Advertisement

Conception, early pregnancy loss, and time to clinical pregnancy: a population-based prospective study

      Abstract

      Objective

      To examine rates of conception and pregnancy loss and their relations with time to clinical pregnancy and reproductive outcomes.

      Design

      A prospective observational study.

      Setting

      Population-based cohort in China.

      Patient(s)

      Five hundred eighteen healthy newly married women who intended to conceive. Upon stopping contraception, daily records of vaginal bleeding and daily first-morning urine specimens were obtained for ≤1 year or until a clinical pregnancy was achieved. Daily urinary hCG was assayed to detect early pregnancy loss (EPL).

      Intervention(s)

      None.

      Main outcome measure(s)

      Conception, pregnancy loss, and time to clinical pregnancy.

      Result(s)

      The conception rate per cycle was 40% over the first 12 months. Of the 618 detectable conceptions, 49 (7.9%) ended in clinical spontaneous abortion, and 152 (24.6%) in EPL. Early pregnancy loss was detected in 14% of all the cycles without clinically recognized pregnancy, but the frequencies were lower among women with delayed time to clinical pregnancy. Early pregnancy loss in the preceding cycle was associated with increased odds of conception (odds ratio [OR], 2.6; 95% confidence interval [CI], 1.8–3.9), clinical pregnancy (OR, 2.0; 95% CI, 1.3–3.0), and EPL (OR, 2.4; 95% CI, 1.4–4.2) but was not associated with spontaneous abortion, low birth weight, or preterm birth in the subsequent cycle.

      Conclusion(s)

      We demonstrated substantial EPL in the non–clinically pregnant cycles and a positive relation between EPL and subsequent fertility.

      Keywords

      Per-cycle conception rates, rates of early pregnancy loss (EPL), and time to clinical pregnancy are of fundamental importance to understanding the reproductive process. To date very few studies have addressed these phenomena and their interrelationships because of the methodological complexity and cost of conducting such studies (
      • Modvig J.
      • Schmidt L.
      • Damsgaard M.T.
      Measurement of total risk of spontaneous abortion the virtue of conditional risk estimation.
      ). To accurately define all conceptions and pregnancy losses in a population, one needs to prospectively follow a cohort of women from the beginning of non-contracepting cycles until the occurrence of the endpoints of interest. The detection of EPL, which is not clinically apparent, requires a sensitive and specific assay of daily urinary hCG. Otherwise, EPL may be mistaken for a nonconceptive cycle.
      Up until the early 1980s, when urine pregnancy tests could not document pregnancies until 6 weeks after the last menstrual period, pregnancy loss occurring before this time could not be documented (
      • Wilcox A.J.
      Surveillance of pregnancy loss in human populations.
      ). The development of sensitive and specific urinary hCG assays has made it possible to detect EPL within a few days of implantation (
      • Wilcox A.J.
      • Weinberg C.R.
      • Wehmann R.E.
      • Armstrong E.G.
      • Canfield R.E.
      • Nisula B.C.
      Measuring early pregnancy loss laboratory and field methods.
      ). Using this newly developed assay, Wilcox and coworkers (
      • Wilcox A.J.
      • Weinberg C.R.
      • O’Connor J.F.
      • Baird D.D.
      • Schlatterer J.P.
      • Canfield R.E.
      • et al.
      Incidence of early loss of pregnancy.
      ) studied 221 healthy women volunteers who were trying to conceive. They observed a per-cycle clinical pregnancy rate of 25% over the first three cycles and a pregnancy loss rate of 32% for all conceptions; two thirds of all pregnancy loss occurred before clinical detection. Zinaman and coworkers (
      • Zinaman M.J.
      • O’Connor J.
      • Clegg E.D.
      • Selevan S.G.
      • Brown C.C.
      Estimates of human fertility and pregnancy loss.
      ) made similar observations in a cohort of 200 women volunteers at a university-based obstetrics and gynecological center.
      Human reproduction is a set of interconnected processes. Time to clinical pregnancy was defined as the number of noncontraceptive cycles that it takes a couple to conceive a clinically recognized pregnancy (
      • Baird D.D.
      • Wilcox A.J.
      • Weinberg C.R.
      Use of time to pregnancy to study environmental exposures.
      ). An increase in time to clinical pregnancy may be due to decreased conception rates, increased EPL, or both. The extent of EPL among cycles without clinically recognized pregnancy and its variability among women with different time to clinical pregnancy is not known. Furthermore, previous reproductive events are often correlated with subsequent reproductive outcomes in a given woman. There are few data on how EPL in previous cycles is associated with reproductive outcomes in the subsequent cycles (
      • Wilcox A.J.
      • Weinberg C.R.
      • O’Connor J.F.
      • Baird D.D.
      • Schlatterer J.P.
      • Canfield R.E.
      • et al.
      Incidence of early loss of pregnancy.
      ).
      We replicated and expanded previous research (
      • Wilcox A.J.
      • Weinberg C.R.
      • O’Connor J.F.
      • Baird D.D.
      • Schlatterer J.P.
      • Canfield R.E.
      • et al.
      Incidence of early loss of pregnancy.
      ,
      • Zinaman M.J.
      • O’Connor J.
      • Clegg E.D.
      • Selevan S.G.
      • Brown C.C.
      Estimates of human fertility and pregnancy loss.
      ) in a large population-based prospective cohort, in which we examined rates of conception and pregnancy loss among women with different times to clinical pregnancy, based on each woman’s daily diary record and the highly sensitive and specific assay of daily urinary hCG to detect EPL. We also investigated EPL in the preceding cycles in relation to conception, pregnancy loss, and birth outcomes in the subsequent cycles.

