Comparison with previous studies

We have previously examined the relationship between folic acid or multivitamin supplement use and infant NTD^{(Reference Gildestad, Øyen and Klungsøyr33)}, oral clefts^{(Reference Gildestad, Bjørge and Vollset34)} and heart defects^{(Reference Leirgul, Gildestad and Nilsen35)} from the same source population (MBRN). There were several differences between the previous three studies and the current one. For example, the current paper separates total and isolated NTD. More than 20 % of infants with NTD had associated anomalies. We observed associations between supplement use and some of the anomalies that co-occurred with NTD (limbs, urinary, digestive defects, abdominal wall defects). Yet, the association measures were nearly identical for total and isolated NTD. A 2015 Cochrane review assessed the effects of folic acid supplementation on birth defects^{(Reference De-Regil, Pena-Rosas and Fernandez-Gaxiola19)} and concluded that folic acid, alone or in combination with vitamins and minerals, prevents NTD, with high-quality evidence according to Grades of Recommendation Assessment, Development and Evaluation (GRADE) criteria^{(Reference Guyatt, Oxman and Vist36)}. However, there was insufficient evidence for other birth defects, motivating efforts to quantify the associations between use of folic acid and birth defects other than NTD^{(Reference Guyatt, Oxman and Kunz37)}. Supplement use was inversely associated with several of the birth defects that co-occurred with oral clefts. A 2018 systematic review from observational studies concluded that folic acid supplementation reduced the risk of non-syndromic oral clefts^{(Reference Jahanbin, Shadkam and Miri38)}, whereas in the 2015 Cochrane report, there were insufficient data from randomised controlled trials to evaluate the effects on oral clefts^{(Reference De-Regil, Pena-Rosas and Fernandez-Gaxiola19)}.

The association between folic acid and multivitamin supplement use and congenital heart defects has been reported in similar studies from Norway and Denmark^{(Reference Øyen, Olsen and Basit20,Reference Leirgul, Gildestad and Nilsen35)} . In the study by Leirgul et al. ^{(Reference Leirgul, Gildestad and Nilsen35)}, individuals with heart defects were identified from four data sources, including MBRN, whereas the present study used data from the MBRN updated with additional 4 years. Yet, the finding in the present study (aRR 1·05, 95 % CI 0·98, 1·12) was similar to the previous paper (aRR 1·10, 95 % CI 1·03, 1·16). Further, in two national birth cohorts in Norway and Denmark, with prospective data from nearly 200 000 births and 2247 individuals with heart defects, folic acid supplement use was not associated with heart defects (aRR 0·99, 95 % CI 0·80, 1·22)^{(Reference Øyen, Olsen and Basit20)}. These findings were also in accordance with the Cochrane review from 2015^{(Reference De-Regil, Pena-Rosas and Fernandez-Gaxiola19)}, albeit with low-quality evidence according to GRADE.

In our study, use of folic acid or multivitamins was associated with a reduced risk of abdominal wall defects, including both gastroschisis and omphalocele. This corresponds well with the literature^{(Reference Botto, Mulinare and Erickson39–Reference Paranjothy, Broughton and Evans41)}, although there are exceptions^{(Reference Godwin, Sibbald and Bedard42,Reference Feldkamp, Carmichael and Shaw43)} .

Use of folic acid supplements before and during pregnancy was associated with reduced risk of infant hypospadias in our study, but we did not observe associations with multivitamins (Table 4). A Hungarian case–control study suggested a dose-dependent relationship between folic acid supplement use and risk for infant hypospadias^{(Reference Mavrogenis, Urban and Czeizel44)}. The folic acid doses were considerably higher than in our study, and however, our findings suggested similar associations. In a systematic review by Goh et al., no associations between multivitamin supplements and hypospadias were found from the selected cohort studies and randomised controlled trials^{(Reference Goh, Bollano and Einarson45)}.

We observed that the use of multivitamins and/or folic acid was associated with reduced risk of limb defects, with the largest risk reduction for the limb reduction subgroup (Table 3). Several studies have shown similar results, with risk reductions of about one-third^{(Reference Shaw, O’Malley and Wasserman46,Reference Wolf, Hegaard and Huusom47)} . However, the quality of the evidence has not been convincing. We found a non-significant, but somewhat lower point estimate for limb reduction defects among infants of mothers using multivitamins only than among infants of folic acid only users (Table 4). The association nearly disappeared for use of folic acid combined with multivitamins. There was limited power to detect associations due to the small number of infants in the multivitamin group. In contrast to limb reduction defects, we found that the use of folic acid combined with multivitamins was associated with a reduced risk of infant clubfoot, while multivitamin use was not. One study has found that elevated plasma homocysteine levels were associated with increased risk of clubfoot^{(Reference Vollset, Refsum and Irgens48)}, which suggest that folate insufficiency contributes to the risk.

