Lifestyle and Pregnancy Loss in a Contemporary Cohort of Women Recruited Prior to Conception, LIFE Study

Abstract

Objective

To estimate pregnancy loss incidence in a contemporary cohort of couples whose lifestyles were measured during sensitive windows of reproduction to identify factors associated with pregnancy loss for the continual refinement of preconception guidance.

Design

Prospective cohort with preconception enrollment.

Setting

16 counties in Michigan and Texas.

Patients

344 couples with a singleton pregnancy followed daily through 7 post-conception weeks of gestation.

Interventions

None. Daily, couples recorded use of cigarettes, caffeinated and alcoholic beverages, and multivitamins. Women used fertility “monitors” for ovulation detection and digital pregnancy tests. Pregnancy loss was denoted by conversion to a negative pregnancy test, onset of menses or clinical confirmation depending upon gestation. Using proportional hazards regression and accounting for right censoring, we estimated adjusted hazard ratios and 95% confidence intervals (aHR; 95% CI) for couples’ lifestyles (cigarette smoking, alcoholic and caffeinated drinks, multivitamins) during three sensitive windows: preconception, early pregnancy and periconception.

Main outcomes measured

Incidence and risk factors for pregnancy loss.

Results

98/344 (28%) women with a singleton pregnancy experienced an observed pregnancy loss. In the preconception window, loss was associated with female age ≥35 years (1.96; 1.13-3.38) accounting for couples’ ages, female and male consumption >2 daily caffeinated beverages (1.74; 1.07-2.81 and 1.73; 1.10-2.72, respectively), and females’ vitamin adherence (0.45; 0.25-0.80). The findings were similar for lifestyle during the early pregnancy and periconception windows.

Conclusions

Couples’ preconception lifestyle factors were associated with pregnancy loss, though women’s multivitamin adherence dramatically reduced risk. The findings support continual refinement and implementation of preconception guidance.

Introduction

Over the past few decades, the detection of pregnancy for many women not receiving infertility treatment has shifted from clinical settings to the privacy of the home. This change is largely attributed to the development of a human chorionic gonadotropin (hCG) beta-subunit radioimmunoassay (1), leading to the availability of the first early pregnancy test in 1976. Currently, there are ≈60 home pregnancy test kits available in the U.S. accounting for approximately $228 million in annual sales (2,3). Home pregnancy tests now generate more revenue than any other over-the-counter home diagnostic test, including those for blood pressure or glucose testing (4), and are used by approximately 85%-90% of women for identifying pregnancy (5).

The shift in pregnancy detection from the clinical setting to the home means that women may now recognize more pregnancies than they did when clinical identification was the norm, usually occurring after 1-2 missed periods. To date, only a few prospective cohort studies with preconception enrollment of women or couples have ever been conducted worldwide to estimate the incidence of pregnancy loss. Overall, loss rates range from a low of 8% among 92 pregnant women in Newcastle, UK (6) to a high of 52% among 118 pregnant women in Southampton, UK (7). Recent data from the U.S. National Survey of Family Growth noted an increasing pattern of self-reported miscarriage between 1970 and 2000, with much of the increase being for losses in the first 7 weeks of gestation, possibly reflecting women’s use of home pregnancy testing (8). While various risk factors have been identified from past research ranging from health and pregnancy history to lifestyle, recent guidance from the American College of Obstetricians and Gynecologists notes the absence of any known interventions to prevent pregnancy loss except for anti-phospholipid syndrome and progesterone supplementation for recurrent losses (9).

Another notable societal change in pregnancy is the shift in thinking from pregnant females to pregnant couples, as evident by colloquial terms such as “we are pregnant” along with the early and sustained involvement of male partners during prenatal care and delivery in many cases. Despite such changes, research has traditionally focused on characteristics of either the female or male partner relative to pregnancy loss but not both in light of few couple-based cohorts.

