Professional Documents
Culture Documents
Premenstrual Symptoms - Severity, Duration and Typology: An International Cross-Sectional Study
Premenstrual Symptoms - Severity, Duration and Typology: An International Cross-Sectional Study
Premenstrual Symptoms - Severity, Duration and Typology: An International Cross-Sectional Study
Menopause International
Clinical diagnost ic crit eria for premenst rual syndrome and guidelines for t heir quant ificat ion f…
Uriel Halbreich, Torbjörn Bäckst röm, Andrea Genazzani
Development and evaluat ion of prot ect ion/risk model based educat ion for prevent ion of adolescent b…
Fazlollah Ghofranipour
Original article
Abstract
Objectives. Determine women’s experiences of premenstrual symptoms.
Study design. Cross-sectional survey.
Sample. In all, 4085 women aged 14–49 years recruited by random telephone digit dialing in France,
Germany, Hungary, Italy, Spain, UK, Brazil and Mexico.
Main outcome measures. Telephone interview checklist of 23 premenstrual symptoms, sociodemographic
variables and lifestyle variables.
Results. The most prevalent symptoms were abdominal bloating, cramps or abdominal pain, breast tender-
ness, irritability and mood swings. Severity of symptoms is directly proportional to duration (R ¼ 0.79).
Hierarchical clustering found the following mental and physical domains and a typology: ‘Mild’ type (40.8%)
with minimal symptoms; ‘Moderate M’ type (28.7%) with moderately severe, mostly mental symptoms;
‘Moderate P’ type (21.9%) with moderately severe, mostly physical symptoms; and ‘Severe’ type (8.6%) with
severe intensity of both mental and physical symptoms. Multiple stepwise regression found significant effects
on symptom duration severity index of age (linear and quadratic effects), current smoking and country.
Conclusions. Further research is needed on the impact of premenstrual symptoms on quality of life,
and whether a brief symptom list could be developed as a valid and reliable tool globally.
n ¼ 531, Italy n ¼ 505, France n ¼ 501, UK n ¼ 500, symptom (during how many of the menstrual cycles in the
Spain n ¼ 500 and Hungary n ¼ 500) and Latin American last 12 months was the symptom experienced). The order of
countries (Brazil n ¼ 548 and Mexico n ¼ 500). symptoms in each group was changed between interviews.
Women who experienced any of the premenstrual symp-
toms were also asked a single question on overall severity
Subjects of these symptoms (mild, moderate or severe).
precision of the proportions is 1.5% and is not mentioned, of Kendall correlation coefficient R ¼ 0.79, mean coeffi-
except for small proportions. cient of determination ¼ 0.570, and range ¼ 0.502–
0.781. Thus, the severity of a symptom is directly pro-
portional to the duration. This relationship is true for all
Results the symptoms studied (minimum R ¼ 0.71). Moreover,
throughout a mixed model assuming correlation between
Sample description all the symptoms, we confirmed that the relationship
between duration and severity was not statistically
A total of 4085 women from Latin American (n ¼ 1048) different among symptoms, and not dependent on any
and European (n ¼ 3037) countries participated in the sociodemographic variables (age, educational level,
study. The response and refusal rates are not known. country).
