The search resulted in a sample of 2,132 titles (353 MEDLINE, 429 Web of Science, 159 Cochrane Database of Systematic/Database of Abstracts of Reviews of Effects (DARE), 937 Embase, 159 PsycINFO, and 95 Cinahl). From these citations, we identified 325 potential controlled trials with data on the association between SSRIs and PMS, premenstrual dysphoria, premenstrual dysphoric disorder, or late luteal phase dysphoric disorder. Crossover trials were excluded for not presenting data at the end of the first phase. After excluding duplicates, we retrieved 163 articles for full-text review, of which 19 were included in the final meta-analysis. For data sets that were presented in multiple publications, we selected those with the most up-to-date results, longest follow-up, or most pertinent outcomes. All data elements relative to the meta-analyses had two reviewers who came to consensus on all items. See Appendix 2
(available online at www.greenjournal.org/cgi/content/full/111/5/1175/DC2
) for a graphic of trial flow. Studies that are included in the meta-analyses are listed in . Other studies with either a drug and/or outcome of interest that did not meet inclusion criteria are listed in Appendix 3
(available online at www.greenjournal.org/cgi/content/full/111/5/1175/DC3
), along with reasons for their exclusion. Nonparametric tests for publication bias resulted in no studies being trimmed or filled (), suggesting a low likelihood of important studies that were missed.
Randomized Controlled Trials Included in Meta-Analysis
Twenty-nine randomized controlled trials (defined as a comparison between an SSRI and placebo for the treatment of PMS/premenstrual dysphoric disorder) in 19 published manuscripts met all inclusion criteria. Meta-analysis of these 29 studies, including 2,964 women, results in an overall odds ratio (OR) of 0.40 (95% CI 0.31-0.51), suggesting a strong association between the use of SSRIs and a reduction in PMS/premenstrual dysphoric disorder symptoms. Heterogeneity (I2=66%) was examined in subsequent meta-regression. A summary of the treatment effect point estimates and 95% confidence intervals for each study included in the meta-analysis is listed in . A forest plot for the overall treatment effect is shown in .
Odds Ratios For Primary Outcome Measurement Instruments
Fig. 2 Pooling of 29 studies favors treatment with selective serotonin reuptake inhibitors over placebo control for premenstrual syndrome and premenstrual dysphoric disorder. The pooled effect size (standardized mean difference) of -0.50 (-0.64 to -0.37) corresponds (more ...)
We conducted meta-regression based on prespecified covariates and on an exploratory basis. Among the exploratory analyses, only the study country showed some evidence of a relationship to the effect size; however, sensitivity analysis excluding the one study conducted outside of North America/Europe showed no appreciable change in the pooled effect size in the meta-analysis. No other variables examined in exploratory analyses (including pharmaceutical sponsorship of the study)34
were found to be both significant and clinically relevant.
Some studies reported on outcomes related to PMS, the less severe version of the disorder compared with premenstrual dysphoric disorder. When results are stratified by PMS compared with premenstrual dysphoric disorder (), the pooled effect size for PMS is OR 0.38 (95% CI 0.22-0.66), whereas for premenstrual dysphoric disorder is OR 0.40 (95% CI 0.30-0.53). There is still significant within-strata heterogeneity (significant I2
of 67% for both PMS and premenstrual dysphoric disorder). also suggests that, with the exception of the studies by Veeninga15
earlier studies tend to report a larger treatment effect for SSRIs in both PMS and premenstrual dysphoric disorder.
Fig. 3 The pooled effect size (standardized mean difference) for studies of premenstrual syndrome is -0.53 (95% CI -0.83 to -0.23), which corresponds to an odds ratio of 0.38 (95% CI 0.22-0.66). The pooled effect size (standardized mean difference) for studies (more ...)
Intermittent dosing studies17-25
yielded a significantly smaller estimate of the treatment effect size (OR 0.55, 95% CI 0.45-0.68, I2
=20%) than continuous dosing studies (OR 0.28, 95% CI 0.18-0.42) with evidence of statistical heterogeneity (I2
Eleven studies allowed participants to adjust the dose of study medication according to their symptoms.17-24
These “flexible” dosing strategies, however, do not conform to the traditional clinical definition of symptomatic dosing. In our analysis, only one study that met inclusion criteria allowed patients to initiate medication upon symptoms, hence we were unable to shed further light on this issue in a pooled analysis. The symptom-onset dosing study reported a smaller effect size (OR 0.67, 95% CI 0.14-3.32) than either intermittent or continuous dosing studies; the wide confidence interval makes it difficult to generalize from this single study.
Fluoxetine, sertraline, and paroxetine were the most common SSRIs studied for PMS/premenstrual dysphoric disorder. The relative effect sizes for these SSRIs are presented in . No significant differences in effect sizes were found, and all were associated with improved symptoms except for fluvoxamine, which has only one small trial that met inclusion criteria, with wide confidence intervals allowing for the possibility of benefit or no benefit. Further subgroup analyses examining drug by dose, dosing regimen, or duration were not conducted, because there were too few studies to provide a comprehensive assessment of these issues.
Odds Ratios By Selective Serotonin Reuptake Inhibitor
lists the pooled treatment effect and odds ratios for the primary outcome assessment instrument used in each study. Eighteen studies used an ordinal scale. The Daily Record of Severity of Problems, used in seven studies, was the most commonly used instrument. A total of nine studies used a visual analog scale (VAS) to assess the primary outcome; the VAS-Mood was used in six of the studies and the VAS-Total in the remaining three studies. The pooled effect size for studies using the Daily Record of Severity of Problems (OR 0.58, 95% CI 0.46-0.75) is smaller than the pooled effect size for studies using the VAS-Mood (OR 0.38, 95% CI 0.27-0.54); however, the three studies using the VAS-Total report a markedly larger pooled effect size (OR 0.13, 95% CI: 0.09, 0.21) than either the Daily Record of Severity of Problems or VAS-Mood. Meta-regression results suggest that the choice of instrument may be associated with the pooled effect size. Stratification by Daily Record of Severity of Problems, VAS-Mood, and VAS-Total eliminates residual statistical evidence of heterogeneity.
All included studies received a Jadad score of 3 or higher. Exploratory meta-regression indicated only one component of the Jadad score was associated with the pooled effect size: whether the study described the method of randomization. Studies that failed to describe the method of randomization had a larger pooled effect size (OR 0.26, 95% CI 0.16-0.44) and greater heterogeneity (I2=73%) than studies that included a description of the method of randomization (OR 0.48, 95% CI 0.38-0.61; I2=43%). Because all studies had a Jadad score of 3 or greater, we did not exclude any studies in sensitivity analyses.