      Materials and methods

       Study population

      This is a part of a prospective reproductive health study funded by the National Institutes of Health among women textile workers in Anhui, China. The study protocols were approved by the human subject committee of the Chinese institutions involved in the study and by the institutional review board of the Harvard School of Public Health.
      The eligibility criteria for women in the field enrollment were as follows: [1] full-time employment; [2] newly married; [3] aged 20 to 34 years; and [4] had obtained permission to have a child. All the women were nulliparous. Women were excluded if [1] they were already pregnant before enrollment; [2] they had tried unsuccessfully to get pregnant for ≥1 year at any time in the past; or [3] they planned to quit or change jobs or to move out of the city over the 1-year course of follow-up.
      Of the 971 women textile workers who met the eligibility criteria, 961 were enrolled and had baseline information. Four hundred forty-three women were excluded from this analysis for the following reasons: 95 continued to use contraceptives; 53 did not begin recording diaries and collecting daily urines immediately after stopping contraception; 121 women declined diary or urine collection; 78 became pregnant because of contraceptive failure; 8 were lost to follow-up; 27 withdrew shortly after enrollment; 34 had inadequate diary or urinary hCG data; 7 had menstrual irregularity (cycle length of >40 days or cycle length of <21 days); 16 had a history of using oral contraceptive pills; and 4 had a history of using an intrauterine device. This report includes 518 women who intended to conceive; who began recording diary and collecting daily urine immediately after stopping contraceptive use; and who had adequate diary and hCG data.

       Field data collection procedures

      After obtaining informed consent, the interviewer administered a baseline questionnaire that collected historical data on menstruation, contraceptive use, reproductive history, sociodemographic characteristics, active smoking and passive smoke exposure, alcohol use, and environmental and occupational exposures. If a woman reported a missed or late period or had early signs or symptoms of pregnancy, she was instructed to go to the affiliated hospital for a check-up and to give a urine sample for hCG assay. Once a woman was confirmed to be pregnant, she received regular prenatal care and delivery services at the designated hospitals according to standard clinical guidelines and was followed for pregnancy outcomes by the research staff.
      Beginning from the date of stopping any contraceptive method and attempting to conceive, each woman kept a daily diary to record sexual intercourse, vaginal bleeding, medication, and medical conditions and collected daily first–morning urine specimens for 12 months or until a pregnancy was clinically confirmed, whichever came first. Each woman was given a 250-mL beaker for collecting urine and a 50-mL, double-seal vial for storing and transferring the urine. Each vial was labeled with a bar code for the appropriate day and date. The woman was instructed to void into the beaker upon awakening each day and then to pour 50 mL of the urine into the double-sealed vial, to place the vial into a plastic bag, and to store this in the refrigerator temporarily. The field coordinators were responsible for collecting urine samples and diaries every day.
      In addition, daily urine specimens were obtained from 37 control women for two known nonconception cycles. Of those, 23 women were not married and not sexually active; 10 women were married, but their husbands were out of town; and 4 women were married, but using contraception during the urine collection cycles. They had no history of fertility problems or chronic illnesses.

