A double-blind randomised placebo controlled trial of postnatal norethisterone enanthate: the effect on postnatal depression and serum hormones

Authors


Correspondence: Professor G. J. Hofmeyr, Department of Obstetrics and Gynaecology, University of the Witwatersrand, Medical School, 7 York Road, Parktown, 2193, South Africa.

Abstract

Objectives To determine the effect of postnatal administration of the long-acting progestogen contraceptive, norethisterone enanthate, on postnatal depression and on serum hormone concentrations, and their association with depression.

Design Double-blind randomised placebo-controlled trial.

Setting A tertiary care hospital in Johannesburg, South Africa.

Population Postnatal women using a non-hormonal method of contraception (n= 180).

Methods Random allocation within 48 hours of delivery to norethisterone enanthate by injection, or placebo.

Main outcome measures 1. Depression scores in the three months postpartum as rated by the Montgomery-Åsberg Depression Rating Scale (MADRS) and the Edinburgh Postnatal Depression Scale (EPDS); 2. serum 17β-oestradiol, progesterone, testosterone and the 17β-oestradiol: progesterone ratio at six weeks postpartum.

Results There was a chance excess of caesarean section deliveries in the progestogen group. Mean depression scores were significantly higher in the progestogen group than in the placebo group at six weeks postpartum (mean MADRS score 8.3 vs 4.9; P = 0.01 11; mean EPDS score 10.6 vs 7.5; P = 0.0022). Mean serum 17β-oestradiol and progesterone concentrations were significantly lower in the progestogen group compared with the placebo group at six weeks postpartum. There were no correlations between any of the hormone parameters and depression at six weeks except in the formula feeding subgroup of the placebo group, where formula feeding and 17β-oestradiol concentrations were positively associated with depression.

Conclusions Long-acting norethisterone enanthate given within 48 hours of delivery is associated with an increased risk of developing postnatal depression and causes suppression of endogenous ovarian hormone secretion.

INTRODUCTION

At least 10% of women become depressed in the first six months postpartum1. Despite the massive fall in circulating progesterone and oestrogen in the days following childbirth2, the ovarian hormone suppression induced by lactation3 and the iatrogenic ovarian suppression induced by postnatal hormonal contraception, researchers have failed to demonstrate a consistent link between hormone levels and postnatal depression4–7. However, failure to demonstrate endocrinological evidence of hormone deficiencies does not exclude them as aetiological factors. Peripheral hormone levels need not necessarily correspond with brain levels, nor are they necessarily an index of brain receptor numbers and affinity8,9.

Both oestrogen and progesterone have psychoactive properties. Whilst much is known about the ability of oestrogen to modulate the synthesis of proteins, neurotransmitters and receptors, little is known about the role of progesterone in the central nervous system. However, certain studies have shown that progesterone and synthetic progestogens have hypnotic, anxiolytic and anaesthetic properties10–13 and progesterone has been shown to modulate serotonergic receptors in the rat brain14. Progesterone and progestogens may therefore have effects on neurotransmitter function and, hence, on psychological symptomatology.

Research into hormonal prophylaxis and treatment of postnatal depression is limited. High dose oestrogen therapy has previously been recommended for the treatment of severe persistent depression in women15, and recently there has been evidence of a role for oestrogen in the treatment of severe postnatal depression. In a double-blind randomised placebo controlled trial of high dose transdermal oestrogen, women receiving oestrogen improved rapidly and to a significantly greater extent than controls16. Dalton17–19 popularised the use of progesterone for prophylaxis against postnatal depression. In an open study18, where women who had experienced previous postnatal depression self-selected to take prophylactic progesterone treatment, Dalton showed a reduction in the recurrence rate of postnatal depression from 68% to 10%. She repeated these findings in a later study19 with similar methodological shortcomings.

