Inflammation and Conception in a Prospective Time-to-Pregnancy Cohort.

The 727 women in the study contributed 2875 cycles and 483 detected pregnancies, for a 67% overall conception rate. The median number of cycles attempting pregnancy at enrollment was 2 (interquartile range [IQR]: 2,3) and 99% of the observations were at seven cycles or fewer. The median CRP level in the sample was 0.92 mg/L and the distribution was right-skewed (25 th , 75 th percentiles: 0.37, 2.52). In the univariate analysis the natural log of CRP was moderately correlated with BMI (Pearson r=0.49) but only weakly correlated with age (r=0.05). When examined in categories, the association between CRP and age showed no dose–response, with women under age 30 having lower levels than women over 30. ( Table 1 ) Compared with white women, unadjusted median CRP levels were higher in women who reported their race as African American, Asian/Pacific Islander, Hispanic, or as “mixed or unknown”. Median CRP levels were also higher in women reporting lower education, current or former smoking, low alcohol use, higher caffeine intake, recent estrogen use, a history of high blood pressure, or a history of diabetes. ( Table 1 ) CRP increased monotonically across categories of BMI. ( Table 1 ) Women with sufficient vitamin D had lower CRP levels, and women who took omega-3 supplements had a higher median CRP. We hypothesized that higher levels of CRP might indicate infection or illness, which might lead to a reduced frequency of intercourse; however, we did not find any evidence that those with higher CRP levels had lower frequency of intercourse in the fertile window. ( Table 1 ) In an unadjusted model, increased CRP was associated with decreased fecundability (per ln-unit increase in CRP: FR(CI): 0.94 (0.88, 0.99). Similarly, estimated fecundability was lower in the highest CRP category (>10 mg/L) versus the lowest (<1 mg/L) (FR(CI): 0.66 (0.45, 0.97)). However, after adjustment, there remained only a small decrease in fecundability per log-unit increase in CRP level (FR(CI): 0.97 (0.91, 1.0)) and CRP categories did not show any clear pattern of association with fecundability. ( Table 2 ) For context, a log-unit increase in CRP is equivalent to multiplying by approximately 2.7. So, an approximate 3-fold increase in CRP was associated with a 3% decrease in fecundability. Given the median CRP level in our sample was 0.92 mg/L, a 3-fold increase in CRP would correspond to an approximate change from 0.92 to 2.76 mg/L. This level of change clinically would bring someone closer to the high-risk category based on AHA criteria, which categorizes values greater than 3 as “high risk”. This suggests that, for approximately 50% of the population, a 3-fold increase in CRP would move them from low risk to borderline high risk. The highest CRP category had lower estimated fecundability (FR(CI): 0.78 (0.52, 1.18)), but the confidence interval was wide. These results were not meaningfully different when women with high blood pressure or diabetes were excluded ( Table 2 ) or when we adjusted for gravidity. ( eTable 1 ) When examining the interaction between CRP and BMI, we again observed no pattern of association between CRP and fecundability. ( Table 2 ) However, among women with obesity and high CRP (>10 mg/L), estimated fecundability was lower than for normal weight women with low CRP, FR(CI): 0.63 (0.35, 1.1). ( Table 2 ) Excluding women with high blood pressure or diabetes did not alter the estimated associations. Within strata defined by vitamin D status, again, for the most part, CRP was not associated with fecundability. ( Table 3 ) Women with high 25(OHD) (>40 ng/ml) and moderately elevated CRP (3 – 10 mg/L) had higher estimated fecundability (FR(CI): 2.1 (1.3, 3.4)), but there was no dose–response across lower categories of CRP. ( Table 3 ) In a sensitivity analysis, we fit the multivariate model using multiply imputed data, which enabled inclusion of adjustment for exercise. The results presented in Tables 2 and 3 were similar to those estimated with the multiply imputed data, and overall interpretations were unchanged from the previously presented results. ( eTable 2 ) We also examined whether the results changed when limiting the analysis to participants who provided their blood sample in the first three cycles of their pregnancy attempt or when adjusting for the number of cycles of attempt at the time of the blood draw but found no differences in the associations ( eTable 3 ). Similarly, when examined among these participants who enrolled early in their attempts, the distribution of CRP across covariates was similar to the study sample overall ( eTable 4 compared with Table 1 ).