The distribution of sociodemographic variables across
countries reveals differences (Table 1). Brazilian and
Mexican samples are characterized by a slightly younger
age. French and German women have a higher education
Classification of symptoms
level than women from other countries. Important dis- We attempt to classify the symptoms according to
parities exist in work participation, in particular the pro- their observed intercorrelation, under the assumption
portions of women in home duties (1% in UK–36% in that correlated prevalence between any two symptoms
Mexico). Income was also very different among countries provides some evidence of a common cause. The
with more than 40% reporting high income in the UK as Kendall correlation matrix of the symptoms was charac-
compared with 1% in Mexico. terized by a mean between-item correlation of R ¼ 0.23
with constantly positive correlations ranging from
R ¼ 0.08–0.69, and all the coefficients are significantly
Premenstrual symptom prevalence, higher than zero (P . 0.001). We carried out a
severity and duration hierarchical clustering (Ward tree algorithm) based
on the correlation matrix. Five groups emerged from
For each individual symptom (Table 2), we report in the tree:
decreasing importance the median of symptom severity
(S) (level 0 ¼ nil and 3 ¼ severe), duration (D) (mean (1) Depressed mood, anger, irritability, mood swings;
number of affected menstrual cycles in the last 12 (2) Anxiety, not in control, restlessness, hopelessness,
months) and the product (severityfrequency). tension;
We found a strong relationship between symptom dur- (3) Confusion, social withdrawal, sleep disturbances,
ation and severity for all the symptoms; mean coefficient poor concentration, lack of energy;
Age (mean) 33.18 32.09 32.03 32.23 32.28 29.64 31.47 28.65 31.44
Parity (mean) 1.20 1.36 1.07 1.10 1.14 0.87 0.97 1.24 1.12
Married (%) 79.5 80.8 79.8 79.7 78.6 65.7 75.8 65.2 75.6
Education (%)
Low 29.9 30.4 27.7 41.4 24.6 16.1 26.2 36.2 28.9
Middle 37.1 51.2 54.1 37.2 33.9 65.0 57.4 42.2 47.4
High 33.0 18.4 18.2 21.4 41.5 19.0 16.4 21.6 23.7
Work (%)
Full-time 31.8 29.8 27.2 35.3 43.7 30.7 52.8 17.2 33.5
Part-time 27.8 33.6 17.1 16.7 17.8 18.4 6.2 16.0 19.2
Student 16.8 16.8 24.0 20.2 22.2 20.3 24.6 29.2 21.7
Unemployed 4.7 18.4 2.2 7.9 4.4 10.9 3.8 2.0 6.8
Housewife 18.9 1.4 29.6 20.0 12.0 19.7 12.6 35.6 18.7
Income (%)
Low 42.0 17.3 37.8 10.0 47.4 55.0 24.2 72.8 40.9
Middle 42.4 39.1 54.4 71.3 43.4 34.8 55.0 26.2 44.0
High 15.6 43.6 7.8 18.7 9.2 10.2 20.8 1.0 15.0
Sport (%)
Never 27.6 21.8 57.1 42.7 51.5 54.2 39.2 43.2 42.2
Occasionally 62.8 53.4 35.5 31.6 39.5 28.3 49.4 33.0 41.7
Daily 9.6 24.8 7.4 25.7 9.0 17.5 11.4 23.8 16.1
Smoking (%)
No 66.5 72.2 72.3 61.4 70.1 83.6 66.8 86.2 72.5
Moderate 25.8 22.0 23.2 29.4 24.4 10.9 23.4 13.2 21.4
Severe 7.7 5.8 4.6 9.2 5.6 5.5 9.8 0.6 6.1
SD ¼ standard deviation
observed as linear, and we added a quadratic term. Only four regions. The majority of women experienced their first
significant predictors were found. The first predictor was age symptom before the age of 20 years (EU 67.7%; Latin
with a linear and quadratic effect, producing an inverse America 70.2%). Only a small percentage of women
u-curve where younger and older women were characterized in both regions experienced their first premenstrual
by a smaller intensity, with a maximum reached around 35 symptom over the age of 40 years (EU 2.2%; Latin
years (Exact value 34.5, 95% confidence interval [32.1, America 1.5%). We carried out a linear stepwise regression
36.9]). After age, the next important predictor was smoking in which no covariate among the available variables
habits: smokers are associated with an increase of 1.8 (1.18, showed a significant effect. No difference was found
2.42). Lastly, we found differences between countries, with between countries.
an increasing order from Germany, Spain, Italy, Hungary,
Mexico, France, Brazil and the UK. The range of change from
Germany (22.57) to the UK (4.04) is 6.67. Our final model is
characterized by a poor determination (R 2 ¼ 0.047), Discussion
underlying that many specific predictors of premenstrual
symptoms were not documented in this study (Table 3). We found that across all countries studied, physical
We studied the difference in symptom profile, calculating symptoms were the most prevalent symptoms reported,
the difference between the sum of mental domain symp- as assessed by severity and number of cycles affected.