       Major outcomes and method of evaluation

      Clinical pregnancy was defined as any pregnancy that lasted ≥6 weeks (42 days) after the onset of the last menstrual period and that was confirmed by hCG assay. Time to clinical pregnancy was defined as the number of menstrual cycles it took from the time that a woman stopped contraceptive methods and began attempts to conceive until she achieved a clinical pregnancy. Clinical spontaneous abortion was defined as a pregnancy loss after a clinical pregnancy that had been <28 weeks of gestation. Early pregnancy loss was defined as clinically unrecognized pregnancy loss detected only by the highly sensitive and specific urinary hCG assay (discussed later in this article). Bayesian methods (
      • Gelman A.
      • Carlin J.B.
      • Stern H.S.
      • Rubin D.B.
      ,
      • Dunson D.B.
      A Bayesian model for fecundability and sterility.
      ) were applied for handling missing hCG data to provide a better estimate of the probability of conception and EPL (see Statistical Analysis section).

       Laboratory assay of urinary hCG

      Urine specimens were stored in our central field laboratory in Anqing, China at −20°C. Urinary hCG levels were analyzed in batches by the immunoradiometric assay (IRMA) developed by O’Connor et al. (
      • O’Connor J.F.
      • Schlatterer J.P.
      • Birken S.
      • Krichevsky A.
      • Armstrong E.G.
      • McMahon D.
      • et al.
      Development of highly sensitive immunoassays to measure human chorionic gonadotropin, its beta subunit and beta core fragment in the urine application to malignancies.
      ) using a combination of capture antibodies for hCG free β subunit and hCG β core fragment (B204) and the intact hCG molecule (B109). This assay is highly sensitive and specific. The lowest hCG concentration detectable by the assay was 0.01 ng/mL (1 mIU = 0.2 ng). The cross-reaction of the assay with either intact LH or LH free β subunit is <1%. All of the urine specimens from each woman were analyzed and tested during a single run of the assay. Each urine specimen during the window of −10 to +5 days of a menstrual cycle was assayed in duplicate. Discrepancies greater than threefold between duplicate assays were presumed to result from technical error, and the assay was repeated. For the remainder, the geometric mean of the replicates was used to summarize the results for each sample. Urine creatinine levels were measured according to the method of Jaffe (
      • Husdan H.
      • Rapoport A.
      Estimation of creatinine by the Jaffe reaction. A comparison of three methods.
      ). All hCG values were normalized to creatinine values to adjust for urine concentration. As a reference value, nonconceptive levels of hCG were determined from the 67 complete cycles of the 37 control women.