Postnatal depression is a disorder in which a high placebo effect can be expected, so it is not surprising that Dalton's methodology has been criticised. In contrast, in at least two double-blind randomised placebo controlled trials20,21 of progesterone for premenstrual syndrome (PMS) which is thought by some to have a similar hormonal aetiology to postnatal depression, no significant differences have been found between the treatment and placebo groups. In a cross-over study22 comparing a long-acting progestogen contraceptive agent containing norethisterone enanthate to a combined oral contraceptive pill for the treatment of PMS, the progestogen significantly decreased symptoms of PMS, including depressive symptoms. However, synthetic progestogens have been implicated in causing depression amongst women using them for contraception23,24, and a premenstrual tension-like syndrome in combined hormone replacement therapy for the climacteric25. There is thus evidence to support the possibility that norethisterone enanthate may either reduce or increase the risk of postnatal depression. Our hypothesis was that postnatal administration of norethisterone enanthate would reduce postnatal depression. Our alternative hypothesis was that norethisterone enanthate might increase postnatal depression.

Many women choose to commence family planning immediately after childbirth and long-acting progestogen contraceptive agents are often commenced at this time, despite the increased bleeding risk26. To date, progestogens have not been studied in a randomised controlled trial, in terms of their effect on postnatal psychological morbidity. In addition, there are few controlled data on the effect of this practice on endogenous hormones. The objective of this trial was to assess the effect of postnatal administration of a synthetic progestogen contraceptive agent, norethisterone enanthate, on postnatal depressive symptomatology. In addition, we sought to determine its effect on serum hormone concentrations at six weeks postpartum and their association with depression.

METHODS

Eligibility

This was a double-blind, randomised placebo-controlled trial that was conducted at Coronation Hospital, an academic Woman and Child hospital, serving a low income urban population in Johannesburg. Between December 1995 and March 1997 women aged 19 years or older, who requested contraceptive methods other than hormonal (eg, tubal ligation, an intrauterine contraceptive device, barrier methods, or if they had undergone tubal ligation at the time of caesarean section) were asked to participate within 48 hours of delivery. All participants were required to use an alternative non-hormonal method of contraception for the duration of the study. Women participated in the study voluntarily and within informed consent was obtained from all participants. Less than one-quarter of the women approached agreed to participate. Reasons given by women choosing not to participate included previous heavy bleeding or migraines using injectable contraceptives, a dislike of injections, an inability to return for follow up, a lack of interest and other unspecified concerns. Although current antidepressant medication and/or psychotherapy were exclusion criteria for the trial, there were no women excluded on this basis.

Intervention

Norethisterone enanthate (200 mg) is a synthetic progestogen that is administered to women as a depot contraceptive agent at approximately eight-weekly intervals by deep intramuscular injection (Nur-Isterate, Schering (Pty) Ltd). In addition to being a progestogen, it has androgenic and weakly oestrogenic properties27. Based on this and the findings of Gunston22, it was selected over depot medroxyprogesterone acetate for the trial.

Participants in the study were randomly allocated to receive a single dose of norethisterone enanthate 200 mg (1 ml) or a 1 ml normal saline placebo by intramuscular injection. Randomisation was done in blocks of four, as described by Altman28, using a random number table. Neither the participant nor the interviewer knew to which group the former had been allocated (double-blind). Preparation of the trial medication and the randomisation code were the responsibility of an author (G.J.H.) not involved in the clinical assessment of the women. The syringes for injection were masked such that the contents could not be ascertained and were administered intramuscularly by another author (M.D.J.) or by a nursing sister not directly involved with the trial. Injections were administered slowly (over two minutes) so as to prevent any guessing of their contents based on the different viscosities of the saline and test medication.

Instruments

The Edinburgh Postnatal Depression Scale (EPDS)29 was used as a subjective report of the women's well being and the Montgomery-Åsberg Depression Rating Scale (MADRS)30 for objective assessment. Both are 10–items scales. The highest score obtainable on the EPDA is 30 and on the MADRS is 60. In a pilot study the EPDS was validated in women from the same population as the trial population31.Reading the questionnaire to participants was shown to give valid results against DSM-IV criteria, and a threshold of 11/12 (ie, a score of > 11) identified 100% of women with major depression and 70.6% of women with minor depression. The MADRS was selected as the single objective measure due to its sensitivity to change and the relatively small emphasis placed on somatic symptoms, which theoretically make it a useful scale for the postnatal period32. MADRS scores of > 18 and > 9 were used to categorise women at risk of major and major/minor depression, respectively. Neither of these depression scales is diagnostic, and so women scoring higher than the thresholds were categorised in this trial as ‘at risk of depression’ and not as ‘depressed’.

