Study participants were members of the North Carolina Early Pregnancy Study, a prospective cohort study conducted from 1982 to 1985 involving 221 women who were attempting pregnancy (Wilcox et al. 1988
). These women had no known fertility problems or major chronic disease. Participation began when women discontinued contraception and continued until 8 weeks past their last menstrual period if they became pregnant, or for 6 months if they did not. During this time, participants collected daily first-morning urine specimens and kept daily diary records of their menstrual bleeding (number of pads and tampons). The study was approved by the Institutional Review Board of the National Institute of Environmental Health Sciences, and participating women provided informed consent.
Ovulation was identified using measurements of estrogen and progesterone metabolites in urine (Baird et al. 1991
). Pregnancy was detected using a highly sensitive and specific polyclonal radioimmunoassay for urinary human chorionic gonadotrophin (hCG) (Armstrong et al. 1984
). This assay measured intact and beta hCG with <1% cross-reaction with human luteinizing hormone (McChesney et al. 2005
). The lowest hCG concentration detectable by the immunoradiometric assay was 0.01 ng per milliliter. Three consecutive days of urinary hCG greater than 0.025 ng/mL was the criteria for pregnancy. This criterion ensured high specificity and was determined from a parallel study of women who had either undergone a tubal ligation. Pregnancy loss was detected by a subsequent fall in urinary hCG. We restricted our study to pregnancies that ended less than 6 weeks after the last menstrual period, with the day of loss considered to be the first day of vaginal bleeding associated with the fall in hCG (Wilcox et al. 1988
). Over-the-counter pregnancy test kits were not available at the time of the study, and only one woman who lost a pregnancy prior to six weeks gestation reported that it was clinically detected.
Following World Health Organization guidelines (World Health Organization Task Force on Adolescent Reproductive Health 1986
), we defined length of a bleed period as the number of days from the start of bleeding up to and including the day before a bleed-free interval of at least 2 days. Menstrual discharge consists of tissue fluid (20–40% of the total discharge) and fragments of the endometrium in addition to blood (50–80% of the total discharge); we refer to this discharge simply as bleeding, as is customary (Oats and Abraham 2005
We used the number of pads and tampons per day as a measure of bleeding quantity, focusing on comparisons within women. Women can differ from one another in their use of pads and tampons for reasons related more to personal habits than to volume of menstrual discharge. Because this variability between women is controlled in a within-woman analysis, it is more informative to conduct within-woman analyses of bleeding (comparing an early loss with the same woman’s regular menses) than to make comparisons across women. In support of this, studies comparing pad and tampon counts with quantitative measurement of blood loss across women have reported considerably higher correlation coefficients when the studies included multiple cycles per woman (0.61 (Higham and Shaw 1999
) and 0.74 (Higham et al. 1990
)) than when they did not (0.14 (Fraser et al. 1981
) and 0.30 (Warner et al. 2004
Women may also differ in their patterns of use of pads versus tampons. However, women in this study were fairly consistent from one period to the next in their use of pads versus tampons. The mean percentage of tampons used per period [100 × tampons/(pads + tampons)] was 62% and the mean within-woman standard deviation in the percentage of tampons used per period was 9%.
Forty-four women had 48 pregnancy losses prior to six weeks gestation during their study participation. One of these women had a single loss for which she reported no associated bleeding; because her reporting was incomplete in other respects, we believed her report of no bleeding could have been a recording error, and we have excluded her from this analysis. Of the remaining women, 36 provided complete data on bleed length and pad-and-tampon use for at least one regular menses in addition to a pregnancy loss bleeding episode. One non-conception cycle was excluded because no bleeding was reported. All but one woman in this sub-sample were white; most were college-educated (72%). The median age was 29 years (range 24–36 years), and 61% were parous at enrollment.
These 36 women in the analysis sample had 38 pregnancy losses prior to six weeks gestation (two women had two losses). Information on bleeding after an ovulatory non-conception cycle was available for 1 to 7 menses per woman (median = 3) for a total of 198 non-conception menses. Analyses were based on differences between the within-woman averages. For each woman, a value for each bleed characteristic was calculated for both the non-conception and the pregnancy loss bleeds. If a woman had more than one non-conception or pregnancy loss bleed, then the mean value of the bleed characteristic was used. A bleed characteristic difference was then computed by subtracting the non-conception bleed value from the pregnancy loss bleed value, and differences were tested with a Wilcoxon signed rank test. Non-parametric correlation statistics and tests for difference were used due to the inherent non-normality of some of the bleed characteristic variables. The 25th and 75th percentile values for the distributions are presented, but means and standard errors are also presented for ease of comparison with other published bleed length values. All P-values cited are two-sided.
The pregnancy losses represented a range of durations and amounts of hCG production, which we hypothesized might also be related to the pattern of bleeding. In order to explore this, we used two markers: maximum level of hCG obtained during the pregnancy, and the length of the failed pregnancy (days from ovulation to onset of bleeding).