Case–Control Study of Nutritional and Environmental Factors and the Risk of Oral Clefts in Thailand

Abstract

Background:

One infant in 700 is born with an oral cleft. Prior studies suggest low micronutrient status is associated with an increased risk of oral clefts. Environmental factors such as passive smoke exposure or supplement use may also affect oral cleft risk. We examined nutrition and environmental related risk factors for oral clefts.

Methods:

We conducted a case–control study in Northeast Thailand in 2012 to 2013. We enrolled 95 cases and 95 controls. We recruited cases with a nonsyndromic cleft lip with or without a cleft palate (CL±P) less than 24 months old. Cases were matched to controls on age and place of conception. We collected survey data, a food frequency questionnaire, and measured zinc concentrations in toenail trimmings. We calculated descriptive statistics by case and control status. We used conditional logistic regression to estimate unadjusted and adjusted associations, 95% confidence intervals (CIs), and p-values.

Results:

Any liver intake (adjusted OR [aOR] for ≥1/week versus none), 10.58; 95%CI, 1.74–64.37, overall p = 0.02) and the presence of food insecurity (aOR, 9.62; 95% CI, 1.52–61.05; p = 0.02) in the periconceptional period increased CL±P risk. Passive smoke exposure increased the risk of CL±P (aOR, 6.52; 95% CI, 1.98–21.44; p < 0.01). Toenail zinc concentrations were not associated with CL±P risk.

Conclusion:

Our findings add to a growing body of knowledge of environmental risk factors for oral clefts from low- and middle-income countries. Our findings on liver are contradictory to prior results. Large multisite studies are needed to identify environmental and genetic risk factors for oral clefts.

Introduction

Globally, 1 infant in every 500 to 700 newborns is born with an oral cleft (World Health Organization, 2004). The vast majority (94%) of children born with an oral cleft are born in a developing country (March of Dimes, 2006). An oral cleft is a complex developmental trait initiated in the first trimester of pregnancy that involves both genetic and environmental risk factors. Environmental risk factors for oral clefts may differ across cultural, socioeconomic, or geographic settings. Identifying environmental risk factors for oral clefts in low- and middle-resource settings may inform on cost-effective, globally relevant, and modifiable risk factors for oral clefts.

Nutritional status may play a role in oral cleft risk. Most although not all studies show that multivitamin use in the periconceptional period reduces the risk of oral clefts (Hayes et al., 1996; Loffredo et al., 2001; Shaw et al., 2006; Badovinac et al., 2007; Johnson and Little, 2008; Jia et al., 2011). However, the specific nutrients responsible for the decrease are unknown. Some studies suggest low zinc concentrations may increase oral cleft risk, although associations may vary by place (Krapels et al., 2004; Tamura et al., 2005; Munger et al., 2009). In a recently published study conducted in Thailand, we found that low intake of liver increased oral cleft risk (McKinney et al., 2013). The only other study to examine liver intake, conducted in Denmark, found no association between liver intake and oral cleft risk but reported that those with the highest levels of vitamin A estimated from liver intake and supplement use combined had a decreased risk of cleft lip with or without cleft palate (CL±P) (Mitchell et al., 2003).

Poor diet quality has been linked to an increased risk of oral cleft (Vujkovic et al., 2007; La et al., 2013). Non-nutritional environmental risk factors are also implicated. For example, in our recently published case–control study in Thailand, we found that menstrual regulation supplements, which may contain alcohol or herbs with estrogenic effects, increase oral cleft risk (McKinney et al., 2013). Environmental or passive smoke exposure may increase oral cleft risk, but it is difficult to separate out the effects of maternal smoke exposure from other sources of smoke exposure (Li et al., 2010).