We sought to estimate the incidence of pregnancy loss and lifestyle risk factors in a contemporary cohort of couples desiring pregnancy who were recruited prior to conception and followed throughout trying and pregnancy. This design facilitated the capture of early pregnancy and losses consistent with the time period in which most losses occur. Pregnancy loss is an endpoint indicative of reproductive and/or developmental toxicity that may be associated with a variety of environmental exposures. Another unique aspect of this design is the ability to assess potential male-mediated risk factors for pregnancy loss such as age, employment and use of cigarettes, alcohol, and caffeine in the context of women’s lifestyle (10-13). As such, this approach goes beyond most of the available research conducted to date that focuses exclusively on female determinants and offers insight for preconception guidance that is increasingly focusing on both men and women.

Material and Methods

Study Design and Population

The study cohort comprised 347/501 (69%) women who had an observed pregnancy while participating in the LIFE Study, which was designed and powered to examine the association between environmental and lifestyle factors and fecundity impairment among couples, including pregnancy loss. We excluded three couples with twin pregnancies resulting in a cohort comprising 344 couples with singleton pregnancies. The LIFE Study utilized population-based sampling frameworks to recruit couples discontinuing contraception for purposes of becoming pregnant from 16 counties in Michigan and Texas. By design, eligibility criteria were minimal and included: 1) in a committed relationship; 2) ability to communicate in English or Spanish; 3) females aged 18-40 and males aged ≥18 years; 4) females’ menstrual cycles between 21-42 days as required by the fertility monitors; 5) no history of injectable hormonal contraception in past year or currently breastfeeding for 6+ months; 6) no clinically diagnosed infertility in either partner; and 7) off contraception <2 months. Prior to enrollment, female partners’ urines were tested to ensure they were not pregnant. Human subjects approval was obtained from participating institutions, and all men and women gave written informed consent before data collection. Complete details about the population-based sampling framework utilized along with response rates have been previously published (14).

Data Collection and Follow-up

Couples were interviewed individually upon enrollment to ascertain socio-demographic, lifestyle, and medical history information, followed by the measurement of height and weight to calculate body mass index (BMI). The couple was then instructed in the completion of daily journals to record their lifestyle behaviors in a manner consistent with how people think about such exposures (e.g., number of cigarettes smoked per day, number of alcoholic and caffeinated beverages consumed per day, taking daily multivitamins). Females also reported sexual intercourse, menses, and home pregnancy test results. Couples completed journals daily until a positive home pregnancy test or up to 12 months of trying. Pregnant women continued journals daily through 7 post-conception weeks gestation then monthly journals until a loss or delivery. Journals were purposefully designed to be short to minimize participant burden and encourage daily adherence.

Standardized prompts were provided in the daily journals regarding lifestyle. Specifically, one alcoholic drink equaled one can or bottle of beer, a glass of wine, a shot of liquor, or a mixed drink. A caffeinated drink equaled a cup of coffee, tea, or can of soda (excluding decaffeinated or caffeine free beverages). Vitamins included over-the-counter multivitamins and prescription prenatal vitamins, and were dichotomized relative to daily usage (yes/no). Couples reported the daily number of cigarettes smoked and alcoholic and/or caffeinated beverages consumed, and were instructed to record a zero if none was used. Couples had the option of reporting data via the web or in returning weekly journal cards consistent with prospective measurement.

Home Fertility and Pregnancy Testing

Women were trained in the use of the Clearblue® Fertility Monitor (Inverness Medical Innovations, Waltham, MA), a urinary test kit that tracks estrone-3-glucuronide and luteinizing hormone (LH) to predict the day of ovulation and help couples time intercourse to maximize their chances of conceiving. In comparison to the gold standard of ultrasonography for the detection of ovulation, the monitor is reported to accurately detect the LH surge (99%) (15). Women also were trained in the use of the Clearblue® digital pregnancy test, which has demonstrated sensitivity and reliability for detecting 25 mIU/mL of hCG, and demonstrated accuracy by women (16). Women tested their urine for pregnancy on the day they expected menstruation consistent with manufacturer’s guidance. Pregnancy loss was defined as conversion to a negative pregnancy test, clinical confirmation, or onset of menstruation depending upon gestational dating and included two women with ectopic pregnancies. Three twin pregnancies were excluded from analysis.