Psychosocial stressors were classified for each participant according to the DSM-III-R Severity of Psychosocial Stressors Scale33. For analysis, codes were grouped to form two categories of stressors: none to moderate (low) and severe to catastrophic (high).

Procedure

Consenting women were interviewed at enrolment and, one week, six weeks and three months postpartum. Baseline data were recorded at enrolment. The EPDS and the MADRS were administered at each visit. The MADRS formed part of a clinical interview with the interviewer (T.A.L.) who was experienced in its application. The EPDS was administered verbally31. The presence of headaches, backache, exhaustion, pain or other symptoms, in addition to the mode of infant feeding, was noted. Participants were asked to keep a daily diary regarding their bleeding after delivery and were instructed to consult a doctor not involved in the study if the bleeding was considered a problem. This precaution ensured that the interviewer remained blind to any potential difference in bleeding patterns between the two groups, as various reports have shown that injectable progestogens cause increased bleeding when given postpartum34,35 At the last interview the women were asked when, if at all, their interest in sex had returned, and this was recorded as weeks postpartum. Blood specimens were taken at the six-week visit day, centrifuged and the serum was stored at −70°C. Women who were unable to return for follow up interviews at the hospital were interviewed, where possible, at home or by telephone. The primary outcome measures were depression scores (as rated by the MADRS and the EPDS) at six weeks and three months postpartum, and serum concentrations of 17-oestradiol, progesterone, testosterone and the 17–oestradiol: progesterone ratio at six weeks postpartum. Secondary outcome measures were mode of infant feeding, libido and vaginal bleeding. Women who were clinically depressed at three months post-partum were referred to a psychiatrist for the appropriate treatment.

Sample size

The sample size was calculated to be 216 using Altman's nomogram28 for continuous variables (α= 0.05, 1–β= 80%), although only 180 women were enrolled due to time constraints. For this sample size calculation the MADRS score was considered the primary outcome measure, and a clinically relevant difference between the treatment and placebo groups in the mean MADRS score was considered to be three. The standard deviation (SD) was calculated using MADRS scores of the first 50 women in this study at 6 weeks postpartum (SD = 7.81).

Statistical methods

There were no post-randomisation exclusions. The randomisation code was broken only after the data were captured and checked. Women were analysed in the groups to which they were allocated (intention-to-treat), including eight women who had subsequently elected to use a progestogenic contraceptive agent. The MADRS and EPDS scores were analysed independently of each other as continuous and categorical variables. Categorical variables were created by allocating thresholds to define women at risk of major depression (MADRS > 18) or major/minor depression (MADRS > 9, EPDS > 11). As the hypothesis was bi-directional, two-tailed tests were used to analyse the data. Continuous data were analysed by Student's t test or, when the subgroup for analysis was < 30, the Mann-Whitney U test. Categorical data were analysed using the χ2 test with Fisher's exact test where appropriate. The significance of primary outcomes and secondary outcomes was assessed at P= 0.05. Relative risks (RR) and 95% confidence intervals (CI) were calculated for outcomes using Statistical Analysis Systems (SAS) version of 6.09.

Due to a chance discrepancy in the mode of delivery between the progestogen and placebo groups, results of the depression scales have been shown separately for mode of delivery. All analyses performed for depression scales, secondary outcomes and somatic complaints have been corrected for this discrepancy, for continuous variables, by performing an analysis of co-variance where mode of delivery is the co-variant, and for categorical variables, by performing the Cochran-Mantel-Haenszel test. Serum hormone calculations are, however, not controlled for mode of delivery, as when only these women were considered (134 women had blood samples taken), the groups were comparable with respect to this variable.

