 REVISED SOY MILK PROJECT

Question Number: 9 Question Type: Category / Hedonics Question Title: What do you think about the AROMA of the samples? Attribute Number: 1 Attribute Title: Q#9.1 Design: T=2, K=2, B=100

Products

Product Code Name 1 - 138 138 control 2 - 643 643 fortified carrageenan

Scale Parameters

Value Descriptor 7 like very much 6 like moderately 5 like slightly 4 neither like nor dislike 3 dislike slightly 2 dislike moderately 1 dislike very much

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation

1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 24 16 15 5 2 1 63 2 - 643 28 20 11 3 1 63 TOTALS 52 36 26 8 3 1 126

Percentage Crosstabulation

1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 38.1 25.4 23.8 7.9 3.2 1.6 100 2 - 643 44.4 31.8 17.5 4.8 1.6 100

Counts, Medians, Means and SD's

Sample Standard Number Count Total Median Mean Deviation 1 - 138 63 367.00 6.00 5.83 1.212 2 - 643 63 386.00 6.00 6.13 0.975

305

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares Squares F Value p-value Samples 1 2.865 2.865 3.58 0.0632 Judges 62 100.429 1.620 2.02 0.0031 Error 62 49.635 0.801 Total 125 152.929 1.223 Standard Error 0.112 (SEM) =

Tukey's HSD = 0.319 (5% Significance Level)

Sample Mean Significantly Different Than Sample 2 - 643 6.13 a 1 - 138 5.83 a

306

Project: 2518 REVISED SOY MILK PROJECT

Question Number: 10 Question Type: Category / Hedonics Question Title: What do you think about the TEXTURE of the samples? Attribute Number: 1 Attribute Title: Q#10.1 Design: T=2, K=2, B=100

Products

Product Code Name 1 - 138 138 control 2 - 643 643 fortified carrageenan

Scale Parameters

Value Descriptor 5 way too thick 4 a little too thick 3 just about right 2 a little thin 1 way too thin

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation

1 2 3 4 5 Sample [5] [4] [3] [2] [1] Total 1 - 138 6 29 17 11 63 2 - 643 1 20 38 4 63 TOTALS 1 26 67 21 11 126

Percentage Crosstabulation

1 2 3 4 5 Sample [5] [4] [3] [2] [1] Total 1 - 138 9.5 46.0 27. 17.5 100 0 2 - 643 1.6 31. 60.3 6.4 100 8

Counts, Medians, Means and SD's

Sample Standard Number Count Total Median Mean Deviation 1 - 138 63 156.00 3.00 2.48 0.895 2 - 643 63 207.00 3.00 3.29 0.607

307

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares Squares F Value p-value Samples 1 20.643 20.643 31.33 0.0000 Judges 62 31.714 0.512 0.78 0.8394 Error 62 40.857 0.659 Total 125 93.214 0.746 Standard Error 0.102 (SEM) =

Tukey's HSD = 0.289 (5% Significance Level)

Sample Mean Significantly Different Than Sample 2 - 643 3.29 a 1 1 - 138 2.48 b

308

Project: 2518 REVISED SOY MILK PROJECT

Question Number: 11 Question Type: Category / Hedonics Question Title: What do you think about the TASTE in your mouth after drinking the sample? Attribute Number: 1 Attribute Title: Q#11.1 Design: T=2, K=2, B=100

Products

Product Code Name 1 - 138 138 control 2 - 643 643 fortified carrageenan

Scale Parameters

Value Descriptor 7 like very much 6 like moderately 5 like slightly 4 neither like nor dislike 3 dislike slightly 2 dislike moderately 1 dislike very much

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation

1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 21 19 13 3 4 3 63 2 - 643 32 18 7 3 2 1 63 TOTALS 53 37 20 6 6 4 126

Percentage Crosstabulation

1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 33.3 30.2 20.6 4.8 6.4 4.8 100 2 - 643 50.8 28.6 11.1 4.8 3.2 1.6 100

Counts, Medians, Means and SD's

Sample Standard Number Count Total Median Mean Deviation 1 - 138 63 356.00 6.00 5.65 1.405 2 - 643 63 387.00 7.00 6.14 1.162

309

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares Squares F Value p-value Samples 1 7.627 7.627 6.40 0.0140 Judges 62 132.159 2.132 1.79 0.0118 Error 62 73.873 1.192 Total 125 213.659 1.709 Standard Error 0.137 (SEM) =

