Content uploaded by Wei Bao
Author content
All content in this area was uploaded by Wei Bao on Apr 12, 2014
Content may be subject to copyright.
Prepregnancy low-carbohydrate dietary pattern and risk of gestational
diabetes mellitus: a prospective cohort study
1–3
Wei Bao, Katherine Bowers, Deirdre K Tobias, Sjurdur F Olsen, Jorge Chavarro, Allan Vaag, Michele Kiely,
and Cuilin Zhang
ABSTRACT
Background: Low-carbohydrate diets (LCDs) have been vastly pop-
ular for weight loss. The association between a low-carbohydrate di-
etary pattern and risk of gestational diabetes mellitus (GDM) remains
unknown.
Objective: We aimed to prospectively examine the association of 3
prepregnancy low-carbohydrate dietary patterns with risk of GDM.
Design: We included 21,411 singleton pregnancies in the Nurses’
Health Study II. Prepregnancy LCD scores were calculated from
validated food-frequency questionnaires, including an overall LCD
score on the basis of intakes of carbohydrate, total protein, and total
fat; an animal LCD score on the basis of intakes of carbohydrate,
animal protein, and animal fat; and a vegetable LCD score on the
basis of intakes of carbohydrate, vegetable protein, and vegetable
fat. A higher score reflected a higher intake of fat and protein and
a lower intake of carbohydrate, and it indicated closer adherence to
a low-carbohydrate dietary pattern. RRs and 95% CIs were estimated
by using generalized estimating equations with log-binomial models.
Results: We documented 867 incident GDM pregnancies during 10 y
follow-up. Multivariable-adjusted RRs (95% CIs) of GDM for com-
parisons of highest with lowest quartiles were 1.27 (1.06, 1.51) for
the overall LCD score (P-trend = 0.03), 1.36 (1.13, 1.64) for the
animal LCD score (P-trend = 0.003), and 0.84 (0.69, 1.03) for the
vegetable LCD score (P-trend = 0.08). Associations between LCD
scores and GDM risk were not significantly modified by age, parity,
family history of diabetes, physical activity, or overweight status.
Conclusions: A prepregnancy low-carbohydrate dietary pattern with
high protein and fat from animal-food sources is positively associated
with GDM risk, whereas a prepregnancy low-carbohydrate dietary
pattern with high protein and fat from vegetable food sources is not
associated with the risk. Women of reproductive age who follow
a low-carbohydrate dietary pattern may consider consuming vegeta-
ble rather than animal sources of protein and fat to minimize their
risk of GDM. Am J Clin Nutr doi: 10.3945/ajcn.113.082966.
INTRODUCTION
Carbohydrate-restricted diets or low-carbohydrate diets
(LCDs)
5
were first introduced w150 y ago (1). These diets re-
main very popular for weight loss because they result in a rapid
reduction in body weight without having to count calories or
compromise the consumption of many palatable foods (2).
However, debates and concerns continue with regard to the long-
term safety and efficacy of these diets (2, 3), and it has been
shown that the weight loss by LCDs may dissipate after 1 y (4, 5).
Moreover, associations between adherence to low-carbohydrate
dietary patterns and cardiometabolic outcomes, such as type 2
diabetes (T2D) (6, 7) and cardiovascular disease (8, 9), remain
controversial.
Gestational diabetes mellitus (GDM), which is a common
pregnancy complication defined as glucose intolerance with onset
or first recognition during pregnancy (10), is a growing health
concern (11). GDM is not only associated with short-term ad-
verse perinatal outcomes (12) but also related to elevated long-
term metabolic risk in both mothersand their offspring (10, 11, 13).
For instance, 35–60% of women who have had GDM will develop
T2D in the next 10–20 y (14). Thus, it is crucial to identify
modifiable risk factors that may contribute to the prevention of
GDM. Low-carbohydrate dietary patterns represent combinations
of a lower content of carbohydrate and higher contents of fat and
protein from the diet. Increased intakes of fat and protein are
naturally needed to compensate energy requirements. In previous
studies, dietary intakes of animal fat and animal protein were
1
From the Epidemiology Branch, Division of Intramural Population
Health Research, Eunice Kennedy Shriver National Institute of Child Health
and Human Development, NIH, Rockville, MD (WB, MK, and CZ); the
Division of Biostatistics and Epidemiology, Department of Pediatrics, Cin-
cinnati Children’s Hospital Medical Center, Cincinnati, OH (KB); the De-
partments of Nutrition and Epidemiology, Harvard School of Public Health,
Boston, MA (DKT and JC); the Centre for Fetal Programming, Department
of Epidemiology Research, Statens Serum Institut, Copenhagen, Denmark
(SFO); the Channing Division of Network Medicine, Department of Medi-
cine, Brigham and Women’s Hospital and Harvard Medical School, Boston,
MA (JC); and the Department of Endocrinology, Rigshospitalet, Copenhagen,
Denmark (AV).
2
Presented in abstract form at the 46th Society for Epidemiologic Re-
search Annual Meeting, in Boston, MA, 18–21 June 2013.
3
Supported by the Intramural Research Program of the Eunice Kennedy
Shriver National Institute of Child Health and Human Development, NIH
(contract HHSN275201000020C). The Nurses’ Health Study II was funded
by the NIH (research grants DK58845, CA50385, P30 DK46200, and UM1
CA176726).
4
Address correspondence to C Zhang, Epidemiology Branch, Division of
Epidemiology, Statistics and Prevention Research, Eunice Kennedy Shriver
National Institute of Child Health and Human Development, NIH, 6100
Executive Boulevard, Rockville, MD 20852. E-mail: zhangcu@mail.nih.gov.
5
Abbreviations used: FFQ, food-frequency questionnaire; GDM, gesta-
tional diabetes mellitus; LCD, low-carbohydrate diet; NHS, Nurses’ Health
Study; T2D, type 2 diabetes.
Received December 31, 2013. Accepted for publication March 19, 2014.
doi: 10.3945/ajcn.113.082966.