Serum analysis

Serum analysis was undertaken only when data collection was complete. 17β-oestradiol, progesterone and testosterone assays were performed by the South African Institute of Medical Research using commercially available chemiluminescent immunoassays (Chiron Diagnostic). Norethisterone was not assayed and, to our knowledge, is not reflected in the measurement of endogenous serum progesterone.

Ethical considerations

Whilst manufacturers of long-acting progestogen contraceptive agents caution against their use in the immediate postnatal period because of the possibility of breakthrough bleeding, the experience at our institution and others26,34 is that post-delivery provides a good opportunity to commence family planning and the option of progestogen contraception should not be withheld. All women participating in the trial were using non-hormonal contraception in addition to the trial medication so that the contraceptive efficacy of the trial medication was not an issue. A written and verbal explanation was given of the fact that the injection might not be of any benefit to the woman and that side effects might be experienced. Ethics approval for the trial was obtained from the University of the Witwatersrand Committee for Research on Human Subjects.

RESULTS

Follow up

Ninety women were enrolled to each group. Three-month follow up was 93.3%, with 96.7% and 90% follow up in the progestogen group and placebo group, respectively. An imbalance in the follow up between the two groups (86 in the progestogen group and 77 in the placebo group) was present at six weeks. There was no significant difference between those who missed and those who attended the six-week visit in the baseline or one-week visit's mean MADRS and EPDA scores. However, the mean EPDS score at three months was significantly higher in the group that missed the six-week visit but who subsequently returned at three months (six women), than the group that attended the six-week visit. This suggests that the imbalance in follow up at six weeks could influence the results presented in the direction of decreasing their significance.

Blinding

Blinding was compromised in only one woman who complained to the interviewer of excessive bleeding at the three-month interview, leading the interviewer (T.A.L.) to suspect that she may belong to the progestogen group. Although this was confirmed when the randomisation code was broken, it is unlikely to introduce bias into the assessment of depression as the hypothesis was bi-directional. The woman scored above the threshold on both depression scales at six weeks and three months.

Baseline characteristics

The women enrolled were black, white, of mixed race and Indian origin. Overall, characteristics of the groups were comparable except for a chance discrepancy in mode of delivery (Table 1). Twenty-four women in the progestogen group and 10 women in the placebo group underwent caesarean section (P= 0.013, Fisher's exact test). Consequently, the appropriate statistical methods were employed to correct for this discrepancy. When the 134 women who had blood samples taken at the six-week interview were considered, there were no significant differences between the randomisation groups with respect to baseline characteristics.

Table 1.  Baseline characteristics of the progestogen and placebo groups expressed as mean [SD] or total number (%).
 Progestogen n= 90Placebo n= 90
  1. *P= 0.013, Fisher's exact test.

Age32.6 [5.9]32.3 [5.0]
Parity4.1 [1.513.8 [1.2]
Married/co-habitation69(76.7)73(90.1)
Employed41(47.7)44(51.2)
Household income
  <R2000/mth54(60.7)61(68.5)
Contraception
  Tubal ligation62(70.5)54(62.1)
  IUCDs and barrier methods21(23.9)30(34.5)
Other progestogens (protocol violations)5 (5.7)3 (3.4)
Caesarean section*24(26.7)10(11.1)
Companion present at delivery27(30.0)18(20.2)
Obstetric complications27(30.0)24(26.7)
History of depressive disorder19(21.1)14(15.6)
Family psychiatric history5 (5.6)6 (6.7)
Previous maternity blues10(11.1)5(5.6)
History of premenstrual syndrome7(7.8)10(11.1)
EPDS score at enrolment13.3[5.8]12.6[5.4]
MADRS score at enrolment6.2[6.6]6.4[7.3]
Psychosocial stressors high19(21.1)17(18.9)

