Neither low salivary amylase activity, cooling cooked white rice, nor

single nucleotide polymorphisms in starch-digesting enzymes reduce

glycemic index or starch digestibility: a randomized, crossover trial in
healthy adults
Thomas MS Wolever,1,2 Ahmed El-Sohemy,1 Adish Ezatagha,2 Andreea Zurbau,1,2 and Alexandra L Jenkins2
1 Department of Nutritional Sciences, Temerty Faculty of Medicine, University of Toronto, Toronto, Ontario, Canada; and 2 INQUIS Clinical Research, Ltd
(formerly GI Labs), Toronto, Ontario, Canada

ABSTRACT Conclusions: The results do not support the hypotheses that low-
Background: It was suggested that low salivary-amylase activity SAA, cooling, and common genetic variations in starch-digesting en-
(SAA) and cooling or stir-frying cooked starch decreases its zymes affect the glycemic index or in vivo carbohydrate digestibility
digestibility and glycemic index. of cooked polished rice. This trial was registered at clinicaltrials.gov
Objective: We determined the effects of SAA, cooling, and as NCT03667963. Am J Clin Nutr 2021;114:1633–1645.
single-nucleotide polymorphisms (SNPs) in the salivary amy-
lase (AMY1), pancreatic amylase (AMY2A, AMY2B), maltase- Keywords: glycemic index, rice, cooling, stir-frying, starch
glucoamylase (MGAM), and sucrase-isomaltase (SI) genes on starch digestibility, salivary amylase activity, genetic polymorphisms,
digestibility and glycemic index of cooked polished rice. maltase-glucoamylase, sucrose-isomaltase
Methods: Healthy subjects [pilot, n = 12; main, n = 20 with low-
SAA (<50 U/mL), and n = 20 with high-SAA (>105 U/mL)]
consumed test meals containing 25 g (pilot) or 50 g (main) available
carbohydrate at a contract research organization using open-label Introduction
(pilot) or assessor-blinded (main), randomized, crossover, Latin- The glycemic index (GI) measures the blood-glucose–raising
square designs (trial registration: NCT03667963). Pilot-trial test ability of the available carbohydrate (avCHO) in foods, defined
meals were dextrose, freshly cooked polished rice, cooked rice conceptually as 100 × Fgr /Ggr where Fgr and Ggr are the glycemic
cooled overnight, stir-fried hot rice, or stir-fried cold rice. Main- responses elicited by the food and glucose, respectively, in the
trial test meals were dextrose, dextrose plus 10 g lactulose, plain hot
rice, or plain cold rice. In both trials, blood glucose was measured
fasting and at intervals over 2 h. In the main trial, breath hydrogen Supported by an anonymous donation to the University of Toronto and by
INQUIS Clinical Research, Ltd. The authors, who are owners or employees
was measured fasting and hourly for 6 h to estimate in vivo starch
of INQUIS, were solely responsible for the design, implementation, analysis,
digestibility. Data were analyzed by repeated-measures ANOVA for
and interpretation of the results.
the main effects of temperature and stir-frying (pilot trial) or the main Supplemental Figures 1–6, Supplemental Methods, and Supplemental
effects of SAA and temperature (main trial) and their interactions. Tables 1–15 are available from the “Supplementary data” link in the online
Effects of 24 single nucleotide polymorphisms (SNPs) were assessed posting of the article and from the same link in the online table of contents at
separately. Means were considered to be equivalent if the 95% CI https://academic.oup.com/ajcn/.
of the differences were within ±20% of the comparator mean for Address correspondence to TMSW (e-mail: twolever@inquis.com).
glucose response/glycemic index or ±7% for digestibility. Abbreviations used: AMY1, salivary amylase gene; AMY2A and AMY2B,
Results: Pilot: neither temperature nor stir-frying significantly pancreatic amylase genes; AMY1CN, AMY1 copy number; avCHO, available
affected glucose incremental AUC (primary endpoint, n = 12). carbohydrate; Cmalab, amount of starch malabsorbed; Dex, dextrose;
Dex+L, dextrose plus lactulose; Fgr , glycemic response elicited by a
Main: mean ± SEM glycemic index (primary endpoint, n = 40)
food; Ggr , glycemic response elicited by glucose; GI, glycemic index;
was equivalent for low-SAA compared with high-SAA (73 ± 3
iAUC, incremental AUC; ICC, intraclass correlation coefficient; L, lactulose;
vs. 75 ± 4) and cold rice compared with hot rice (75 ± 3 vs. MGAM, maltase-glucoamylase; R, sum of breath hydrogen concentrations
70 ± 3). Estimated starch digestibility (n = 39) was equivalent for after rice; SAA, salivary amylase activity; SI, sucrase-isomaltase; SNP, single
low-SAA compared with high-SAA (95% ± 1% vs. 92% ± 1%) nucleotide polymorphism; TH, sum of breath hydrogen concentrations.
and hot rice compared with cold rice (94% ± 1% vs. 93% ± 1%). Received March 15, 2021. Accepted for publication June 14, 2021.
No meaningful associations were observed between genotypes and First published online July 22, 2021; doi: https://doi.org/10.1093/ajcn/
starch digestibility or glycemic index for any of the SNPs. nqab228.

Am J Clin Nutr 2021;114:1633–1645. Printed in USA. © The Author(s) 2021. Published by Oxford University Press on behalf of the American Society for
Nutrition. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com 1633
1634 Wolever et al.

same subjects (1). Potato and rice are usually considered to have of Toronto and the protocols of both studies were approved
a high GI, but, in fact, their GIs vary widely (2–4). The high SD by the University of Toronto Health Sciences Research Ethics
of reported GI values for potato, (mean ± SD: 68 ± 20; n = 29) Board and were carried out at INQUIS (formerly Glycemic
and rice (61 ± 21; n = 87) (5) is not explained by interlaboratory Index Laboratories, Inc.; Toronto, ON, Canada). Prior to their
variation (SD = 9) (6, 7) and likely reflects differences in variety, participation all subjects provided informed consent by signing
maturity, or methods of processing/cooking that influence starch the approved consent form. The pilot trial was not registered;
structure and its rate of hydrolysis (8–12). The high variation in however, the main trial was registered on clinicaltrials.gov as
GI estimates in individuals (13–15) is explained by day-to-day NCT03667963, where GI was declared as the primary endpoint
variation in glycemic responses within subjects (16, 17), with with secondary endpoints carbohydrate malabsorbed, glycemic
age, sex, ethnicity, BMI, and diabetes having no effect on GI response, and breath hydrogen response. Breath methane was
because they affect Fgr and Ggr to the same extent (7, 18, 19). not included because we overlooked that the breath analyzer
For a personal characteristic to affect GI, it would have to change measured both hydrogen and methane simultaneously.
Fgr or Ggr but not both.
Starch must be digested to elicit a glycemic response,
but glucose is directly absorbed; thus, differences in starch Pilot trial
digestion could result in interindividual variation in GI. The We studied n = 12 healthy participants [5 males, 7 females;
enzymes involved in starch digestion include salivary amylase mean ± SD (range) age: 46 ± 16 (21–66) y; BMI (kg/m2 ):
(AMY1), pancreatic amylase (AMY2), sucrase-isomaltase (SI), 29.6 ± 5.9 (22.9–40.7)] on 5 separate occasions between
and maltase-glucoamylase (MGAM). Single nucleotide polymor- 17 February and 26 March 2016 in the morning after 10–12-
phisms (SNPs) in AMY1, AMY2A, and AMY2B are associated h overnight fasts using an open-label, randomized, crossover
with differences in AMY1 copy number (AMY1CN) (20), which, design (study flow chart is shown in Supplemental Figure 1).
in turn, determines salivary amylase activity (SAA) and varies Test meals consisted of 25 g avCHO from glucose-monohydrate
markedly in different individuals (21). However, the role of (27.5 g dextrose) dissolved in 250 mL water, or polished long-
salivary amylase in starch digestion is not fully understood (22). grain white rice (32.1 g dry weight; NuPack; Shah Trading Co.).
Recent evidence suggests that high-AMY1CN and high-SAA Rice was cooked by placing 128.4 g (4 portions), 320 g water, and
lead to more complete starch digestion and higher GI (23). a pinch of salt into a rice cooker and cooked until done (∼18 min);
However, other studies suggest that high-SAA has no effect (24) the weight of the cooked rice divided by 4 was the weight of
or reduces glycemic responses (25). The brush-border enzymes each individual serving. Rice was either cooked on the morning
SI and MGAM hydrolyze disaccharides. Congenital SI deficiency, of the test and served hot without stir-frying (Plain-Hot) or after
a rare condition due to a mutation in the SI gene, causes sucrose stir-frying with 5 mL sesame oil for 2.5 min (SFHot), or cooked
malabsorption (26). SI and MGAM are also involved in starch the previous afternoon, placed in a refrigerator at 4◦ C overnight,
digestion (27), but the effect of common SNPs on their activity is and either served cold the next morning either without stir-frying
unknown. (Plain-Cold) or after stir-frying with 5 mL sesame oil for 2.5 min
Factors affecting the GI of rice are relevant because rice is (SFCold). On the first day, each participant chose to drink 1 or
a staple food in East and South East Asia, where high rice 2 cups of water, tea, or coffee (with or without 30 mL 2% milk)
consumption is associated with increased diabetes risk in some with the test meal, and the same drink was given with the other 4
(28–30) but not all studies (31). One study showed that the test meals. Test meals were served in randomized order.
association of rice intake with diabetes risk factors varied in Participants came to INQUIS between 08:00 and 10:00 h; after
different regions of China (32), a result the authors speculated being weighed and providing 2 fasting finger-stick blood samples
might be due to the practice in some areas of China of cooling ∼5 min apart, they started to consume 1 of the 5 test meals.
rice overnight and stir-frying prior to consumption. Cooling may Additional finger-stick blood samples were obtained at 15, 30,
reduce the glycemic impact of cooked rice (33, 34) and other 45, 60, 90, and 120 min after starting to eat.
starchy foods (35–39), but the effect is not consistent. Exposing
starch to fatty acids in vitro creates starch/lipid complexes with
reduced digestibility (40, 41); however, the effect of stir-frying Main trial
rice on glycemic response is unknown. Healthy participants were recruited from individuals who had
Thus, our objectives were to determine the effects of SAA, previously had saliva collected to measure SAA and variations
cooling, stir-frying, and variation in AMY1, AMY2A, AMY2B, in 24 SNPs in AMY1, AMY2A, AMY2B, MGAM, and SI genes
MGAM, and SI genes on the glycemic impact and digestibility (methods described in the Supplemental Methods). Those
of polished white rice in healthy subjects without diabetes. recruited for this study were in the lower or upper or tertiles of
SAA (<50 U/mL vs. >10 5 U/mL) and selected in such a way that
the sex, age, BMI, and ethnicity of the low- and high-SAA groups
Methods would be similar. We recruited n = 41 subjects; 1 in the high-SAA
We conducted 2 separate studies: a pilot trial to determine the group dropped out for personal reasons after the first test meal
effect on glycemic response of cooling and stir-frying polished and was replaced; thus, 20 subjects with low-SAA [10 males,
long-grain white rice and a main trial to determine the effects 10 females; aged 39.0 ± 13.7 y; BMI (kg/m2 ), 25.1 ± 3.4; n = 10
of cooling, SAA, and SNPs in genes encoding for starch- Whites] and 20 subjects with high-SAA (11 males, 9 females;
digesting enzymes on the GI and carbohydrate digestibility of aged 45.1 ± 15.2 y; BMI, 27.6 ± 4.5; n = 14 Whites) completed
polished medium-grain white rice. The procedures followed the trial (study flow chart shown as Supplemental Figure 2).
were in accordance with the ethical standards of the University Additional details about the participants are shown in Table 1.
 GI of rice by salivary amylase activity and cooling 1635
TABLE 1 Details of participants in the main trial1