We set out to identify environmental risk factors for oral clefts in northeast Thailand. We conducted a case–control study to examine associations between nutritional status (zinc, liver intake, supplement use) and risk of CL±P. We used a unique biomarker that measured zinc concentrations in toenails, hypothesizing that low zinc concentrations increase the risk of an oral cleft. We collected detailed liver consumption information to examine the association between type of liver intake and CL±P and examined food security as a marker of diet quality. We also examined the degree to which menstrual regulation supplements, passive smoke exposure, and a fever increases risk of CL±P.

Materials and Methods

We conducted a case–control study from March 2012 to December 2013 at the Center of Cleft Lip-Cleft Palate and Craniofacial Deformities (“Tawanchai Center”) at Khon Kaen University in northeast Thailand. We recruited cases with a nonsyndromic CL±P under 24 months of age seen in the Center. Infants with a cleft palate only or a syndrome were excluded. Cases with anomalies that were not part of a syndrome were eligible. We collected clinic photographs and a study-trained craniofacial nurse completed a physical exam form to classify the case’s oral cleft phenotype. We recruited cases during a clinic visit to the Center. Those who spoke Thai or an Isaan dialect of Thai were eligible for the study.

We enrolled one control per case. Controls were matched to cases on age within 4 months of birth. Because we anticipated differences regarding access to different types of foods and prenatal care, we matched on place of conception. We classified case mothers as living in a small town, semi-urban, or urban area. We recruited matched controls from one of three well-baby clinics located in hospitals in a (1) small town, (2) semi-urban area, or (3) urban area to broadly account for the environment. Because well-baby clinics are free and commonly based in hospitals, most parents take their children to well-baby visits, and controls recruited from these clinics are likely to be representative of the population that gave rise to the cases.

Mothers completed an interviewer-administered questionnaire, a food frequency questionnaire (FFQ), and provided toenail trimmings. We elicited demographic, behavioral, nutritional, and maternal health data using the questionnaire, which was adapted from the Centers for Disease Control and Prevention’s National Birth Defects Prevention Study (Shaw et al., 2006). We measured each type of liver intake (chicken/duck, pork, and beef) using an FFQ developed specifically for the northeast Thai diet by our Thai nutritionist (B. Muktabhant). We used the Fred Hutchinson Cancer Research method for quantifying intake from an FFQ (Fred Hutchinson Cancer Research Center, 2015). Briefly, the FFQ measured frequency of intake in nine categories: (1) never, (2) once per month, (3) two to three times per month, (4) once per week, (5) twice per week, (6) three to four times per week, (7) five to six times per week, (8) once per day, and (9) two or more times per day. We weighted each frequency of intake for size with small, medium, and large portions given weights of 0.5, 1.0, and 1.5, respectively, to generate intake in medium servings per week and per year. Because intake in controls was not normally distributed, we measured liver intake as none, less than one, and one or more medium servings per week. We measured food insecurity, a measure of diet quality and hunger, using the validated Household Food Insecurity Access Scale (HFIAS), which is used globally (Coates et al., 2007; Humphries et al., 2015). The periconceptional period for our survey was typically defined as between the 3 months before to the 3 months after conception.

Toenail zinc concentrations are easy to measure in toenail trimmings, and toenail zinc concentrations correlate with dietary zinc intake (Gonzalez et al., 2008). Because toenails grow on average at a rate of 1 millimeter per month, trimmings typically represent zinc exposure 3 to 12 months prior (Garland et al., 1993). For cases recruited into the study within 3 months of birth, this means our measure of zinc may approximate exposure during the etiologically relevant time point in the first trimester of pregnancy. For sample collection, mothers clipped their toenails using stainless steel toenail clippers. Trimmings were collected on a plastic tray, funneled into small paper envelopes, and labeled for analysis. Zinc in toenail trimmings is stable for extended periods and at room temperatures found in Thailand, and no special handling was required. Toenail zinc concentrations were measured in parts per million using instrumental neuron activation analysis at the Missouri University Research Reactor Center (Garland et al., 1993).