Depression scores

In comparison with the placebo group, those women receiving the progestogen injection were at a significantly greater risk of developing depressive symptomatology by six weeks postpartum according to mean depression scores on objective (MADRS) (8.3 [7.5] vs 4.9 [9.7] p= 0.0111, analysis of co-variance) and subjective (EPDS) (10.6 [6.5] vs 7.5 [7–0]; P= 0.0022, analysis of co-variance) measures (Table 2). The relative risk (RR) of scoring > 9 on the MADRS and > 11 on the EPDS for women in the progestogen group at the six-week interview was 2.556 (95% CI 1.262–5.175) and 3.035 (95% CI 1 515–6.080), respectively. Significantly fewer women in the caesarean section group were at risk of depression at six weeks postpartum, compared with those who delivered vaginally. No significant difference in depressive symptomatology was evident at one week or three months on either of the depression scales.

Table 2.  Primary outcome measures in the progestogen and placebo group expressed as mean score [SD] and total number (%) of women with depressive symptomatology, defined by MADRS > 18 (major depression), MADRS > 9 (major and minor depression) and EPDA > 11 (major and minor depression). Results shown are controlled for mode of delivery. A P < 0.05 is considered a significant difference between the two groups. P values were calculated using the Cochran-Mantel-Haenszel test except for continuous variables where analysis of co-variance was used.
 Vaginal deliveryCaesarean sectionTotal   
 ProgestogenPlaceboProgestogenPlaceboProgestogenPlaceboPRR95% CI
  1. *95% CI for the mean difference between the randomisation groups.

  2. P= 0.0075, vaginal vs caesarean section delivery.

Six weeks (n)63682398677   
  Mean MADRS9.1 [8.1]5.9 [6.8]7.5 [9.0]3.7[6.7]8.3 [7.5]4.9 [9.7]0.0111 0.720–5.473*
  MADRS > 189(14.5)4(5.9)2(8.7)l(11.1)11(13.0)5(6.5)0.1582.1300.703–6.455
  MADRS > 928 (45.2)17(25.0)7 (30.4)1(11.1)35(41.2)18(23.4)0.0082.5561.262–5.175
  Mean EPDS11.6 [6.0]8.4 [5.6]9.6 [6.5]6.8 [5.1]10.6 [6.5]7.5 [7.0]0.0022 1.089–4.735*
  EPDS>1134 (54.8)19 (27.9)5(21.7)tl(11.1)39 (45.9)20 (26.0)0.0023.0351.515–6.080
Three months (n)637124108781   
  Mean MADRS7.3 [8.3]5.8 [6.9]5.3 [7.2]8.8[7.9]6.6 [8.4]6.1 [9.0]0.573 −1.10–2.10*
  MADRS > 187(11.1)5(7.0)1 (4.2)2(20.0)8 (9.2)7(8.6)0.8951.0910.368–3.231
  MADRS > 920(31.7)20(28.2)4(16.7)3 (30.0)24(27.6)23 (36.4)0.9301.0260.519–2.027
  Mean EPDS9.9 [6.7]8.4 [5.6]8.2 [6.5]10.7 [6.2]9.3 [6.5]8.5 [7.2]0.659 −1.180–2.610*
  EPDS > 1124(38.1)20 (28.2)4(16.7)4 (40.0)28 (32.2)24(29.6)0.5731.2030.619–2.339

Hormone parameters

Of the 134 participants who had blood taken at the six-week interview, 73 had received the progestogen after delivery and 61 had received the placebo. All serum hormone concentrations were within normal physiological ranges. Table 3 shows the mean serum hormone values, standard deviations and P-values for each group overall and controlled for mode of infant feeding.

Table 3.  Mean serum hormone values in the progestogen and placebo groups at six weeks postpartum, controlled for mode of infant feeding. P c 0.05 was considered statistically significant. P values were calculated using Student's t-test unless n < 30, in which case the Mann-Whitney Utest was used.
 Progestogen (A)Placebo (B) Total (A+B)**
GroupnMean [SD]nMean [SD]P(A vs B)nMean [SD]
  1. *Comparison between mode of infant feeding groups were performed using analysis of covariance, controlling for randomisation group.

  2. **Comparisons between randomisation groups were performed using analysis of covariance, controlling for mode of infant feeding.