Am J Clin Nutr doi: 10.3945/ajcn.113.082966. Printed in USA. Ó2014 American Society for Nutrition 1of7
AJCN. First published ahead of print April 9, 2014 as doi: 10.3945/ajcn.113.082966.
Copyright (C) 2014 by the American Society for Nutrition
positively associated with GDM risk, whereas intake of vegetable
protein was inversely associated with risk (15, 16). Theoretically,
long-term adherence to low-carbohydrate dietary patterns, partic-
ularly those that are mainly based on animal foods, may have
detrimental effects on GDM risk because they result in an increase
in animal fat intakes and a decrease in the consumption of whole
grains, dietary fiber, fruits, and vegetables. However, the effect of
low-carbohydrate dietary patterns on the development of GDM
remains unknown. With the use of data from a large cohort study,
we aimed to prospectively examine the association between 3
prepregnancy low-carbohydrate dietary patterns and risk of GDM.
SUBJECTS METHODS
Study population
The Nurses’ Health Study II (NHS II) is an ongoing, pro-
spective cohort study of 116,671 female nurses aged 25–44 y at
study inception in 1989 (17). Participants receive biennial
questionnaires regarding disease outcomes and lifestyle behav-
iors, such as smoking status and medication use. The follow-up
for each questionnaire cycle is.90%. This study was approved
by the Partners Human Research Committee (Boston, MA) with
participant consent implied by the return of questionnaires.
We included NHS II participants in this analysis if they
reported at least one singleton pregnancy that lasted .6mo
between 1991 and 2001. The 1991 questionnaire was the first
time dietary information was administered. Thus, we set this
year as the baseline for this analysis, and we only included
pregnancies after the return of the 1991 questionnaire. The 2001
questionnaire was the last time GDM incidence was ascertained
because the majority of NHS II participants had passed re-
productive age by then; thus, the follow-up was through the
return of the 2001 questionnaire. Pregnancies became eligible if
there was no GDM reported in a previous pregnancy or a pre-
vious diagnosis of T2D, cardiovascular disease, or cancer. We
excluded pregnancies if the participant did not return a pre-
pregnancy food-frequency questionnaire (FFQ), left .70 FFQ
items blank, or reported unrealistic total energy intake (,600
or .3500 kcal/d). Women with GDM in a previous pregnancy
were not included because they may have changed their diets and
lifestyles during the next pregnancy to prevent recurrent GDM.
Exposure assessment
Beginning in 1991 and every 4 y thereafter, we asked par-
ticipants to report their food intakes by using a semiquantitative
FFQ. We computed intake of individual nutrients including
protein by multiplying the frequency of consumption of each
food by the nutrient content of the specified portion on the basis
of food-composition data from USDA (18). The reproducibility
and validity of the FFQ has been extensively documented (19–
21). In a previous validation study that compared energy-adjusted
macronutrient intake assessed by using a FFQ with four 1-wk diet
records, Pearson’s correlation coefficients were 0.61 for total
carbohydrate, 0.52 for total protein, and 0.54 for total fat (20).
Missing exposure data were carried forward from the most recent
FFQ for which data were captured. Overall, missing exposure
existed in w6% of pregnancies.
To represent the adherence to various low-carbohydrate dietary
patterns, we calculated 3 LCD scores (ie, overall LCD, animal
LCD, and vegetable LCD scores) for each participant as pre-
viously described (8). Briefly, we divided study participants into
11 strata according to each of fat, protein, and carbohydrate
intakes expressed as percentages of energy. We assigned the
participants 0–10 points for increasing intake of total fat, 0–10
points for increasing intake of total protein, and, inversely, 10–0
points for increasing intake of carbohydrate. We summed points
for the 3 macronutrients to create the overall LCD score, which
ranged from 0 to 30. Similarly, we also created an animal LCD
score on the basis of the percentage of energy of carbohydrate,
animal protein, and animal fat and a vegetable LCD score on
the basis of the percentage of energy of carbohydrate, vegeta-
ble protein, and vegetable fat. A higher score reflected higher
intake of fat and protein and lower intake of carbohydrate, and
it indicated closer adherence to a low-carbohydrate dietary
pattern. LCD scores have been used in previous studies in as-
sociation with risk of T2D (6, 7), cardiovascular disease (8),
and mortality (22).
Covariate assessment
Participants reported their heights and weights in 1989 and
updated their weights on each biennial questionnaire. The self-
reported weight was highly correlated with the measured weight
(r= 0.97) in a previous validation study (23). BMI (in kg/m
2
)
was computed as weight divided by the square of height. Total
physical activity was ascertained by the frequency that partici-
pants engaged in common recreational activities from which
metabolic equivalent task hours per week were derived. Ques-
tionnaire-based estimates correlated well with detailed activity
diaries in a previous validation study (r= 0.56) (24).
Outcome ascertainment
Incident GDM was ascertained by a self-report on each bi-
ennial questionnaire through 2001. In the case of more than one
pregnancy that lasted .6 mo and reported within a 2-y ques-
tionnaire period, GDM status was attributed to the first preg-
nancy. In a previous validation study in a subgroup of the NHS
II cohort, 94% of GDM self-reports were confirmed by medical
records (17). In a random sample of parous women without
GDM, 83% of subjects reported a glucose screening test during
pregnancy, and 100% of subjects reported frequent prenatal
urine screenings, which suggested a high level of GDM sur-
veillance in this cohort (17).
Statistical analysis
We divided women into quartiles according to their pre-
pregnancy LCD scores. To represent the long-term habitual diet
and reduce measurement error (25), we calculated a cumulative
average LCD score on the basis of the information from 1991,
1995, and 1999 FFQs. Generalized estimating equations, which
allowed us to account for correlations in repeated observations
(pregnancies) contributed by a single participant (26), with log-
binomials models (27) were used to estimate RRs and 95% CIs.
In a few instances, models did not converge, and log-Poisson
models (28), which provide consistent but not fully efficient risk
estimates, were used.