Low-SAA (<50 U/mL) High-SAA (>105 U/mL)

Sex (M:F), n 10:10 11:9
Age, y 39.1 ± 13.7 (22–61) 45.1 ± 15.2 (19–68)
Ethnicity,2 n
Black 0 1
White 10 14
Chinese 2 0
Indigenous 1 2
Latin American 3 0
South Asian 1 3
Southeast Asian 3 0
Height, m 1.70 ± 0.09 (1.51–1.89) 1.70 ± 0.10 (1.46–1.90)
Weight, kg 66.5 ± 3.6 (58.9–73.6) 79.3 ± 14.6 (56.8–74.1)
BMI, kg/m2 25.1 ± 3.4 (20.6–33.4) 27.6 ± 4.5 (21.5–35.0)
Lean:overweight:obese,3 n 8:8:4 5:8:7
SAA at screening (U/ml) 29.5 ± 14.6 (5.6–49.2) 162.6 ± 90.5 (105.6–526.1)
SAA during the study (U/ml) 50.4 ± 43.6 (7.5–194.2) 171.7 ± 164.8 (17.7–665.8)∗
1 Values are means ± SD (range) unless otherwise indicated. ∗ Different from Low-SAA group: P = 0.004

(2-tailed t-test). SAA, salivary amylase activity.

2 Ethnicity was determined by asking subjects to select their “Ethnicity” from the following list: Caucasian, Black

(e.g., African, Haitian, Jamaican, Somali), Indigenous, South Asian (eg, East Indian, Pakistani, Punjabi, Sri Lankan),
Arab (eg, Egyptian, Lebanese, Moroccan), West Asian (e.g., Iranian), Filipino, Southeast Asian (e.g., Cambodian,
Indonesian, Vietnamese), Latin American, Chinese, Japanese, Korean, or Other. "Caucasian" is referred to in the table
and text as "White" as requested by AJCN editors.
3 For White and Black participants; lean = BMI <25; overweight = BMI 25–29.99; obese = BMI ≥30. For

others: lean = BMI <23; overweight = BMI 23–26.99; obese = BMI ≥27.

Each subject was studied on 4 separate occasions between 22 Subjects continued to sit quietly until 210 min after starting the
August and 18 December 2018 after 10–12-h overnight fasts test meal, at which time they started to consume lunch, which
using a single-blind, randomized, crossover design (those doing consisted of the subject’s choice of 1 or 2 cheese sandwiches with
the biochemical analysis, data entry, and statistical analysis were juice and 2 cookies [2470 or 3550 kJ (590 or 850 kcal), 18 or 28 g
not aware of the identity of the test meals until after the database protein, 21 or 31 g fat, 86 or 124 g carbohydrate, and 2 or 3 g
had been cleaned and locked). The interval between tests was 3– dietary fiber]. The lunch chosen was repeated on each test day.
14 d for 117 of the 120 intervals; on 1 occasion each of 3 subjects After the 6-h breath sample, subjects were offered a snack and
had intervals of 17, 22, or 24 d. were free to leave.
On each test occasion subjects consumed 1 of 4 different test
meals containing 50 g avCHO: 55 g dextrose (Dex) dissolved in
250 mL water; Dex plus 10 g lactulose (15 mL lactulose syrup) Blood and breath sample collection and analysis
dissolved in 250 ml water (Dex+L); 58 g (dry weight) of hot rice Each finger-stick blood sample consisted of 2–3 drops
or cold rice. Rice was cooked by placing 58 g polished medium- collected into fluoro-oxalate vials, which were twirled to mix
grain white rice (Calrose variety), 174 mL water, and a pinch of the fluoro-oxalate with the blood and stored at 4◦ C until the
salt into a rice cooker and cooking for ∼22 min. Hot rice was end of the test at which time the set of samples was bundled
cooked on the morning of the test and served hot; cold rice was together with a rubber band and stored at −20◦ C prior to
cooked the previous afternoon, placed in a refrigerator at 4◦ C glucose analysis using an automatic analyzer (2300 STAT; Yellow
overnight, and served cold the next morning. Rice test meals Springs Instruments) within 5 d. Breath samples (10 mL) were
contained 0 g fat, 3 g protein, 50 g carbohydrate, and 0 g dietary collected into glass tubes using the AlveoSampler™ breath test
fiber. On the first day, each participant chose to drink 1 or 2 cups kit (QuinTron Instrument Co.) and analyzed for hydrogen (H2 )
of water, tea, or coffee (with or without 30 mL 2% milk) with and methane (CH4 ) using the BreathTracker Analyzer (QuinTron
the test meal, and the same drink was given with the other 3 test Instrument Co.).
meals. Test meals were served in randomized order.
Participants came to INQUIS between 08:00 and 10:00 h; after
being weighed, providing 2 fasting finger-stick blood samples Calculations
∼5 min apart, and providing a fasting breath sample, they started Incremental AUCs (iAUCs), ignoring area below fasting, were
to consume 1 of the 4 test meals. Additional finger-stick blood calculated using the trapezoid rule (42). Fasting glucose was
samples were obtained at 15, 30, 45, 60, 90, and 120 min after taken to be the mean of the concentrations in the 2 fasting
starting to eat, and additional breath samples were collected at samples. GI was not calculated in the pilot trial since each subject
hourly intervals for 6 h after starting to eat. On 1 occasion, after tested the glucose control only once. In the main trial, GI was
the 120-min blood and breath samples had been collected, 2 mL calculated by expressing the iAUC after each rice test meal as
of saliva was collected as described in the Supplemental Methods. a percentage of the mean iAUC after Dex and Dex+L taken by
1636 Wolever et al.

