Obesity at age 20 years was in this large study of women associated with increased risk of irregular periods and failing to become pregnant even if trying for one straight year, but not with the risk of experiencing at least one miscarriage.
As we are exploring relationships between body mass index at age 20 and reproductive problems, we have restricted the analytical sample to women at an age (aged 40 and above) when they most likely will have experienced reproductive problems if they will ever do so, particularly if these problems should have clinical consequences. As detailed above, we did not include irregular periods or failing be become pregnant before the age of 20 as an outcome in the study, only problems after the age of 20. We do not, however, know when the women experienced her (first) miscarriage; this may have happened as a teenager. Due to the mean age of the included women (nearly 60 years), modern treatment for infertility (like in vitro fertilization) has played a minor role for our findings.
The relationship between BMI at age 20 and irregular periods and problems of becoming pregnant may be explained by an increased risk of oligo- and anovulation in obese women and a number of other adverse effects of obesity on reproductive physiology in women [
1–
3,
25]. Our results support previous findings of a U-formed relationship between body mass index at age 23 and problems of becoming pregnant and menstrual problems before age 33. Even obesity at age 7 may be a predictor for the latter [
18].
Data from a self-administered questionnaire hamper the possibilities for further discussion with regard to etiology. One explanation may be that a relatively high percentage of the obese women may have had polycystic ovary syndrome (PCOS). Polycystic ovary syndrome affects 5–10 % of women in a general population, and many (at least one out of three) of the patients with PCOS are obese [
26,
27].
No relationship was found between obesity and the odds of reporting one or more miscarriages. Previous population-based studies have not given a consistent picture with regard to body mass index as a risk factor for miscarriages; underweight may be just as important as obesity [
6–
12]. Traditionally, PSOS has been thought to play a major role also for the risk of recurrent miscarriages [
28], but this has recently been questioned [
29].
The lack of relationship between body mass at age 20 and miscarriages may to some extent be due to misclassification as early miscarriages often are overlooked and any relationship will tend to be attenuated. However, the positive, direct relationship between the number of miscarriages and the odds of reporting failing to become pregnant indicates that information about the miscarriages has some validity and may further suggest that some women have interpreted the question “Did you ever try for one straight year or more to become pregnant and, during that time, not become pregnant?” less precisely than was the intention, answering that they for one straight year or more were not able to get pregnant or give birth to a live born child.
Given that there is no, or only a very weak, relationship between obesity and the risk of a miscarriage and that some women have interpreted the question as suggested above, this points to a possible stronger relationship between obesity (and possible underweight) and failure to become pregnant than the results presented in Table may indicate.
Whereas irregular periods and failing to become pregnant reflect problems of becoming pregnant, the women who have a miscarriage have conceived and are therefore fertile. Thus, according to our findings, weight may be more important for becoming pregnant than for remaining pregnant. When the relationship with failing to become pregnant was adjusted for irregular periods and miscarriages, the odds ratio associated with obesity was attenuated. However, it is debatable whether it is correct to adjust for these variables which most likely are on the causal pathway. If not, obesity (BMI ≥ 30 kg/m2) increases the odds of having problems of becoming pregnant with approximately 45 %, even after adjustments for other likely confounders.
Table shows that women who were obese at age 20 were less likely to ever have been married. We adjusted for marital status in all analyses (Table ) and the stratified analyses by ever married status (Table ) clearly demonstrate that the relationship we found is not due to women who never were married and therefore may not have tried to become pregnant. One might assume that experiencing irregular periods, the variable most strongly related to obesity, was independent of marital status, but the risk was found to be higher in never married women. We have, however, adjusted all relationships for marital status.
Our study has some limitations. There is a positive relationship between BMI in spouses [
15,
30]. Obese men have reduced fertility [
1,
31]. Thus, the higher odds for reporting problems of becoming pregnant in obese women may be due to obesity in the male partner. We are not able to link spouses in our database. It is however unlikely that the positive (and stronger) association (Table ) between obesity and irregular periods is related to male obesity.
The information from the women did not make it possible to differentiate between a miscarriage (a spontaneous loss of a fetus before the 20th week of pregnancy) and a stillbirth (a delivery after 20 completed weeks’ gestation of a fetus showing no signs of life) [
32]. The former is much more frequent, and our results will pertain largely to miscarriages. Wilcox et al. [
5] found that 31 % of pregnancies were lost, two out of three before the pregnancy was detected clinically. Currently <1 % of all pregnancies in the US end as a stillbirth, but the percentage is higher in blacks than in other ethnic groups [
32]. The risk of a stillbirth was higher during the childbearing years of the women included in our analysis, though. Obesity has in most studies been found to increase the risk of stillbirths [
32,
33].
