Missing date of last menstrual period: Impact on assessment of hCG for women at risk for an ectopic pregnancy

Abstract

Describing the initial rise in hCG in early pregnancy is critical for the diagnosis of women at risk for ectopic pregnancy (EP). When date of last menstrual period (LMP) is unknown, the result is an undefined time axis for describing the profile. Estimates of expected rise, which can only be defined for the subset of subjects who have known LMP date, could potentially be biased. We utilize questionnaire data about the certainty of recall of LMP date to evaluate this bias in women ultimately diagnosed with a viable intrauterine pregnancy (IUP). Retrospective Cohort. Women presenting with pain/bleeding in first trimester of pregnancy, with no definitive diagnosis at first visit, were selected for study. Women were prospectively followed until establishment of a definitive diagnosis of IUP. At presentation, women were asked to classify their recall of LMP date as certain, uncertain, or unknown. The relationship between the hCG levels and LMP dates were modeled using regression methods that accommodate covariate measurement error. Of the 293 women who were ultimately diagnosed with IUP, 18.4 percent were uncertain or had unknown date of LMP. The 5 percent of women with no LMP information were assigned the average number of days since last LMP from the sample. Each category of LMP certainty was initially modeled separately. We found that the shape of the hCG profile for women of uncertain LMP was markedly similar to the profile derived from those women who were certain about the date. However, the amount of measurement error in the LMP values was much larger and dominated the profiles. This lead to very wide confidence bounds around the estimated hCG profiles. Thus, limiting their clinical relevance. Similar results were found using a single model for all the women, with varying errors for the certainty categories. While the measurement error in LMP was large, evaluation of hCG using the date of presentation (DP) for care as the time axis proved much more useful. Within the subset of women with known LMP, the slope for change in hCG determined using PD as the time axis was 0.402 95% CI (0.389,0.415) and gave similar result as using LMP which had slope 0.407 95% CI(0.395,04.19). Therefore, it is appropriate to make decisions about normal rise in hCG using PD. Many women who present with symptoms of pain and bleeding early in pregnancy are unsure of their LMP. The hCG rise of women with an IUP is very similar in those with certain or uncertain LMP. Estimating the LMP introduced far too much variability to be clinically useful. However, given that no bias in the estimated profile for hCG was detected for subjects with uncertain and unknown LMP we conclude that it is safe to use the date of presentation (and not LMP) to make decisions about hCG profiles for all women who present in early pregnancy who are at risk for EP.