Time to Conceive ( 19 , 20 ) is a prospective study of time to pregnancy that enrolled women ages 30–44, who reported being within three months of beginning their pregnancy attempt. Women were excluded if they had diagnosed fertility issues (history of PCOS, use of fertility medications, endometriosis). Women enrolled in the study resembled the population of women giving birth in Chapel Hill and Raleigh with respect to age, race, and education ( 21 ). The original Time to Conceive study was approved by the University of North Carolina IRB; this sub-study was approved by the Duke University IRB. Time to Conceive participants completed an online baseline questionnaire that encompassed reproductive history, supplement use, diet, and behaviors. For the first four months in the study, participants completed online daily diaries and monthly diaries where they recorded supplement use, menstrual bleeding, and intercourse. If they did not conceive in the first four months, they were asked to complete just the online monthly diaries for the next eight months. We scheduled an in-person clinic visit for day 2, 3, or 4 of their first menses following enrollment. At this visit a venous blood sample was drawn (“baseline” sample). Serum and blood spots from this sample were stored frozen at −30°C. At the in-person clinic visit, women were given pregnancy test kits and asked to test on menstrual cycle days 28, 31, 34, etc. until they tested positive or menses began. We considered conception to have occurred if the participant reported a positive home pregnancy test. We defined pregnancy attempt time as the number of menstrual cycles from the beginning of the attempt until the occurrence of either a positive pregnancy test or a censoring event (12 total cycles of attempt, withdrawal from the study, or beginning fertility treatment). This total attempt time was the sum of: 1) the number of cycles of trying to conceive before the in-person visit, 2) the number of menstrual cycles prospectively quantified in her daily menstrual diary, and 3) the number of menstrual cycles that occurred after the daily diaries ended, and the woman completed monthly diaries only. The first quantity, the number of cycles a woman had been trying to conceive at the time of the in-person visit, was estimated by dividing the time since discontinuing contraception or time since last pregnancy by the usual cycle length. In some cases, this assessment of attempt time was inconsistent with the participant’s self-report of being within three cycles of attempt when she was originally screened by telephone; in all cases our assessment was used. This analysis was limited to outcomes in cycles one to twelve of their pregnancy attempt based on our quantification of her attempt time. ( Figure ) We estimated the third quantity, the number of cycles that occurred after the daily diary ended, by dividing the time from the last cycle in the daily diary up to the time of the positive pregnancy test date by the usual reported cycle length. CRP is highly stable in stored specimens, with storage of up to 11 years having negligible impact on levels ( 22 ). The Biomarkers Core at Duke University measured CRP in stored serum samples (N=759 women, Figure ) using a high sensitivity immunoturbidimetric assay. They ran anonymous human serum samples in duplicate on each plate and a random sample of test specimens were also chosen as blinded duplicates. The mean intra-assay coefficient of variation was 5.6% and the mean inter-assay coefficient of variation was 8.9%. The lower limit of quantification was 1.3 pg/ml. We defined four clinically relevant categories of CRP based on CDC and AHA criteria ( 23 ): 3.0 – 10.0 mg/L, and >10.0 mg/L. We extracted 25(OH)D from 6mm punches from stored blood spots using previously described methods ( 24 ). 25(OH)D 3 and 25(OH)D 2 were quantified through liquid chromatography-tandem mass spectrometry. 25(OH)D measured in dried blood spots shows good agreement with plasma measures ( 25 ). Blinded samples indistinguishable from test samples were also sent to the lab. Based on these samples, the intra-assay coefficient of variation was 6.3% and the inter-assay coefficient of variation was 7.7%. Because we did not collect blood spots in the first years of the study, 25(OH)D measurements were only available for a subset of women (N=507). We chose covariates based on their associations with both CRP and fecundability. We calculated age at the beginning of each menstrual cycle (using birth date and the date the menstrual cycle started). Women reported their daily use of omega-3 supplements on their daily diary. At baseline, participants self-reported their race (“How would you best describe your race? African American, Caucasian, Asian/Pacific Islander, American Indian/Alaskan, Hispanic, Other (mixed or unknown)”), gravidity (0, 1, or 2 or more previous pregnancies), height, weight, education, the date they discontinued contraception and the type of contraception used. We used this information, combined with menses dates recorded in the daily diary, to calculate the time since hormonal contraception use at the start of each menstrual cycle of attempt. This was classified as use in the previous month, 2 months prior, or 3 months or more prior. In monthly diaries, participants also reported their number of alcoholic drinks per month, number of caffeinated drinks