Premenstrual syndrome (PMS) is characterized by a range of physical and emotional symptoms that occur during the luteal phase and resolve with menses.1,2 Symptoms can vary from mild to severe and are typically associated with reduced occupational productivity, decreased health-related quality of life, and impairment in social relationships.3–6 Epidemiologic studies suggest that moderate or severe PMS is relatively common, with estimates as high as 24–32%.4,7 Milder physical and emotional symptoms associated with menses may occur in up to 80% of women.5
Current guidelines for the diagnosis of premenstrual disorders require prospective daily symptom charting for two cycles.1,2 However, prospective charting can be difficult and is not commonly performed.8 Alternatively, retrospective assessments are more practical, but currently available methods have low specificity and are considered unreliable.9–11 An effective screening tool for PMS would identify the population of women more likely to have clinically significant PMS from the larger group of women with milder symptoms. A positive screening test could then trigger prospective symptom charting and enhance the likelihood that these receive an appropriate diagnostic work-up, including prospective symptom charting.
As with PMS, the more severe variant of the condition, premenstrual dysphoric disorder, is also characterized by variability in the timing of onset, severity, and duration of premenstrual symptoms.12 In this study, 34–46% of women continued to have moderate or severe symptoms the day after menses. Another study found that symptoms peaked on the first day of menses.12,13 Given that the peak period of symptom expression occurs around the onset of menses, this interval may constitute an ideal point in the menstrual cycle to administer a screening instrument. The onset of menses also provides an easy-to-identify marker that can be used to standardize scores among women.
The Daily Record of Severity of Problems (DRSP), a validated daily symptom chart widely used for the diagnosis of PMS and premenstrual dysphoric disorder, includes 21 items grouped into 11 domains that address the criterion symptoms of premenstrual dysphoric disorder (depression, anxiety, lability, anger, interest in activities, concentration, lethargy, appetite, sleep, control, and physical symptom).1,14 Each item is rated on a scale of 1 (“not at all”) to 6 (“extreme”). We evaluated the potential use of DRSP scores from day one of menses as a screening instrument for PMS by comparing these scores to the standard of two cycles of prospective daily symptom ratings.
MATERIALS AND METHODS
This study was approved by the Institutional Review Board of the Cedars-Sinai Health System. The study population was comprised of a non–treatment-seeking group of women, 18–45 years of age, who were enrolled in a large medical group in southern California from July 1, 1998, through December 31, 1999. The study methodology has been described in detail elsewhere.15 Briefly, women were initially contacted by telephone, invited to participate in the study, and provided informed consent. A screening survey and the Medical Outcomes Study Short Form-36 (SF-36) were then administered. The screening survey collected information regarding demographics, menstrual cycle length, recent pregnancy, current oral contraceptives use, and use of nonprescription medications or calcium to treat premenstrual symptoms. Women who had irregular menses, were pregnant, or were within 1 year postpartum were excluded. Eligible participants were asked to complete a daily diary, which included the 21-item DRSP symptom instrument and three occupational productivity questions. Diaries were kept for two consecutive cycles (64 days). For purposes of this analysis, participants who did not complete the daily questionnaire for two cycles were excluded.
The sample was randomly divided into a “model-building” set (approximately 70%) and a “testing” set (approximately 30%) before data analysis. Based on the total number of items in the DRSP and the total sample size of respondents, sample sizes for both the model-building and testing sets were sufficiently large enough to allow for the reliable assessment of covariates. Using a random number generator in SAS (SAS Institute Inc., Cary, NC), each woman was assigned a random number drawn from a uniform distribution ranging from 0 to 1; those assigned a number greater than 0.7 were assigned to the testing set. The creation of the models and determination of appropriate cutoff scores were conducted on the model-building set.