Use of folic acid supplements was associated with an increased risk of urinary system defects, mainly driven by the large proportion of infants with hydronephrosis in this category. This corresponds well with the literature^{(Reference Godwin, Sibbald and Bedard42,Reference Groen In ’t Woud, Renkema and Schreuder49,Reference Canfield, Collins and Botto50)} . However, the association was not present for infants where mothers used multivitamins only, which does not correspond with previous findings^{(Reference Botto, Olney and Erickson17,Reference Wolf, Hegaard and Huusom47)} . According to EUROCAT, there are discrepancies in diagnostic criteria for hydronephrosis cases^{(Reference Morris, Springett and Greenlees51)}, and registration of hydronephrosis is recommended only if the renal pelvis is 10 mm or more after birth, while the MBRN does not specify this distinction. Further, the ascertainment of hydronephrosis in the MBRN may be lower than that of birth defects that are visible after birth.

Study strengths and limitations

Major strengths of our study include the prospective design as well as the large study population of almost 900 000 infants, with detailed information on maternal periconceptional supplement use, more than 18 000 infants with birth defects, and information on several potential confounders. With a study of this magnitude, the effect size across all birth defects may be highly statistically significant even if associated with small deviations of the relative risk from 1. For example, it is possible to detect a 10 % risk reduction of a rare event, which may seem small at the level of the individual if the focus is on the clinical importance of the prevention strategy. One should bear in mind, however, that a small relative risk may have an important impact from a public health perspective.

Our study had some limitations. First, misclassification of exposure may have biased the association measures. The exposure information was in most cases gathered early in pregnancy, which reduces the risk of maternal recall bias, because mothers do not know at this point the birth defect status of their babies. Errors in reporting by the hospitals concerning supplement type and timing may have occurred. The group of non-users was a mixture of mothers who did not use supplements and mothers where the hospitals did not report the use (missing). This non-differential misclassification of the exposure is considered to be present and could have biased the association measures towards the null. According to a recent study from the MBRN that explored the epidemiology of limb reduction defects, including their relation with maternal use of folic acid/multivitamin supplements, approximately 12 % of all births (excluding TOPFA) had missing information on folic acid and multivitamin supplement use during 1999–2016^{(Reference Klungsøyr, Nordtveit and Kaastad52)}. When multiple imputation was applied to handle missing data on folic acid and multivitamin supplement use, the estimate was close to our finding for limb reduction defects (aRR 0·7, 95 % CI 0·6, 0·9)^{(Reference Klungsøyr, Nordtveit and Kaastad52)}.

There is a risk of misclassification of birth defects in the MBRN, with varying ascertainment for the different defects. Validation studies have been performed for certain birth defects (oral clefts^{(Reference Kubon, Sivertsen and Vindenes53)}, gastroschisis^{(Reference Kazaura, Lie and Irgens54)} and Down syndrome^{(Reference Melve, Lie and Skjaerven55)}) showing satisfactory results. However, for oral clefts, the ascertainment depended on the visibility and severity of the cleft, indicating that some birth defects may be under-reported to the MBRN. For example, it is likely that the diagnosis of hydronephrosis reported to the MBRN depends on a prenatal diagnosis by ultrasound, and many women only have the routine ultrasound around 18 gestational weeks.

Second, the lack of information on supplement use in TOPFA could have introduced selection bias. If women who terminated a pregnancy for a birth defect were less likely to use folic acid, the associations would be underestimated. However, if women who used folic acid were more likely to opt for a termination of pregnancy due to birth defects, it would bias the risk ratio towards the null. This situation could be plausible if increased folic acid supplement use was associated with better prenatal care, which in turn could be associated with a prenatal diagnosis of birth defects. However, in Norway, prenatal screening with ultrasound is free of charge and offered to all pregnant women at 18 gestational weeks or earlier for women with a high-risk pregnancy. Therefore, such differential bias is not likely. Except for nervous system defects (51·4 % TOPFA) and omphalocele (37·2 %), most pregnancies affected by the described birth defects were continued to delivery. Selection bias is thus mostly a concern for the association between supplement use and NTD and omphalocele. Women who use IVF are more likely to have a multiple gestations and are, by definition, pregnancy planners and therefore much more likely to use folic acid^{(Reference Vollset, Gjessing and Tandberg56)}. Birth defects occur more frequently in infants from IVF^{(Reference Hansen, Kurinczuk and Milne57)} and multiple^{(Reference Glinianaia, Rankin and Wright23)} pregnancies as compared with naturally conceived infants and singletons. However, birth defects from IVF pregnancies only accounted for 3·6 % of total birth defects, and the associations were nearly unchanged in analyses restricted to singletons. Thus, selection bias due to IVF and multiple gestations could be ruled out.