per day, and the number of cigarettes smoked per day. We asked participants about their use of other nicotine products, but none were reported. Since participants rarely reported cigarette smoking in the prospective monthly diaries, we used smoking status at baseline (current, former, never). We matched monthly reports with each menstrual cycle of attempt. We single imputed missing monthly reports by taking a previous month’s values. If a previous month was not available, we used a subsequent month’s values. On the monthly diary participants also reported, in categories, their average amount of vigorous exercise (0 hours, 7 hours per week). If a woman did not have at least one prospective monthly report of caffeine or alcohol her baseline report was used (exercise was not captured on the baseline questionnaire). We defined frequency of intercourse during the fertile window, using several daily diary variables, as the number of days on which intercourse occurred during the six days up to and including ovulation, and ovulation based on data from an ovulation predictor kit, basal body temperature, cervical mucus monitoring, or imputed at the modal value of cycle day 15, as previously described ( 26 ). We describe the univariate distribution of CRP and bivariate distribution with covariates using medians and the 25 th and 75 th percentiles. We estimated fecundability ratios (FRs) and 95% confidence intervals (CI) using log-binomial regression ( 27 ). In this model conception (yes/no) is the outcome for each cycle during follow up and an intercept is estimated for each cycle of attempt by including attempt cycle number as a categorical predictor. Including the attempt cycle number allows fecundability to decline across cycles and accommodates delayed entry, which accounts for left truncation ( 27 ). For example, a woman whose blood draw occurred in her third cycle, would only contribute information to the model beginning at cycle three. To explore the association between CRP and fecundability, we fit a model with restricted cubic splines ( 28 ) of CRP adjusted for age, race, and BMI. The result of this analysis showed a small, linear decrease in fecundability across CRP level ( eFigure 1 ). We used the Akaike Information Criterion (AIC) to compare the model fit with splines to a model using log-transformed and mean-centered CRP as a linear term. The AIC was slightly smaller for the latter, indicating that the linear model was the preferred parameterization. Thus, we examined CRP in two ways: first, continuously, and natural log transformed and mean-centered; and second, in four levels consistent with professional society guidance, as previously described. The multivariable model included: age, race, body mass index, education, recent estrogen use, alcohol intake, caffeine intake, and smoking history. To examine effect modification by BMI we estimated interaction terms with BMI in four categories: <18.5 kg/m 2 , 18.5-<25, 25-<30, and ≥30, using the multivariable log-binomial model described previously. In these analyses each BMI and CRP category is compared to a single referent: BMI of 18.5-<25 and CRP <1 mg/L ( 29 ). We considered allowing the baseline fecundability to vary by BMI stratum by testing an interaction term between cycle number and BMI but found no evidence of interaction and the interaction term was not included in the final model. To examine the associations between CRP and fecundability while stratifying by vitamin D status, we used three strata of vitamin D: vitamin D insufficiency (25(OH)D 40ng/ml) ( 19 , 30 ). In this analysis, each vitamin D and CRP category was compared with a single referent category: sufficient vitamin D (30–40 ng/ml) and low CRP (<1 mg/L). We considered allowing the baseline fecundability to vary by vitamin D stratum by testing an interaction term between cycle number and vitamin D but found no evidence of interaction; thus, we did not include the interaction term in the final model. Numbers were too small in the low and high categories of 25(OH)D for us to fully stratify these models also by BMI. Instead, as a sensitivity analysis, we also present the model with interactions between CRP and vitamin D level additionally limited to women with a BMI of 18.5 - <25 kg/m 2 . We performed several additional sensitivity analyses. Gravidity may be a confounder if it causes changes in CRP levels; however, if CRP levels cause subfecundity, then low gravidity would be a descendant of subfecundity, and inclusion of gravidity in the model would be an inappropriate over-adjustment. Thus, we examined the additional adjustment for gravidity in a sensitivity analysis. Self-reported diabetes (N=8) and high blood pressure (N=39) were rare in this population but are often associated with higher levels of CRP. Because the number was small, we examined whether the primary multivariable results changed after exclusion of women with either condition. Also, to address missing covariate data, we performed multiple imputation by fully conditional specification with predictive mean matching ( 31 ). We ran the multivariable model across the 20 imputed data sets. Finally, we estimated the associations between CRP and fecundability after adjustment for the attempt time at enrollment (blood draw), or when the sample was limited to those who had been trying for less than four cycles at the time of the blood draw.