Statistical Analysis

The cohort was first characterized by pregnancy outcome to identify factors associated with loss, with significance based upon either the Chi-square or the Kruskal-Wallis tests. We estimated the cumulative incidence of pregnancy loss following conception using Kaplan-Meier techniques, while accounting for right censoring (withdrawals), and the time required to become pregnant. For pregnancy cycles with missing fertility monitor data required for estimating the day of ovulation (n=59; 17%), we assumed ovulation to be 14 days prior to the first positive pregnancy test. This assumption is consistent with the secretory phase being less variable than the proliferative phase of the cycle (17). We defined time-to-pregnancy loss as the number of observed days between ovulation or the monitor’s peak fertility day (LH surge) and the date of the reported loss. The peak (LH) monitor day was assumed to be the day of conception, given the ovum’s short (≈24 hours) interval for fertilization.

A priori, we defined three sensitive windows for ascertainment of couples’ lifestyles and time-to-pregnancy loss: 1) preconception window, time from enrollment to estimated conception; 2) early pregnancy window, the day following conception (or peak LH) through 7 post-conception weeks and 3) periconception window, inclusive of the two windows as an overall summary window. Thereafter, women converted to monthly reporting for the remainder of pregnancy through loss or delivery.

To ensure proper model specification, we thoroughly inspected spaghetti plots for each lifestyle by partner during each window and modeled lifestyles as daily averages over the observed time in each window. Since only women completed early pregnancy journals, we assumed male partners’ preconception lifestyle remained the same during early pregnancy. To maintain a couple based analytic framework, we included the male partner in the estimation of HRs for early pregnancy and periconception along with the prospectively measured female exposures.

We first modeled each lifestyle behavior separately during each of the sensitive windows then included potential confounders that were identified a priori based on a review of the literature: female age (18), couples ages modeled as the difference (years) given their known correlation and reports of male age being associated with loss (10), couples’ BMI (continuous) (19,20) sexual frequency (21), and prior history of loss conditional on gravidity (22). Using Cox’s proportional hazards model (23), we modeled time to loss (post-conception days) to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) after first assessing potential non-linear relationships using splines. Despite assessing non-linear relations for all behaviors and characteristics (e.g., age, alcohol, BMI) with pregnancy loss, we only observed threshold effects only for female age (≥35 years) and male and female caffeinated beverage consumption at ≈2 daily cups depending upon the sensitive window. We modeled these risk factors, accordingly. We estimated the average number of daily cigarettes smoked and alcoholic drinks consumed to fully utilize all daily information in keeping with pregnancy being cycle based. Also, averages were thought to be amenable to guidance rather than to simply model totals during the windows. Multivitamin usage was modeled in terms of adherence or the proportion of days during each sensitive window the female or male partner reported taking vitamins, ranging from 0 (no vitamins taken) to 1 (vitamins taken daily). Intercourse frequency was modeled as the average number of acts during the pregnancy cycle. Separate models were run for each of the three time intervals in anticipation that couples may change their behaviors or lifestyles when becoming pregnant. We empirically assessed such lifestyle changes using paired T-tests.

Results

Ninety-eight (28%) of the 344 couples who became pregnant during the study experienced a pregnancy loss. Among women becoming pregnant, the cumulative incidence of pregnancy loss following the estimated day of conception increased during early pregnancy then plateaued (≈72 days) with 1% of losses occurring in the first 10 days, and increasing to 15% by 30 days, and 28% by 126 days. All losses occurred before 22 weeks gestation. Notably, time-to-pregnancy as prospectively ascertained was not associated with pregnancy loss (Figure 1, panels a and b, respectively; p=0.97).

