Differences in symptom scores and health outcomes in premenstrual syndrome.

Background: In studies of premenstrual syndrome (PMS), a significant response to treatment is commonly defined as a 50% reduction in symptom scores, but empirical support for this definition is lacking. We compared healthcare utilization in women with and without PMS according to the Daily Record of Severity of Problems (DRSP) scores in order to determine the degree of symptomatic variation in premenstrual symptoms associated with differences in the burden of illness.

Methods: Participants were women aged 18-45 years enrolled in a medical group in southern California. Respondents completed the Medical Outcomes Study Short Form-36 (SF-36) at baseline and the DRSP symptom and occupational productivity items daily. Luteal phase DRSP scores were averaged over two consecutive cycles. Respondents were categorized as having mild/no and moderate/severe PMS or premenstrual dysphoric disorder (PMDD) using a previously published, validated algorithm. Medical costs were estimated from medical claims data over the 2 years prior to study entry.

Results: Compared with women with moderate/severe PMS/PMDD (n = 117), those with mild/no PMS/PMDD (n = 271) had 43% lower DRSP scores (29.7 and 52.4, p < 0.05), higher SF-36 mental component summary (49.9 and 40.5, p < 0.0001) and physical component summary (52.6 and 50.8, p = 0.04) scores, and fewer workdays per month with reduced productivity (13.3 and 22.0, p < 0.0001) and workdays missed due to health reasons (1.2 and 2.7, p = 0.001). Women with moderate/severe PMS/PMDD had greater odds of having >10 office visits (OR = 1.80, 95% CI 1.01, 3.22) and of accumulating >$500 in medical charges (OR = 1.9, 95% CI 1.2, 3.0).

Conclusions: A 43% difference in premenstrual vs. postmenstrual symptom scores is associated with a significant difference in healthcare burden. These data support the use of a 50% reduction in symptom ratings as a clinically relevant improvement in PMS/PMDD treatment trials, although smaller differences may also be meaningful.