The hypothesis that the many non-specific changes normally

associated with cyclical ovarian activity are the primary aetiological

factors in the premenstrual syndrome was tested by

suppressing ovulation with subcutaneous oestradiol implants.

Sixty eight women with proved premenstrual syndrome were

treated under placebo controlled conditions for up to 10 months

in a longitudinal study. Active treatment was combined with

cyclical oral norethisterone to produce regular withdrawal

periods. Symptoms were monitored with daily menstrual distress

questionnaires, visual analogue scales, and the 60 item general

health questionnaire. Of the 35 women treated with placebo 33

improved, giving an initial placebo response rate of 94%. The

placebo effect gradually waned, but the response to the active

combination was maintained for the duration of the study.

Analysis of the prospective symptom ratings showed a significant

superiority of oestradiol implants over placebo after two months

for all six symptom clusters in the menstrual distress questionnaire.

Changes seen in the retrospective assessments were less

significant but the trend was the same.

Treatment with oestradiol implants and cyclical progestogen

was well tolerated and appears to be both rational and effective

for severe cases of the premenstrual syndrome.


All aspects of the premenstrual syndrome remain subjects of much

controversy.' Nevertheless, in the confusion concerning definition,

diagnosis, aetiology, and treatment five observations appear to be

salient. The premenstrual syndrome is associated with cyclical

ovarian activity and does not occur before puberty, during pregnancy,

or after the menopause. Menstruation itself is incidental, as

cyclical symptoms continue after hysterectomy, provided that

ovarian function is conserved.' Most healthy women report adverse

symptoms before menstruation.4 Extensive metabolic and psychological

studies have failed to find a specific abnormality in the

premenstrual syndrome.' Lastly, the condition shows a strong

response to placebo. While the factors concerned in responsiveness

to placebo are extremely complex,6 the evidence none the less

suggests a primary role for cyclical ovarian activity in pathogenesis.

That the ovarian cycle is associated with such profound effects

should not be surprising. Many physical, psychological, and

behavioural changes affecting all body systems occur during the

normal menstrual cycle. Furthermore, the premenstrual syndrome

is only one of many disorders that regularly fluctuate with ovarian

activity; examples include migraine, epilepsy, rheumatoid arthritis,

diabetes, and asthma. The corollary of considering cyclical ovarian

activity as fundamental in the aetiology of the premenstrual

syndrome is that suppression of ovulation should abolish symptoms.'

Anovulation may be achieved by subcutaneous implants of

oestradiol (Organon Laboratories)9 combined with cyclical oral

progestogen to ensure regular withdrawal periods and prevent

endometrial hyperplasia. "' The benefits of that approach have been

reported in a retrospective study." This paper presents the results of

a prospective randomised controlled study of oestradiol implants

and cyclical oral norethisterone in the management of the premenstrual


Patients and methods

Study design-As there is no standard treatment for the premenstrual

syndrome and the condition shows a strong response to placebo,' the study

treatment was compared against a placebo. A parallel study was appropriate,

as subcutaneous oestradiol implants are long acting preparations with

biochemical activity lasting over 12 months." We find that the therapeutic

benefit is maximal two to four months after implantation and that the

response generally wanes after five to six months, when the implant is

repeated.' Hence we planned to monitor treatment over two consecutive

implants with special emphasis on months 2-4 and 8-10. Patients were

reviewed every three months and given fresh implants when symptoms were

returning. All those who were not sterilised were advised to use nonhormonal

contraception throughout.

Patients-Women who fulfilled the following criteria were recruited from

the premenstrual syndrome clinic at Dulwich Hospital: (a) age 25-45 with

regular periods and a cycle length of 24-32 days; (b) not using drugs that

affect ovarian function; (c) not suffering from other medical or psychiatric

condition; (d) family complete; (e) normal pelvic organs on clinical

examination; (f) agreeable to treatment with hormone implants; (g) prepared

to monitor symptoms daily during the study; (h) a complaint of premenstrual

distress for at least the past six months; and (i) confirmation of the

diagnosis of the premenstrual syndrome on prospective daily symptom

rating for at least one ofthe six symptom clusters studied. Of 132 consecutive

patients referred to the clinic during the study period, 68 fulfilled these

criteria and were entered into the study.