The DRSP includes 21 items that describe both emotional and physical premenstrual symptoms. These 21 items are grouped into 11 questions that represent the 11 premenstrual dysphoric disorder symptom domains described in the Fourth Edition of the Diagnostic and Statistical Manual of Mental Disorders of the American Psychiatric Association (DSM IV).1 The scores for any single domain are calculated as the sum of the individual items that comprise the domain. For example, domain 1 of the DRSP has three items that assess depressive mood symptoms. To score domain 1, the ratings on each of the three items (1a-1c) are summed (range 3–18). Total DRSP scores are then calculated as the sum of the individual domain scores (range 21–126). We alternatively calculated DRSP scores by retaining only the score of the highest-ranked item within domains. Thus, if the ratings for 1a, 1b, and 1c were “2,” “4,” and “5,” respectively, the score for domain 1 would be “5”. The range of possible total DRSP scores using this alternative method is 11–66.
We further explored the performance of shortened versions of the DRSP. A “best subsets” selection process was used to identify a subset of the 21 DRSP items that fit the data sufficiently. Model fitting statistics and a priori knowledge were used to identify a reduced covariate model. Best subsets regression is a model-building technique that helps determine which subset or subsets of predictor (independent) variables perform well at predicting responses on a dependent variable; thus, all possible subsets of independent variables are examined to determine the best subset of covariates. This method involves examining all possible subsets of models created from all possible combinations of covariates. Model fitting statistics including Mallows Statistic (C-p) (for assessing fit), the adjusted R2 (determining if the model was overfit), and the standard deviations are evaluated to find the most parsimonious model. A priori knowledge was also used to identify a reduced covariate model. We then used backward step-wise regression to further reduce the number of items included in the model. Using backwards regression, the covariate with the least significant coefficient was removed using the t statistic with a P<.10 criterion until all remaining variables had a significance greater than 0.10.
The reference standard for diagnosis was the presence or absence of PMS based on the two cycles of daily symptom charting. The luteal phase was defined as the 5 days before menses and the follicular phase as days 6–10 after the onset of menses. Participants were considered to have moderate or severe PMS based on daily diaries when they met the following criteria in at least one of two consecutive cycles: 1) mean follicular phase scores were 30% or less of mean luteal phase scores for at least one item among any of three different symptoms and 2) mean luteal phase scores were 3 or more for at least three items. This definition of PMS was previously shown to have construct validity.15 After applying this algorithm, participants were categorized as either PMS positive or PMS negative during the two menstrual cycles of observation.
Statistical analyses were conducted using SAS 9 statistical software (SAS Institute Inc, Cary, NC). Differences in categorical variables were assessed using two-tailed χ2 tests; Fisher exact test was used when expected cell counts were 5 or less. Differences in continuous variables were calculated using two-tailed Student t tests.
A receiver operating characteristics curve analysis was used to ascertain the reliability and predictive value of DRSP scores on the first day of menses in comparison with a diagnosis of PMS established using the reference standard.16 Likelihood ratios were used to select “optimal” cutoff scores, thus maximizing the area under the receiver operating characteristics curve.
Of 1,578 women contacted, 697 (44.2%) met eligibility criteria agreed to participate, of whom 388 (55.7%) completed the daily questionnaire for two cycles (Fig. 1). Among these participants, 313 (80.7%) rated 2 or more symptoms at least 3 (“mild”) or greater on the first day of menses. There were no statistically significant differences in SF-36 scores, demographics, oral contraceptive use, or the use of prescription and over-the-counter medications to treat premenstrual symptoms between study completers and noncompleters (Table 1).
Baseline characteristics of participants categorized by PMS screening status and by PMS diagnosis are reported in Table 2. A greater proportion of participants who screened positive for PMS reported the use of prescription medications to relieve PMS symptoms compared with those who screened negative. No other differences in baseline symptoms associated with screening status were observed. The prevalence of PMS based on the reference standard of two cycles of prospective daily symptom charting was 30.4%. The “best-fit” model included 16 of the original 21 DRSP items; backwards regression then reduced this number to seven items.
The receiver operating characteristics curves comparing DRSP scores on the first day of menses with a diagnosis of moderate or severe PMS based on two cycles of daily symptom ratings are presented in Figure 2 A and B. For the 21-item DRSP, optimal cutoff values (see MATERIALS AND METHODS for description of how an optimal cutoff was identified) were a score 50 using the standard scoring method and a score of 31 using the alternative scoring method.
The score for the seven-item abbreviated version of the DRSP are derived from the equation:
(–0.7524)*hopeless + (0.3944)*worthless + (0.257)*mood swings + (0.2546)*sensitive + (0.3968)*concentration + (0.2009)*cravings + (0.2148)*bloating + (–3.6348).