the same subject; the mean of the resulting values was the GI of coefficient (ICC) calculated as (sb 2 − sw 2 )/(sb 2 + sw 2 ), where
the rice meal. Individual values >2 SDs above the mean were sb 2 and sw 2 were the between-subjects mean square and within-
excluded as outliers. subjects mean square terms, respectively, from the ANOVA.
Breath H2 and CH4 increments were calculated using the The significance of the differences in frequency distributions
lowest value within the first 3 h as the baseline (43). For each was assessed using the chi-square test. The significance of
test meal we calculated total H2 (TH) as the sum of the breath H2 associations between outcomes and alleles/genotypes was not
(H) concentrations from the lowest value within the first 3 h to adjusted for multiple comparisons.
the value at 6 h. For each participant we then calculated TH per For equivalence testing, the mean iAUC for cold rice and stir-
g of lactulose (L) as (DL-D)/10, where, DL = TH after Dex+L, fried rice in the pilot trial and the mean GI for the low-SAA group
D = TH after Dex, and 10 = the grams of lactulose in Dex+L. and for cold rice in the main trial were taken to be equivalent
Since lactulose is not digestible and is rapidly fermented when it to those for hot rice and plain rice, respectively, if the upper
enters the colon, L represents the amount of breath H2 excreted and lower bounds of the 95% CI of the differences were within
by each participant per gram of rapidly fermented carbohydrate ±20% of the mean for the comparator. The rationale for this is
entering the colon. The estimated amount of starch malabsorbed that Health Canada considers that iAUC differences of <20% are
after each rice test meal (Cmalab) was calculated as R/L, where physiologically insignificant and do not qualify for a glucose-
R is TH after each rice test meal (39, 43). Therefore, estimated reduction claim (44). Carbohydrate digestibility (main trial) in
starch digestibility (%) = 100 − (2 × Cmalab). the low-SAA group and for cold rice was taken to be equivalent
to that for the high-SAA group and hot rice, respectively, if the
upper and lower bounds of the 95% CI of the differences were
Statistical power and statistical analysis within ±7% of the mean for the comparator. The rationale for
The primary endpoint for the pilot trial was glucose iAUC this is that, classically, fat malabsorption is diagnosed if fecal fat
and in the main trial was GI. Power analysis was based on an output is ≥7 g while consuming 100 g fat/d (45).
assumed within-individual CV of 23% (7) and used the 2-tailed
t-distribution. For the pilot trial, n = 12 provided 80% power to
detect a 29% difference in iAUC between individual test meals Results
and 80% power to detect a 20% difference for the main effects
Pilot trial
of hot compared with cold or plain compared with stir-fried
[Health Canada considers 20% to be the minimal physiologically All test meals were well received and consumed within the
significant difference in glycemic response (44)]. For the main prescribed time. There was no significant difference in mean
trial, n = 40 provided 80% power to detect a 15% difference in fasting glucose among the 5 test meals (Table 2). Mean glucose
GI. increments after Dex were significantly higher than those after
In the pilot trial, to assess the difference between the rice all 4 rice test meals at 15, 30, and 45 min and significantly less
test meals and dextrose, glucose increments were subjected to than all 4 rice test meals at 90 and 120 min (Figure 1A). Mean
repeated-measures ANOVA using the general linear model for glucose iAUC after Dex was significantly greater than those after
the main effects of time and test meal and the time × test all 4 rice test meals (Table 2, Supplemental Table 1).
meal interaction using Prism 9.1.0 (GraphPad Software). After When only the 4 rice test meals were considered, there was
demonstrating a significant time × test meal interaction [F(28,308) no significant time × temperature interaction [F(7,231) = 1.47,
= 19.00, P < 0.0001], values for iAUC and glucose increments P = 0.18]. Mean glucose increments were similar for hot and cold
at each time were analyzed by ANOVA. Glucose increments for rice at all time points (Figure 1B, C). There was a significant main
the 4 rice test meals were subjected to ANOVA examining for effect of stir-frying on glucose increments at 30 and 120 min,
the main effects of time, test meal, temperature, and stir-frying with the mean for stir-fried rice being significantly less than
and their interactions. Since there were significant time × test plain rice at 30 min (1.49 ± 0.15 vs. 1.79 ± 0.14 mmol/L,
meal [F(21,231) = 1.99, P = 0.007] and time × stir-frying [F(7,231) P = 0.006) and greater than plain rice at 120 min (0.33 ± 0.08
= 2.44, P = 0.020] interactions, values for iAUC and glucose vs. 0.24 ± 0.08 mmol/L, P = 0.043; Figure 1B, D). There was
increments at each time were assessed by ANOVA. a significant temperature × stir-frying interaction on glucose
A similar approach was taken for the main trial; first, increments at 120 min, with the difference between plain hot
increments of glucose and breath H2 and CH4 were subjected to and cold rice being greater than that between stir-fried hot and
ANOVA for the main effects of time, SAA, and test meal and cold rice (0.29 ± 0.09 vs. −0.11 ± 0.13 mmol/L, P = 0.007;
their interactions. After demonstrating significant interactions, Figure 1B). Mean iAUC after cold rice did not differ significantly
differences among individual means were assessed by ANOVA. from that after hot rice, and mean iAUC after stir-fried rice
Separate ANOVAs were run to assess the differences between did not differ significantly from that after plain rice (Table 2,
Dex and Dex+L, between hot and cold rice, and between the Supplemental Table 1).
3 genotypes for each SNP (homozygous for the minor allele,
heterozygous, and homozygous for the major allele). Differences
between individual means were assessed using Tukey’s test to Main trial
adjust for multiple comparisons with the criterion for significance The salivary amylase concentrations measured before the
being a 2-tailed P < 0.05. study started were used to classify subjects a priori into
Univariate regression analysis was performed using the low-SAA (<50 U/mL) and high-SAA (>105 U/mL) groups; the
LINEST function of Excel (Office 16; Microsoft Corp.) to deter- means ± SDs (range) of the salivary amylase concentrations
mine the slope and Pearson’s correlation coefficient (r). Between- measured before and during the study are shown in Table 1. The
individual variation was assessed using the intraclass correlation SAA concentration measured during the study was between 51
 GI of rice by salivary amylase activity and cooling 1637
TABLE 2 Pilot trial: results for fasting glucose and glucose iAUC1

Dextrose Plain rice Stir-fried rice Mean

Fasting glucose, mmol/L
Hot rice 4.44 ± 0.12 4.52 ± 0.13 4.56 ± 0.15 4.54 ± 0.14
Cold rice 4.59 ± 0.14 4.54 ± 0.15 4.56 ± 0.14
Mean 4.56 ± 0.13 4.55 ± 0.14 P = 0.49∗
Glucose iAUC, mmol × min/L
Hot rice 148.3 ± 16.6 74.1 ± 6.6∗∗ 68.1 ± 5.6∗∗ 71.1 ± 5.6
Cold rice 80.2 ± 9.8∗∗ 72.9 ± 9.4∗∗ 76.5 ± 8.9
Mean 77.1 ± 7.3 70.5 ± 5.9 P = 0.88∗
1 Values are means ± SEMs for n = 12 participants. None of the differences for fasting glucose and none of the

differences between hot and cold rice or plain and stir-fried rice are statistically significant. The details of the ANOVA
are shown in Supplemental Table 1. ∗ Significance of temperature × stir-frying interaction among rice test meals.
∗∗ Different from Dextrose by Tukey’s test (2-tailed P < 0.0001). iAUC, incremental AUC.

and 104 U/mL, inclusive, for 7 subjects in the low-SAA group prior to the study was significantly correlated to those measured
and 3 in the high-SAA group while 1 low-SAA subject and during the study, ICC = 0.733. All results involving SAA
3 high-SAA subjects had results that placed them in the opposite presented below are based on the a priori values unless otherwise
group. The log of the salivary amylase concentrations measured indicated.
All test meals were well received and consumed within the
prescribed time. Two (2) of 1280 (0.16%) values for glucose
were missing due to clotted blood and they were imputed as the
mean of the surrounding values; 1 implausible fasting glucose
concentration (6.18 mmol/L) was replaced by the mean of the
other 3 values for the same subject (4.25 mmol/L). There were
no missing values for breath H2 and breath CH4 . There were no
significant differences between low- and high-SAA, or among the
4 test-meals for mean fasting glucose (Table 3) or for mean breath
H2 or breath CH4 (Figure 2).

Blood glucose increments.