We classified participants as being food secure or food insecure based on cut points prespecified for this scale (Coates et al., 2007). We classified mothers as having “any multivitamin” if they reported periconceptional intake of a prenatal or general multivitamin. Mothers were categorized as having taken a folic acid supplement if they reported periconceptional intake of folic acid, an iron/B-vitamin complex, which includes folic acid, or a prenatal or general multivitamin, which contain folic acid. For multivitamins and folic acid supplements, we classified participants into three groups: those who took these supplements before the second month of pregnancy in the periconceptional period, the second to third month of pregnancy, and those who took them later or not at all. We classified women as having taken a menstrual regulation supplement if they reported intake in the periconceptional period. We collected information on passive smoke exposure from a husband, family member living in the home, or from someone at work or school. Covariates were analyzed as categorized in Tables 1 and 2. The proportion missing for each variable is indicated in Tables 1 and 2.

TABLE 1.

Infant Birth and Demographic Characteristics by Case and Control Status

	Control (n = 95)		Case (n = 95)
	n	%	n	%
Matching variables
Place of birth
Rural (Poo wiang)	37	39.0	37	39.0
Semi-urban (Ubon rat)	38	21.0	38	21.0
Urban	20	40.0	20	40.0
Infant age at recruitment (matched control ± 4 months age of case)
<3	43	45.3	47	49.5
3–6	27	28.4	29	30.5
6–12	15	15.8	7	7.4
12–30	10	10.5	12	12.6
Infant characteristics
Infant sex
Male	55	57.9	51	55.4
Female	40	42.1	41	44.6
Birth weight (in grams)
≥2500	85	89.5	84	88.4
<2500	10	10.5	11	11.6
Infant gestational age at birth (weeks)
Full or post-term	60	63.8	69	72.6
Preterm	34	36.2	26	27.4
Infant has an anomaly other than cleft
No	95	100.0	95	100.0
Yes	0	0.0	0	0.0
Infant has family history of oral cleft (first or second degree)
No	94	99.0	85	89.5
Yes	1	1.1	10	10.5
Mother and family characteristics
Time of first prenatal care visit
<4 week gestation	12	12.9	5	5.6
4–8 weeks gestation	29	31.2	19	21.4
8–12 weeks gestation	27	29.0	29	32.6
>12 weeks gestation	25	26.9	36	40.5
Mother’s age at time of infant’s birth (in years)
<25	32	34.0	37	39.0
25–34	50	53.2	43	45.3
≥35	12	12.8	15	15.8
Mother’s highest level of education
Primary school/none	16	16.8	20	21.5
Junior high school	23	24.2	26	28.0
High school	25	26.3	27	29.0
Any college	31	32.6	20	21.5
Household income (monthly, in baht)a
Quartile 1 (<99,600)	22	25.9	16	18.8
Quartile 2 (99,600–216,000)	23	27.1	29	34.1
Quartile 3 (228,000–348,000)	19	22.4	20	23.5
Quartile 4 (≥348,000)	21	24.7	20	23.5
Marital status1
Married	68	76.4	76	85.4
Not married	21	23.6	13	14.6
Mother’s total number of pregnancies
1	36	37.9	32	35.2
2	33	34.7	36	39.6
≥3	26	27.4	23	25.3
Wanted to be pregnant at the time
No	36	37.9	33	35.1
Yes	59	62.1	61	64.9
Mother drank alcohol in first trimester
No	81	86.2	83	88.3
Yes	13	13.8	11	11.7
Mother ever smoked >100 cigarettes
No	0	0.0	1	1.1
Yes	92	100.0	93	98.9

% is percent, all variables missing <5% unless otherwise noted.

^aVariable missing 5 to 10%

TABLE 2.