In our sample of over 700 women aged 30 and older, and after adjustment for age, race, BMI, and several other covariates, we found no evidence for an association between CRP and fecundability. Also, when examining the interaction of CRP and BMI, CRP was not consistently associated with fecundability. Moderately high CRP (3–10 mg/L) was associated with higher fecundability among women with high vitamin D, but there was no dose–response across lower levels of CRP and vitamin D. CRP is an acute phase protein, a family of proteins found in the plasma in response to infection or inflammation ( 32 ). BMI and CRP in ovarian follicular fluid are correlated, suggesting that inflammation in ovarian follicles increases with increasing BMI ( 32 ). However, BMI is not a perfect marker of adiposity and the correlation between CRP and triglycerides in follicular fluid is stronger than that with BMI, which might indicate that dyslipidemia is a stronger cause of inflammation and ovarian dysfunction than BMI ( 32 , 33 ). High levels of follicular fluid triglycerides have been associated with unexplained infertility ( 34 ). If this is true, the lack of association between circulating CRP and fecundability may be the result of BMI being an imperfect proxy for lipid levels. Future research should explore the associations between CRP and fecundability while accounting for serum lipid levels, rather than BMI. Our a priori hypothesis was that higher levels of CRP would be associated with reduced fecundability, which has been previously reported for levels of CRP between 1.95 and 9.9 mg/L ( 17 ), although that estimate was weakened with adjustment for adiposity. In our analysis, fecundability was not lower among women with a similar CRP level of 3 – 10 mg/L. It tended to be lower, although imprecisely, among women with high CRP (>10 mg/L), a level excluded from the previous study. It has been suggested that CRP levels >10 mg/L be excluded from research studies as potentially reflective of acute infection. However, women came into the study office for a blood draw and were unlikely to have been markedly ill when their CRP was measured. Further, we saw no association between high CRP and intercourse frequency, which might also be affected by acute illness. Thus, while elevated CRP may be a marker of acute illness, our findings indicate that high levels of CRP may also be otherwise meaningful. In our study high CRP was imprecisely associated with reduced fecundability in women with obesity. Future studies could examine these high levels of CRP and fecundability. While stratification creates small samples, we also examined CRP and fecundability within levels of vitamin D. Our hypothesis was that higher vitamin D status might ameliorate the effects of inflammation, bringing estimates for high CRP towards the null. This hypothesis stemmed from previously reported associations between higher vitamin D and lower risk of abnormal menstrual cycles ( 30 ), higher fecundability ( 19 ), and increased live birth rates ( 35 ). We found no evidence that CRP underlies the associations between vitamin D and reproduction. Our study is the first prospective investigation of CRP in a community-based sample of women of older reproductive age. While our study is large, there were small numbers in some of the strata of interest. Time to pregnancy was measured with the gold standard design: women were enrolled early in their attempt to become pregnant and followed for conception. Women who had a previous diagnosis of infertility, or conditions that may lead to infertility, were excluded from the study. However, we did not ask women if they had autoimmune disease, which might influence both CRP and fecundability. It is possible that some participants had undiagnosed PCOS. PCOS may be a cause of CRP levels, or a descendent of CRP levels. If the latter, PCOS is on the causal pathway between CRP and fecundability, while if the former, undiagnosed PCOS could lead to residual confounding if PCOS also affects fecundability independently of CRP. We would expect potential confounding by PCOS to lead to an observed association between CRP and fecundability. Given our essentially null findings, we hypothesize that residual confounding by PCOS is unlikely to be important. We did not have any measures of adiposity other than BMI, and other measures of adiposity such as serum lipids or waist-to-hip ratio may be more closely correlated with CRP levels then BMI. For internal validity, the analysis assumes that women who enrolled in the study at cycle 2 or cycle 3 are not systematically different in their fecundability from women who joined at cycle 1 and were still not pregnant at cycle 2 or cycle 3. We found no evidence for associations between CRP levels and fecundability. High CRP tended to be associated with reduced fecundability, but further research is needed with a larger sample that includes, rather than excludes, higher levels of CRP. We did not see any consistent associations between CRP and fecundability within categories of BMI. Further research related to CRP and fecundability should include additional markers of inflammation. While CRP is an informative biomarker, it is part of a larger inflammatory pathway that deserves further exploration. Exploring other biomarkers, such as inflammatory cytokines, in conjunction with CRP might provide greater insight into the potential role of inflammation in fertility.