There were significant main effects of time and test meal and
significant SAA × test meal and time × test meal interactions for
glucose increments (Supplemental Table 2). When only Dex and
Dex+L were included in the ANOVA, there were no significant
interactions; when only hot rice and cold rice were included in
the ANOVA, there were significant SAA × test meal, time × test
meal, and SAA × time × test meal interactions (Supplemental
Table 2).
In the low-SAA group, mean glucose increments for both rice
test meals were significantly below both dextrose test meals at
15 and 30 min; at 45 and 60 min, hot rice, but not cold rice,
differed significantly from both of the dextrose test meals; at
FIGURE 1 Pilot trial: postprandial blood glucose responses. Values are 120 min, both rice test meals were significantly greater than
mean ± SEM blood glucose increments for n = 12 participants. (A) Mean Dex, but not Dex+L. Hot rice differed from cold rice only at
of 4 rice test meals compared with Dex. (B) Comparison of plain hot, plain 45 min (Figure 2A, Supplemental Table 3). In the high-SAA
cold, SF hot, and SF cold rice. (C) Main effect of temperature (mean of group, mean glucose increments for both rice test meals were
plain hot and SF hot vs. mean of plain cold and SF cold). (D) Main effect
of stir-frying (mean of plain hot and plain cold vs. mean of SF hot and SF significantly below both dextrose test meals at 15, 30, and 60 min;
cold). ∗ Mean for Dex differs from mean of each of the 4 rice test meals by at 45 min, hot rice was significantly less than both dextrose test
Tukey’s test (P < 0.0001). † Mean for Dex differs from Hot (P = 0.0024), meals, whereas cold rice was less than Dex only; at 120 min,
Cold (P = 0.026), SF Hot (P = 0.005), and SF Cold (P = 0.003) by Tukey’s hot rice was significantly greater than both dextrose test meals
test. ‡ Mean for Dex differs from Hot (P < 0.0001), Cold (P = 0.021), SF
Hot (P < 0.0001), and SF Cold (P = 0.0001) by Tukey’s test. Cold, cooked (Figure 2B, Supplemental Table 3).
rice cooled overnight in a refrigerator; Dex, dextrose; F, significant main When only the 2 rice meals were included in the ANOVA, there
effect of stir-frying by ANOVA (stir-fried rice < plain rice, P = 0.007); were significant SAA × test meal interactions for the rice test
Hot = freshly cooked hot rice; S, stir-fried rice differs significantly from plain meals at 60 min (P = 0.033) and 120 min (P = 0.027). At 60 min,
rice by ANOVA (P = 0.015); SF, stir-fried just before serving (long-grain
polished white rice); T × F, significant temperature × stir-frying interaction mean glucose increment after hot rice was less than after cold
by ANOVA (P = 0.007). rice in the low-SAA group (P = 0.018), but similar to cold rice
1638 Wolever et al.

TABLE 3 Effects of SAA and temperature on glycemic responses elicited by rice1

Dextrose Medium-grain polished rice

Endpoint and SAA group Alone + Lactulose Mean Hot Cold Mean
Fasting glucose, mmol/L
Low 4.35 ± 0.07 4.31 ± 0.07 4.33 ± 0.07 4.40 ± 0.06 4.34 ± 0.07 4.37 ± 0.06
High 4.48 ± 0.11 4.52 ± 0.11 4.50 ± 0.11 4.49 ± 0.10 4.53 ± 0.10 4.51 ± 0.09
Mean 4.41 ± 0.07 4.41 ± 0.07 P = 0.3392 4.44 ± 0.06 4.44 ± 0.06 P = 0.3203
Glucose peak riseT , mmol/L
Low 4.26 ± 0.24 4.41 ± 0.27 4.34 ± 0.24 3.06 ± 0.17 3.50 ± 0.24 3.28 ± 0.18
High 4.55 ± 0.40 4.33 ± 0.35 4.44 ± 0.35 3.17 ± 0.16 3.29 ± 0.21 3.23 ± 0.17
Mean 4.40 ± 0.23A 4.37 ± 0.22A P = 0.2772 3.11 ± 0.12B 3.40 ± 0.16∗,B P = 0.2293
Peak timeT × G , min
Low 33.0 ± 1.4b 35.3 ± 2.0 34.1 ± 1.4 33.8 ± 1.5 39.8 ± 2.3a 36.8 ± 1.6
High 39.8 ± 2.0 36.0 ± 2.0 37.9 ± 1.3 37.5 ± 1.7 37.5 ± 1.7 37.5 ± 1.2
Mean 36.4 ± 1.3 35.6 ± 1.4 P = 0.1072 35.6 ± 1.2 38.6 ± 1.4 P = 0.0653
Glucose iAUCT , mmol × min/L
Low 236 ± 17 245 ± 23 240 ± 19 162 ± 14 198 ± 23 180 ± 17
High 262 ± 27 255 ± 25 259 ± 24 197 ± 21 181 ± 17 189 ± 18
Mean 249 ± 16A 250 ± 17A P = 0.5192 180 ± 13B 189 ± 14B P = 0.0223
Glycemic indexF × G,4 %
Low — — — 68.6 ± 3.5 81.3 ± 6.7 75.0 ± 3.9
High — — — 76.3 ± 5.8 69.7 ± 4.1 73.0 ± 4.1
Mean — — — 72.5 ± 3.4 75.5 ± 4.0 P = 0.0433
Glycemic index,5 %
Low — — — 66.8 ± 3.2 (19) 77.3 ± 5.7 (19) 72.5 ± 3.2 (38)
High — — — 72.5 ± 4.5 (19) 72.2 ± 3.4 (19) 75.1 ± 3.7 (38)
Mean — — — 69.7 ± 2.8 (38) 74.7 ± 3.3 (39) P = 0.2173
1 Values are means ± SEM for n = 20 with low SAA and n = 20 with high SAA (except where indicated). SAA, salivary amylase activity (low = <50

U/mL, high = >105 U/mL). T Significant main effect of test meal by ANOVA (P < 0.0001). AB Means with different superscript letters differ significantly by
Tukey’s test (P < 0.0001). T × G Significant test meal × SAA-group interaction by ANOVA (P < 0.0001). a Differs from Low-SAA dextrose (P = 0.017) and
Hot rice (P = 0.042, Tukey’s test); b Differs from High-SAA dextrose (P = 0.041, Bonferroni test). F × G Significant test meal × SAA-group interaction by
ANOVA for the 2 rice test meals, P < 0.05. ∗ Hot rice differs from cold rice by ANOVA including only rice test meals (P = 0.037). iAUC, incremental AUC;
SAA, salivary amylase activity.
2 P value for SAA group × lactulose interaction by ANOVA for Dextrose test meals.
3 P value for SAA group × temperature interaction by ANOVA for Rice test meals.
4 Values shown are for n = 20 in each of the 4 groups.
5 Values shown are means ± SEMs after excluding outliers (>2 SDs from the mean). (19, 38, 39) Numbers in parentheses indicate the number of

subjects included in the mean ± SEM.

(P > 0.99) in the high-SAA group; at 120 min, hot rice was higher There were no significant differences between test meals or
than cold rice in the high-SAA group (P = 0.0098), but similar SAA groups for peak time, but there was a SAA × test meal
to cold rice (P > 0.99) in the low-SAA group (Figure 2A, B; interaction (P = 0.034), with the negative difference (−6.8 min,
Supplemental Table 4). NS) between Dex and cold rice in the low-SAA group differing
In all 40 subjects, mean glucose increments after both rice test from the positive difference (+2.3 min, NS) in the high-SAA
meals were significantly less than those after both dextrose test group (Table 3, Supplemental Table 5).
meals at 15, 30, 45, and 60 min, and significantly greater than Glucose iAUC was positively related to age (P = 0.046),
both dextrose test meals at 120 min (Figure 2C). In addition, the and there was high between-subject variation in glucose iAUC
glucose increment after cold rice was significantly greater than (ICC = 0.811; Supplemental Table 6). Mean glucose iAUC
that after hot rice at 45 min (Figure 2C). increments after both rice test meals were less than those after
The overall mean glucose increments in the low-SAA group both dextrose test meals (P < 0.0001; Table 3). When comparing
did not differ significantly from those in the high-SAA group at only the 2 rice test meals, mean iAUC was similar between
any time (Figure 2D). hot rice and cold rice (P = 0.39), and between SAA groups
(P = 0.71); however, there was an SAA × test meal interaction
(P = 0.022), with hot rice eliciting a lower response than
Blood glucose peak rise, peak time, iAUC, and GI. cold rice in the low-SAA group (162 vs. 198 mmol × min/L;
Mean glucose peak rises after both rice test meals were less P = 0.053) but a higher response in the high-SAA group (197 vs.
than those after both dextrose test meals (P < 0.0001). When 181 mmol × min/L; P = 0.58; Table 3, Supplemental Table 3).
only the 2 rice test meals were compared, hot rice elicited a There was negligible between-subject variation in GI
lower peak rise than cold rice (P = 0.037), with no SAA × test (ICC = 0.162; Supplemental Table 6) and no correlation
meal interaction (P = 0.23; Table 3, Supplemental Table 5). between SAA and GI (Figure 3E and Supplemental Figure
 GI of rice by salivary amylase activity and cooling 1639

FIGURE 2 Main trial: postprandial blood glucose and breath H2 and CH4 responses. Values are means ± SEM for n = 20 with low SAA and n = 20
with high SAA after CR, Dex, Dex+L, and HR. Panels show results for glucose increments (A, B, C, D), breath H2 (E, F, G, H), and breath CH4 (I, J, K,
L) for test meals in low-SAA (→), high-SAA (→), all subjects (n = 40, •→◦), and overall means for low-SAA () vs. high-SAA (), respectively.
abcdef Significance of differences by Tukey’s test (P < 0.05) as follows: a Dex vs. Dex+L; b Dex vs. HR; c Dex vs. CR; d Dex+L vs. HR; e Dex+L vs. CR; f HR
vs. CR. Exact P values are given in Supplemental Tables 3 and 5. ∗ Significant SAA × test-meal interaction for ANOVA comparing hot and cold rice only:
P = 0.033 at 60 min and P = 0.027 at 120 min. CR, cold rice; Dex, dextrose; Dex+L, dextrose plus lactulose; HR, Hot rice; SAA, salivary amylase activity.