Figure 1.

Cumulative incidence and 95% confidence intervals of pregnancy loss by gestation (panel A), and the incidence and 95% confidence intervals of pregnancy loss by time-to-pregnancy in observed menstrual cycles (panel B).

Few significant differences emerged regarding baseline characteristics and pregnancy loss status among couples achieving an observed pregnancy with the exception of maternal age ≥35 years after accounting for the difference in couples’ ages given their correlated nature (r=0.70), and paternal education (≤high school versus higher) (Table 1). Certain lifestyles were significantly (p<0.05) associated with pregnancy loss across various sensitive windows (Figure 2). During the preconception window, male but not female partner’s caffeinated beverage consumption was positively (62% and 41%; p<0.001) associated with pregnancy loss, while female vitamin adherence was negatively associated (59% and 69%; p=0.02). During the early pregnancy and periconception windows, partners’ caffeine consumption was positively associated with pregnancy loss, while female partner’s vitamin adherence was negatively associated (p<0.001).

Table 1.

Comparison of couples’ socio-demographic characteristics by observed pregnancy loss, the LIFE Study (n=344).

Socio-demographic Characteristic	Pregnancy Loss (n=98) # (%)	No Pregnancy Loss (n=222) # (%)	Lost To Follow-up (n=24) # (%)
Female Partner
Age (years):*
≤29	48 (49)	118 (53)	18 (75)
30-34	31 (32)	82 (37)	3 (13)
≥35	19 (19)	22 (10)	3 (13)
Mean (±SD)	30.3 (±4.1)	29.6 (±3.8)	29.3 (±3.7)
Race/ethnicity:
White, Non-Hispanic	82 (85)	180 (82)	23 (96)
Black, Non-Hispanic	3 (3)	3 (1)	0 (0)
Hispanic	7 (7)	21 (10)	1 (4)
Other, Non-Hispanic	5 (5)	16 (7)	0 (0)
Education:
≤ High school	6 (6)	9 (4)	0 (0)
Some/college graduate	91 (94)	211 (96)	23 (100)
Health insurance:
No	6 (6)	6 (3)	3 (13)
Yes	91 (94)	214 (97)	21 (88)
Household income:
<50,000	11 (12)	30 (14)	3 (12)
50,000 - 99,999	53 (56)	96 (44)	14 (58)
≥100,000	30 (32)	91 (42)	7 (29)
Employed:
No	22 (24)	44 (20)	4 (17)
Yes	76 (78)	178 (80)	20 (83)
BMI:
Under/healthy ≤24.9	44 (45)	111 (50)	15 (63)
Overweight 25.0 – 29.9	23 (24)	61 (28)	4 (17)
Obese ≥30	31 (32)	49 (22)	5 (21)
Mean (±SD)	27.8 (±6.8)	26.6 (±6.6)	26. (±7.3)
Prior history of pregnancy loss:
Nulligravid	37 (38)	86 (39)	10 (42)
Gravid, with no prior history	36 (37)	93 (42)	12 (50)
Gravid, with prior history	24 (25)	42 (19)	2 (8)
Male Partner
Age (years):
≤29	28 (29)	86 (39)	9 (38)
30-34	40 (41)	84 (38)	9 (38)
≥35	30 (31)	52 (23)	6 (25)
Mean (±SD)	32.3 (±4.7)	31.4 (±4.6)	31.0 (±3.7)
Race/ethnicity:
White, Non-Hispanic	84 (87)	182 (82)	19 (79)
Black, Non-Hispanic	2 (2)	5 (2)	0 (0)
Hispanic	4 (4)	20 (9)	3 (13)
Other, Non-Hispanic	7 (7)	14 (6)	2 (8)
Education:**
≤ High school	7 (7)	5 (2)	6 (25)
Some/college graduate	90 (93)	215 (98)	18 (75)
Health insurance:
No	8 (8)	10 (5)	3 (13)
Yes	89 (92)	211 (96)	21 (88)
Household income:
<50,000	9 (10)	26 (12)	2 (8)
50,000 - 99,999	50 (54)	102 (46)	13 (54)
≥100,000	34 (37)	92 (42)	9 (38)
Employed:
No	3 (3)	6 (3)	0 (0)
Yes	95 (97)	216 (97)	24 (100)
Male BMI:
Under/healthy ≤24.9	14 (15)	46 (21)	3 (13)
Overweight 25.0 – 29.9	43 (45)	85 (39)	9 (39)
Obese ≥30	38 (40)	86 (40)	11 (48)
Mean (±SD)	29.8 (±5.3)	29.1 (±4.8)	30.2 (±4.9)
Previously fathered a pregnancy:
No	39 (40)	86 (39)	12 (50)
Yes	59 (60)	136 (61)	12 (50)