       Statistical analysis

      The major outcomes of interest were conception and pregnancy loss (clinical spontaneous abortion and EPL). A central focus of our analysis was the development and application of methods to determine conception status based on observed hCG values. There were several challenges. First, there were considerable variations in daily urine hCG values even among nonconception cycles (e.g., for the 37 control women). In addition, 21% of the days in cycles with no clinical pregnancy had missing hCG values (primarily due to missing urine collection), and many of those days with missing hCG values fell within the critical window of −10 to +5 days of the menstrual cycle. To distinguish normal variation from a true hCG rise due to conception and to address missing hCG values, we employed Bayesian methods (
      • Gelman A.
      • Carlin J.B.
      • Stern H.S.
      • Rubin D.B.
      ,
      • Dunson D.B.
      A Bayesian model for fecundability and sterility.
      ) to model daily conception status among all the women subjects including the control women. We assumed that the square roots of hCG values followed a normal distribution with constant mean and variance before conception. For postconception, we modeled the square root of hCG values with a normal distribution with a quadratic mean function and constant variance, with the mean and variance differing for clinical pregnancy and EPL. We choose noninformative proper prior distributions for each parameter and fit the model using Markov Chain Monte Carlo methods. This model allowed us to calculate a probability of conception for each observed cycle. Our data showed that this model was 100% sensitive and specific for those cycles with known conception status: the probability was 0 for all the control cycles and 1.0 for all the cycles with clinical pregnancy. Among 1,095 cycles without clinical pregnancy, 929 cycles had a probability of conception of 0; 14 cycles had a probability between 0 and 0.9; 18 cycles had a probability of between 0.9 and 1.0; and 134 cycles had a probability of 1.0. For the purpose of subsequent analysis, we defined conception as a probability of ≥0.9. However, our inference is quite robust to the choice of cutoff. These methods will be described in more details in a subsequent statistical paper.
      We used standard life table methods to estimate the cycle-specific rates of clinical pregnancy and total conception (clinical pregnancy plus EPL). We also estimated the rates of total conception and EPL among women with different times to clinical pregnancy. Furthermore, we examined the relationships between EPL in the preceding cycles and reproductive outcomes in the subsequent cycles, using logistic regression models. Because our analysis included multiple cycles of the same woman, the standard errors were estimated by using a generalized estimation equation to accommodate correlations in reproductive outcomes among cycles (
      • Liang K.-Y.
      • Zeger S.L.
      Longitudinal data analysis using generalized linear models.
      ).

      Results

      This report includes 518 women who contributed a total of 1,561 menstrual cycles. This is a young, nulliparous cohort. All the women were newly married and intended to have a baby over the course of the study. These women began recording a diary and collecting daily urine immediately after stopping contraceptive use and had adequate diary and hCG data. These women did not smoke or drink alcohol, but 65% of them were exposed to passive smoke. Prior contraceptive methods used by the 518 women were condom, 30.3%; rhythm, 7.4%; withdrawal, 10.5%; and none, 51.8% (newly married couples began diary and urine collection immediately after cohabitation). As shown in Table 1, the women who were excluded from the analysis, and the women who served as controls were similar to those included in the analysis in terms of age, height, weight, body mass index, age of menarche, education, passive smoke exposure, and occupational exposures.
      TABLE 1Characteristics of women textile workers at Anqing, China.
      Maternal characteristicsWomen included in the analysis (n = 518)Women excluded from the analysis (n = 443)Women serving as controls (n = 37)
      Age, y (mean ± SD)24.9 ± 1.724.8 ± 1.624.9 ± 2.3
      Height, m (mean ± SD)1.58 ± 0.051.58 ± 0.051.59 ± 0.05
      Weight, kg (mean ± SD)49.3 ± 5.649.4 ± 5.550.5 ± 6.5
      BMI, kg/m2 (mean ± SD)19.8 ± 2.019.9 ± 2.020.1 ± 2.4
      Age of menarche, y (mean ± SD)14.7 ± 1.414.7 ± 1.614.7 ± 1.1
      Education (%)
      Middle school64.565.067.6
      High school34.733.632.4
      College or above0.81.40
      Job-related stress (%)
      Low or no64.865.873.0
      Moderate29.429.224.3
      High5.85.02.7
      Dust exposure (%)
      Light or no33.737.421.6
      Moderate37.342.956.8
      High29.019.721.6
      Noise exposure (%)
      Light or no25.628.88.1
      Moderate level36.036.451.4
      High level38.534.840.5
      Passive smoking at home (%)64.768.461.5
      Passive smoking at work (%)0.60.92.7
      Wang. Conception, early pregnancy loss, time to clinical pregnancy. Fertil Steril 2003.