2of7 BAO ET AL
Covariates in multivariable models included age (mo), parity
(0, 1, 2, or $3), race-ethnicity (white, African American, His-
panic, Asian, and others), family history of diabetes (yes or no),
cigarette smoking (never, past, or current), alcohol intake (0.0,
0.1–5.0, 5.1–10.0, or .10.0 g/d), physical activity (quartiles),
total energy intake (quartiles), and BMI (9 categories as
follows: ,21.0, 21.0–22.9, 23.0–24.9, 25.0–26.9, 27.0–28.9,
29.0–30.9, 31.0–32.9, 33.0–34.9, and $35.0). We updated all these
covariates, except race-ethnicity and family history of diabetes that
were reported in 1989. We conducted tests of linear trend across
quartiles of the LCD score by assigning the median value for each
quartile and fitting this as a continuous variable in models.
To evaluate a potential effect modification, we performed
stratified analyses according to age (,35 compared with $35),
parity (nulliparous compared with parous), family history of
diabetes (yes compared with no), physical activity (higher than
median compared with lower than median), and overweight
(BMI ,25 compared with $25). We also conducted interaction
tests via multiplicative interaction terms in multivariable models.
To explore potential dietary contributors for the association, we
additionally adjusted for each nutrient component of LCD scores
(eg, animal fat, animal protein, vegetable fat, and vegetable pro-
tein), other nutrients (eg, saturated fat, dietary cholesterol, heme
iron, dietary fiber, and glycemic load), and foods or food groups
(eg, red meat, poultry, fish, eggs, dairy food, fruits, vegetables,
whole grains, nuts, and legumes), as previously described (7). To
minimize the potential effects of changes in diet during pregnancy,
we also conducted a sensitivity analysis by excluding current
pregnancies at the time of each FFQ. To further address the pos-
sibility of residual confounding, we additionally adjusted for
a propensity score that reflected associations of LCD scores with
the other variables, as previously mentioned, in the multivariate-
adjusted model (29). All statistical analyses were performed with
SAS software (version 9.2; SAS Institute Inc.). P,0.05 was
considered statistically significant.
RESULTS
We documented 867 incident GDM pregnancies in 21,411
singleton pregnancies in 15,265 women during 10 y of follow-up.
At baseline, women with higher LCD scores were more likely to
be current smokers, reported less physical activity, had higher
BMI, and consumed more heme iron, red meat, poultry, and high-
fat dairy but less total calories, dietary fiber, magnesium, vitamin
C, vitamin E, low-fat dairy, fruit, vegetables, whole grains, and
sugar-sweetened beverages (Table 1). We observed similar re-
sults for the animal LCD score. For the vegetable LCD score,
participants with higher scores consumed more nuts, legumes,
fruit, and whole grains but less calcium than did women with
a lower score. Each of these 3 LCD scores was inversely asso-
ciated with the dietary glycemic index and glycemic load.
Overall and animal LCD scores were positively associated
with GDM risk, whereas the vegetable LCD score was not as-
sociated with the risk. Multivariable-adjusted RRs (95% CIs) of
GDM for comparisons of highest with lowest quartiles were 1.53
(1.28, 1.82) for the overall LCD score (P-trend ,0.001), 1.63
(1.36, 1.96) for the animal LCD score (P-trend ,0.001), and
0.91 (0.74, 1.11) for the vegetable LCD score (P-trend = 0.39)
(Table 2). The significant association of overall and animal LCD
scores with GDM risk remained after additional adjustment for
BMI, with corresponding RRs (95% CIs) of 1.27 (1.06, 1.51)
(P-trend = 0.03) and 1.36 (1.13, 1.64) (P-trend = 0.003), re-
spectively. When LCD scores were modeled as a continuous
variable, we showed 6% higher (RR: 1.06; 95% CI: 1.02, 1.11)
risk of GDM associated with each 5-unit increment of the
overall LCD score and 8% higher (RR 1.08; 95% CI 1.03, 1.12)
risk of GDM associated with each 5-unit increment of the ani-
mal LCD score. Associations between LCD scores and GD risk
were not significantly differentiated by overweight status (see
Supplementary Figures 1–3 under “Supplemental data” in the
online issue.). In addition, associations were not significantly
modified by other risk factors of GDM such as age, parity,
family history of diabetes, or physical activity.
Associations between LCD scores and GDM risk were robust
in multiple sensitivity analyses. First, similar results were ob-
served in a propensity score analysis; adjusted RRs (95% CIs) of
GDM for comparisons of highest with lowest quartiles were 1.24
(1.04, 1.49) for the overall LCD score, 1.33 (1.10, 1.60) for the
animal LCD score, and 0.85 (0.69, 1.03) for the vegetable LCD
score. Second, a sensitivity analysis in which missing exposure
data were not carried forward also yielded similar results com-
pared with those in our main analysis; adjusted RRs (95% CIs) of
GDM risk for comparisons of highest with lowest quartiles were
1.33 (1.10, 1.61) for the overall LCD score, 1.48 (1.21, 1.80) for
the animal LCD score, and 0.83 (0.68, 1.03) for the vegetable
LCD score. Third, we observed similar results in a sensitivity
analysis by excluding current pregnancies at the time when
women completed the FFQ; adjusted RRs (95% CIs) of GDM for
comparisons of highest with the lowest quartiles were 1.17 (0.87,
1.57) for the overall LCD score, 1.38 (1.02, 1.88) for the animal
LCD score, and 0.81 (0.57, 1.15) for the vegetable LCD score. In
addition, we conducted a sensitivity analysis by dividing LCD
scores into more refined categories (ie, deciles). Adjusted RRs
(95% CIs) of GDM risk for comparison of highest with lowest
deciles were 1.46 (1.08, 1.95) for the overall LCD score, 1.67
(1.25, 2.24) for the animal LCD score, and 0.76 (0.55, 1.05) for
the vegetable LCD score.