Liver by Type and Frequency and Its Association for Cleft Lip with or without Cleft Palate

	Control		Case		Unadjusted		Adjusteda
	n	%	n	%	OR (95% CI)	p-Value	OR (95% CI)	p-Value
All types of liver intake, frequency of medium sized portions
None	15	16.3	8	8.9	Ref	0.07	Ref	0.02
>0 & <1 time per week	14	15.2	22	24.4	4.60 (1.25, 16.84)		21.62 (2.48, 188.20)
≥1 time per week	63	68.5	60	66.7	2.45 (0.87, 6.90)		10.58 (1.74, 64.37)
Chicken or duck liver intake, frequency of medium sized portions
None	17	18.5	8	8.7	Ref	0.05	Ref	0.03
>0 & <1 time per week	16	17.4	24	26.1	4.88 (1.34, 17.77)		12.29 (1.66, 91.18)
≥1 time per week	59	64.1	60	65.2	3.30 (1.08, 10.09)		10.09 (1.77, 57.62)
Pork liver intake, frequency of medium sized portions
None	33	35.5	35	38.0		0.86		0.58
>0 & <1 time per week	27	29.0	28	30.4	1.11 (0.56, 2.21)		1.26 (0.51, 3.08)
≥1 time per week	33	35.5	29	31.5	0.91 (0.46, 1.78)		0.78 (0.34, 1.76)
Beef liver intake, frequency of medium sized portions
None	57	61.3	58	63.0	Ref	0.16	Ref	0.04
>0 & <1 time per week	18	19.4	24	26.1	1.42 (0.71, 2.82)		2.60 (1.03, 6.53)
≥1 time per week	18	19.4	10	10.9	0.51 (0.20, 1.28)		0.56 (0.19, 1.62)

All variables missing <5%.

^aAdjusted OR accounts for matching factors (age of infant at recruitment and town size of residence at conception), income, education, number of pregnancies, and passive smoke exposure.

% = percent, n = count, OR = odds ratio.

STATISTICAL ANALYSIS

We calculated descriptive statistics for infant birth, maternal pregnancy, demographic characteristics, and factors of interest across case and control status. Because cases were matched to controls on place of conception and age at recruitment, we used conditional logistic regression to estimate unadjusted and adjusted odds ratios (ORs) between proposed risk factors and CL±P along with 95% confidence intervals (CI) and p-values. We generated separate models for each potential risk factor, adjusting for factors identified a priori as potential confounders. We examined liver intake by all types combined and by type of liver (chicken/duck, pork, or beef). We assessed each potential confounder separately and in combination with other confounders. Our modeling strategy was to drop confounders that did not change estimates to a meaningful degree (generally less than 10%) from the initial and combined models. Adjustment variables are noted for each model (Tables 2 and 3).

TABLE 3.

Nutritional and Environmental Factors by Case–Control Status and Unadjusted and Adjusted Associations

	Control		Case		Unadjusted		Adjusteda
	n	% or $\overline{x}$	n	% or $\overline{x}$	OR (95% CI)	p-Value	OR (95% CI)	p-Value
Zinc concentration in toenails (mean in ppm)b	89	60.0	92	62.9	1.01 (0.99, 1.04)	0.19	1.01 (0.99, 1.04)	0.19
Food insecurity (%)c
Secure	82	87.2	69	75.8	Ref		Ref
Insecure	12	12.8	22	24.2	3.00 (1.19, 7.56)	0.02	9.62 (1.52, 61.05)	0.02
Any multivitamin (%)d
Did not take	67	72.0	61	67.8	Ref	0.63	Ref	0.79
Started 2nd or 3rd month of pregnancy	17	18.3	18	20.0	1.44 (0.60, 3.45)		1.02 (0.20, 5.20)
Started prior to 2nd month of pregnancy	9	9.7	11	12.2	1.48 (0.52, 4.21)		0.63 (0.14, 2.75)
Folic acid containing supplement (%)d
Did not take	44	46.3	45	50.0	Ref	0.83	Ref	0.52
Started 2nd or 3rd month of pregnancy	31	32.6	31	34.4	1.03 (0.54, 1.98)		0.61 (0.18, 2.03)
Started prior to 2nd month of pregnancy	20	21.1	14	15.6	0.80 (0.34, 1.86)		0.58 (0.22, 1.56)
Menstrual regulation herbal supplement use (%)d
No	82	90.1	79	85.9	Ref		Ref
Yes	9	9.9	13	14.1	1.67 (0.61, 4.59)	0.32	7.36 (0.76, 71.05)	0.08
Fevere
No	83	87.4	71	74.7	Ref		Ref
Yes	12	12.6	24	25.3	3.40 (1.25, 9.22)	0.02	1.52 (0.42, 5.51)	0.52
Passive smoke exposures (husband, family member, or at work)f
No	75	79.0	54	56.8	Ref		Ref
Yes	20	21.1	41	43.2	3.10 (1.52, 6.32)	0.002	6.52 (1.98, 21.44)	0.002