3). The mean GI for rice was similar between Hot and Cold did not differ significantly from each other at any point in time
(P = 0.51) and between SAA (P = 0.73) groups, but the (Figure 2E–G and Supplemental Table 7).
SAA × test meal interaction was significant (P = 0.043), with Mean breath H2 nadir over the first 3 h and mean breath H2
hot rice having a lower GI than cold rice in the low-SAA group sum after Dex, hot rice, or cold rice did not differ significantly
(69 vs. 81, P = 0.12) but a higher GI in the high-SAA group (76 from each other; Dex+L had a significantly higher H2 nadir
vs. 70, P = 0.63; Table 3, Supplemental Table 5). However, after than hot and cold rice and a significantly higher H2 sum than
excluding outliers, as is a requirement of the ISO (International after all 3 other test meals (Table 4, Supplemental Table 8).
Standards Organization) method (46), the SAA × test meal Between-subject variation was moderate for overall mean breath
interaction became nonsignificant (Table 3, Supplemental H2 concentration (ICC = 0.513) and carbohydrate digestibility
Table 5). (ICC = 0.565), with carbohydrate digestibility being higher in
non-Whites than Whites (P = 0.011; Supplemental Table 6).
Estimated mean carbohydrate digestibility after rice was ∼93%
Breath gases and carbohydrate digestibility.
and was similar between hot and cold rice (P = 0.82) and
There were significant main effects of time and test meal, and between the low- and high-SAA groups (P = 0.12) and with no
significant SAA × test meal and time × test meal interactions significant SAA × test meal interaction (Table 4, Supplemental
for breath H2 concentrations (Supplemental Table 2). In the low- Table 8).
SAA group, the high-SAA group, and both groups combined, There was a significant main effect of time and significant
breath H2 increments were higher after Dex+L compared with SAA × time, SAA × test meal, and time × test meal interactions
all the other 3 test-meals at 2, 3, 4, 5, and 6 h (all P < 0.0001), for breath CH4 concentrations (Supplemental Table 2). Overall
whereas breath H2 increments after Dex, hot rice, and cold rice mean breath CH4 differed between subjects (ICC = 0.887;
1640 Wolever et al.

FIGURE 3 Equivalence testing and correlation between GI and SAA. (A–D) Values are means (95% CIs) of differences in panel A (pilot): glucose iAUC
for cold compared with hot rice (means for plain and stir-fried; ) and SF compared with plain rice (means for hot and cold; ), n = 12 subjects; panel B
(main): GI of cold vs. hot rice for n = 40 () or n = 38 ( , outliers excluded); panel C (main): GI of rice (mean of cold and hot) in the low-SAA compared
with high-SAA groups for n = 40 () or n = 38 (, outliers excluded); and panel D (main): estimated carbohydrate digestibility (n = 39) for the low-SAA
compared with high-SAA groups () or for cold compared with hot (). Vertical dashed lines represent differences of: panel A, ±20% of the mean iAUC for
hot rice (black lines) and plain rice (gray lines); panel B, ±20% of the mean GI for high-SAA for n = 40 (black lines) and n = 38 (gray lines); panel C, ±20%
of the mean GI for hot rice for n = 40 (black lines) and n = 38 (gray lines); and panel D, ±7% of the mean for digestibility high-SAA (black lines) and hot rice
(gray lines). (E) Correlations between the GI of rice (mean of hot and cold) and SAA measured before the study (•, r = 0.09, P = 0.57, n = 40) and during
the study (◦, r = 0.01, P = 0.93, n = 40). Regression lines for SAA measured before (solid line) or during (dashed line) the study. CHO, carbohydrate; GI,
glycemic index; iAUC, incremental AUC; OE, outliers excluded; SAA, salivary amylase activity; SF, stir-fried.

Supplemental Table 6), but there was no significant difference to 3.7 ppm in 31 subjects (low-CH4 group) and 10.9 to 47.5 ppm
among test meals (Figure 2C, Table 4). in the remaining 9 subjects (high-CH4 group). The percentage of
Since we did not correct breath CH4 for room air CH4 , we subjects with high CH4 did not differ in the low-SAA compared
could not officially determine which participants were non– with high-SAA groups (30% vs. 15%, respectively; P = 0.26).
methane producers. However, the overall average breath CH4 The median (IQR) overall breath H2 concentration in the low-
concentration over the four 6-h collection periods ranged from 1.7 CH4 group, 15.9 (12.6, 23.3) ppm, was significantly greater than
 GI of rice by salivary amylase activity and cooling 1641
TABLE 4 Effects of SAA and temperature on breath H2 and CH4 and carbohydrate digestibility1

Dextrose Medium-grain polished rice

Endpoint and SAA group Alone + Lactulose Hot Cold Mean for group
Breath hydrogen nadir,2 ppm
Low 8.8 ± 3.5 11.5 ± 2.3 5.7 ± 1.6 5.1 ± 1.0 7.8 ± 1.6
High 7.7 ± 2.7 14.7 ± 2.9 5.2 ± 0.8 5.1 ± 1.0 8.6 ± 1.5
Mean 8.3 ± 2.2 13.1 ± 1.9A 5.5 ± 0.9 5.9 ± 1.0 P = 0.6753
Breath hydrogen sum,4 ppm
Low 51.4 ± 11.6 219.5 ± 26.1 36.1 ± 6.0 28.0 ± 4.6 83.7 ± 10.0
High 41.2 ± 9.4 196.8 ± 22.4 38.2 ± 5.2 54.7 ± 11.4 82.7 ± 8.3
Mean 46.3 ± 7.4 208.1 ± 17.1B 37.2 ± 3.9 41.3 ± 6.4 P = 0.2423
Breath methane nadir,2 ppm
Low 9.1 ± 3.0 7.9 ± 2.3 7.8 ± 3.3 9.9 ± 3.5 8.7 ± 2.8
High 2.9 ± 0.7 4.0 ± 1.0 4.6 ± 2.6 3.4 ± 1.0 3.7 ± 1.2
Mean 6.0 ± 1.6 5.9 ± 1.3 6.2 ± 2.1 6.7 ± 1.9 P = 0.5153
Breath methane sum,4 ppm
Low 65.4 ± 22.6 68.9 ± 22.6 44.6 ± 17.9 52.4 ± 16.7 57.8 ± 17.9
High 19.2 ± 5.0 32.8 ± 8.6 30.9 ± 16.3 25.1 ± 7.1 27.0 ± 8.8
Mean 42.3 ± 12.0 50.8 ± 12.3 37.8 ± 12.0 38.8 ± 9.2 P = 0.2843
Carbohydrate digestibility,5 %
Low — — 93.9 ± 1.4 95.7 ± 0.9 94.8 ± 1.0
High — — 92.5 ± 1.8 91.2 ± 1.4 91.9 ± 1.3
Mean — — 93.2 ± 1.1 93.5 ± 1.0 P = 0.1263
1 Values are means ± SEMs for n = 20 with low SAA and n = 20 with high SAA (except where indicated). The details of the ANOVA are shown in

Supplemental Table 8. Low, <50 U/mL; High, >105 U/mL; SAA, salivary amylase activity.
2 Lowest concentration within the first 3 h. A Mean for Dextrose + Lactulose differs from Hot rice (P = 0.0008) and Cold rice (P = 0.0020) by Tukey’s

test.
3 P value for SAA group × test-meal interaction by ANOVA.
4 Sum of concentrations from the nadir during the first 3 h to the concentration at 6 h. B Mean for Dextrose + Lactulose differs from all other test meals

(P < 0.0001) by Tukey’s test.