Tukey's HSD = 0.389 (5% Significance Level)

Sample Mean Significantly Different Than Sample 2 - 643 6.14 a 1 1 - 138 5.65 b

310

Project: 2518 REVISED SOY MILK PROJECT

Question Number: 12 Question Type: Ranking Question Title: Which sample did you like the BEST? Design: T=2, K=2, B=100

Products

Product Code Name 1 - 138 138 control 2 - 643 643 fortified carrageenan

Crosstabulation

Sample 1 2 Total 1 - 138 20 43 63 2 - 643 43 20 63 TOTALS 63 63 126

Percentage Crosstabulation

Sample 1 2 Total 1 - 138 31.8 68.3 100 2 - 643 68.3 31.8 100

Friedman Analysis of Rank This procedure is valid for Complete Block Experimental Designs with no missing data only. This is a Complete Block Design.

Calculated Degrees Friedman of p- Statistic Freedom value 8.39 1 0.004

Critical values corresponding to specific levels of significance: 10%=2.71,5%=3.84,1%=6.63 The samples differ at the 10% level. (8.39 >= 2.71) The samples differ at the 5% level. (8.39 >= 3.84) The samples differ at the 1% level. (8.39 >= 6.63)

Tukey's HSD = 15.547 (5% Significance Level) Rank Sample Total Significantly Different Than Sample 1 - 138 106.00 a 2 2 - 643 83.00 b

311

Project: 2518 REVISED SOY MILK PROJECT

Question Number: 13 Question Type: Category / Hedonics Question Title: If you were to receive a bottle of this soy milk, would you drink all of it? Attribute Number: 1 Attribute Title: Q#13.1 Design: T=2, K=2, B=100

Products

Product Code Name 1 - 138 138 control 2 - 643 643 fortified carrageenan

Scale Parameters

Value Descriptor 5 Certainly 4 Maybe 3 I don't know 2 Maybe not 1 Certainly not

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation

1 2 3 4 5 Sample [5] [4] [3] [2] [1] Total 1 - 138 34 18 6 5 63 2 - 643 48 8 5 2 63 TOTALS 82 26 11 7 126

Percentage Crosstabulation

1 2 3 4 5 Sample [5] [4] [3] [2] [1] Total 1 - 138 54.0 28.6 9.5 7.9 100 2 - 643 76.2 12.7 7.9 3.2 100

Counts, Medians, Means and SD's

Sample Standard Number Count Total Median Mean Deviation 1 - 138 63 270.00 5.00 4.29 0.941 2 - 643 63 291.00 5.00 4.62 0.771

312

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares Squares F Value p-value Samples 1 3.500 3.500 5.29 0.0248 Judges 62 50.714 0.818 1.24 0.2024 Error 62 41.000 0.661 Total 125 95.214 0.762 Standard Error 0.102 (SEM) =

Tukey's HSD = 0.29 (5% Significance Level)

Sample Mean Significantly Different Than Sample 2 - 643 4.62 a 1 1 - 138 4.29 b

313

Ecuador Sensory Panel – All complete child ballots

Demographics

Project: 2518 CHILDREN BALLOT ANALYSIS - ALL INCOMPLETE BALLOTS EXCLUDED

Question Number: 2 Question Type: Multiple Choice (Demographic) Question Title: How old are you?

Choices

Number Value Choices 1 [1] 5-17 years old 2 [2] 18+ years old

Crosstabulation

1 2 Sample [1] [2] Total n/a 23 23 TOTALS 23 23

Percentage Crosstabulation

1 2 Sample [1] [2] Total n/a 100 100

314

Project: 2518 CHILDREN BALLOT ANALYSIS - ALL INCOMPLETE BALLOTS EXCLUDED

Question Number: 3 Question Type: Multiple Choice (Demographic) Question Title: Are you male or female?

Choices

Number Value Choices 1 [2] Male 2 [1] Female

Crosstabulation

1 2 Sample [2] [1] Total n/a 9 14 23 TOTALS 9 14 23

Percentage Crosstabulation

1 2 Sample [2] [1] Total n/a 39.1 60.9 100

315

Project: 2518 CHILDREN BALLOT ANALYSIS - ALL INCOMPLETE BALLOTS EXCLUDED

Question Number: 4 Question Type: Multiple Choice (Demographic) Question Title: Have you had soy milk before?