All variables missing <5%.

^aAll adjusted ORs account for matching variables (age of infant at recruitment and town size of residence at conception).

^bNo factors confounded the association so the adjusted OR is the same as the unadjusted OR.

^cAdjusted OR also accounts for income, number of pregnancies, passive smoke exposure.

^dAdjusted OR also accounts for income, number of pregnancies, passive smoke exposure, food insecurity, and marital status.

^eAdjusted OR also accounts for income, passive smoke exposure, food insecurity, liver intake (yes/no), and marital status.

^fAdjusted OR also accounts for income, number of pregnancies, mother’s age at infant birth, and marital status.

Ppm = parts per million; % = percent; $\overline{x}$ = mean; n = count; OR = odds ratio.

In sub-analysis, we examined liver as a continuous variable and in quartiles. We examined the potential interaction between multivitamin or folate supplement use with liver intake (yes/no) on CL±P risk. We examined zinc concentrations in quartiles and among cases and controls recruited within 3 months of birth because the measure of zinc at this time likely approximates zinc concentrations near time of conception. We estimated the association between family history of an oral cleft and CL±P risk. To assess whether the subset of cases with a genetic etiology could have biased risk estimates, we estimated unadjusted and adjusted associations for those without a family history of an oral cleft, adjusting for the same confounders in our primary analysis. All analyses were performed in Stata 13.0. We obtained human ethics approval from the University of Washington and Khon Kaen University.

Results

We enrolled 95 cases with any type of oral cleft and 95 matched controls. Most cases recruited were less than 6 months old (Table 1). With regard to infant characteristics, cases and controls were similar on sex, birth weight, and presence of other anomalies. In terms of maternal characteristics, fewer case than control mothers had prenatal care in the first month of pregnancy. Fewer case mothers had a college education than control mothers, although a smaller proportion of cases relative to controls were in the lowest income quartile. Mothers of cases and controls were similar in terms of their desire to be pregnant and their periconceptional alcohol and smoking behaviors. Only one mother in the study reported smoking (Table 1).

More than 60% of study participants reporting eating liver more than once per week; chicken or duck liver was the type of liver eaten most frequently (Table 2). In adjusted analysis, mothers who reported eating any type of liver once or more per week were more than 10 times as likely to have a child with a CL±P than those who did not (95% CI, 1.7–64.4). In stratified analysis by type of liver, mothers who reported eating chicken or duck liver once or more per week had a 10-fold increased risk of having a child with a CL±P (95% CI, 1.8–54.6). Associations between pork or beef liver intake were not as strong as those observed for chicken or duck liver intake, and at the most frequent levels of intake (≥1 time per week) showed inverse associations, and for pork, were not statistically significant.

Zinc concentrations were similar for cases and controls (Table 3). More control than case mothers reported being food secure and reported taking a folic acid containing supplement before the second month of pregnancy. More case than control mothers reported taking a menstrual regulation supplement. Fever and passive smoke exposure were more common in cases than controls. Mothers who were food insecure were 9.6 times more likely to have a child with a CL±P than those who reported being food secure (95% CI, 1.5–61.1). Passive smoke exposure increased the risk of having a child with a CL±P more than sixfold (95% CI, 1.9–21.4). Associations for folic acid containing supplement use, menstrual regulation herbal supplement use, and fever were in the direction expected but not statistically precise (Table 3).