To examine which dietary variable was responsible for these
associations between LCD scores and GDM risk, we conducted
additional adjustments for several foods, food groups, or nutri-
ents. The association of the animal LCD score with GDM risk for
comparisons of highest with lowest quartiles was no longer
significant after additional adjustment for quartiles of red meat
(servings/d) (RR: 1.08; 95% CI: 0.88, 1.33), animal fat (per-
centage of energy) (RR: 1.03; 95% CI: 0.76, 1.40), or heme iron
(mg/d) (RR: 1.06; 95% CI: 0.83, 1.36), which indicated that red
meat, animal fat, and heme iron may be the main contributors to
the observed association between the animal LCD score and
GDM risk. We performed similar analyses for the vegetable LCD
score by adjusting for dietary sources of vegetable protein and
vegetable fat; however, these adjustments did not substantially
alter the association.
DISCUSSION
In this prospective cohort study, we observed that a prepreg-
nancy dietary score that represented a low-carbohydrate, high
animal protein and animal fat dietary pattern was significantly
and positively associated with GDM risk. Conversely, a pre-
pregnancy dietary score that represented a dietary pattern low in
LOW-CARBOHYDRATE DIETARY PATTERN AND GD RISK 3of7
TABLE 1
Age-adjusted characteristics of the study population in 1991 according to Qs of prepregnancy LCD scores in 15,265 women
1
Characteristic
Overall LCD score Animal LCD score Vegetable LCD score
Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4
Participants (n) 4404 4205 3268 3388 3976 4315 3582 3391 5044 4004 3484 2732
Age in 1991 (y) 31.9 63.3
2
32.1 63.3 32.0 63.3 31.9 63.2 32.0 63.3 32.1 63.3 31.9 63.3 31.9 63.1 31.7 63.2 32.1 63.3 32.1 63.3 32.3 63.4
White (%) 92 94 94 93 92 93 94 93 92 93 94 94
Family history of
diabetes (%)
10 11 11 13 10 11 10 13 11 11 10 12
Nulliparous (%) 40 35 34 34 43 35 32 34 33 36 38 41
Current smoking (%) 7 8 10 12 7 9 10 11 8 8 10 11
Alcohol (g/d) 2.8 64.8 3.1 65.6 3.3 65.5 3.0 64.9 2.8 64.5 3.1 65.3 3.2 65.7 3.1 65.3 2.5 64.9 3.1 65.4 3.3 65.2 3.6 65.2
BMI (kg/m
2
) 22.7 63.8 23.2 64.0 23.7 64.3 24.5 65.0 22.6 63.8 23.2 64.1 23.7 64.2 24.5 65.0 23.3 64.2 23.3 64.1 23.5 64.3 23.8 64.7
Physical activity,
(MET-h/wk)
27.2 632.7 23.6 629.4 21.8 626.1 19.1 623.9 28.1 634.7 23.2 627.7 21.1 624.6 19.8 625.8 24.2 630.4 23.4 628.1 23.2 629.0 21.3 626.0
Total calories (kcal/d) 1906 6567 1866 6538 1813 6543 1714 6534 1886 6575 1879 6540 1815 6540 1731 6531 1900 6551 1832 6544 1813 6552 1732 6540
Carbohydrate (% of energy) 58.9 64.6 51.8 62.3 47.3 62.2 41.6 63.8 58.8 65.2 52.1 63.4 47.8 63.2 42.1 64.2 54.6 66.8 50.6 66.8 48.5 66.6 46.0 65.4
Protein (% of energy) 16.6 62.4 19.1 62.7 20.1 62.9 21.9 62.8 16.3 62.4 18.7 62.4 20.2 62.6 22.2 62.7 18.9 63.5 19.7 63.3 19.3 63.2 18.9 63.0
Animal protein (%
of energy)
11.3 62.6 14.0 62.7 15.2 63.0 17.4 62.9 10.7 62.4 13.7 62.2 15.4 62.4 18.0 62.8 14.6 63.5 14.7 63.6 14.1 63.6 13.3 63.3
Vegetable protein (%
of energy)
5.3 61.4 5.1 61.0 4.8 60.8 4.4 60.8 5.6 61.4 5.0 60.9 4.7 60.8 4.3 60.7 4.4 60.9 5.0 61.1 5.2 61.1 5.6 61.0
Total fat (% of energy) 26.0 63.8 30.0 63.7 33.0 63.9 36.6 63.8 26.5 64.6 30.0 64.3 32.5 64.3 35.6 64.3 27.6 64.6 30.4 64.7 32.7 64.6 35.6 64.4
Animal fat (% of energy) 13.4 63.2 16.6 62.8 18.7 63.0 22.0 63.7 12.5 62.7 16.3 62.3 18.8 62.4 22.6 63.3 17.1 64.4 17.6 64.7 17.5 64.7 17.0 64.0
Vegetable fat (% of energy) 12.6 63.4 13.4 63.7 14.3 64.2 14.5 63.9 14.0 64.0 13.8 64.0 13.6 63.8 13.0 63.4 10.4 62.3 12.8 62.2 15.2 62.3 18.7 63.4
Saturated fat (% of energy) 9.2 61.8 10.8 61.7 12.0 61.9 13.4 62.0 9.1 61.8 10.8 61.7 11.9 61.9 13.3 62.0 10.5 62.3 11.1 62.4 11.6 62.3 12.1 62.2
Monounsaturated fat (%
of energy)
9.7 61.7 11.2 61.7 12.5 61.8 13.9 61.8 10.0 62.0 11.3 62.0 12.2 62.0 13.4 61.9 10.2 61.9 11.4 62.0 12.4 62.0 13.7 62.0
Polyunsaturated fat (%
of energy)
4.8 61.1 5.3 61.1 5.8 61.3 6.1 61.3 5.1 61.3 5.4 61.3 5.6 61.3 5.7 61.2 4.5 60.8 5.3 60.8 5.9 60.9 6.9 61.4
trans Fat (% of energy) 1.3 60.5 1.5 60.5 1.7 60.6 1.9 60.6 1.4 60.6 1.6 60.6 1.7 60.6 1.8 60.6 1.3 60.4 1.5 60.5 1.7 60.6 2.0 60.7
Cholesterol (mg/d)
3
187 649 231 647 255 650 293 661 179 645 226 642 256 646 299 658 232 662 243 667 242 667 233 660
Glycemic index
3
55.1 63.2 54.0 63.1 53.5 63.0 53.1 63.3 55.1 63.1 54.1 63.1 53.6 63.1 53.0 63.3 54.2 63.6 53.9 63.1 53.8 63.1 53.8 62.9
Glycemic load
3
146 616 126 610 114 69 100 611 146 617 127 612 116 611 101 612 133 621 123 619 118 619 112 615
Total fiber (g/d)
3
19.9 66.8 18.6 65.1 17.3 64.1 15.7 63.7 20.6 66.8 18.5 65.0 17.2 64.1 15.5 63.7 16.9 65.5 18.5 65.8 18.7 65.2 18.7 64.6
Magnesium (mg/d)
3