       Conception rates

      Table 2 summarizes the pregnancies and outcomes cycle by cycle. Among the 518 women attempting to conceive, the average probability of conceiving a clinical pregnancy within a given cycle over the first 12 months was 29.9%. However, the probability of conception declined in later cycles as the proportion of subfertile women among those still not pregnant gradually increased: 32.2% for 1–3 cycles, 28.2% for 4–6 cycles, 17.4% for 7–9 cycles, and 12.2% for 10–14 cycles. In combining clinical pregnancy and EPL, the total conception rate was 39.6% per cycle.
      TABLE 2Pregnancies and outcomes according to menstrual cycle.
      ParameterMenstrual cycleTotal
      Sum of a given row. For example, in the first row, 1,561 represents total number of observed menstrual cycles.
      1234567891011121314
      No. of women starting the cycle518352225141986949372318149441,561
      Total conception
      No.220149955139281312323201618
      Conception rate (%)42.542.342.236.239.840.626.532.415.3
      Percentage for cycles 9 through 14.
      39.6
      Early pregnancy loss (EPL)
      No.64291711101044111000152
      EPL rate (%)
      Number of EPL divided by total number of pregnancies.
      29.119.517.921.625.635.730.833.327.3
      Percentage for cycles 9 through 14.
      24.6
      Clinical pregnancy
      No.1561207840291898212201466
      Clinical pregnancy rate (%)30.134.134.728.429.626.118.421.611.1
      Percentage for cycles 9 through 14.
      29.9
      No. of women who dropped out after completing the cycle10763023622210044
      No. of women who completed the study without a clinical pregnancy
      The number of cycles during the 12-month study period varied from 9 to 14.
      1112038
      Wang. Conception, early pregnancy loss, time to clinical pregnancy. Fertil Steril 2003.
      a Number of EPL divided by total number of pregnancies.
      b Sum of a given row. For example, in the first row, 1,561 represents total number of observed menstrual cycles.
      c Percentage for cycles 9 through 14.
      d The number of cycles during the 12-month study period varied from 9 to 14.

       Pregnancy loss rates

      There was a total of 618 conceptions (either noted as clinical pregnancies or EPL, detected by hCG assay). Of 618 total conceptions, 373 (60.4%) ended as live births, 49 (7.9%) as spontaneous abortions, 152 (24.6%) as EPL, 6 as induced abortion, 2 as ectopic pregnancies or moles, 4 as still births, 31 as ongoing clinical pregnancies, and 1 as lost to follow-up. In combining clinical spontaneous abortion and EPL, the total rate of pregnancy loss was 32.5%.

       Conception, EPL, and time to clinical pregnancy

      In this cohort of women, approximately 50% became clinically pregnant during the first two cycles, and >90%, during the first six cycles. As shown in Figure 1, with increasing times to clinical pregnancy there was a sharp decline in conception rate per cycle; furthermore, on conception, the proportion of all the conceptions ending in EPL was higher among women with delayed times to clinical pregnancy. Furthermore, EPL was detected in 14.0% of all the cycles without clinical pregnancy over the first 12 months. Table 3 shows EPL among women subgroups defined by time to clinical pregnancy. Women with delayed time to clinical pregnancy tend to have lower frequency of EPL (P value for trend test, .205).
      Figure thumbnail GR1
      FIGURE 1Per-cycle conception rate among women with different time to clinical pregnancy. ■, early pregnancy loss; □, clinical pregnancy.
      Wang. Conception, early pregnancy loss, time to clinical pregnancy. Fertil Steril 2003.
      TABLE 3Early pregnancy loss among nonclinical pregnancy cycles by women with different times to clinical pregnancy.
      Time to clinical pregnancy
      Time to clinical pregnancy is the number of menstrual cycles it takes for a woman from the time of stopping contraceptive methods and attempting to conceive until achieving a clinical pregnancy.
      No. of womenEarly pregnancy loss
      Early pregnancy loss is clinically unrecognized early pregnancy detected only by the highly sensitive and specific urinary hCG assay.
      among nonclinical pregnancy cycles, n (%)
      Total,
      Trend test for the Total column: P=.205.
      n (%)
      12345
      1156CP
      212022 (18.3)CP22 (18.3)
      37816 (20.5)11 (14.1)CP27 (17.3)
      4408 (20.0)6 (15.0)7 (17.5)CP21 (17.5)
      5298 (27.6)4 (13.8)3 (10.3)4 (13.8)CP19 (16.4)
      6183 (16.7)2 (11.1)1 (5.6)1 (5.6)3 (16.7)10 (11.1)
      Total44157 (20.0)23 (13.9)11 (12.6)5 (10.6)3 (16.7)99 (16.4)
      Note: CP, clinical pregnancy; any pregnancy that lasted 6 weeks (42 days) or more after the onset of last menstrual period. If a woman’s time to clinical pregnancy is 2, then she has one nonclinical pregnancy cycle before CP; if a woman’s time to clinical pregnancy is 3, then she has two nonclinical pregnancy cycles before CP; and so on. If we look at the table by row, the data in each row show the number and the percentage of early pregnancy loss in each of the nonclinical pregnancy cycles among women with a given time to pregnancy. Data in each column show the number and the percentage of early pregnancy loss in a given nonclinical pregnancy cycle by women with different times to clinical pregnancy.
      Wang. Conception, early pregnancy loss, time to clinical pregnancy. Fertil Steril 2003.
      a Time to clinical pregnancy is the number of menstrual cycles it takes for a woman from the time of stopping contraceptive methods and attempting to conceive until achieving a clinical pregnancy.
      b Early pregnancy loss is clinically unrecognized early pregnancy detected only by the highly sensitive and specific urinary hCG assay.
      c Trend test for the Total column: P=.205.