NOTE: Three twin pregnancies were uniformly excluded from all analyses. Pregnancy losses all occurred before 21 weeks.

Categorical measures were tested with chi-squared test, and continuous measures with the Kruskal-Wallis test. All p-values are two-sided.

^*P < 0.05

^**P < 0.0001

Figure 2.

Comparison of male and female partner’s lifestyle by pregnancy loss and sensitive window (preconception, early pregnancy, periconception).

** p 0.02 * p < 0.001

Several significant adjusted hazard ratios emerged for lifestyle in each sensitive window along with female age ≥35 years accounting for the couples’ ages (Table 2). Most notably, female caffeine consumption >2 daily caffeinated drinks remained significantly associated with pregnancy loss [preconception (aHR=1.74; 1.07-2.81), early pregnancy (aHR=3.05; 1.75-5.34), and periconception (aHR=2.58; 1.56-4.27)]. Similarly, male partners’ caffeinated beverage consumption also was significantly associated with loss [preconception (aHR=1.73; 1.10-2.72), early pregnancy (aHR=1.88; 1.21-2.93), and periconception (aHR=1.75; 1.12-2.74)]. Female vitamin adherence was associated with a reduced risk of loss in every sensitive window [preconception (aHR=0.45; 0.25-0.80), early pregnancy (aHR=0.21; 0.10-0.43), and periconception (aHR= 0.22; 0.11-0.44)]. None of the other lifestyle characteristics were significantly associated with pregnancy loss in any of the final adjusted models.

Table 2.

Lifestyle during sensitive windows of human reproduction and development and pregnancy loss - proportional hazard ratios, the LIFE Study (n=344).