17β-oestradiol (pmol/L)
  Breastfeeding5476.84 [53.01]49152.58 [114.31]0.0001103115.10 [107.58]
  Formula feeding18179.06 [154.4919339.12 [177.25]0056927247.70 [108.761
  *P (breast vs formula) 0.0095 0.0021  0.0001
  Whole group73135.54 [119.27]61227.25 [126.99]0.0001134138.74 [124.73]
Progesterone (nmol/L)
  Breastfeeding540.56 [0.64]490.60 [0.56]0.74211030.58 [3.15]
  Formula feeding180.83 [0.65]97.07 [12.11]0.0394273.95 [3.27]
  *P (breast vs formula) 0.0236 0.0014  0.0001
  Whole group730.70 [3.61]613.835 [4.4]0.00011341.15[3.59]
Testosterone (nmol/L)
  Breastfeeding540.94 [1.44]491.62 [2.85]0.12231031.27 [2.15]
  Formula feeding181.47 [2.15]90.84 [0.54]0.4072271.33 [2.18]
  *P (breast vs formula) 0.0077 0.1679  0.8996
  Whole group731.08 [2.38]611.52 [2.54]0.25601341.25 [2.11]
E2: progestogen
  Breastfeeding542078.2 [1701.9]493639.7 [3162.6]0.00301032851.7 [3011.2]
  Formula feeding183634.0 [4312.5]93614.7 [4836.8]04404273837.9 [3043.4]
  *P (breast vs formula) 0.2494 0.2013  0.1345
  Whole group732713.7 [3336.4]613975.9 [3552.9]0.01971342996.0 [3066.0]

1. 17β-oestradiol

Mean serum 17β-oestradiol levels were significantly lower in the progestogen group compared with the placebo group (135.54 [119.27] pmol/L vs 227.25 [126.99] pmol/L; P= 0.0001, Student's t test). As would be expected, serum 17β-oestradiol was significantly higher in formula-feeding women than breastfeeding women (247.70 [108.76] pmol/L vs 115.10 [107.581 pmol/L; P= 0.0016, Mann-Whitney Utest).

2. Progesterone

Mean serum progesterone levels were significantly different between the progestogen and placebo groups as a whole (0.70 [3.614] nmol/L vs 3.84 [4.40] nmol/L; P= 0.0001 Student's t test). This effect was due to a significant difference in the formula-feeding women only (0.83 [0.65] nmol/L in the progestogen group compared with 7.07 [12.11] nmollL in placebo group; P= 0.0394, Mann-Whitney Utest).

3. Testosterone

There were no significant differences in mean serum testosterone levels between the progestogen and placebo groups overall, or when controlled for mode of infant feeding.

4. 17β-oestradiol:progesterone ratio

The ratio of 17β-oestradiol to progesterone was significantly lower in the progestogen group compared with the placebo group (2713.7 vs 3975.9; P= 0.0245, Student's t test), the main difference occurring amongst breastfeeding participants (2078.2 vs 3639.7; P= 0.0030, Student's t test).

Secondary outcomes

The number of days of vaginal bleeding during the three month study period was significantly greater in the progestogen group when compared with the placebo group (25.5 vs 14.0; P= 0.0001, Student's t test), as was the number of women who considered the bleeding troublesome (27 vs 8; P= 0.001, Fisher's exact test). There was no correlation between number of bleeding days and depression score on either scale. No significant differences were found in mode of infant feeding or libido (Table 4).

Table 4.  Secondary outcome measures in the progestogen and placebo group expressed as total number (%) of women or as mean [SD]. P < 0.05 is considered a significant difference between the two groups.
 ProgestogenPlacebo 
 n n P
  1. 2 test.

  2. **Fisher's exact test.

  3. †Student's r-test.