326 684 326 673 315 666 303 664 329 685 321 672 316 667 305 665 317 678 324 674 319 670 313 668
Heme iron (mg/d)
3
0.8 60.3 1.0 60.3 1.2 60.3 1.4 60.4 0.7 60.3 1.0 60.3 1.2 60.3 1.5 60.4 1.0 60.4 1.1 60.4 1.1 60.4 1.1 60.4
Potassium (mg/d)
3
2915 6579 2932 6505 2862 6460 2802 6435 2898 6583 2905 6507 2881 6474 2839 6436 2929 6549 2930 6499 2865 6471 2757 6462
Calcium (mg/d)
3
1048 6419 1117 6432 1077 6418 1037 6414 1012 6409 1082 6417 1107 6431 1087 6428 1185 6471 1086 6402 1003 6361 928 6368
Vitamin C (mg/d)
3
299 6302 253 6267 222 6248 198 6254 301 6312 248 6261 227 6253 204 6250 266 6267 255 6287 233 6259 220 6284
Vitamin E (mg/d)
3
22.4 648.1 20.2 644.0 21.0 647.6 17.4 634.4 23.4 650.1 20.2 644.6 19.8 643.4 17.8 636.5 19.9 643.2 20.6 643.8 20.1 642.0 21.6 649.5
Red meat (servings/d) 0.5 60.4 0.7 60.5 0.8 60.5 1.0 60.6 0.5 60.4 0.7 60.5 0.8 60.5 1.0 60.6 0.7 60.5 0.8 60.6 0.8 60.6 0.7 60.5
Poultry (servings/d) 0.4 60.2 0.5 60.3 0.5 60.3 0.6 60.3 0.3 60.2 0.5 60.3 0.5 60.3 0.6 60.3 0.5 60.3 0.5 60.3 0.5 60.3 0.4 60.3
Fish (servings/d) 0.2 60.2 0.2 60.2 0.2 60.2 0.2 60.2 0.2 60.2 0.2 60.2 0.2 60.2 0.2 60.2 0.2 60.2 0.2 60.2 0.2 60.2 0.2 60.2
Eggs (servings/d) 0.2 60.2 0.2 60.2 0.2 60.2 0.2 60.2 0.1 60.2 0.2 60.2 0.2 60.2 0.2 60.2 0.2 60.2 0.2 60.2 0.2 60.2 0.2 60.2
Low-fat dairy (servings/d) 1.6 61.3 1.7 61.3 1.5 61.2 1.3 61.2 1.4 61.1 1.7 61.3 1.6 61.3 1.5 61.3 2.0 61.4 1.6 61.2 1.3 61.0 0.9 60.8
High-fat dairy (servings/d) 0.8 60.7 1.0 60.9 1.1 61.0 1.2 61.1 0.8 60.6 1.0 60.9 1.1 61.0 1.2 61.1 1.0 61.0 1.0 60.9 1.0 60.9 1.0 60.9
(Continued)
4of7 BAO ET AL
carbohydrate and high in vegetable protein and vegetable fat was
not significantly associated with GDM risk. To our knowledge,
the current study is the first attempt to examine the association
between a low-carbohydrate dietary pattern and risk of GDM
incidence in a large prospective cohort. Although we are unaware
of previous studies that specifically evaluated a prepregnancy
low-carbohydrate dietary pattern and risk of GDM, our results were
largely consistent with previous findings of a low-carbohydrate
dietary pattern in association with T2D risk in the Health Pro-
fessionals Follow-Up Study (7).
To interpret associations between a low-carbohydrate dietary
pattern and risk of GDM, each of the macronutrients and their
major food sources should be considered, because an individual
with a low-carbohydrate dietary pattern tends to have a relatively
higher intake of fat and protein to compensate energy re-
quirements. Observed divergent associations of animal compared
with vegetable LCD scores with GDM risk indicated that as-
sociations may not have been the result of a lower quantity of
carbohydrate intake. A previous study (30) has shown a null
association of total carbohydrate intake but significant associa-
tion of the quality of carbohydrate with GDM risk. The positive
association of GDM risk with the LCD score, in particular the
animal LCD score, could have been attributable to detrimental
effects of animal fat and animal protein. The relation between
dietary fat, especially animal fat, and impaired glucose metab-
olism has been well documented (31). For dietary protein, an
animal protein–rich meal compared with a vegetable protein-
rich meal resulted in higher plasma concentrations of branched-
chain amino acids (32), which have been positively linked to the
development of insulin resistance and incident diabetes in recent
metabolomics studies (33–35). Higher intakes of animal fat (15)
and animal protein (16) were previously associated with in-
creased risk of GDM, whereas higher intake of vegetable protein
was associated with lower risk (16). Red meat, which is a major
dietary source of animal protein and animal fat that was asso-
ciated with GDM risk (16, 36), was shown in the current study
to be responsible for the association between the animal LCD
score and GDM risk. Besides animal fat, we also showed that
heme iron was a contributor to the association, which was
consistent with previous findings (37, 38). Other aspects of red
meat, such as advanced glycation end products formed during
grilling red meat (39) and nitrite and nitrate preservatives in
processed red meat (40), may also contribute to the association.
However, we were unable to assess their roles in our current
analysis because of the lack of such data.
Our study has several strengths, including the prospective
design that established the temporal direction of associations,
large sample size, long-term follow-up, high response rates
(.90%) of each questionnaire cycle, and detailed prospective
dietary assessments with extensively validated FFQs (19–21).