Characteristic	Preconception		Early Pregnancy		Periconception (Preconception & Early Pregnancy)
	Unadjusted	Adjusted	Unadjusted	Adjusted	Unadjusted	Adjusted
	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)
Female age (≥35 vs. < 35 yrs.)	1.81 (1.09, 3.01)	1.96 (1.13, 3.38)	1.81 (1.09, 3.01)	1.85 (1.10, 3.13)	1.81 (1.09, 3.01)	1.95 (1.13, 3.35)
Difference in partners’ ages	1.02 (0.96, 1.08)	1.02 (0.96, 1.08)	1.02 (0.96, 1.08)	1.02 (0.96, 1.07)	1.02 (0.96, 1.08)	1.03 (0.97, 1.08)
Prior pregnancy, no prior loss1	0.92 (0.58, 1.45)	0.78 (0.48, 1.26)	0.92 (0.58, 1.45)	0.68 (0.42, 1.11)	0.92 (0.58, 1.45)	0.69 (0.43, 1.13)
Prior pregnancy, prior loss2	1.31 (0.79, 2.20)	1.15 (0.66, 2.00)	1.31 (0.79, 2.20)	1.05 (0.61, 1.81)	1.31 (0.79, 2.20)	1.03 (0.59, 1.80)
Female BMI	1.02 (0.99, 1.05)	1.01 (0.98, 1.04)	1.02 (0.99, 1.05)	1.00 (0.97, 1.03)	1.02 (0.99, 1.05)	1.01 (0.98, 1.04)
Male BMI	1.02 (0.98, 1.06)	1.01 (0.97, 1.06)	1.02 (0.98, 1.06)	1.01 (0.97, 1.06)	1.02 (0.98, 1.06)	1.02 (0.98, 1.06)
Female average cigarette smoking3	1.06 (0.99, 1.14)	1.01 (0.92, 1.10)	1.11 (1.03, 1.19)	1.02 (0.92, 1.13)	1.09 (1.01, 1.16)	1.01 (0.91, 1.11)
Male average cigarette smoking3	1.03 (0.99, 1.07)	1.01 (0.95, 1.07)	1.03 (0.99, 1.07)	1.00 (0.94, 1.07)	1.03 (0.99, 1.07)	0.99 (0.93, 1.06)
Female average alcohol consumption3	1.10 (0.79, 1.53)	1.12 (0.70, 1.79)	2.16 (1.16, 4.02)	1.65 (0.77, 3.54)	1.35 (0.86, 2.12)	1.19 (0.64, 2.23)
Male average alcohol consumption3	1.01 (0.82, 1.24)	0.97 (0.73, 1.28)	1.01 (0.82, 1.24)	0.95 (0.76, 1.20)	1.01 (0.82, 1.24)	0.97 (0.73, 1.27)
Female caffeinated beverage consumption (>2 vs. ≤2 daily cups)	2.25 (1.44, 3.53)	1.74 (1.07, 2.81)	3.58 (2.18, 5.86)	3.05 (1.75, 5.34)	2.79 (1.79, 4.36)	2.58 (1.56, 4.27)
Male caffeinated beverage consumption (>2 vs. ≤2 daily cups)	2.02 (1.34, 3.05)	1.73 (1.10, 2.72)	2.02 (1.34, 3.05)	1.88 (1.21, 2.93)	2.02 (1.34, 3.05)	1.75 (1.12, 2.74)
Female vitamin adherence4	0.58 (0.34, 0.96)	0.45 (0.25, 0.80)	0.25 (0.13, 0.49)	0.21 (0.10, 0.43)	0.35 (0.19, 0.63)	0.22 (0.11, 0.44)
Male vitamin adherence4	0.99 (0.57, 1.71)	1.27 (0.70, 2.30)	0.99 (0.57, 1.71)	1.15 (0.65, 2.03)	0.99 (0.57, 1.71)	1.36 (0.75, 2.47)
Average intercourse frequency5	1.00 (0.96, 1.05)	1.01 (0.96, 1.05)	1.00 (0.96, 1.05)	0.99 (0.95, 1.04)	1.00 (0.96, 1.05)	0.99 (0.95, 1.04)

NOTE: Models adjust for all characteristics simultaneously. Restricted to women with singleton pregnancies; 3 twin pregnancies were excluded from analysis. Males’ lifestyles for the early pregnancy and periconception windows were modeled based upon his reported usage in the preconception window (as he did not record lifestyle during early pregnancy).

¹No prior pregnancy loss versus no prior pregnancy.

²Prior pregnancy loss versus no prior pregnancy.

³Average daily number of cigarettes smokes or alcoholic drinks consumed.

⁴Adherence denotes the proportion of days vitamins taken, ranging from 0 (none) to 1 (daily), during each sensitive window.

⁵Average number of intercourse acts during the pregnancy cycle.