       Early pregnancy loss in relation to subsequent reproductive outcomes

      Figure 2 shows the Kaplan-Meier curves of cumulative conception rates by the number of menstrual cycles, stratified by EPL status. Women with EPL (line 1) had a higher cumulative conception rate than did those women without EPL (line 2). The conditional risk ratio (line 2 as reference) estimated by Cox model is 1.6 (95% confidence interval [CI], 1.2–2.1; P=.0017). As shown in Table 4, , EPL in the preceding cycle was associated with increased odds of conception (odds ratio [OR], 2.6; 95% CI, 1.8–3.9), clinical pregnancy (OR, 2.0; 95% CI, 1.3–3.0), and EPL (OR, 2.4; 95% CI, 1.4–4.2) but was not associated with increased odds of spontaneous abortion (OR, 1.1; 95% CI, 0.4–3.3), low birth weight (<2,500 g; OR, 1.7; 95% CI, 0.6–4.7), and preterm birth (at <37 weeks; OR, 1.4; 95% CI, 0.4–4.7) in the subsequent cycle.
      Figure thumbnail GR2
      FIGURE 2Kaplan-Meier curves of cumulative conception rates by the number of menstrual cycles, stratified by early pregnancy loss status.
      Wang. Conception, early pregnancy loss, time to clinical pregnancy. Fertil Steril 2003.
      TABLE 4Early pregnancy loss in previous cycles in relation to pregnancy outcomes in subsequent cycles.
      Current cycleEarly pregnancy loss
      Early pregnancy loss, clinically unrecognized early pregnancy detected only by the highly sensitive and specific urinary hCG assay; clinical pregnancy, any pregnancy that lasted 6 weeks (42 days) or more after the onset of last menstrual period; total conception, both early pregnancy loss and clinical pregnancy.
      in previous cycle
      No (Total cycles, 954)Yes (Total cycles, 149)OR
      Adjusted for age, body mass index, noise exposure (low/moderate/high), dust exposure (low/moderate/high), husband’s cigarette smoking status (yes/no), and perceived stress during work (low/moderate or high).
      95% CI
      Conception status
      No conception (reference)589561.0
      Total conception
      Early pregnancy loss, clinically unrecognized early pregnancy detected only by the highly sensitive and specific urinary hCG assay; clinical pregnancy, any pregnancy that lasted 6 weeks (42 days) or more after the onset of last menstrual period; total conception, both early pregnancy loss and clinical pregnancy.
      315832.61.8–3.9
      P<.01.
      Clinical pregnancy
      Early pregnancy loss, clinically unrecognized early pregnancy detected only by the highly sensitive and specific urinary hCG assay; clinical pregnancy, any pregnancy that lasted 6 weeks (42 days) or more after the onset of last menstrual period; total conception, both early pregnancy loss and clinical pregnancy.
      259512.01.3–3.0
      P<.01.
      Outcome of conception
      Live birth (reference)208461.0
      Early pregnancy loss56322.41.4–4.2
      P<.01.
      Clinical spontaneous abortion2151.10.4–3.3
      Outcome of clinical pregnancy
      Birth weight
      ≥2,500 g (reference)184391.0
      <2,500 g1761.70.6–4.7
      Gestational age
      ≥37 weeks (reference)191421.0
      <37 weeks1541.40.4–4.7
      Note: Generalized Estimation Equation (GEE) was used to accommodate correlations among cycles from a given woman.
      Wang. Conception, early pregnancy loss, time to clinical pregnancy. Fertil Steril 2003.
      a Early pregnancy loss, clinically unrecognized early pregnancy detected only by the highly sensitive and specific urinary hCG assay; clinical pregnancy, any pregnancy that lasted 6 weeks (42 days) or more after the onset of last menstrual period; total conception, both early pregnancy loss and clinical pregnancy.
      b Adjusted for age, body mass index, noise exposure (low/moderate/high), dust exposure (low/moderate/high), husband’s cigarette smoking status (yes/no), and perceived stress during work (low/moderate or high).
      * P<.01.