Third, we were able to adjust for a priori selected confounders known to be associated with birth defects and supplement use. Adjustment did not change the estimates much. However, stratification by time period showed that from 2006 onwards, associations were stronger for some defects. A possible explanation for this finding could be that more women complied with the recommendations in the latest time period. The reported prevalence of supplement use before and during pregnancy in the two time periods was 13·8 % (1999–2005) and 28 % (2006–2013). Analyses conditioned on complete smoking information produced similar results as when assigning missing to a separate category. Adjusting for year of birth slightly weakened the association measure (aRR total birth defects without adjusting for year of birth: 0·90, 95 % CI 0·87, 0·94), possibly due to differences in the prevalence of supplement use, and natural variations in outcome prevalence over the study period. The observed associations in our study could be due to unmeasured confounding. We had no information on maternal BMI^{(Reference Persson, Cnattingius and Villamor58)} or baseline diet^{(Reference Carmichael, Yang and Feldkamp59)}.

The exact mechanism of why folic acid supplementation prevents NTD or other birth defects is not known, but associations have been established with erythrocyte folate concentration^{(Reference Bailey and Hausman60)}. Women below the cut-off for what is considered optimal erythrocyte are thought to have elevated risk of having babies with NTD^{(Reference Botto and Mastroiacovo61)}. Folate functions as a coenzyme in one-carbon metabolism during the methylation cycle and is crucial for DNA synthesis, and during pregnancy, there are higher rates of one-carbon transfer reactions^{(Reference Gernand, Schulze and Stewart4)}. The randomised controlled trials establishing a preventive role for NTD were specific to folic acid. Dietary folate is less bioavailable than folic acid, and consumption of dietary folates is considered to result in a significantly smaller increase in erythrocyte folate concentration than folic acid^{(Reference Cuskelly, McNulty and Scott62,Reference McNulty and Pentieva63)} . Since food fortification with folic acid is not mandatory in Norway, the prevention strategy with a supplementation programme requires a high proportion of planned pregnancies and high compliance with the programme. Many Norwegian women are not aware of the importance of supplementation before pregnancy and have diets that do not meet the recommendations. According to MBRN online statistics (http://statistikkbank.fhi.no/mfr/), maternal use of folic acid supplements before pregnancy was reported for around a third of infants in 2018. The proportion of women who use folic acid supplements before conception is even lower in some subgroups of the population^{(Reference Nilsen, Daltveit and Iversen64)}. Since the prevalence of folic acid supplement use remains low, efforts to increase supplement compliance are needed.

Our findings suggest that some birth defects may be prevented by maternal use of folic acid or multivitamins, but that the extent of the risk reduction differs for the different types of birth defects and type of exposure (multivitamin or folic acid). Better definitions of the exposures regarding specific supplement types, timing and dosage would be beneficial in understanding the different roles of specific micronutrients in the aetiology of birth defects. Thus, in unplanned pregnancies, even though the time window to prevent NTD is closed, initialising supplementation during pregnancy could be beneficial to reduce defects that occur later in the first trimester. Since our study suggests that folic acid or multivitamin supplementation could have protective associations for other birth defects besides NTD, the prevention potential may be larger than anticipated. With an annual birth rate of nearly 60 000 and a birth defect prevalence of about 2 %, folic acid supplements used by the 40 000 women reported as non-users could prevent birth defects in nearly fifty infants annually in Norway, assuming the protective associations of 6 % risk reduction in our study were causal.

In conclusion, use of folic acid and/or multivitamin supplements before and during pregnancy was associated with a small reduction in the risk of total birth defects among live births and stillbirths. More specifically, the risk reduction was observed for abdominal wall defects, genital defects and limb defects. Folic acid supplement use was associated with increased risk for hydronephrosis. Protective associations were suggested for supplement use and NTD, respiratory system defects and digestive system defects although CI included the null value of 1.