Discussion

In our couple-based prospective cohort that was enrolled prior to conception, the incidence of pregnancy loss among couples who achieved a recognized pregnancy during the study was 28%, which is comparable (25% to 31%) to earlier estimates from prior preconception cohort studies that utilized urinary hormonal profiles to define pregnancy loss (24-26). However, our incidence is notably higher than that reported for a prospective occupational cohort of couples recruited prior to conception who were followed for 6 months (17%; 12%-22%) (27). Time-to-pregnancy was not associated with pregnancy loss in our study affirming recent data from a large pregnancy cohort study (28). Our incidence estimate relative to earlier cohorts does not support an increased incidence of pregnancy loss as suggested by survey data (8). Still, the relatively high incidence underscores the inefficiency of human reproduction (29).

While female age has long been associated with pregnancy loss, we observed a threshold effect at ≥35 years even when accounting for the difference in the couples’ ages. The extent to which this finding reflects aged gametes or cumulative environmental exposures inclusive of lifestyle requires further investigation. Couples’ lifestyle behaviors were associated with pregnancy loss, most notably consumption of >2 daily caffeinated beverages. The relation between consumption of caffeinated beverages and pregnancy loss has received considerable research, with much of the available evidence relying upon retrospective ascertainment of caffeine consumption including after pregnancy loss that may induce reporting bias (30). Interpreting our findings within the context of available literature that utilized prospective cohort designs with preconception enrollment is challenging in light of few such studies. However, a recent preconception cohort study that recruited women trying for pregnancy via the internet reported a positive association between females’ caffeine consumption during early pregnancy but not prior to conception and pregnancy loss (31). However, there are noteworthy differences between the methods used in this study and ours that may impact findings, including a lower incidence of pregnancy loss in the internet based cohort study in comparison to ours (14% and 28%, respectively), reliance on last menstrual period for gestational dating versus capture of the peak LH day as an estimate of conception in our study, and reliance on a single measurement of recalled caffeine intake prior to and during pregnancy in the internet based study in comparison to our capture of daily consumption before and during early pregnancy. Pregnancy based cohorts also have reported affirming other studies focusing on caffeine consumption during early pregnancy (32,33), but contrasting with a negative finding in another preconception cohort study (34). Our observation that preconception caffeine consumption >2 daily beverages is also associated with pregnancy loss fails to support reverse causation as a possible explanation, where one assumes that women with healthy pregnancy may avoid caffeine consumption due to nausea/vomiting or food aversions believed associated with a healthy pregnancy.

With regard to alcohol consumption whose prevalence was much higher than that for cigarette smoking in our cohort, earlier authors reported couples’ alcohol consumption (≥10 weekly drinks) in the conception cycle to be associated with a higher risk of spontaneous abortion (35) as was consuming ≥4 weekly drinks in a cohort of pregnant women (36). We did not observe alcohol consumption to be associated with pregnancy loss after covariate adjustment, which may reflect differences in the sampling framework (e.g., recruitment of occupational and pregnant women versus population based sampling), differences in the baseline prevalence of alcohol consumption or its timing and frequency during sensitive windows, model specification that includes multivitamin usage, or residual confounding by other lifestyle factors such as smoking or caffeine consumption.

An important and previously unreported lifestyle behavior is the tremendous reduction in pregnancy loss associated with vitamin adherence irrespective of sensitive window. For example, a 55% reduction in pregnancy loss was found for women taking vitamins daily (100% adherence) in the preconception periods versus those who did not (0% adherence), increasing to a 79% reduction for adherence during early pregnancy. In an extended analysis, we assessed whether preconception multivitamin usage is independent of early pregnancy usage and vice versa with the important caveat that women’s behaviors were correlated over windows resulting in very inefficient estimators for effect decomposition and findings that may be susceptible to collider bias in the presence of unmeasured confounders. The unadjusted and adjusted hazard ratios for daily preconception multivitamin usage controlling for similar usage in the early pregnancy window reflected the importance of the preconception window (HR=0.58; 95% CI 0.34, 0.96 and HR=0.83; 95% CI 0.34, 2.04, respectively) (data not shown). Similarly when modeling early pregnancy vitamin adherence controlling for similar preconception usage, a reduced risk of loss was observed in the unadjusted and adjusted analyses (HR=0.25; 95% CI 0.13, 0.49 and 0.23; 95% CI 0.07, 0.73), respectively. These data may support usage in both sensitive windows.