Return of sexual interest    0.604*
  < 8 weeks postpartum7716(20.8)7219 (26.4) 
  > 8–12 weeks postpartum7710(13.0)7211(15.3) 
  None7751 (66.2)7242(58.3) 
Breastfeeding (exclusive or partial)
  On day 19085 (94.4)9084(93.3)1.000**
  At 6 weeks postpartum8663 (73.3)8064 (80.0)0.361**
  At 12 weeks postpartum8759 (67.8)8161 (74.4)0.398**
  Not enough milk8724(26.7)8114(17.2)0.374**
Vaginal bleeding (days)
  After delivery8525.1 [2.0]7913.8 [24.9]0.001†
  At 6 weeks8526.3 [1.5]7917.0 [17.810.0003‡
  At 12 weeks8534.9 [2.1]7921.0 [25.8]0.0001
  Troublesome bleeding8527(31.8)808(10.0)0.001**

Somatic complaints

Somatic complaints including headaches, backache and pain occurred with similar frequency in both groups. However, significantly more women in the progestogen group reported being exhausted at one week (19 vs 8; P= 0.035, Cochran-Mantel-Haenszel test) and three months (16 vs 6; P= 0.038, Cochran-Mantel-Haenszel test) than in the placebo group when controlled for mode of delivery.

Other findings

Overall, there was no significant difference in the mean serum progesterone, 17β-oestradiol, testosterone or 17β-oestradiol: progesterone ratio between the women at risk of depression and those not at risk of depression at six weeks postpartum (Table 5). However, in the placebo group, women who were at risk of depression according to the MADRS at the six-week interview had a significantly higher serum 17β-oestradiol level (233.04 pmoVL vs 168.48 pmol/L; P= 0.0467, Mann-Whitney U test) than those women not at risk of depression. Findings were similar using the EPDS. When controlled for mode of infant feeding, further analysis of the placebo group showed this significant difference to occur in formula-feeding women only. Hormone parameters at six weeks were not associated with depression scores at any of the other visits, and Pearson's correlation coefficients were consistently not significant.

Table 5.  Serum hormone values for women at risk of depression (MADRS > 9) and those not at risk at six weeks postpartum, overall and for the progestogen and placebo groups separately. Values are expressed as n or mean [SD]. A < 0.05 was considered statistically significant. P values were calculated using Student's t-test unless n < 30, in which case the Mann-Whitney Utest was used.
 Progestogen (A)Placebo (B)P(A vs B)Total (A + B)
17β-Oestradiol (pmol/L)
  At risk29107.67 [94.69]15233.04 [144.22]0.000344150.40(132.66)
  Not at risk4497.98 [94.69]45168.48 [135.21]0.005589133.60(121.51)
  P (at risk vs not at risk) 0.5420 0.0467  0.4688
Progesterone (nmol/L)
  At risk290.56 [0.51]151.84 [4.27]0.2329441.00(2.55)
  Not at risk440.69 [0.73]451.70 [5.62]0.2383891.20(4.04)
  P (at risk vs not at risk) 0.6079 0.5308  0.7580
Testosterone (nmol/L)
  At risk291.12 [1.74]152.06 [4.16]0.2050441.44(2.79)
  Not at risk441.04 [1.571451.02 [0.43]0.9318891.03(1.14)
  P (at risk vs not at risk) 0.9412 0.9249  0.3561
E, progestogens
  At risk292700.4 [2837.5]154607.9 [3527.5]0.0766443350.7(3182.8)
  Not at risk442269.9 [2539.4]453420.1 [3348.5]0.0718892851.5(3015.3)
  P (at risk vs not at risk) 0.4007 0.2492  0.3794

Overall, the mode of infant feeding was not associated with depression on either depression scale. However, in the placebo group breastfeeding women (57/76) were significantly less prone to depressive symptoms at six weeks postpartum as rated by the EPDS (P= 0.007, Fisher's exact test; RR = 0.215, 95% CI 0.070–0.660) and the MADRS (P = 0.018, Fisher's exact test; RR = 0.093, 95% CI 0.013–0.667) compared with those formula-feeding (18/76) their babies. By contrast, in the progestogen group there was no association between the mode of infant feeding and depression at six weeks postpartum.