We acknowledge that there were several limitations. First, the
misclassification of dietary intakes of carbohydrate, fat, and
protein was possible. However, the random ,within-person error
would have been nondifferential because the prepregnancy di-
etary information was captured prospectively; therefore, our
observed associations may have underestimated the true RRs.
Furthermore, the use of cumulative averages of dietary intakes
for participants with more than one prepregnancy FFQ reduced
the random error. Second, our study population consisted mostly
of white American women in whom we showed a high correlation
TABLE 1 (Continued )
Characteristic
Overall LCD score Animal LCD score Vegetable LCD score
Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4
Nuts (servings/d) 0.2 60.3 0.3 60.3 0.3 60.3 0.2 60.4 0.3 60.4 0.3 60.4 0.2 60.3 0.2 60.2 0.2 60.2 0.2 60.2 0.3 60.3 0.4 60.6
Legumes (servings/d) 0.4 60.4 0.4 60.3 0.3 60.3 0.3 60.3 0.4 60.4 0.4 60.3 0.3 60.3 0.3 60.3 0.3 60.3 0.4 60.3 0.4 60.3 0.4 60.3
Vegetables (servings/d) 3.4 62.2 3.3 62.0 3.1 61.8 2.9 61.7 3.5 62.3 3.3 62.0 3.1 61.8 2.9 61.6 2.9 61.8 3.3 62.1 3.4 62.1 3.3 62.0
Fruit (servings/d) 1.5 61.2 1.3 60.9 1.1 60.7 0.8 60.6 1.5 61.2 1.3 60.9 1.1 60.8 0.9 60.7 1.4 61.1 1.3 60.9 1.2 60.9 0.9 60.7
Whole grains (servings/d) 1.3 61.2 1.2 61.1 1.0 60.9 0.9 60.9 1.4 61.2 1.2 61.0 1.0 60.9 0.8 60.8 0.9 60.9 1.1 61.0 1.2 61.1 1.3 61.2
SSBs (servings/d) 1.0 61.3 0.5 60.7 0.3 60.5 0.2 60.3 0.9 61.2 0.5 60.8 0.4 60.6 0.2 60.4 1.0 61.2 0.4 60.6 0.3 60.4 0.2 60.3
1
Values were standardized to the age distribution of the study population. All comparisons were significant in trend tests across categories except for the following: nuts for the overall LCD score and
vitamin E and high-fat dairy for the vegetable LCD score. LCD, low-carbohydrate diet; MET-h, metabolic equivalent task hours; Q, quartile; SSB, sugar-sweetened beverage.
2
Mean 6SD (all such values).
3
Values were energy adjusted.
LOW-CARBOHYDRATE DIETARY PATTERN AND GD RISK 5of7
between the overall LCD score and animal LCD score (R= 0.94,
P,0.001), which indicated that most of the women who had
a low-carbohydrate dietary pattern consumed animal rather than
plant foods as their major sources of protein and fat. Thus, the
direct generalization of our findings to other populations whose
major food sources of macronutrients are different (41) may be
limited. Indeed, inconsistent associations of long-term effects of
LCDs on adverse health outcomes, such as cardiovascular disease
(8, 9) and mortality (42), have been reported in European and US
populations. The association between LCD scores and risk of
GDM across different race-ethnic groups warrants additional
evaluations. Third, the entire population in this study was aged
$25 y. Because advanced maternal age is a known risk factor for
GDM (43), future studies are needed to examine associations
between LCD scores and GDM risk in women ,25 y of age.
In conclusion, our findings indicate that a prepregnancy dietary
pattern relatively low in carbohydrate and high in protein and fat
from animal-food sources is positively associated with GDM risk,
whereas a prepregnancy dietary pattern relatively low in car-
bohydrate and high in protein and fat from vegetable-food
sources was not associated with the risk. Women of reproductive
age who follow a low-carbohydrate dietary pattern may consider
consuming vegetable rather than animal sources of protein and fat
(in particular red meat) to minimize their risk of GDM. Because
of the observational study design, our study cannot confirm the
causation between adherence to a low-carbohydrate dietary
pattern and risk of GDM. Future studies with a randomized
controlled trial design are warranted.
We thank Thomas L Halton (Departments of Nutrition, Harvard School of
Public Health) for devising LCD scores.
The authors’ responsibilities were as follows—WB: contributed to the
design and analysis of the study and wrote the manuscript; KB: conducted
a technique review and reviewed and edited the manuscript; DKT: contrib-
uted to the data analysis and reviewed and edited the manuscript; SFO, JC,
AV, and MK: interpreted results and reviewed and edited the manuscript; CZ:
contributed to the design and analysis of the study and reviewed and edited
the manuscript; and WB and CZ: are the guarantors of this work and, as
such, had full access to all study data and took responsibility for the integrity
of data and the accuracy of the data analysis. None of the authors had
a personal or financial conflict of interest.
REFERENCES
1. Kennedy RL, Chokkalingam K, Farshchi HR. Nutrition in patients with
type 2 diabetes: are low-carbohydrate diets effective, safe or desirable?
Diabet Med 2005;22:821–32.
2. Malik VS, Hu FB. Popular weight-loss diets: from evidence to practice.
Nat Clin Pract Cardiovasc Med 2007;4:34–41.
3. Bravata DM, Sanders L, Huang J, Krumholz HM, Olkin I, Gardner CD.
Efficacy and safety of low-carbohydrate diets: a systematic review.
JAMA 2003;289:1837–50.
4. Foster GD, Wyatt HR, Hill JO, McGuckin BG, Brill C, Mohammed
BS, Szapary PO, Rader DJ, Edman JS, Klein S. A randomized trial of
a low-carbohydrate diet for obesity. N Engl J Med 2003;348:2082–90.
5. Davis NJ, Tomuta N, Schechter C, Isasi CR, Segal-Isaacson CJ, Stein D,
Zonszein J, Wylie-Rosett J. Comparative study of the effects of a 1-year
dietary intervention of a low-carbohydrate diet versus a low-fat diet on
weight and glycemic control in type 2 diabetes. Diabetes Care 2009;32:
1147–52.