Treatment-Eligible patients were treated under double blind conditions,

such that neither the patient nor the assessing physician was aware of

the allocation. Treatment was randomised by a second physician, whose

role was to dispense the appropriate implant and progestogen. Allocation

was based on random numbers, equating odd numbers with active

and even numbers with placebo treatment. Blocking in groups of 10

was incorporated to minimise imbalance in allocation. Treatment was either

a 100 mg subcutaneous oestradiol implant and 5 mg oral norethisterone for

seven days per cycle (oestradiol group) or placebo implant and 5 mg placebo,

identical in appearance with the active drug, for seven days per cycle

(placebo group). Implants were inserted deep into the subcutaneous tissue

of the anterior abdominal wall using the technique of Thom and Studd. "

The timing of tablets was adjusted individually depending on the length of

the menstrual cycle, such that the last tablet was taken two days before the

next expected period. Implants were repeated if the assessing physician

judged on interviewing the patient that symptoms were returning or had

failed to respond to the first implant.

Assessment of symptoms-Symptoms before and during treatment were

monitored using three parameters. Daily symptom ratings were obtained

using a modified Moos menstrual distress questionnaire consisting of 34

adverse symptoms in six symptom clusters (pain, concentration, behavioural

change, autonomic reactions, water retention, and negative affect) scored

daily on a scale of 0-3 (no symptoms to severe symptoms).'4 While

recognising that conclusions drawn from the analysis of symptom clusters

cannot be applied directly to isolated symptoms, this approach is appropriate

for a condition typically associated with numerous non-specific

complaints, as it reduces the need for multiple statistical testing. A 100 mm

visual analogue scale was also completed at each clinic visit to describe how

the women -had been feeling since their last attendance (O=very well to

'100=very'unwell).'5 Lastly, the 60 item general health questionnaire was

applied at each visit as a non-specific screening tool for current psychiatric

morbidity'6; a score was obtained by counting the number of items for which

morbidity was said-to be increasing. Weight, blood pressure, menstrual

pattern, changes in medication, side effects, complications, and reasons for

stopping treatment prematurely were also recorded by the assessing


Statistical methods-The total daily ratings for each of the six symptom

clusters in the menstrual distress questionnaire during the pretreatment

cycle and every second cycle after treatment (cycles 2, 4, 6, 8, and 10) were

assessed by trend analysis. '7 This provides both qualitative and quantitative

statistics concerning chronological data, in this case symptoms during the

menstrual cycle. Records were analysed provided that no more than three

days of values were missing in a cycle. Qualitative analysis of the pattern of

symptom trends before treatment was used for diagnosing the premenstrual

syndrome (eligibility criterion (i)); the syndrome was defined as a condition

associated during the menstrual cycle with (a) significant (p<005) positive

(worsening) symptom trends during the 14 days before menstruation and at

no other time in the cycle; and (b) significant (p<005) negative (improving)

symptom trends some time after the onset of menstruation and at no time

after the presence of significant positive trends (fig 1). Quantitative analysis

was used to provide three indices of the severity of symptoms before

and after treatmnent-namely, the maximum and minimum exponentially

smoothed averages;'representing peaks and troughs in symptoms, and the

mean symptom score during the cycle (fig 1). These indices were calculated

for the six symptom clusters in the menstrual distress questionnaire and for

all symptoms recorded. The.effect of treatment on these three variables was

assessed by the Kruskal-Wallis test (within groups) and, using the difference

between scores before and after treatment, Wilcoxon's rank sum test

(betweengroups).-Data fromthe visual analogue scale (without transformation)

and general health questionnaire were analysed by similar nonparametric

techniques. Other variables were assessed using one way analysis

of variance (within groups) and the unpaired t test and x2 test with Yates's

correction (between groups). All statistics were two tailed. A significant

difference was defined as p<0 05.

Power calculation-Trial size was estimated by considering that 90% of

the cases would improve with active treatment and 50% with placebo and

accepting a type I error of 5% and a type II error of 10%; with this calculation

the required number of patients in each treatment group would be 23.

Allowing for the length of the study and the pharmacokinetics of implant

treatment, we decided to continue recruitment until roughly that number of

cases had been treated for at least four months.


The characteristics of the two treatment groups, including baseline scores

from the menstrual distress questionnaire, visual analogue scale, and general

health questionnaire, were similar except for a significantly higher mean

diastolic blood pressure in the placebo group (unpaired t test: df=66;

t=2-96; p<001) (table I). The average duration of follow up was also

comparable, with 5-5 (SD 3-2) and 4-9 (2-7) months respectively for the

women treated with oestradiol and placebo. There were no drop outs from

the group receiving oestradiol. Conversely, two women given placebo

stopped treatment after three months because they thought that the implant

had failed to help their symptoms, while a third failed to attend for follow up

at six months.