DISCUSSION

This randomised placebo-controlled trial of a contraceptive agent was made possible because contraceptive efficacy was not an outcome measure and participants were using other (non-hormonal) methods of contraception. The primary outcome measure was a screening (rather than diagnostic) test. When selecting thresholds for the scales we chose those that achieved 1000/0 sensitivity in an earlier study31. Women choosing to participate in the trial were likely to be sensitised to the outcome measure and may have exaggerated depressive symptoms (as mentioned in the introduction, one can expect a high placebo effect in studies on depression). For these reasons, the number of women in the study who were assessed to be at risk of depression was quite high. As this trial had a randomised placebo-controlled double-blind design, this should not interfere with the interpretation of the results.

Our findings suggest that women receiving injectable progestogens after delivery are at an increased risk of depression in the postnatal period. That this effect had subsided by three months postpartum in this study is probably related to the fact that only a single dose of norethisterone enanthate was administered. Had the women received another dose, as they would have for contraceptive purposes, it is possible that depressive symptoms may have persisted. Dalton and Holton suggest that only naturally occurring progesterone has mood-elevating properties and that the synthetic progestogens as found in contraceptives, by suppressing natural progesterone possibly cause depression. For this reason, the results of this trial, whilst probably generalisable to other long-acting progestogen contraceptive agents, are not generalisable to progesterone.

Lactation, depending on the frequency and duration of breastfeeding, has a profound depressant effect on circulating oestrogen and progesterone levels3 in the postnatal period. Synthetic progestogens in contraceptive dosages also suppress ovarian hormone secretion by inhibiting pituitary secretion of follicle stimulating hormone and luteinisg hormone. There is surprisingly little information available on serum hormone levels after postnatal administration of long-acting progestogen contraceptive agents. Molland et al.36, in a study on immediate postpartum insertion of long-acting levonorgestrel implants in 14 women formula-feeding their infants, reported mean serum 17β-oestradiol levels to be 83.6 pmol/L and mean serum progesterone levels to be 0.95 nmol/L at eight weeks postpartum. To our knowledge, the present study is the only randomised controlled trial comparing serum hormone profiles between women receiving immediate postpartum long-acting progestogen contraception and those receiving placebo. Overall, serum 17β-oestradiol was significantly lower in the progestogen group compared with the placebo group, with the lowest 17β-oestradiol concentration occurring in breastfeeding women receiving the progestogen. Mean serum 17β-oestradiol levels in the group of women comparable to those in the study by Molland et al.36 (formula-feeding women receiving the progestogen) were twice those found in their study, although sample sizes are small. Serum progesterone concentrations in the present study were also significantly lower in the progestogen group compared to the placebo group. As in other studies on postnatal depression, hormone parameters failed to correlate with depression scores.

Most contraceptive methods have side effects that need to be weighted against the advantage of avoiding unwanted pregnancy. The potential for mood disturbance is a feature common to hormonal methods of contraception in general. Depot progestogen contraceptives constitute an important contraceptive option for women worldwide, particularly during lactation. The benefits of these methods may be optimised by careful counselling of clients, inquiry about specific risk factors, such as previous depression, and monitoring for individual adverse responses to that appropriate changes may be made. Replication of the results of Gregoire et al.16 on oestrogen therapy and research into oestrogen prophylaxis for postnatal depression are warranted, as are randomised controlled trials on the use of progesterone for postnatal depression.

CONCLUSION

Depot norethisterone enanthate given within 48 hours of delivery is associated with an increased risk of developing postnatal depression and causes suppression of endogenous 17β-oestradiol and progesterone secretion. Progestogen contraceptives should be used with caution in the postnatal period and in women with a history of depression.

Acknowledgements

The authors would like to thank all the women who participated in the trial, the Coronation Hospital postnatal ward staff, and the department secretary, Ms N. Bennett, for assistance. Research grants from Schering (Pty) Ltd, the Iris Ellen Hodges Trust of the University of the Witwatersrand, the South African Medical Research Council and the South African Institute for Medical Research supported this study. We would also like to thank Schering (Pty) Ltd for supplying the trial medication.

Ancillary