The optimal cutoff values (see MATERIALS AND METHODS for description of how an optimal cutoff was identified) for the seven-item abbreviated DRSP was determined to be 0.82 (range –3.15 to 2.62). The area under the curve, sensitivity and specificity, positive predictive value, and negative predictive values for the different 21-item DRSP scoring methods and the seven-item DRSP version derived from the model-building set are shown in Table 3. Testing set results are presented in Table 4.
In the testing set, compared with participants with DRSP symptom scores below the cutoff values, participants who screened positive for PMS had statistically significantly lower mean SF-36 mental component summary scores but not physical component summary scores (Tables 5 and 6). All four SF-36 mental domain scores (vitality, social functioning, role-emotional and mental health) and one of four physical domain scores (bodily pain) were significantly lower among those who screened positive for PMS, irrespective of the scoring method or screening instrument used. Participants who screened positive for PMS also recorded a greater mean number of days absent per month and mean number of workdays with at least a 50% reduction in productivity.
A diagnosis of PMS based on current guidelines requires the prospective daily charting of symptoms and impairment for at least two cycles. However, this requirement places a significant burden on the patient and the health care provider and is rarely completed in clinical settings. The persistence of premenstrual symptoms into the follicular phase in the majority of women with PMS presents the opportunity to obtain diagnostic information at one readily identifiable point in time. The results of this analysis suggest that inspection of DRSP scores on the first day of menses may help to identify a population of women with a relatively high probability of having clinically significant PMS or premenstrual dysphoric disorder. These women should then undergo further diagnostic assessment to rule out other emotional or physical disorders.17,18
The negative predictive value of scores of both the 21-item and seven-item DRSP on the first day of menses were high, although the positive predictive values were only modest. The concordance of the model-building and testing set results support the reliability of these findings. DRSP scores above optimal cutoff values were associated with reduced occupational productivity, higher rates of absenteeism, and lower SF-36 mental component summary but not physical component summary scores. We have previously demonstrated that the difference in SF-36 physical component summary scores among women with and without PMS based on two cycles of prospective daily symptom charting to be small, although significant.19
The results of these analyses suggest that 21-item or seven-item DRSP scores below optimal cutoff values significantly reduce the likelihood participants would meet criteria for clinically significant PMS were they to complete daily symptom diaries for two consecutive cycles. An appropriate evaluation of women with DRSP scores above threshold values would include the use of daily ratings to ascertain symptom and impairment severity as well as symptom cyclicity within the menstrual cycle.
Unlike the cross-sectional analysis of symptoms we used in this study, previous attempts to develop screening procedures for PMS have used retrospective instruments.20–22 However, retrospective reports of premenstrual symptoms are often not confirmed by ratings made during the symptomatic phase of the menstrual cycle.23 Use of a daily rating instrument, such as the DRSP, during a period of active symptoms obviates the need for the recall of symptom severity and timing. Daily Record of Severity of Problems ratings made daily may then be used to establish a diagnosis of PMS or premenstrual dysphoric disorder and subsequently evaluate response to treatment.
This analysis has several important limitations. The study data used to test the usefulness of the DRSP as a screening instrument for PMS were collected as part of an epidemiologic study and were not originally intended for the purpose. Although the finding of significant functional impairment and reduced health-related quality of life in women who screened positive for PMS using the DRSP on the first day of menses is reassuring, the results of this analysis require prospective confirmation. Moreover the generalizability and reproducibility of these findings are not known. Only 55.7% of those deemed eligible for the study contributed data to the analysis. The screening performance characteristics of the DRSP in comparison to other daily rating forms were not evaluated. Moreover, this study did not determine the discriminative ability of the screener to identify women with PMS compared with various other emotional conditions.
In conclusion, a screening procedure for PMS based on either 21-item or seven-item DRSP symptom ratings on the first day of menses was found to have acceptable test characteristics. These preliminary findings support the use of the DRSP for the identification of women who are most likely to benefit from a more thorough diagnostic evaluation.