6. Halton TL, Liu S, Manson JE, Hu FB. Low-carbohydrate-diet score
and risk of type 2 diabetes in women. Am J Clin Nutr 2008;87:339–46.
TABLE 2
Risk of GDM according to quartile of prepregnancy LCD scores
1
LCD scores
P-trendQ1
2
Q2 Q3 Q4
Overall LCD score —
Median score 6 12 18 24 —
GDM/pregnancies 227/6273 223/5973 164/4574 253/4591 —
Model 1 1.00 1.05 (0.88, 1.26)
3
1.03 (0.85, 1.26) 1.60 (1.34, 1.90) ,0.001
Model 2 1.00 1.06 (0.89, 1.27) 1.04 (0.85, 1.27) 1.53 (1.28, 1.82) ,0.001
Model 3 1.00 1.00 (0.84, 1.20) 0.92 (0.76, 1.13) 1.27 (1.06, 1.51) 0.03
Animal LCD score
Median score 5 12 18 25 —
GDM/pregnancies 196/5659 230/6050 186/5060 255/4642 —
Model 1 1.00 1.14 (0.95, 1.38) 1.14 (0.93, 1.38) 1.70 (1.42, 2.04) ,0.001
Model 2 1.00 1.15 (0.96, 1.39) 1.13 (0.93, 1.38) 1.63 (1.36, 1.96) ,0.001
Model 3 1.00 1.09 (0.90, 1.31) 1.01 (0.82, 1.23) 1.36 (1.13, 1.64) 0.003
Vegetable LCD score
Median score 9 13 17 22 —
GDM/pregnancies 290/7175 234/5692 201/4859 142/3685 —
Model 1 1.00 0.99 (0.83, 1.17) 0.97 (0.81, 1.15) 0.88 (0.72, 1.07) 0.22
Model 2 1.00 1.03 (0.87, 1.22) 1.02 (0.85, 1.21) 0.91 (0.74, 1.11) 0.39
Model 3 1.00 1.02 (0.86, 1.21) 0.95 (0.80, 1.14) 0.84 (0.69, 1.03) 0.08
1
Model 1 was adjusted for updated age (mo) and parity (0, 1, 2, or $3). Model 2 was adjusted as for model 1 and for
race-ethnicity (white, African American, Hispanic, Asian, or other), family history of diabetes (yes or no), cigarette
smoking (never, past, or current), alcohol intake (0.0, 0.1–5.0, 5.1–10.0, or .10.0 g/d), physical activity (Qs), and total
energy intake (Qs). Model 3 was adjusted as for model 2 and for BMI (9 categories as follows: ,21.0, 21.0–22.9, 23.0–
24.9, 25.0–26.9, 27.0–28.9, 29.0–30.9, 31.0–32.9, 33.0–34.9, and $35.0 kg/m
2
). RRs (95% CIs) were estimated using
generalized estimating equations with log-binomial models. Trend tests across Qs of LCD scores were performed by
assigning the median value for each Q and fitting this as a continuous variable in the models. GDM, gestational diabetes
mellitus; LCD, low-carbohydrate diet; Q, quartile.
2
Reference.
3
RR; 95% CI in parentheses (all such values).
6of7 BAO ET AL
7. de Koning L, Fung TT, Liao X, Chiuve SE, Rimm EB, Willett WC,
Spiegelman D, Hu FB. Low-carbohydrate diet scores and risk of type 2
diabetes in men. Am J Clin Nutr 2011;93:844–50.
8. Halton TL, Willett WC, Liu S, Manson JE, Albert CM, Rexrode K,
Hu FB. Low-carbohydrate-diet score and the risk of coronary heart
disease in women. N Engl J Med 2006;355:1991–2002.
9. Lagiou P, Sandin S, Lof M, Trichopoulos D, Adami HO, Weiderpass E.
Low carbohydrate-high protein diet and incidence of cardiovascular
diseases in Swedish women: prospective cohort study. BMJ 2012;344:
e4026.
10. Reece EA, Leguizamon G, Wiznitzer A. Gestational diabetes: the need
for a common ground. Lancet 2009;373:1789–97.
11. American Diabetes Association. Gestational diabetes mellitus. Diabetes
Care 2004;27(suppl 1):S88–90.
12. Metzger BE, Lowe LP, Dyer AR, Trimble ER, Chaovarindr U, Coustan
DR, Hadden DR, McCance DR, Hod M, McIntyre HD, et al. Hyper-
glycemia and adverse pregnancy outcomes. N Engl J Med 2008;358:
1991–2002.
13. Bellamy L, Casas JP, Hingorani AD, Williams D. Type 2 diabetes mellitus
after gestational diabetes: a systematic review and meta-analysis. Lancet
2009;373:1773–9.
14. Centers for Disease Control and Prevention. National diabetes fact
sheet: national estimates and general information on diabetes and
prediabetes in the United States, 2011. Atlanta, GA: US Department of
Health and Human Services, Centers for Disease Control and Pre-
vention, 2011.
15. Bowers K, Tobias DK, Yeung E, Hu FB, Zhang C. A prospective study
of prepregnancy dietary fat intake and risk of gestational diabetes. Am
J Clin Nutr 2012;95:446–53.
16. Bao W, Bowers K, Tobias DK, Hu FB, Zhang C. Prepregnancy dietary
protein intake, major dietary protein sources, and the risk of gestational
diabetes mellitus: a prospective cohort study. Diabetes Care 2013;36:
2001–8.
17. Solomon CG, Willett WC, Carey VJ, Rich-Edwards J, Hunter DJ,
Colditz GA, Stampfer MJ, Speizer FE, Spiegelman D, Manson JE. A
prospective study of pregravid determinants of gestational diabetes
mellitus. JAMA 1997;278:1078–83.
18. Hu FB, Stampfer MJ, Manson JE, Rimm E, Colditz GA, Speizer FE,
Hennekens CH, Willett WC. Dietary protein and risk of ischemic heart
disease in women. Am J Clin Nutr 1999;70:221–7.