5 Means ± SEMs for n = 39 (values of 1 subject, 4% for hot rice and –8% for cold rice, excluded).

that in the high-CH4 group, 11.9 (10.6, 14.4) ppm (P = 0.044 by There was no significant association between any of the
the Mann-Whitney test). SNPs in AMY1, AMY2A, and AMY2B and GI (Supplemental
Table 11) and no association between the changes in GI and
changes in SAA per minor allele (Supplemental Figure 5A).
Effect of SNPs in AMY1, AMY2A, and AMY2B genes on SAA, There were significant associations between 3 of the SNPs and
GI, and carbohydrate digestibility. carbohydrate digestibility, with the mean ± SEM change in
The 12 SNPs in AMY1, AMY2A, and AMY2B genes we digestibility per minor allele being 1.7% ± 0.6% for rs10881197
measured are shown in Supplemental Table 9. The minor allele (P = 0.011), −1.8% ± 0.8% for rs1157739 (P = 0.041), and
frequencies in our study population did not differ significantly 1.6% ± 0.6% for rs6696797 (P = 0.014) (Supplemental Table
from those reported in the literature (20) for 11 of the 12 SNPs; 12). However, the SNPs associated with increased carbohydrate
the only difference was a lower frequency for rs1930212 in digestibility were, if anything, associated with decreased SAA
our population. Minor allele frequency for subjects in the low- activity (r = −0.166, P = 0.61 for the slopes and r = −0.588,
compared with high-SAA groups did not differ significantly for P = 0.059 for the correlation coefficients; Figure 4C and D,
any of the 12 SNPs, and genotype frequency only differed for respectively).
rs1566154 with n for GG:GA:AA being 1:9:10 compared with
4:2:14 in the low- and high-SAA groups, respectively (P = 0.03;
Supplemental Table 9). None of the 10 SNPs in the AMY1 gene Effect of SNPs in MGAM and SI genes on GI and
was significantly associated with a difference in measured SAA, carbohydrate digestibility.
but each minor allele in AMY2A and AMY2B was associated with There were no significant associations between any of the
a 104-U/mL increase in SAA (P = 0.009; Supplemental Table 9 MGAM SNPs or 3 SI SNPs and either GI (P = 0.26 to 1.00)
10). The relations between minor allele number and SAA for or carbohydrate digestibility (P = 0.14 to 0.97) (Supplemental
each of the 12 SNPs in AMY1, AMY2A, and AMY2B in this study Tables 13 and 14).
(Supplemental Table 10) correlated with the relations between
minor allele number and AMY1CN for the same SNPs reported in
the literature (20) for the change in SAA or AMY1CN per minor Equivalence testing
allele (P = 0.055) and the respective correlation coefficients The 95% CIs of the differences in iAUC between hot and
(P = 0.003) (Figure 4A and B, respectively). cold rice and stir-fried and plain rice in the pilot trial were not
1642 Wolever et al.

FIGURE 4 Interrelations among AMY1CN, SAA, GI, and carbohydrate digestibility for 12 SNPs in the AMY1, AMY2A, or AMY2B genes. (A) Points
are change in SAA per MA (y-axis) on change in AMY1CN per MA for n = 12 SNPs (x-axis; r = 0.57, P = 0.055). (B) Points are correlation coefficients
for associations between SAA and MAN (y-axis) on correlation coefficients between AMY1CN and MAN for 12 SNPs (x-axis; r = 0.77, P = 0.003). (C)
Points are change in GI per MA (◦) and %CD (multiplied by 4) per MA (• or •), respectively (y-axis), on change in SAA per MA, n = 12 SNPs (x-axis; r =
−0.04, P = 0.91, and r = −0.17, P = 0.61). (D) Points are correlation coefficients for associations between GI and MAN (◦) and between %CD (multiplied
by 4) and MAN (• or •), respectively (y-axis), on the correlation coefficients between SAA and MAN for n = 12 SNPs (x-axis; r = 0.19, P = 0.56, and r =
−0.56, P = 0.059). Panels A and B: data are from Supplemental Table 10 (y-axes) and reference (20) (x-axes). Panels C and D: data are from Supplemental
Tables 10–12; black circles are for SNPs where the changes in CD per MA and the correlation between CD and MAN are significant (P < 0.05). AMY1,
salivary amylase gene; AMY2A and AMY2B, pancreatic amylase genes; AMY1CN, AMY1 gene copy number; CD, carbohydrate digestibility estimated from
breath hydrogen; Corr, correlation coefficient; GI, glycemic index; MA, minor allele; MAN, minor allele number; SAA, salivary amylase activity; SNP, single
nucleotide polymorphism.

within ±20% of the mean iAUC for hot rice and stir-fried rice, Discussion
respectively (Figure 3A). Thus, the pilot trial failed to show The pilot trial showed that neither cooling nor stir-frying
that the glycemic responses elicited by cold rice and stir-fried cooked long-grain polished rice significantly reduced glucose
rice were equivalent to those after hot rice and stir-fried rice, iAUC. The main trial showed that the GI of cold medium-
respectively. However, with more statistical power, the results of grain polished rice was equivalent to that of freshly cooked hot
the main trial showed that the GI of cold rice was equivalent to rice and that the GI was equivalent in subjects with high-SAA
that of hot rice (Figure 3B) and the GI of rice in the low-SAA versus low-SAA. Although it is generally considered that cooling
group was equivalent to that in the high-SAA group (Figure 3C). cooked (gelatinized) starch creates slowly digested and resistant
The main trial results also showed that the digestibility of rice in starch (retrogradation) (11), studies show that cooling cooked
the low-SAA group was equivalent to that in the high-SAA group starchy foods often (33–39), but not always (33, 37–39), elicits
and that the digestibility of cold rice was equivalent to that of hot a significant reduction in glycemic response. Furthermore, the
rice (Figure 3D). literature on the effect of differences in AMY1CN and SAA on
The mean GIs of rice in participants with different genotypes glycemic response or GI is inconsistent (23–25). Thus, the current
of SNPs in AMY1, AMY2A, and AMY2B were equivalent for results, while perhaps unexpected, are not unique.
rs10881197, rs1330403, and rs6696796, but not for any of the In the pilot trial, although mean glucose iAUC was similar
other SNPs tested (Supplemental Figure 4A) and equivalent for among the 4 rice test meals, stir-fried rice elicited a lower glucose
only 1 (rs4132774) of the MGAM or SI SNPs (Supplemental increment at 30 min and a higher increment at 120 min compared
Figure 5A). However, estimated starch digestibility was equiv- with plain rice (Figure 1D). This could be due to the formation of
alent for all of the AMY1, AMY2A, and AMY2B SNPs, except for resistant starch when cooled rice is reheated (34) or to starch/fatty
rs1930212 (Supplemental Figure 4B) and for all of the MGAM acid complexes, which occur when starch is heated with fatty
and SI SNPs (Supplemental Figure 5B). acids. However, only small amounts of such complexes form
 GI of rice by salivary amylase activity and cooling 1643
when starch is heated with triglycerides (40, 41). It is known on SAA, the results can be compared meaningfully because
that adding fat to carbohydrate reduces the glycemic response Atkinson et al. attributed the difference in GI to differences in
by reducing gastric emptying (47) and increasing postprandial SAA. They showed that AMY1CN was related to SAA, while we
insulin relative to glucose (48). Stir-frying involved adding 5 mL showed that the associations between 12 amylase SNPs and SAA
(4.6 g) sesame oil to the 25 g avCHO portion of rice (0.184 g were significantly related to the reported associations between
fat/g avCHO). Based on 15 studies we could find in which the same SNPs and AMY1CN. Atkinson et al. included only
various doses and types of fat were added to bread, rice, potato, Europeans in their low- and high-AMY1CN groups, while we
cornstarch, or a mixed meal consumed by subjects without included White (n = 25) and non-White (n = 15) participants;
diabetes (Supplemental Table 15), regression analysis suggested however, their mean ± SD GIs were similar (74.1 ± 15.4 vs.
that mean (95% CI) glucose iAUC is reduced by 54% (35%, 74.5 ± 22.1, respectively). Only 3 of 201 subjects studied by
67%) for each gram of fat/gram avCHO (n = 27, P < 0.001; Atkinson et al. (23) had SAA >100 U/mL, whereas all 20 of our
Supplemental Figure 6). The observed relative response of stir- high-SAA subjects had SAA >105 U/mL. The mean ± SD SAA
fried rice in this trial, 70.5/77.1 = 0.91 (Table 2), is within the in our low-SAA and high-SAA groups was ∼3 times that of the
95% CI of that expected from its fat content (0.184 g/g avCHO), respective groups reported by Atkinson et al., a difference that
0.82 to 0.92 (mean, 0.87; Supplemental Figure 6). may reflect different methods of analysis.
In the pilot trial, cooling cooked rice had no significant effect The breath H2 method of assessing carbohydrate digestibility
on glycemic response. Post hoc power analysis showed the study relies on the following facts: the only source of H2 in humans
had 88% power to detect a 20% reduction in glucose iAUC, is the colonic fermentation of carbohydrate; the H2 produced
suggesting that any effect, if present, is small. GI was not diffuses into blood, equilibrates with alveolar air, and is expired
calculated because subjects did not repeat the dextrose test meal. in breath; and the amount of H2 in breath is proportional to the
However, mean glucose iAUC after plain hot rice was ∼50% of amount produced in the colon, which, in turn, is proportional
that after dextrose, suggesting it had a low-GI (<55). This is not to the amount of carbohydrate fermented (56). Evidence that
unusual since, of the 31 GI values for polished rice listed in the breath H2 is a valid estimate of starch malabsorption includes
2008 International GI Tables (49), n = 15 (48%) had a low GI, studies showing that fermentable nonstarch polysaccharides do
n = 8 (26%) had a medium GI, and only n = 11 (35%) had a high not increase breath H2 (57) and that starch digestibility assessed
GI. The GI of rice is inversely related to its amylose content (50, by breath H2 was similar to that measured in ileal effluent (58).
51), which, in turn, is inversely related to the grain length (52). We used the breath H2 method to show that cooling cooked
We wondered if cooling might only reduce glycemic response in potatoes reduced in vivo carbohydrate digestibility from 97% to
rice with a high GI. Thus, in the main trial, we used medium-grain 95% (P = 0.02), an effect associated with a significant reduction
polished rice (expected to have a high GI) and included n = 40 in mean GI from 78 to 49 (39). This suggests that the method
subjects to have sufficient statistical power to detect a difference is sensitive enough to detect small differences in carbohydrate
of 15%. digestibility. Here we only measured breath H2 for 6 h instead of
The main trial confirmed that medium-grain Calrose rice the 8–10 h previously used; however, in the potato study (39),
had a high GI. Nevertheless, the GI and digestibility of cold a difference in breath H2 concentration between hot and cold
compared with hot-rice were statistically equivalent. These potatoes was readily apparent by 6 h, whereas here, after 6 h,
findings disagree with some studies showing that cooling white breath H2 after cold rice was virtually identical to that after hot
rice reduced its glycemic impact (33, 34). However, the effect rice. Furthermore, the estimated mean (95% CI) percentage of
of thermal or other types of processing varies in different rice rice starch entering the colon, 6% (4%, 8%), is consistent with
varieties; for example, cooling reduced the glycemic impact of the amounts of resistant starch measured in white rice (59, 60)
white but not red rice (33), and parboiling reduced the glycemic and the amounts of starch recovered in ileal effluent from rice
impact of long-grain white rice (53) and Italian Fine Ribe rice (4–5%) and other low-fiber starchy foods such as white bread,
(54), but not BR-16 rice (55) or Pelde rice (55). Similarly, the pasta, and breakfast cereals (2–5%) (58, 61).
effect of cooling on the glycemic response and digestibility of We previously found that the minor alleles of MGAM and
cooked potatoes varied in different genetic selections (39). SI SNPs were associated with significant differences in nutrient
A large study by Atkinson et al. (23) showed that, compared intake in the Toronto Nutrigenomics and Health Study (Oseni
with n = 20 subjects with high AMY1CN (mean ± SD, L, El-Sohemy A, 2012; unpublished data). These effects could
10.4 ± 1.6) and high SAA (47 ± 25 U/mL), n = 20 subjects be due to differences in starch digestion, which might cause GI
with low AMY1CN (3.1 ± 0.9) and low SAA (11 ± 8 U/mL) symptoms or alter the colonic microbiome in such a way as to
had significantly lower GIs for white bread (mean ± SEM GI, influence food intake. However, we were unable to detect any
70 ± 3 vs. 81 ± 3) and pasta (44 ± 2 vs. 50 ± 2) (23). Our significant difference in GI or starch digestibility associated with
findings that the GI of rice was equivalent in subjects with low- these, or any, of the other SNPs we assessed.
SAA (72 ± 3) and those with high-SAA (75 ± 3) are not It is concluded that these results do not support the hypotheses
incompatible with those of Atkinson et al., in that our ±20% that low-SAA and cooling affect the GI or in vivo carbohydrate
equivalence limits do not rule out the GI differences of 14% for digestibility of cooked polished rice. However, small effects of
white bread and 12% for pasta found by Atkinson et al. However, stir-frying and of 11 of the 12 SNPs in MGAM and SI we
we found no correlation between SAA and GI (Figure 3E, assessed cannot be ruled out due to a failure to demonstrate
Supplemental Figure 3), a result not consistent with Atkinson equivalence.
et al. (23) who found linear relations between AMY1CN and
GI for 6 different starchy foods. Although Atkinson et al. (23) The authors’ responsibilities were as follows—TMSW and AE-S:
selected subjects based on AMY1CN, whereas we selected based conceived of the studies; TMSW: obtained Institutional Review Board
1644 Wolever et al.