1. Premenstrual dysphoric disorder. In: First MB, editor. Diagnostic and statistical manual of mental disorders: DSM-IV-TR 4th ed. Washington (DC): American Psychiatric Association; 2000. p. 771–4.
2. American College of Obstetricians and Gynecologists. Clinical management guidelines for obstetricians–gynecologists: premenstrual syndrome. ACOG Practice Bulletin 15. Washington, DC: ACOG; 2000.
3. Halbreich U, Borenstein J, Pearlstein T, Kahn LS. The prevalence, impairment, impact, and burden of premenstrual dysphoric disorder (PMS/PMDD). Psychoneuroendocrinology 2003;28:1–23.
4. Campbell EM, Peterkin D, O’Grady K, Sanson-Fisher R. Premenstrual symptoms in general practice patients. Prevalence and treatment. J Repro Med 1997;42:637–46.
5. Hylan TR, Sundell K, Judge R. The impact of premenstrual symptomatology on functioning and treatment-seeking behavior: experience from the United States, United Kingdom and France. J Womens Health Gend Based Med 1999;8:1043–52.
6. Chawla A, Swindle R, Long S, Kennedy S, Sternfeld B. Premenstrual dysphoric disorder: is there an economic burden of illness? Medical Care 2002;40:1101–12.
7. Wittchen HU, Becker E, Lieb R, Krause P. Prevalence, incidence and stability of premenstrual dysphoric disorder in the community. Psychol Med 2002;32:119–32.
8. Kraemer GR, Kraemer RR. Premenstrual syndrome: diagnosis and treatment experiences. J Womens Health 1998;7:893–907.
9. Nogueira Pires ML, Calil HM. Clinical utility of the premenstrual assessment form as an instrument auxiliary to the diagnosis of premenstrual dysphoric disorder. Psychiatry Res 2000;94:211–9.
10. Ainscough CE. Premenstrual emotional changes a prospective study of symptomatology in normal women. J Psychosom Res 1990;34:35–45.
11. Taylor RJ, Fordyce ID, Alexander DA. Relationship between personality and premenstrual symptoms: a study in five general practices. Br J Gen Pract 1991;41:55–7.
12. Pearlstein T, Yonkers KA, Fayyad R, Gillespie JA. Pretreatment pattern of symptom expression in premenstrual dysphoric disorder. J Affect Disord 2005;85:275–82.
13. Meaden PM, Hartlage SA, Cook-Karr J. Timing and severity of symptoms associated with the menstrual cycle in a community-based sample in the Midwestern United States. Psychiatry Res 2005;134:27–36.
14. Endicott J, Nee J, Harrison W. Daily Record of Severity of Problems (DRSP): reliability and validity. Arch Womens Ment Health 2006;9:41–9.
15. Borenstein JE, Dean BB, Endicott J, Wong J, Brown C, Dickerson V, et al. Health and economic impact of the premenstrual syndrome. J Reprod Med 2003;48:515–24.
16. Metz CE. Basic principals of ROC analysis. Semin Nucl Med 1978;8:283–98.
17. Freeman EW. Premenstrual syndrome and premenstrual dysphoric disorder: definitions and diagnosis. Psychoneuroendocrinology 2003; 28 suppl:25–37.
18. Dell DL. Premenstrual syndrome, premenstrual dysphoric disorder, and premenstrual exacerbation of another disorder. Clin Obstet Gynecol 2004;47:568–75.
19. Dean BB, Borenstein JE, Knight K, Yonkers K. Evaluating the criteria used for identification of PMS. J Womens Health (Larchmt) 2006;15:546–55.
20. Steiner M, Macdougall M, Brown E. The premenstrual symptoms screening tool (PSST) for clinicians. Arch Womens Ment Health 2003;6:203–9.
21. Halbreich U, Endicott J, Schacht S, Nee J. The diversity of premenstrual changes as reflected in the Premenstrual Assessment Form. Acta Psychiatr Scand 1982;65:46–65.
22. Moos RH. The development of a menstrual distress questionnaire. Psychosom Med 1968;30:853–67.
© 2007 by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. All rights reserved.
23. Endicott J, Halbreich U. Retrospective report of premenstrual depressive changes: factors affecting confirmation by daily ratings. Psychopharmacol Bull l982;18:109–12.