In sub-analysis, associations between liver consumption and oral cleft risk varied depending on how liver intake was classified. Associations between total liver intake and chicken and duck intake and oral clefts were strongest when the referent group was no liver intake (Supplemental Table 1, which is available online). We observed no interaction between liver intake (yes/no) and multivitamin or folate supplement use (data not shown). As anticipated, those with a family history of an oral cleft (N = 11) were more likely to have a CL±P (OR, 10.0; 95% CI, 1.3–78.1; p = 0.03) than those without this family history. In sub-analyses, our adjusted estimates were similar when we restricted to the subset of matched cases and controls in which the case had no family history of oral clefts (data not shown). Even when restricting to matched cases and controls both enrolled at less than 3 months of age (N = 80), we observed no association between zinc concentrations and oral cleft risk (OR, 1.0; 95% CI, 0.99–1.1; p = 0.28). We observed no association when examining zinc concentrations in quartiles (data not shown).

Discussion

We found that overall liver intake and food insecurity in the periconceptional period increased CL±P risk. Preliminary findings suggestion associations may vary by type of liver. Passive smoke exposure increased the risk of CL±P. Toenail zinc concentrations were not associated with CL±P risk. The association between menstrual regulation supplement use and CL±P risk was in the direction expected but was statistically imprecise. These findings add to a growing body of knowledge about nutritional and environmental risk factors for CL±P globally.

Our findings on liver consumption in this and our prior study were both strong, statistically significant associations. However, our prior study showed a strong protective effect of liver consumption (McKinney et al., 2013). Our current study shows that liver consumption increases CL±P risk. Associations in both studies should be interpreted with caution because the confidence intervals suggest a wide range of estimates were plausible. There are other possible explanations for the observed differences. Our prior study was conducted in 40 different hospitals in six provinces across Thailand, and controls were matched on hospital at the time of birth. Our current study recruited cases from a single hospital with controls selected from well-baby clinics from three hospitals based on place of conception. Differences in place or control selection may show different associations.

Our findings suggest the possibility that differences in associations between type of animal liver intake and CL±P risk may exist. The micronutrient concentrations present in animal livers vary widely. For example, beef liver contains less than 17,000 IU/100 g of vitamin A, whereas duck liver contains nearly 40,000 IU/100 g (United States Department of Agriculture [USDA], 2015a,b). A diet containing 200 g/week of duck liver (80,000 IU/week) exceeds one threshold of an increase in cranial neural crest defects with supplemental vitamin A >70,000 IU/week (Rothman et al., 1995). However, the extent to which supplementation versus liver intake yields biologically active all-trans retinoic acid and retinoid metabolites is unclear. Evidence suggests it may vary by dose or source of liver (Buss et al., 1994; van Vliet et al., 2001). There is also the possibility that teratogenic effects are more potent in individuals that are not vitamin A deficient.

Food insecurity was associated with an increased risk of CL±P. The HFIAS is a measure developed to assess not only hunger but uncertainty and worry, diet quality, and social unacceptability in food consumption (Coates et al., 2006b, 2007). The HFIAS is an 18-item scale designed to capture universal experiences related to food insecurity cross-culturally and in different countries (Coates et al., 2006b, 2007). Its development was facilitated by an evaluation of food insecurity in 22 countries that identified universal concepts of food insecurity, which are incorporated in the HFIAS (Coates et al., 2006a). Since its dissemination in 2007, it has been widely used globally to assess food insecurity (Nepal Nutrition and Food Security Bulletin, 2010; Humphries et al., 2015). As far as we know, we are the first to use the HFIAS in a study of oral clefts. The only other study on food insecurity and oral clefts used a five-item survey and was inconclusive (Carmichael et al., 2007).