approval, did the initial statistical analysis, wrote the first draft of the 18. Lan-Pidhainy X, Wolever TMS. Are glycemic and insulinemic index
manuscript, and takes overall responsibility for its content; ALJ and AE: were values of foods similar in control, hyperinsulinemic and type 2 diabetic
responsible for executing the studies; AE-S: provided advice about the SNPs patients? Eur J Clin Nutr 2011;65(6):727–34.
to analyze; AZ: provided important assistance with the statistical analysis; 19. Wolever TMS, Jenkins AL, Vuksan V, Campbell J. The GI values of
foods containing fructose are affected by metabolic differences between
and all authors: reviewed, read, and approved the final manuscript. TMSW,
subjects. Eur J Clin Nutr 2009;63(9):1106–14.
his wife, and ALJ are owners and employees of INQUIS. AE is a full-time 20. Usher CL, Handsaker RE, Esko T, Tuke MA, Weedon MN, Hastie AR,
employee of INQUIS. AZ is a part-time employee of INQUIS. AE-S is the Cao H, Moon JE, Kashin S, Fuchsberger C, et al. Structural forms of
Founder and holds shares in Nutrigenomix, Inc. human amylase locus and their relationships to SNPs, haplotypes and
obesity. Nat Genet 2015;47(8):921–5.
21. Perry GH, Dominy NJ, Claw KG, Lee AS, Fiegler H, Redon R,
Data Availability Werner J, Villanea FA, Mountain JL, Misra R, et al. Diet and the
evolution of human amylase gene copy number variation. Nat Genet
Data described in the manuscript, code book, and analytic code 2007;39(10):1256–60.
will be made available upon request. 22. Butterworth PJ, Warren FJ, Ellis PR. Human α-amylase and starch
digestion: an interesting marriage. Starch Strke 2011;63(7):395–405.
23. Atkinson FS, Hancock D, Petocz P, Brand-Miller JC. The physiologic
and phenotypic significance of variation in human amylase gene copy
References number. Am J Clin Nutr 2018;108(4):737–48.
1. Jenkins DJA, Wolever TMS, Taylor RH, Barker HM, Fielden H, 24. Tan VM, Ooi DS, Kapur J, Wu T, Chan YH, Henry CJ, Lee YS. The
Baldwin JM, Bowling AC, Newman HC, Jenkins AL, Goff DV. role of digestive factors in determining the glycemic response in a
Glycemic index of foods: a physiological basis for carbohydrate multiethnic Asian population. Eur J Nutr 2016;55(4):1573–81.
exchange. Am J Clin Nutr 1981;34(3):362–6. 25. Mandel AL, Breslin PAS. High endogenous salivary amylase activity
2. Henry CJK, Lightowler HJ, Strik CM, Storey M. Glycaemic index is associated with improved glycemic homeostatsis following starch
values for commercially available potatoes in Great Britain. Br J Nutr ingestion in adults. J Nutr 2012;142(5):853–8.
2005;94(6):917–21. 26. Nichols BL Jr, Auricchio S. 50 Years of progress since congenital
3. Brand-Miller JC, Pang E, Bramall L. Rice: a high or low glycemic index sucrose-isomaltase deficiency recognition. J Pediatr Gastroenterol Nutr
food? Am J Clin Nutr 1992;56(6):1034–6. 2012;55(Suppl 2):S2–7.
4. Panlasigui LN, Thompson LU, Juliano BO, Perez CM, Yiu SH, 27. Lin AHM, Hamaker BR, Nichols BL Jr. Direct starch digestion by
Greenberg GR. Rice varieties with similar amylose content differ in sucrase-isomaltase and maltase-glucoamylase. J Pediatr Gastroenterol
starch digestibility and glycemic response in humans. Am J Clin Nutr Nutr 2012;55(Suppl 2):S43–5.
1991;54(5):871–7. 28. Villegas R, Liu S, Gao YT, Yang G, Li H, Zheng W, Shu XO.
5. Wolever TMS. The glycaemic index: a physiological classification of Prospective study of dietary carbohydrates, glycemic index, glycemic
dietary carbohydrate. Wallingford (UK): CABI Publishing; 2006. p. load and incidence of type 2 diabetes mellitus in middle-aged Chinese
120–3. women. Arch Intern Med 2007;167(21):2310–6.
6. Wolever TMS, Vorster HH, Björck I, Brand-Miller J, Brighenti F, Mann 29. Zuniga YL, Rebello SA, Oi PL, Sheng H, Lee J, Tai ES, Van Dam RM.
JI, Ramdath DD, Granfeldt Y, Holt S, Perry TL, et al. Determination of Rice and noodle consumption is associated with insulin resistance and
the glycaemic index of foods: interlaboratory study. Eur J Clin Nutr hyperglycemia in an Asian population: the Hong Kong dietary survey.
2003;57(3):475–82. Public Health Nutr 2011;14:1133–41.
7. Wolever TMS, Brand-Miller JC, Abernethy J, Astrup A, Atkinson F, 30. Nanri A, Mizoue T, Noda M, Takahashi Y, Kato M, Inoue M, Tsugane
Axelsen M, Björck I, Brighenti F, Brown R, Brynes A, et al. Measuring S. Rice intake and type 2 diabetes in Japanese men and women: the
the glycemic index of foods: interlaboratory study. Am J Clin Nutr Japan public health center-based prospective study. Am J Clin Nutr
2008;87(1):247S–57S. 2010;92(6):1468–77.
8. Liu Q, Tarn R, Lynch D, Skjodt NM. Physicochemical properties of 31. Soriguer F, Colomo N, Olivera G, Garica-Fuentes E, Esteva I, Ruiz de
dry matter and starch from potatoes grown in Canada. Food Chem Adana MS, Morcillo S, Porras N, Valdés S, Rojo-Martínez G. White rice
2007;105(3):897–907. consumption and risk of type 2 diabetes. Clin Nutr 2013;32(3):481–4.
9. Liu Q, Weber E, Currie V, Yada R. Physicochemical properties of 32. Dong F, Howard A-G, Herring AH, Popkin BM, Gordon-Larsen P.
starches during potato growth. Carbohydr Polym 2003;51(2):213–21. White rice intake varies in its association with metabolic markers of
10. Englyst HN, Cummings JH. Digestion of polysaccharides of potato in diabetes and dyslipidemia across region among Chinese adults. Ann
the small intestine of man. Am J Clin Nutr 1987;45(2):423–31. Nutr Metab 2015;66(4):209–18.
11. Reed MO, Ai Y, Leutcher JL, Jane J. Effects of cooking methods 33. Ananda D, Zuhairini Y, Sutadipura N. Resistant starch in cooled white
and starch structures on starch hydrolysis rates of rice. J Food Sci rice reduce glyaemic index. Obesity Res Clin Pract 2013;7:38.
2013;78(7):H1076–81. 34. Sonia S, Witjaksono F, Ridwan R. Effect of cooling cooked rice on
12. Lu LW, Monro J, Lu J, Rush E. Effect of cold storage, reheating and resistant starch content and glycemic response. Asia Pacific J Clin Nutr
particle size on in-vitro glucose release and starch digestibility among 2015;24:620–5.
five rice products in Auckland, New Zealand. Rice Res Open Access 35. Burton P, Lightowler HJ. The impact of freezing and toasting on the