toms (duration severity index) (DSIM) and the sum of Wittchen et al. 13 used data from a prospective longitudi-
physical domain symptoms DSIP divided by TDSI following nal community survey of 1488 German women aged
the expression (DSIM–DSIP)/TDSI. This ratio is always ,1, 14–24 years and found that physical complaints were the
and is positive or negative reflecting that the symptom most frequently endorsed premenstrual symptoms at
profile is predominantly physical or mental. We used a baseline and over all three waves of assessments (experi-
multiple stepwise linear regression to identify the main enced by 45% of the women). Physical symptoms were
predictors. The only variables with an effect on the mental– the most frequently reported premenstrual symptoms in a
physical domain difference were the following: (a) age with Japanese sample5 and among Australian twin registry
a decrease of 0.03 [20.05, 20.01], thus a relative increase of sample women,14 but not among two USA samples.11,15
physical symptoms compared with mental symptoms when The most prevalent symptom (abdominal bloating or
age increases; (b) work with an increase of 0.16 [0.05, 0.28] abdominal swelling) was greater than 10 times more
of mental symptoms for no paid work or housewife and, at important than the least prevalent (confusion). The
the opposite end, an increase in physical symptoms for standard deviations of the DSI index are large, which
women with full-time employment; (c) age at first suggests a large between-women variability. It remains to
symptoms seems to have an important increase of 0.40 be understood how this variability is distributed.
[0.21, 0.60] for each increasing year of the first symptoms; We found a strong relationship between severity and
(d) after adjustment for these first covariates, the country duration for all the symptoms. This indicates that severity
effect remains significant (P , 0.001) with mental com- of a symptom is directly proportional to the duration or
ponents highest for the UK and lowest for Germany. how many cycles were affected. In principle, a woman
reporting severe symptoms is also suffering from these
symptoms for most menstrual cycles. Thus, when
Variables associated with age at first approximation is needed, the severity level can be used
alone. It should be noted that in this survey women were
symptoms not asked how many days per cycle the symptoms lasted.
The age at which women experienced a premenstrual A discussion on the effects of predictors must take into
symptom for the first time was comparable between two account how the considered endpoint of the summed
or total TDSI has been constituted and measured to
appreciate the clinical relevance of possible effects. In this
Table 3 Predicting total duration severity index (TDSI)
context, there was only a very limited influence of the
Coeff. CI DF MS P value available covariates. It should be noted that oral contra-
ceptives prescribed in 2003 were not associated with the
Constant 2.847 [20.792, 6.486] 1 322.630 0.121 prevalence of premenstrual symptoms. Age has an
Age 0.560 [0.312, 0.808] 1 2688.552 ,0.001 obvious effect, with less symptoms found before 20 and
Age2 20.008 [20.012, 20.004] 1 2381.495 ,0.001
after 45 and a maximum around 35 years. However, the
Smoker 1.808 [1.185, 2.431] 1 4438.913 ,0.001
Country 7 3225.058 ,0.001
clinical relevance of the difference is modest, with a
Germany 22.539 [23.477, 21.601] slightly larger contribution from heavy smoking.
Spain 22.483 [23.447, 21.520] The residual variability is mainly taken into account by
Italy 21.549 [22.505, 20.594] country, but from Germany to Mexico the difference is
Hungary 20.705 [21.666, 0.256] not clinically significant. In this respect, Brazil and the UK
Mexico 20.390 [21.362, 0.581] are the only countries to have a significantly larger impact
France 0.178 [20.781, 1.137] on symptom prevalence. It is apparent that countries
Brazil 3.333 [2.406, 4.260] cannot be separated in their effect on premenstrual
UK 4.157 [3.198, 5.116] symptoms by cultural or geographical classifications.