Choices

Number Value Choices 1 [3] Yes 2 [2] No 3 [1] I don't know

Crosstabulation

1 2 3 Sample [3] [2] [1] Total n/a 16 7 23 TOTALS 16 7 23

Percentage Crosstabulation

1 2 3 Sample [3] [2] [1] Total n/a 69.6 30.4 100

316

Project: 2518 CHILDREN BALLOT ANALYSIS - ALL INCOMPLETE BALLOTS EXCLUDED

Question Number: 5 Question Type: Multiple Choice (Demographic) Question Title: What do you think about soy milk?

Choices

Number Value Choices 1 [1] I like it 2 [2] I don't know 3 [3] I don't like it

Crosstabulation

1 2 3 Sample [1] [2] [3] Total n/a 16 7 23 TOTALS 16 7 23

Percentage Crosstabulation

1 2 3 Sample [1] [2] [3] Total n/a 69.6 30.4 100

317

Hedonics Project: 2518 CHILDREN BALLOT ANALYSIS - ALL INCOMPLETE BALLOTS EXCLUDED

Question Number: 6 Question Type: Category / Hedonics Question Title: IN GENERAL, how do you feel about the samples? Attribute Number: 1 Attribute Title: Q#6.1 Design: T=2, K=2, B=100

Products

Product Code Name 1 - 138 138 control 2 - 643 643 fortified carrageenan

Scale Parameters Value Descriptor 7 like very much 6 like moderately 5 like slightly 4 neither like nor dislike 3 dislike slightly 2 dislike moderately 1 dislike very much

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation

1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 14 5 3 1 23 2 - 643 18 3 2 23 TOTALS 32 8 5 1 46

Percentage Crosstabulation

1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 60.9 21.7 13.0 4.4 100 2 - 643 78.3 13.0 8.7 100

Counts, Medians, Means and SD's

Sample Standard Number Count Total Median Mean Deviation 1 - 138 23 146.00 7.00 6.35 1.027 2 - 643 23 154.00 7.00 6.70 0.635

318

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares Squares F Value p-value Samples 1 1.391 1.391 1.64 0.2130 Judges 22 13.478 0.613 0.72 0.7723 Error 22 18.609 0.846 Total 45 33.478 0.744 Standard Error 0.191 (SEM) =

Tukey's HSD = 0.563 (5% Significance Level)

Sample Mean Significantly Different Than Sample 2 - 643 6.70 a 1 - 138 6.35 a

319

Project: 2518 CHILDREN BALLOT ANALYSIS - ALL INCOMPLETE BALLOTS EXCLUDED

Question Number: 7 Question Type: Category / Hedonics Question Title: What do you think about the TASTE of the samples? Attribute Number: 1 Attribute Title: Q#7.1 Design: T=2, K=2, B=100

Products

Product Code Name 1 - 138 138 control 2 - 643 643 fortified carrageenan

Scale Parameters

Value Descriptor 7 like very much 6 like moderately 5 like slightly 4 neither like nor dislike 3 dislike slightly 2 dislike moderately 1 dislike very much

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 13 7 1 1 1 23 2 - 643 17 5 1 23 TOTALS 30 12 1 2 1 46

Percentage Crosstabulation 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 56.5 30.4 4.4 4.4 4.4 100 2 - 643 73.9 21.7 4.4 100

Counts, Medians, Means and SD's

Sample Standard Number Count Total Median Mean Deviation 1 - 138 23 144.00 7.00 6.26 1.214 2 - 643 23 153.00 7.00 6.65 0.714

320

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares Squares F Value p-value Samples 1 1.761 1.761 2.31 0.1424 Judges 22 26.913 1.223 1.61 0.1365 Error 22 16.739 0.761 Total 45 45.413 1.009 Standard Error 0.181 (SEM) =

Tukey's HSD = 0.534 (5% Significance Level)

Sample Mean Significantly Different Than Sample 2 - 643 6.65 a 1 - 138 6.26 a

321

Project: 2518 CHILDREN BALLOT ANALYSIS - ALL INCOMPLETE BALLOTS EXCLUDED

Question Number: 8 Question Type: Category / Hedonics Question Title: What do you think about the COLOR of the samples? Attribute Number: 1 Attribute Title: Q#8.1 Design: T=2, K=2, B=100