19. Willett WC, Sampson L, Stampfer MJ, Rosner B, Bain C, Witschi J,
Hennekens CH, Speizer FE. Reproducibility and validity of a semi-
quantitative food frequency questionnaire. Am J Epidemiol 1985;122:51–65.
20. Willett WC, Sampson L, Browne ML, Stampfer MJ, Rosner B, Hennekens
CH, Speizer FE. The use of a self-administered questionnaire to assess diet
four years in the past. Am J Epidemiol 1988;127:188–99.
21. Salvini S, Hunter DJ, Sampson L, Stampfer MJ, Colditz GA, Rosner B,
Willett WC. Food-based validation of a dietary questionnaire: the ef-
fects of week-to-week variation in food consumption. Int J Epidemiol
1989;18:858–67.
22. Fung TT, van Dam RM, Hankinson SE, Stampfer M, Willett WC,
Hu FB. Low-carbohydrate diets and all-cause and cause-specific
mortality: two cohort studies. Ann Intern Med 2010;153:289–98.
23. Rimm EB, Stampfer MJ, Colditz GA, Chute CG, Litin LB, Willett WC.
Validity of self-reported waist and hip circumferences in men and
women. Epidemiology 1990;1:466–73.
24. Wolf AM, Hunter DJ, Colditz GA, Manson JE, Stampfer MJ, Corsano KA,
Rosner B, Kriska A, Willett WC. Reproducibility and validity of a self-
administered physical activity questionnaire. Int J Epidemiol 1994;23:
991–9.
25. Hu FB, Stampfer MJ, Rimm E, Ascherio A, Rosner BA, Spiegelman D,
Willett WC. Dietary fat and coronary heart disease: a comparison of
approaches for adjusting for total energy intake and modeling repeated
dietary measurements. Am J Epidemiol 1999;149:531–40.
26. Hanley JA, Negassa A, Edwardes MD, Forrester JE. Statistical analysis
of correlated data using generalized estimating equations: an orienta-
tion. Am J Epidemiol 2003;157:364–75.
27. Wacholder S. Binomial regression in GLIM: estimating risk ratios and
risk differences. Am J Epidemiol 1986;123:174–84.
28. Zou G. A modified poisson regression approach to prospective studies
with binary data. Am J Epidemiol 2004;159:702–6.
29. D’Agostino RB Jr. Propensity score methods for bias reduction in the
comparison of a treatment to a non-randomized control group. Stat
Med 1998;17:2265–81.
30. Zhang C, Liu S, Solomon CG, Hu FB. Dietary fiber intake, dietary
glycemic load, and the risk for gestational diabetes mellitus. Diabetes
Care 2006;29:2223–30.
31. Lichtenstein AH, Schwab US. Relationship of dietary fat to glucose
metabolism. Atherosclerosis 2000;150:227–43.
32. Brandsch C, Shukla A, Hirche F, Stangl GI, Eder K. Effect of proteins
from beef, pork, and turkey meat on plasma and liver lipids of rats
compared with casein and soy protein. Nutrition 2006;22:1162–70.
33. Newgard CB, An J, Bain JR, Muehlbauer MJ, Stevens RD, Lien LF,
Haqq AM, Shah SH, Arlotto M, Slentz CA, et al. A branched-chain
amino acid-related metabolic signature that differentiates obese and
lean humans and contributes to insulin resistance. Cell Metab 2009;9:
311–26.
34. Tai ES, Tan ML, Stevens RD, Low YL, Muehlbauer MJ, Goh DL,
Ilkayeva OR, Wenner BR, Bain JR, Lee JJ, et al. Insulin resistance is
associated with a metabolic profile of altered protein metabolism in
Chinese and Asian-Indian men. Diabetologia 2010;53:757–67.
35. Wang TJ, Larson MG, Vasan RS, Cheng S, Rhee EP, McCabe E, Lewis
GD, Fox CS, Jacques PF, Fernandez C, et al. Metabolite profiles and
the risk of developing diabetes. Nat Med 2011;17:448–53.
36. Zhang C, Schulze MB, Solomon CG, Hu FB. A prospective study of
dietary patterns, meat intake and the risk of gestational diabetes mel-
litus. Diabetologia 2006;49:2604–13.
37. Bowers K, Yeung E, Williams MA, Qi L, Tobias DK, Hu FB, Zhang C.
A prospective study of prepregnancy dietary iron intake and risk for
gestational diabetes mellitus. Diabetes Care 2011;34:1557–63.
38. Qiu C, Zhang C, Gelaye B, Enquobahrie DA, Frederick IO, Williams
MA. Gestational diabetes mellitus in relation to maternal dietary heme
iron and nonheme iron intake. Diabetes Care 2011;34:1564–9.
39. Cai W, Ramdas M, Zhu L, Chen X, Striker GE, Vlassara H. Oral ad-
vanced glycation endproducts (AGEs) promote insulin resistance and
diabetes by depleting the antioxidant defenses AGE receptor-1 and
sirtuin 1. Proc Natl Acad Sci USA 2012;109:15888–93.
40. Virtanen SM, Jaakkola L, Rasanen L, Ylonen K, Aro A, Lounamaa R,
Akerblom HK, Tuomilehto J. Nitrate and nitrite intake and the risk for
type 1 diabetes in Finnish children. Childhood Diabetes in Finland
Study Group. Diabet Med 1994;11:656–62.
41. Willett WC. Low-carbohydrate diets: a place in health promotion?
J Intern Med 2007;261:363–5.
42. Noto H, Goto A, Tsujimoto T, Noda M. Low-carbohydrate diets and
all-cause mortality: a systematic review and meta-analysis of obser-
vational studies. PLoS ONE 2013;8:e55030.
43. Makgoba M, Savvidou MD, Steer PJ. An analysis of the in-
terrelationship between maternal age, body mass index and racial or-
igin in the development of gestational diabetes mellitus. BJOG 2012;
119:276–82.
LOW-CARBOHYDRATE DIETARY PATTERN AND GD RISK 7of7