Error 4074 137.110
The only conclusion is that premenstrual symptoms
CI ¼ confidence interval; DF ¼ degrees of freedom; MS ¼ mean square severity is likely influenced by many other variables not
(R 2 ¼ 0.054) documented here.
The finding that certain symptoms vary significantly by can be developed that will be a valid and reliable tool for
country suggests that cultural or other factors may sig- all countries studied.
nificantly contribute to awareness of these symptoms or
to variability in perception of severity of some symptoms.
Of interest, the UK was more likely to report mental Acknowledgments: We would like to acknowledge the
symptoms and to report more severe symptoms than assistance of TNS/EMNID, Germany, in the development of the
other European countries. In this respect, UK women are questionnaire and the following fieldwork agencies: OPERA, the
similar to US women as there are reports that US women UK; PLM, France; Demoscopia, Spain; Nomesis, Italy; MASMI,
Hungary; AC Nielsen, Brazil; Analitica, Mexico. Institutional
are also more likely to report affective symptoms.11 These
review board approval was not sought as this was a questionnaire
findings suggest an area for further research.
survey with no intervention. Women were asked to consent
A major strength of the current study is that the same verbally to answer questions in the telephone-administered
sampling, method and measures were used across conti- questionnaire survey about their menstrual cycle experience.
nents and that the results were not constrained to one of All results were stored without any identification data. This study
the classificatory systems. Another important strength of was supported by Bayer Schering Pharma, Women’s Health Care,
the current study is that it is a population-based study Germany. Data were analysed independently of the company by
with racial diversity. Of the numerous studies previously the first two authors.
conducted on premenstrual symptoms, only a few have
been population based and these have usually been Competing interests: Lorraine Dennerstein: expert advisory
board/speaker honoraria from Boehringer Ingelheim and Wyeth
limited to one particular country. Many studies have used
Pharmaceuticals, speaker honoraria Bayer Schering; Philippe
convenience samples such as those of white women
Lehert: regular Senior Consulting Statistician for the following
attending a US university or seeking treatment at health laboratories: Merck Kgaa, Sanofi-Aventis, Ipsen, Serono, Bayer
clinics. Results from such studies are not generalizable to Schering; Torbjorn Backstrom: Bayer Schering Pharma speaker
other more diverse populations of women. The current honoraria and Schering Plough grant application referee; Klaas
study also incorporated all symptoms listed by the current Heinemann: Bayer Schering Pharma employee.
classificatory systems and did not limit symptom experi-
ence to those required for PMDD, which is very much Accepted: 16 June 2009
weighted to psychological symptoms with only one
somatic group of symptoms among 11 symptom groups.
A major limitation to the study is reliance on retro-
spective self-report. Self-report measures can introduce References
bias, as interpretations can vary between women. The
retrospective approach to PMS studies has been criticized 1 American Psychiatric Association. Diagnostic and Statistical
as discrepancies have been found between retrospective Manual of Mental Disorders (DSM). 4th edn. Washington, DC: APA,
1994
and prospective ratings in the same population of 2 American College of Obstetricians and Gynecologists (ACOG).