Products

Product Code Name 1 - 138 138 control 2 - 643 643 fortified carrageenan

Scale Parameters Value Descriptor 7 like very much 6 like moderately 5 like slightly 4 neither like nor dislike 3 dislike slightly 2 dislike moderately 1 dislike very much

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 9 8 3 2 1 23 2 - 643 12 10 1 23 TOTALS 21 18 3 3 1 46

Percentage Crosstabulation 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 39.1 34.8 13.0 8.7 4.4 100 2 - 643 52.2 43.5 4.4 100

Counts, Medians, Means and SD's

Sample Standard Number Count Total Median Mean Deviation 1 - 138 23 136.00 6.00 5.91 1.276 2 - 643 23 148.00 7.00 6.43 0.728

322

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares Squares F Value p-value Samples 1 3.130 3.130 4.97 0.0364 Judges 22 33.609 1.528 2.42 0.0217 Error 22 13.870 0.630 Total 45 50.609 1.125 Standard Error 0.165 (SEM) =

Tukey's HSD = 0.486 (5% Significance Level)

Sample Mean Significantly Different Than Sample 2 - 643 6.43 a 1 1 - 138 5.91 b

323

Project: 2518 CHILDREN BALLOT ANALYSIS - ALL INCOMPLETE BALLOTS EXCLUDED

Question Number: 9 Question Type: Category / Hedonics Question Title: What do you think about the AROMA of the samples? Attribute Number: 1 Attribute Title: Q#9.1 Design: T=2, K=2, B=100

Products Product Code Name 1 - 138 138 control 2 - 643 643 fortified carrageenan

Scale Parameters Value Descriptor 7 like very much 6 like moderately 5 like slightly 4 neither like nor dislike 3 dislike slightly 2 dislike moderately 1 dislike very much

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 10 3 6 3 1 23 2 - 643 12 8 2 1 23 TOTALS 22 11 8 4 1 46

Percentage Crosstabulation 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 43.5 13.0 26.1 13.0 4.4 100 2 - 643 52.2 34.8 8.7 4.4 100

Counts, Medians, Means and SD's

Sample Standard Number Count Total Median Mean Deviation 1 - 138 23 133.00 6.00 5.78 1.278 2 - 643 23 146.00 7.00 6.35 0.832

324

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares Squares F Value p-value Samples 1 3.674 3.674 4.80 0.0393 Judges 22 34.304 1.559 2.04 0.0510 Error 22 16.826 0.765 Total 45 54.804 1.218 Standard Error 0.182 (SEM) =

Tukey's HSD = 0.535 (5% Significance Level)

Sample Mean Significantly Different Than Sample 2 - 643 6.35 a 1 1 - 138 5.78 b

325

Project: 2518 CHILDREN BALLOT ANALYSIS - ALL INCOMPLETE BALLOTS EXCLUDED

Question Number: 10 Question Type: Category / Hedonics Question Title: What do you think about the TEXTURE of the samples? Attribute Number: 1 Attribute Title: Q#10.1 Design: T=2, K=2, B=100

Products

Product Code Name 1 - 138 138 control 2 - 643 643 fortified carrageenan

Scale Parameters

Value Descriptor 5 way too thick 4 a little too thick 3 just about right 2 a little thin 1 way too thin

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation

1 2 3 4 5 Sample [5] [4] [3] [2] [1] Total 1 - 138 2 12 4 5 23 2 - 643 4 17 2 23 TOTALS 6 29 6 5 46

Percentage Crosstabulation

1 2 3 4 5 Sample [5] [4] [3] [2] [1] Total 1 - 138 8.7 52.2 17.4 21.7 100 2 - 643 17.4 73.9 8.7 100

Counts, Medians, Means and SD's

Sample Standard Number Count Total Median Mean Deviation 1 - 138 23 57.00 3.00 2.48 0.947

326

Sample Standard Number Count Total Median Mean Deviation 2 - 643 23 71.00 3.00 3.09 0.515

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares Squares F Value p-value Samples 1 4.261 4.261 5.96 0.0232 Judges 22 9.826 0.447 0.62 0.8616 Error 22 15.739 0.715 Total 45 29.826 0.663 Standard Error 0.176 (SEM) =

Tukey's HSD = 0.518 (5% Significance Level)