Daily symptom ratings-Six of the 91 (6 6%) menstrual distress questionnaires

in the oestradiol group and eight of the 81 (9 9%/o) in the placebo group

were inadequate for analysis. Symptoms generally improved at the start of

treatment with both oestradiol and placebo, as denoted by a fall in peak and

mean daily ratings (figs 2 and 3). While the benefits of active implants

were sustained, resulting in a significant improvement in all symptom

clusters during treatment (Kruskal-Wallis tests: df=5; 12 9<X'<40S5;

00001<p<0-05), the placebo response tended to wane during the study

period and the overall effect was not statistically significant for any of the

clusters (Kruskal-Wallis tests: df=5; 1 l<X2<9-0; p>01). Changes in

trough symptom scores were variable after both treatments (fig 4), a

significant and maintained reduction occurring only for negative affect after

oestradiol implants (Kruskal-Wallis test: df=5; j'=11-9; p<005). Comparison

of the two treatments showed superiority of the active combination

over placebo at all times and for all variables (table II). The improvement in

total scores in the menstrual distress questionnaire were consistantly

significant after the first two months of the study for peak and mean daily

symptom ratings (Wilcoxon's rank sum tests: 0-001 <p<0-05). Differences

in the minimum exponentially smoothed average were smaller and reached

significance only at four months (Wilcoxon's rank sum test: p<001).

Individual comparisons of the six symptom clusters showed that the efficacy

of oestradiol applied to all complexes, with greatest benefit on pain,

autonomic reactions, water retention, and negative affect (for example,

Wilcoxon's rank sum tests 0O001<p<Ol at four months for maximum

exponentially smoothed average and mean daily ratings). Once more

changes in the minimum exponentially smoothed average generally failed to

reach the 5% level of significance compared with placebo.

Visual analogue scale-Irrespective of the treatment all but four women

reported an improvement in symptoms throughout the study, scores on the

visual analogue scale being reduced by on average 41-54o (table III). The

exceptions included two members of the oestradiol group (one at six months

and one at nine months) and two treated with placebo (both at three

months). Overall, the reductions in visual analogue scores were significant

for both oestradiol (Kruskal-Wallis test: df=3; x2-=503; p<0001) and

placebo (Kruskal-Wallis test: df=3; x2=55 6; p<0001). Although active

treatment was associated with the greater fall in visual analogue scores at all

times, the large interpatient variability in measurements and the relatively

small number of observations later in the study meant that the comparison

with placebo failed to reach statistical significance (Wilcoxon's rank sum

tests: p>O-1).

General health questionnaire-Similar results were obtained from analysis

of the 60 item general health questionnaire. Both treatments were followed

by significant reductions in scores early on, but by nine months results in the

placebo group were worse than before treatment (table III). Nevertheless,

the treatment effects were not significantly different because of the large

scatter of the data (Wilcoxon's rank sum tests: p>0. 1).

Other variables-Of the women who had tried conventional treatments,

30 out of 30 receiving oestradiol and 27 out of 32 (84-4%) placebo considered

implants to be more effective (Yates's X2 test: df= 1; x'= 1-92; p<02). The

average number ofimplants dispensed during the study was similar for both

treatment groups, being 2-1 (SD 0-51) for oestradiol and 2-0 (0-48) for

placebo. For the 67 women who had more than one implant during the study

the average interval between implants was 5-6 (SD 1 -6) and 5-0 (1-6) months

for oestradiol and placebo respectively (unpaired t test: df=65; t= 15;

p<O2). There were no significant changes in weight or blood pressure

during the study.

Side effects-Both treatments were well tolerated and side effects were

usually mild and transient. Most common complaints in the oestradiol and

placebo groups included mastalgia (nine and two cases respectively), nausea

(three and four), weight gain (five and none), and headache (two and three).

Menstrual pattern remained unchanged or improved for most women. In

particular, there was a tendency for period pain and flow to decrease with

active treatment. Onlv two women receiving oestradiol and four placebo

reported episodes of irregular bleeding.


Several important observations are evident from our study.