      Discussion

      This study has several strengths. It is a large population-based prospective study using the highly sensitive and specific hCG assay to detect EPL. The study subjects consisted of a cohort of newly married women who planned to become pregnant over the course of the study. Therefore these pregnancies were not interrupted or confounded by unplanned pregnancies, induced abortions (except for medical reasons), or contraceptive use. These women began to record daily diaries and to collect urine samples immediately after they stopped contraception.
      Missing data is a methodological challenge that has not been adequately addressed in previous studies. Detecting EPL requires measurements of daily urinary hCG over the entire menstrual cycle. In reality, even with the most well-executed study, a portion of observed menstrual cycles will have missing hCG values, primarily because of missing urine collection. Although these missing values will not affect the estimate for the rate of clinical pregnancy, they can affect the estimate for the rates of conception and EPL. We applied Bayesian methods (
      • Gelman A.
      • Carlin J.B.
      • Stern H.S.
      • Rubin D.B.
      ,
      • Dunson D.B.
      A Bayesian model for fecundability and sterility.
      ) for handling missing hCG data; otherwise, we would have thrown out 30% of the cycles with ≥3 consecutive days with missing hCG data in the critical window (−10 to +5 days). Detailed methods will be described in a subsequent statistical paper. Our approach is a clear improvement over the standard approach that ignores the missing data and thus likely underestimates the true probability of total conception and pregnancy loss.
      We found that the overall per cycle conception rate was 40% over the first 12 months. The rate of clinical pregnancy was 30%. Approximately one third of all the conceptions detected by urinary hCG assay failed to survive to delivery. More than two thirds of these losses occurred before the pregnancy had been clinically recognized. Our findings are in general agreement with the observations made by Wilcox et al. (
      • Wilcox A.J.
      • Weinberg C.R.
      • O’Connor J.F.
      • Baird D.D.
      • Schlatterer J.P.
      • Canfield R.E.
      • et al.
      Incidence of early loss of pregnancy.
      ) and Zinaman et al. (
      • Zinaman M.J.
      • O’Connor J.
      • Clegg E.D.
      • Selevan S.G.
      • Brown C.C.
      Estimates of human fertility and pregnancy loss.
      ), the two prospective studies that used the highly sensitive and specific urinary hCG assay.
      With its larger sample size, this study further extended the previous understanding of conception and EPL. As shown in Figure 1, delayed time to clinical pregnancy was, to a great extent, associated with decreased conception rate, and, to a much lesser extent, with increased frequencies of EPL. There are few data depicting how early EPL in previous cycles could affect subsequent reproductive outcomes (
      • Wilcox A.J.
      • Weinberg C.R.
      • O’Connor J.F.
      • Baird D.D.
      • Schlatterer J.P.
      • Canfield R.E.
      • et al.
      Incidence of early loss of pregnancy.
      ). We found that on average, EPL was detected in 14% of all the cycles without clinical pregnancy. However, those women with delayed time to clinical pregnancy appeared to have lower frequencies of EPL. We also found that EPL in the preceding cycles was associated with a significant increase in the odds of total conception, clinical pregnancy, and EPL in the subsequent cycles. Our data lend support to a previous notion (
      • Wilcox A.J.
      • Weinberg C.R.
      • O’Connor J.F.
      • Baird D.D.
      • Schlatterer J.P.
      • Canfield R.E.
      • et al.
      Incidence of early loss of pregnancy.
      ) that EPL is apparently a positive indicator that the stages of reproduction leading to implantation are intact. Furthermore, our data showed that EPL was not associated with increased odds of clinical spontaneous abortion, low birth weight, or preterm birth in the subsequent cycles.
      Several limitations need to be taken into account when interpreting the results of this study. We were only able to detect EPL in which detectable levels of hCG were produced and excreted in the women’s urine. The true rate of EPL may be higher. All the women subjects were shift workers in a textile industry; the other major occupational exposures in this cohort were dust, noise, heat, and high humidity. The healthy-worker effect often exists in occupation-based studies; that is, persons of good general health status are selected to be employed in industry (
      • Monson R.R.
      ). However, >90% of all reproductive-aged women in urban China work outside the home, and job placement is not dependent on reproductive status. Furthermore, these women were young and nulliparous. Therefore, caution is needed before generalizing our findings to other populations.