Our finding is consistent with other studies that have found that poor diet quality increases the risk of oral cleft (Vujkovic et al., 2007; La et al., 2013). Based on other research, our finding is consistent with a body of evidence that suggests poor micronutrient status increases CL±P risk (van Rooij et al., 2003; Munger et al., 2004; Tamura et al., 2005). An alternative explanation is that food insecurity is a marker of stress. Stressful life events have been linked to an increased risk of oral clefts, but there is little research on the effects of chronic stress for which food insecurity could be a proxy (Carmichael and Shaw, 2000; Goenjian et al., 2011).

Our finding that passive smoke exposure increases oral cleft risk is consistent with a recent meta-analysis that found passive smoke is associated with a 1.5-fold increase in risk of oral clefts (Sabbagh et al., 2015). Maternal smoking is the best established environmental risk factor for oral clefts. That passive smoke exposure increases oral cleft risk is not as widely appreciated. Public health approaches to reducing environmental exposure may help reduce the incidence of oral clefts. The national prevalence of smoking in Thai women over 15 years of age (1.9%) (World Health Organization, 2015) and the smoking prevalence in the subset of Thai mothers with infants (0.3%) (Anuntaseree et al., 2008) is consistent with the proportion who reported maternal smoking (1/190 or 0.5%) in our study, suggesting underreporting of maternal smoking was likely minimal. Future research should examine other types of smoke exposure, including e-cigarettes or cooking smoke prevalent in many low-resource settings.

Toenail zinc is a unique biomarker previously used in studies of cardiovascular disease, cancer, and neural tube defects (Milunsky et al., 1992; Platz et al., 2002; Martin-Moreno et al., 2003). We are the first to use this measure in a study of oral clefts. We found no association between toenail zinc concentrations and CL±P risk. Studies conducted in the Philippines, Poland, and the Netherlands report that low blood zinc concentrations increased oral cleft risk (Krapels et al., 2004; Tamura et al., 2005; Hozyasz et al., 2009). The only biomarker study in the United States, conducted in Utah, showed plasma zinc concentrations were not associated with oral cleft risk. The other U.S. studies, which estimated concentrations from an FFQ, lacked statistical precision (Shaw et al., 2006; Munger et al., 2009). The type of zinc measure (plasma, whole blood, toenail, FFQ) and time since the oral cleft developed, which varies across studies ranging from approximately 1 to 6 years, may introduce variation into measures that yield different findings, as could other methodological differences (Krapels et al., 2004; Tamura et al., 2005; Hozyasz et al., 2009). Another possible explanation is that zinc measures in the periconceptional period (which our measure likely approximates), do not approximate zinc concentrations after birth.

There are strengths and limitations to our study. Few studies have examined risk factors we study, such as various types of liver intake, food insecurity, and passive smoke exposure outside high-resource countries. Our findings provide avenues for future research on understudied environmental risk factors that may inform future prevention. In terms of limitations, several of the observed associations in our study lacked statistical precision as evidenced by the wide confidence intervals around our estimates, likely because of our relatively small sample size. Income and number of pregnancies tended to be the strongest confounders, and we observed large differences in unadjusted and adjusted estimates, which adds uncertainty to our estimates. We are unable to make inference from findings such as menstrual regulation supplement and vitamin use and suggest caution in interpreting estimates. Certain associations observed in our prior study (e.g., diabetes, maternal smoking) were unable to be examined in the current study because we had too few mothers in our study who reported having these conditions or behaviors.

With emphasis and funding placed on new, unexplored areas of study, there is often little replication of studies in similar contexts. Replication in similar populations and settings in identifying risk factors for oral clefts is an essential and currently an underappreciated part of the scientific process that is needed to ensure the inference we make and public health measures taken are well-justified. Large studies are needed to sort out the role of liver consumption, menstrual regulation supplement use, and other nutritional and environmental risk factors for oral clefts. Large studies with a genetic component could potentially identify gene–environment interactions. A multisite study that uses the same, rigorous methodology in different contexts could address the heterogeneity that exists across single site study findings, and illuminate environmental and genetic risk factors for oral clefts on a global scale.