Subfecundity (difficulty conceiving a pregnancy) is a pressing public health problem. Approximately 6.7 million (11%) U.S. women of reproductive age report impaired fecundity and 1.5 million are estimated to be infertile ( 1 ). Moreover, women are delaying child-bearing to older ages ( 2 ). It is well known that risk of subfecundity increases with age ( 3 – 5 ), as do other adverse reproductive outcomes such as spontaneous miscarriage ( 4 , 6 – 8 ), preterm birth ( 4 , 7 ), and fetal growth restriction ( 7 ). With more women attempting pregnancy for the first time at older ages, it is increasingly important to identify modifiable risk factors for adverse outcomes in this population. Inflammation plays an important part in female reproduction ( 9 ). Local inflammation facilitates ovarian follicle development and ovulation; however, excess inflammation can lead to an increase in reactive oxygen species and oocyte damage ( 9 ). C-reactive protein (CRP) is a known marker of systemic inflammation. Macrophages and adipocytes release IL-6, which stimulates the liver to produce CRP, which then activates the complement system ( 9 ). Both the CDC and the American Heart Association recommend that circulating CRP be used to identify people at higher risk for cardiovascular events ( 10 ). This is particularly true for women. Among apparently healthy women, higher CRP predicted a 5- to 7-fold increase in risk of vascular events ( 11 ). Data from women with polycystic ovary syndrome (PCOS) suggest that systemic low-grade inflammation, as measured by CRP, may contribute to ovulatory disorders or subfertility, even independently of obesity ( 12 ) ( 13 , 14 ). Higher serum CRP levels were reported among fertility treatment patients with unsuccessful intrauterine insemination cycles ( 15 ) or euploid pregnancy loss ( 16 ). Circulating CRP levels were higher in women with unexplained infertility compared with fertile women and also higher in women with diminished ovarian reserve compared with women with normal ovarian reserve. Most of these studies are cross-sectional, and do not directly examine the association of CRP on the probability of conceiving. Only one previous study has examined fecundability and, in this study, higher levels of CRP measured with a high sensitivity assay were associated with reduced fecundability, but the association was attenuated with adjustment for body mass index ( 17 ). That sample only included women with a history of pregnancy loss and over 60% of them were under 30 years of age. In total, this literature suggests that higher CRP is associated with adverse reproductive function, but it is unclear if the association exists outside of populations with infertility, if there is an association with fecundability or just with infertility, or if it exists in older women who may be, due to their age, at higher risk of both subfecundity and chronic inflammation. Our objective was to examine the association between pre-conception CRP levels, measured with a high sensitivity assay, and fecundability in a cohort of pregnancy planners aged 30 and over. We also examined the interaction between CRP and BMI. Antioxidants may lessen the damage caused by inflammation and reactive oxygen species. A recent Cochrane review suggests that antioxidant consumption may improve fertility outcomes in subfertile women ( 18 ). Vitamin D has antioxidant properties, and has been associated with fecundability ( 19 ). As an exploratory analysis, we assessed whether the association between CRP and fecundability differed by vitamin D levels.

eFigure 1. Predicted mean fecundability across CRP level, adjusted for age, race, and BMI and presented at the reference levels of covariates. The predicted mean was estimated with a discrete-time fecundability model with restricted cubic splines of CRP.