The women were asked to state their current height. We have used this height when computing the body mass earlier in life, at age 20. Thus, our analyses have most likely somewhat overestimated the BMI at age 20. However, as the associations we found were basically independent of age at enrollment (aged 40–54, 55–69, or ≥70, and thus time since the women were 20 years old), little bias is introduced when applying current height when computing BMI earlier in life.
The main weakness of our study is that weight is self-reported and recalled. Underweight women tend to overestimate the self-reported weight whereas obese women underestimate it [
34]. Thus, in women with BMI < 18.5 kg/m
2, the reported BMI is probably higher than the true BMI and the opposite is true for obese women. However, the most important in our context is the ability to rank the women according to BMI and measured and self-reported BMI has been found to be highly correlated (r
s = 0.94) in this population [
35] as in the previous Adventist Health Study (AHS-1) [
36].
The mean age at enrolment was 59.9 years, and the women were asked to recall their weight nearly 40 years earlier, at age 20. Data from the Nurses’ Health Study [
37] indicate that women are able to recall their weight at age 18, the correlation coefficient between recalled and measured weight was 0.87. The women in the NHS cohort (aged 25–42) were, however, significantly younger than in our study. Data from women who took part in both this Adventist Health Study (AHS-2) and the former one (AHS-1 in 1976) demonstrate strong correlations (r = 0.82 for women of all ages) between recalled weight in the 1970s and weight stated in the questionnaires in 1976 [
38]. Thus, misclassification of recalled BMI at least in terms of relative rank appears to be quite small for recall of 25–30 years.
We do not find it likely that our results can be explained by differential recall of weight at age 20 as this would imply a strong correlation between reporting problems of becoming pregnant or, in particular, irregular periods and falsely recalled overweight and obesity when aged 20 years old.
We only have data concerning weight and height and thus BMI. It is probable that a more relevant measure is the percentage of body fat, a measure that was strongly correlated (r = 0.84) with BMI in US women aged 20–39 [
39]. Information about adipose tissue distribution, like waist circumference or waist/hip-ratio, may have given additional information, although the correlation between BMI and waist circumference in relatively young women is high (r = 0.93) according to recent NHANES data [
39].
One possible source of bias would be that women who complete the lifestyle questionnaire are survivors. Obese, relatively young, women have higher mortality than women with normal weight [
40,
41]. However, the mortality in women aged less than 40 is low, particularly in this relatively healthy group of subjects with low smoking prevalence, and the relationships did not depend on the age of the women when completing the questionnaire (Table and web tables 1–3). Thus, it is unlikely that survival bias has impacted on our results to any measurable degree.
The prevalence of obesity at age 20 (2.4 %) is relatively low, but it is for instance similar to the prevalence of obesity in women included in the SWAN cohort which was based on women aged 17–18 years old in the late 1960s [
42], Furthermore, the study population is somewhat selected as all the women were Adventists. It could be that obesity is associated with irregular periods and miscarriages differently in this group of women than in the general population. However, we find this unlikely and the stratified analyses did not suggest any interaction with lifestyle.
The study has, however, significant strengths. It is large in terms of women included, which has allowed detailed stratified analyses. The main findings were found to be very consistent in the different strata of the population. Another related strength is that this study has been conducted in a rather unique US population with a relatively high proportion of women who have abstained from alcohol and smoking for their entire life.
Additionally, 25 % of the analytical population are black Adventists of US and Caribbean origin; approximately 90 % of the remaining 75 % are white, non-Hispanic women. Both underweight and obesity was associated with being black (Table ). After adjustment for age and marital status and compared to other women, black women were only slightly more likely to report irregular periods (2 %) or failing to become pregnant even if trying for least 1 year (6 %). However, blacks were more likely to report at least one miscarriage [OR = 1.34 (95 % CI: 1.27–1.40)]. However, as detailed in Table and the web appendix, stratified analyses demonstrate that there are few indications that ethnicity has influenced our findings significantly.
In summary, this large study found that women who were obese when they were 20 years old were at a significantly increased risk of failing to become pregnant even if trying for one straight year. One of the explanations for this seems to be that obese women have difficulties to conceive due to irregular periods, rather than increased risk of miscarriages.
Synnøve F. Knutsen, Keiji Oda, and Gary E. Fraser