2016;4(3):3. glycaemic response of white bread. Eur J Clin Nutr 2008;62(5):594–9.
13. Wolever TMS, Nuttall FQ, Lee R, Wong GS, Josse RG, Csima A, 36. Fernandes G, Velangi A, Wolever TMS. Glycemic index of
Jenkins DJA. Prediction of the relative blood glucose response of potatoes commonly consumed in North America. J Am Diet Assoc
mixed meals using the white bread glycemic index. Diabetes Care 2005;105(4):557–62.
1985;8(5):418–28. 37. Robertson TM, Brown JE, Fielding BA, Robertson MD. The cumulative
14. Zeevi D, Korem T, Zmora N, Israeli D, Rothschild D, Weinberger effects of chilling and reheating a carbohydrate-based pasta meal on
A, Ben-Yacov O, Lador D, Avnit-Sagi T, Lotan-Pompan M, et al. the postprandial glycaemic response: a pilot study. Eur J Clin Nutr
Personalized nutrition by prediction of glycemic responses. Cell 2021;75(3):570–2.
2015;163(5):1079–94. 38. Hodges C, Archer F, Chowdhury M, Evans BL, Ghelani DJ, Mortoglou
15. Matthan NR, Ausman LM, Meng H, Tighiouart H, Lichtenstein AH. M, Guppy FM. Method of food preparation influences blood glucose
Estimating the reliability of glycemic index values and potential response to a high carbohydrate meal: a randomised cross-over trial.
sources of methodological and biological variability. Am J Clin Nutr Foods 2020;9(1):23.
2016;104(4):1004–13. 39. Kinnear T, Wolever TMS, Murphy AM, Sullivan A, Liu Q, Bizimungu
16. Wolever TMS, Jenkins DJA, Jenkins AL, Josse RG. The glycemic B. Effect of preparation method on the glycaemic index of novel potato
index: methodology and clinical implications. Am J Clin Nutr clones. Food Funct 2011;2(8):438–44.
1991;54(5):846–54. 40. Nebesny E, Rosicka J, Tkaczyk M. Effect of enzymatic hydrolysis
17. Wolever TMS. Personalized nutrition by prediction of glycemic of wheat starch on amylose-lipid complexes stability. Starch Strke
responses: fact or fantasy? Eur J Clin Nutr 2016;70(4):411–3. 2002;54(12):603–8.
 GI of rice by salivary amylase activity and cooling 1645
41. Chao C, Yu J, Wang S, Copeland L, Wang S. Mechanisms underlying 51. Boers HM, ten Hoorn JS, Mela DJ. A systematic review of the
the formation of complexes between maize starch and lipids. J Agric influence of rice characteristics and processing methods on postprandial
Food Chem 2018;66(1):272–8. glycaemic and insulinaemic responses. Br J Nutr 2015;114(7):
42. Wolever TMS. Effect of blood sampling schedule and method 1035–45.
calculating the area under the curve on validity and precision of 52. Goodman DE, Rao RM. Amylose content and puffed volume of
glycaemic index values. Br J Nutr 2004;91(2):295–300. gelatinized rice. J Food Sci 1984;49(4):1204–5.
43. Wolever TMS, Cohen Z, Thompson LU, Thorne MJ, Jenkins MJA, 53. Wolever TMS, Jenkins DJA, Kalmusky J, Jenkins AL, Giordano C,
Prokipchuk EJ, Jenkins DJA. Ileal loss of available carbohydrate in man: Giudici S, Josse RG, Wong GS. Comparison of regular and parboiled
comparison of a breath hydrogen method with direct measurement using rices: explanation of discrepancies between reported glycemic
a human ileostomy model. Am J Gastroenterol 1986;81:115–22. responses to rice. Nutr Res 1986;6(4):349–57.
44. Health Canada. Draft guidance document on food health claims 54. Casiraghi MC, Brighenti F, Pellegrini N, Leopardi E, Testolin G. Effect
related to the reduction in post-prandial glycaemic response. of processing on rice starch digestibility evaluated by in vivo and in vitro
[Internet]. June 2013. [Accessed 2020 Mar 26]. Available from: methods. J Cereal Sci 1993;17(2):147–56.
https://www.canada.ca/en/health-canada/services/food-nutrition/publi 55. Larsen HN, Christensen C, Rasmussen OW, Tetens IH, Choudhury
c-involvement-partnerships/technical-consultation-draft-guidance-d NH, Thilsted SH, Hermansen K. Influence of parboiling and
ocument-food-health-claims-related-post-prandial-glycaemia.html. physico-chemical characteristics of rice on the glycaemic index in
45. Erchinger F, Engjom T, Jurmy P, Tjora E, Gilja OH, Demcevski G. non-insulin-dependent diabetic subjects. Eur J Clin Nutr 1996;50:
Fecal fat analysis in chronic pancreatitis importance of fat ingestion 22–7.
before stool collection. PLoS One 2017;12(1):e0169993. 56. Bond JH, Levitt MD. Use of pulmonary hydrogen (H2 ) measurements
46. International Standards Organization. ISO 26642:2010. Food to quantitate carbohydrate absorption. Study of partially gastrectomized
products—determination of the glycaemic index (GI) and patients. J Clin Invest 1972;19:771–4.
recommendation for food classification. [Internet]. Published 57. Wolever TMS, ter Wal P, Spadafora P, Robb P. Guar, but not psyllium,
23 September 2010. [Accessed 2021 Jan 15]. Available from: increases beath methane and serum acetate concentrations in human
http://www.iso.org/iso/catalogue_detail.htm?csnumber=43633. subjects. Am J Clin Nutr 1992;55(3):719–22.
47. Cunningham KM, Read NW. The effect of incorporating fat into 58. Liljeberg Elmstål H. Resistant starch content in a selection of
different components of a meal on gastric emptying and postprandial starchy foods on the Swedish market. Eur J Clin Nutr 2002;56(6):
blood glucose and insulin responses. Br J Nutr 1989;61(2):285–90. 500–5.
48. Collier GR, Greenberg GR, Wolever TMS, Jenkins DJA. The 59. Patterson MA, Malya M, Stewart ML. Resistant starch content in foods
acute effect of fat on insulin secretion. J Clin Endocrinol Metab commonly consumed in the United States: a narrative review. J Acad
1988;66(2):323–6. Nutr Diet 2020;120:2, 230–44.
49. Atkinson FS, Foster-Powell K, Brand-Miller JC. International tables 60. Jenkins DJA, Cuff D, Wolever TMS, Knowland D, Thompson L,
of glycemic index and glycemic load values: 2008. Diabetes Care Cohen Z, Prokipchuk. Digestibility of carbohydrate foods in an
2008;31(12):2281–3. ileostomate: relationship to dietary fiber, in vitro digestibility, and
50. Goddard MS, Young G, Marcus R. The effect of amylose content on glycemic response. Am J Gastroenterol 1987;82:709–17.
insulin and glucose responses to ingested rice. Am J Clin Nutr 1984;39 61. Englyst HN, MacFarlane GT, Cummings JH. New concepts in starch
(3):388–92. digestion in man. Proc Nutr Soc 1988;47(1):64A.