This strong protective effect for preconception multivitamin usage is somewhat consistent with findings from a preconception cohort of female Chinese textile workers, where women in the highest versus lowest quartile of vitamin B₆ had a lower odds of pregnancy loss (aOR=0.5; CI 0.3-1.0) (37). Related support stems from previous studies that involved self-reported vitamin or folate usage and pregnancy loss. Pregnant women were queried about the combined use of multi- and prenatal-vitamins, which were associated with a significantly reduced odds of loss and self-reported pregnancy loss corroborated by medical records (38). In a cohort of women with prospectively collected folate intake and self-reported pregnancy loss, women in the highest versus lowest quintile of folate intake had a 19% reduction in risk of loss (39). Collectively, these data are supportive of a beneficial association between vitamin usage and pregnancy loss.

Our study has many strengths including its unique population-based couple design that captured both partners’ lifestyles in advance of and during early pregnancy or time intervals that are typically missed when relying on pregnancy cohorts. Our use of fertility monitors enabled us to accurately estimate conception using ovulation as its proxy, and use of digital pregnancy tests with demonstrated validity and user accuracy (16). We also were able to quantify small but highly significant (P ≤0.001) reductions in women’s daily mean number of cigarettes smoked (−0.15) or alcoholic (−0.34) and caffeinated (−0.51) beverages consumed between the preconception and early pregnancy sensitive windows, underscoring the need for continual guidance aimed at promoting healthy behaviors.

Still, important limitations need to be considered when interpreting the findings. We chose to assess pregnancy loss without further categorization relative to gestational age in light of no established standard endocrine criteria for defining loss (40). We recognize that we may have missed some losses if hCG concentrations were below analytical sensitivity limits or given the molecular heterogeneity of hCG and its variants. For instance, despite our standardized protocol for testing for pregnancy, it is possible that some women tested too late and thereby missed an hCG rise and fall. We also did not have information on the specific types of beverages consumed or specific vitamins taken that would be helpful to delve into potential biologic mechanisms underlying the association between multivitamin behavior and pregnancy loss, nor did we have information on decaffeinated or caffeine free beverages or other aspects of lifestyle such as physical activity, diet, recreational drugs or sleep habits. As such, we cannot rule out the potential for residual confounding.

Conclusions

In summary, our data suggest that couples’ lifestyles are associated with the risk of pregnancy loss, including during the preconception period, emphasizing the need for continued efforts to promote healthy lifestyles before pregnancy begins, viz., preconception guidance. Our findings underscore the feasibility and utility of prospective cohort designs with preconception recruitment of couples for assessing exposures associated with reproductive and developmental toxicants in support of an earlier conclusion (41). Also, couples might continue to be advised that advanced age increases the risk of pregnancy loss and to limit caffeinated intake to fewer than 3 daily beverages irrespective of source, and that women should continue to take daily multivitamins before and during pregnancy consistent with clinical guidance. Our findings are not intended to suggest substituting decaffeinated or caffeine free beverages as a safer alternative, as we did not query couples on these products. Collectively, our findings are supportive of the need for ongoing efforts to ensure couples receive preconception guidance, and our findings await corroboration from other preconception couple based cohorts to ensure preconception guidance issued for the U.S. population (http://www.cdc.gov/preconception) are responsive to discoveries to the extent possible.

In conclusion, our findings are highly relevant for contemporary cohorts of couples at risk for or planning pregnancy who might want to adopt or maintain lifestyles that minimize the risk of pregnancy loss. All such lifestyles are amenable to change, and such guidance has been shown to influence healthy behaviors (42).