Firstly, even carefully selected women suffering from the premenstrual

syndrome usually show a significant response to placebo,

certainly early in treatment. The immediate response to placebo in

our study was 94%, all but two of the 35 women receiving a placebo

implant reporting improvement during the first three months. This

is much higher than the often quoted 50-60% and, indeed, larger

than the maximum rate reported (89%). The duration of the

response varied, but on average the daily symptom ratings, which

are generally considered to be the most accurate index of symptoms,

returned to pretreatment scores by six months. The response to the

subsequent placebo implant appeared to be weaker and of shorter

duration (figs 2-4). Though the "surgical" nature of the treatment

may in some way account for such a potent placebo response, we

must assume that psychosocial factors play an important part in this

condition. Many of these patients admit to problems with personal,

family, and sexual relationships in particular, and it is this aspect of

their distress that we believe responds to placebo.

Despite the extent of the placebo response, subcutaneous

oestradiol implants with cyclical oral norethisterone proved to be

the progestogen. Controlled studies have failed to find that

norethisterone, or indeed other progestogens, are successful for

any but isolated symptoms of the premenstrual syndrome.' Conversely,

implants of 100 mg oestradiol have been shown to be

highly effective in the management of the syndrome. Though the

retrospective ratings (visual analogue and general health questionnaire

scores) failed to distinguish between the treatments, their

infrequent application necessarily means that they represent the

weakest index of treatment effect. Conversely, the improvement

seen in daily symptom ratings with oestradiol was statistically

superior after the first two months, and this difference applied to all

symptom clusters. To our knowledge this is the first treatment

subjected to a large scale controlled study to have been found to be

successful in relieving the typical complaints of the syndrome. Our

results support those of the pilot retrospective study, which, in

addition, showed that the treatment was well tolerated after

prolonged use." In that study lower doses of oestradiol were

generally dispensed after the initial 100 mg pellet without loss of

effect, and this is to be recommended in the long term to avoid


We attribute the efficacy of our treatment to suppression of

cyclical ovarian activity by the oestrogen rather than any effect of

contraceptive-though not immediately, as pregnancies may occur

during the first six months.' We have also found that follicular

activity and luteinisation continue in some cases during the first few

months (unpublished observations). Incomplete suppression of

ovulation early on may partly explain the inferior response at two


Ovulation may be suppressed by other therapeutic manoeuvres.

The combined oral contraceptive pill may not only be contraindicated

in this age group but, though the definitive study has still

to be done, is probably effective in only half or so of cases.' The

reason may be that the benefits of anovulation secondary to the

oestrogen component are negated by progestogenic side effects

which are similar to the typical complaints of the premenstrual

syndrome.''" Furthermore, the cyclical nature of the pill regimen

is itself associated with hormonal swings. Crystalline implants of

"natural" oestradiol not only provide a continuous source of

oestrogen but do not have the potential cardiovascular complications

of "synthetic" oestrogens; implants are already widely and

safely used for the treatment of climacteric complaints.'9 Giving

norethisterone for a shorter duration each cycle may account for the

better tolerance with our treatment, though we have seen women

who are extremely sensitive to progestogen. Such patients may

improve with reduced dosage or other progestational agents.

Ovarian function may also be inhibited by danazol. Recent trials

have reported favourably on certain symptoms (D H Gilmore et al,

paper presented to 11th World Congress of Gynecology and

Obstetrics, West Berlin, 1985). Danazol, however, appears to be

poorly tolerated and is associated with a considerable drop out rate

because of androgenic and other side effects. Gonadotrophin

releasing hormone agonist has been used to produce reversible

"medical ovariectomy" and relief of premenstrual distress but, as

pointed out by the authors, the low levels of oestradiol release

induced precludes its long term use.-' Oestradiol implants are both

well tolerated and also maintain plasma oestrogen concentrations in

the normal to high follicular range."

In conclusion we have shown that subcutaneous implants of

oestradiol combined with cyclical oral norethisterone is of definite

benefit for the premenstrual syndrome. Taken in conjunction with

the degree of placebo responsiveness shown by these patients, we

hypothesise that the primary event in this condition is ovarian

activity with its associated physical, psychological, and behavioural

changes, which, in this group ofwomen, are amplified secondary to

psychosocial factors or as yet undetermined central mechanisms.

Despite arguments about the importance of"reproductive biology"

in the genesis of psychiatric morbidity2' 2 there seems little doubt

that the hormonal and other changes associated with the ovarian

cycle have profound effects on both the soma and the psyche, effects

that may logically be controlled by manipulating ovulation.