Sample Mean Significantly Different Than Sample 2 - 643 3.09 a 1 1 - 138 2.48 b

327

Project: 2518 CHILDREN BALLOT ANALYSIS - ALL INCOMPLETE BALLOTS EXCLUDED

Products Product Code Name 1 - 138 138 control 2 - 643 643 fortified carrageenan

Scale Parameters Value Descriptor 7 like very much 6 like moderately 5 like slightly 4 neither like nor dislike 3 dislike slightly 2 dislike moderately 1 dislike very much

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 13 4 4 1 1 23 2 - 643 16 6 1 23 TOTALS 29 10 5 1 1 46

Percentage Crosstabulation 1 2 3 4 5 6 7 Sample [7] [6] [5] [4] [3] [2] [1] Total 1 - 138 56.5 17.4 17.4 4.4 4.4 100 2 - 643 69.6 26.1 4.4 100

Counts, Medians, Means and SD's

Sample Standard Number Count Total Median Mean Deviation 1 - 138 23 141.00 7.00 6.13 1.290 2 - 643 23 153.00 7.00 6.65 0.573

328

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares Squares F Value p-value Samples 1 3.130 3.130 4.63 0.0426 Judges 22 28.957 1.316 1.95 0.0628 Error 22 14.870 0.676 Total 45 46.957 1.043 Standard Error 0.171 (SEM) =

Tukey's HSD = 0.503 (5% Significance Level)

Sample Mean Significantly Different Than Sample 2 - 643 6.65 a 1 1 - 138 6.13 b

329

Project: 2518 CHILDREN BALLOT ANALYSIS - ALL INCOMPLETE BALLOTS EXCLUDED

Question Number: 12 Question Type: Ranking Question Title: Which sample did you like the BEST? Design: T=2, K=2, B=100

Products Product Code Name 1 - 138 138 control 2 - 643 643 fortified carrageenan

Crosstabulation

Sample 1 2 Total 1 - 138 8 15 23 2 - 643 15 8 23 TOTALS 23 23 46

Percentage Crosstabulation

Sample 1 2 Total 1 - 138 34.8 65.2 100 2 - 643 65.2 34.8 100

Friedman Analysis of Rank This procedure is valid for Complete Block Experimental Designs with no missing data only. This is a Complete Block Design. Calculated Degrees Friedman of p- Statistic Freedom value 2.13 1 0.144

Critical values corresponding to specific levels of significance: 10%=2.71 5%=3.84 1%=6.63 No difference between the samples at the 10% level. (2.13 < 2.71) No difference between the samples at the 5% level. (2.13 < 3.84) No difference between the samples at the 1% level. (2.13 < 6.63)

Tukey's HSD = 9.394 (5% Significance Level) Rank Sample Total Significantly Different Than Sample 1 - 138 38.00 a 2 - 643 31.00 a

330

Project: 2518 CHILDREN BALLOT ANALYSIS - ALL INCOMPLETE BALLOTS EXCLUDED

Products

Product Code Name 1 - 138 138 control 2 - 643 643 fortified carrageenan

Scale Parameters

Value Descriptor 5 Certainly 4 Maybe 3 I don't know 2 Maybe not 1 Certainly not

Note: Numbers shown in brackets are the 'values' associated with the category selected.

Crosstabulation

1 2 3 4 5 Sample [5] [4] [3] [2] [1] Total 1 - 138 14 5 2 2 23 2 - 643 21 1 1 23 TOTALS 35 6 3 2 46

Percentage Crosstabulation

1 2 3 4 5 Sample [5] [4] [3] [2] [1] Total 1 - 138 60.9 21. 8.7 8.7 100 7 2 - 643 91.3 4.4 4.4 100

Counts, Medians, Means and SD's

Sample Standard Number Count Total Median Mean Deviation 1 - 138 23 100.00 5.00 4.35 0.982

331

Sample Standard Number Count Total Median Mean Deviation 2 - 643 23 112.00 5.00 4.87 0.458

This is a Complete Block Design.

Analysis of Variance This analysis does not compensate for missing data or lack of balance. Sum of Mean of D.F. Squares Squares F Value p-value Samples 1 3.130 3.130 4.6 0.0426 3 Judges 22 10.957 0.498 0.7 0.7601 4 Error 22 14.870 0.676 Total 45 28.957 0.643 Standard Error 0.171 (SEM) =

Tukey's HSD = 0.503 (5% Significance Level)

Sample Mean Significantly Different Than Sample 2 - 643 4.87 a 1 1 - 138 4.35 b

332