      References

        • Modvig J.
        • Schmidt L.
        • Damsgaard M.T.
        Measurement of total risk of spontaneous abortion.
        Am J Epidemiol. 1990; 132: 1021-1038
        • Wilcox A.J.
        Surveillance of pregnancy loss in human populations.
        Am J Ind Med. 1983; 4: 285-291
        • Wilcox A.J.
        • Weinberg C.R.
        • Wehmann R.E.
        • Armstrong E.G.
        • Canfield R.E.
        • Nisula B.C.
        Measuring early pregnancy loss.
        Fertil Steril. 1985; 44: 366-374
        • Wilcox A.J.
        • Weinberg C.R.
        • O’Connor J.F.
        • Baird D.D.
        • Schlatterer J.P.
        • Canfield R.E.
        • et al.
        Incidence of early loss of pregnancy.
        N Engl J Med. 1988; 319: 189-194
        • Zinaman M.J.
        • O’Connor J.
        • Clegg E.D.
        • Selevan S.G.
        • Brown C.C.
        Estimates of human fertility and pregnancy loss.
        Fertil Steril. 1996; 65: 503-509
        • Baird D.D.
        • Wilcox A.J.
        • Weinberg C.R.
        Use of time to pregnancy to study environmental exposures.
        Am J Epidemiol. 1986; 124: 470-480
        • Gelman A.
        • Carlin J.B.
        • Stern H.S.
        • Rubin D.B.
        Bayesian data analysis. Chapman and Hall, London1997
        • Dunson D.B.
        A Bayesian model for fecundability and sterility.
        J Am Stat Assoc. 2000; 95: 1054-1062
        • O’Connor J.F.
        • Schlatterer J.P.
        • Birken S.
        • Krichevsky A.
        • Armstrong E.G.
        • McMahon D.
        • et al.
        Development of highly sensitive immunoassays to measure human chorionic gonadotropin, its beta subunit and beta core fragment in the urine.
        Can Res. 1988; 48: 1361-1366
        • Husdan H.
        • Rapoport A.
        Estimation of creatinine by the Jaffe reaction. A comparison of three methods.
        Clin Chem. 1968; 14: 222-238
        • Liang K.-Y.
        • Zeger S.L.
        Longitudinal data analysis using generalized linear models.
        Biometrica. 1986; 73: 13-22
        • Monson R.R.
        Occupational epidemiology. CRC Press, Boca Raton, FL1990