We thank all our colleagues who helped in the premenstrual syndrome

clinic, including Laila Hanna, Tim O'Dowd, Maggie Blott, and Julia

Montgomery. We are also grateful to Linda Abbott and Trish Waghorn,

whose administrative and secretarial work was essential; and, finally, to

Sister P A Fryer and other members of the nursing staff who prepared the

implant sets for the clinic. This work was supported by a grant from King's

Voluntary Research Trust.


I Magos AL, Studd JWW. The premenstrual syndrome. In: Studd JWW, ed. Progress in obstetics

and naecolorj. Vol 4. Edinburgh: Churchill Livingstone, 1984:334-50.

2 Magos AL, Studd JWW. PMS-a new approach to cause and cure. Contemporay Obserics and

Gywcology 1984;24:85-91.

3 Backstrom T, Boyle H, Baird DT. Persistence of symptoms of premenstrual tension in

hysterectomized women. BrJObset Gynaecol 1981;88:530-6.

4 Taylor JW. The timin of menstruation-related symptoms assessed by a daily symptom rating

scale. Acia PsychiasrScand 1979;60:87-105.

5 Mattson B, von Schoultz B. A comparison between lithium, placebo and a diurectic in

premenstrual tension. Acwa PsychiaarScand 1974;255(suppl):75-84.

6 Grunbaum A. The phcebo concpt in medkine and psychiatry. PsycholMed 1986;16:19-38.

7 Magos Al, Studd JWW. Effects of the menstrual cycle on medical-disorders. BrJ Hosp Med


8 Studd JWW. Premeasne tension syndrome. BrMedJ 1979;i:410.

9 Greenblatt RB, Asch RH, MSahesh VB, Bryner JR. Implantation of pure crystalline pellets of

estradiol for conception control. AmJ Obst Gynecol 1977;127:520-4.

10 Studd JWW, Thom MH. Oestrogens and endometrial cancer. In: Studd JWW, ed. Progress in

obstetnrcsand gnaeceidag. \'ol 1. Edinburgh: Churchill l.ivingstone, 1981:182-98.

11 Magos AL, Collins WP, Studd JWW. M n t of the premenstrual syndrome by

subcutaneous implants of oestradiol. journal of Psychosomatic Obsttics and Gynaecologv


12 ThomMH, Cqlins WP, Studd JWW. Hormonal profiles in postmenopausal women after therapy

with subcutaneous implants. BrJ7 Obsm Gynoecol 1981;88:426-33.

13 Tlon MH, Studd JWW. Practical procedures-hormone implantation. Br Med _ 1980;280:


14 MoosAR. Typology ofmenstrual cycle symptoms. Amj Obstet Gynecol 1969;103:390-402.

15 Maxwell C. Sensitivity and accuracy of the visual anaiogue scale: a psycho-physical classroom

experiment. BrJ7 Ci Phor}acol 1978;6:15-24.

16 Goldberg DP, Kay C,Thompson L. Psychiatric morbidity in general practice and the community.

PsycholMed 1976;6:565-9:

17 Magos AL, Studd JWW Assessment ofmenstrual cycle symptoms by trend analysis. Am7Obsws

Gynecol(in press).

18 Hammarback S, Backstrom T, Holst J, von Schoultz B, Lyrens S. Cyclical mood changes as in the

premenstrual tension syndrome during sequential estroeprogestagen poatmenopausal replhement therapy. Acta Obsiet Gywcol Scand 1985;64:393-7.

18& Magos AL, Brewster E, Singh R, O'Dowd T, Brincat M, Studd JWW. The effects of

norethisterone in postmenopausal women on oestrogen replacement therapy: a model for the

premenstrual syndrome. BrJ Obstet Gyaecol (in press.J

19 Studd JWW. Hormone implats in the climacteric syndrome. In: Campbel S, ed. The managementofthemenopsenadposmenopausalyears. London: MTPPress, 1976:383-5.

20 Muse KN, Cetel NS, Futterman LA, Yen SSC. The premenstrual syndrome: effects of "medical

ovariectomy."NEgl Med 1984;311:1345-9.

21 Jenkins RJ, Clare AW. Women and mental illness. BrMedJ 1985;291:1521-2.

22 Studd JWW, Brincat M, Magos A, Montgomery J. Women and mental illness. Br Med J 1986;292:201.

(Acceped30April 1986)