patients.16 However, daily charting is difficult with up to Premenstrual syndrome: clinical management guidelines for
50% of women refusing.7,17 Some recent studies have Obstetricians – Gynecologists. Int J Gynaecol Obstet 2001;73:183
shown that retrospective reports can be accurate, even for 3 World Health Organisation, The ICD-10 classification of mental,
behavioural and developomental disorders. 10th rev. Geneva,
PMDD diagnosis.18,19
Switzerland, 1987
The current study was also limited in the information 4 Dean BB, Borenstein JE, Knight K, et al. Evaluating the criteria
collected on variables that may affect the prevalence of used for identification of PMS. J Womens Health (Larchmt)
premenstrual symptoms. A further limitation is the lack of 2006;15:546
record keeping on response rates, which limits the infer- 5 Takeda T, Tasaka K, Sakata M, et al. Prevalence of premenstrual
syndrome and premenstrual dysphoric disorder in Japanese
ences that can be drawn about the generalizability of the women. Arch Womens Ment Health 2006;9:209
data. However, other surveys have found similar results to 6 Adewuya AO, Loto OM, Adewumi TA. Premenstrual dysphoric
the present survey: that over 90% of women have at least disorder amongst Nigerian university students: prevalence,
one premenstrual symptom20 and that around 6–9% of comorbid conditions, and correlates. Arch Womens Ment Health
2008;11:13
women are reported as having severe symptoms.21,22
7 Sveinsdóttir H, Bäckström T. Prevalence of menstrual cycle
Thus, this suggests that the level of symptom severity symptom cyclicity and premenstrual dysphoric disorder in a
reported in our study does not differ from that of other random sample of women using and not using oral contracep-
major studies. tives. Acta Obstet Gynecol Scand 2000;79:405
8 Cohen LS, Soares CN, Otto MW, et al. Prevalence and predictors of
premenstrual dysphoric disorder (PMDD) in older premenopausal
women. The Harvard Study of Moods and Cycles. J Affect Disord
Conclusions 2002;70:125
9 Masho SW, Adera T, South-Paul J. Obesity as a risk factor for pre-
Physical symptoms are the most prevalent premenstrual menstrual syndrome. J Psychosom Obstet Gynaecol 2005;26:33
10 Pearlstein T, Halbreich U, Batzar ED, et al. Psychosocial function-
symptoms experienced by women. Thus, these symptoms
ing in women with premenstrual dysphoric disorder before and
are an important component of premenstrual syndromes after treatment with sertraline or placebo. J Clin Psych 2000;
that are psychobiological in nature. A significant number 61:101
of women in both continents report moderate/severe 11 Endicott J. Is premenstrual dysphoric disorder a distinct clinical
premenstrual symptoms, and these vary with age, country entity? J Womens Health Gend Based Med 1999;8:663
12 Linhart H, Zucchini W. Model Selection. New York: Wiley, 1986
and smoking status. Further research should establish the 13 Wittchen HU, Becker E, Lieb R, et al. Prevalence, incidence and
impact on quality of life, the relationship to existing stability of premenstrual dysphoric disorder in the community.
classificatory systems and whether a brief symptom list Psychol Med 2002;32:119
14 Treloar SA, Heath AC, Martin NG. Genetic and environmental from prospective and retrospective trials. Can J Psychiatry
influences on premenstrual symptoms in an Australian twin 2005;50:327
sample. Psychol Med 2002;32:25 19 Bertone-Johnson ER, Hankinson SE, Hankinson SE, et al. A simple
15 Singh BB, Berman BM, Simpson RL, et al. Incidence of pre- method of assessing premenstrual syndrome in large prospective
menstrual syndrome and remedy usage: a national probability studies. J Reprod Med 2007;52:779
sample study. Alt Ther Health Med 1998;4:75 20 Rasheed P, Al-Sowielem LS. Prevalence and predictors of pre-
16 Marvan ML, Cortes-Iniestra S. Women’s beliefs about the menstrual syndrome among college-aged women in Saudi Arabia.
prevalence of premenstrual syndrome and biases in recall of Ann Saudi Med 2003;23:381
premenstrual changes. Health Psychol 2001;20:276 21 Adiguzel HE, Taskin OE, Danaci AE. The symptomatology and preva-
17 Sternfeld B, Swindle R, Chawla A, et al. Severity of premenstrual lence of symptoms of premenstrual syndrome in Manisa, Turkey.
symptoms in a health maintenance organization population. Turk psikiyatri derg (Turkish Journal of Psychiatry) 2007;18:215
Obstet Gynecol 2002;99:1014 22 Angst J, Sellaro R, Stolar M, et al. The epidemiology of peri-
18 Steiner M, Streiner DL. Validation of a revised visual menstrual psychological symptoms. Acta Psychiatry Scand
analog scale for premenstrual mood symptoms: results 2001;104:10