ARTICLE IN PRESS The Breast (2007) 16, 503–512 THE BREAST www.elsevier.com/locate/breast ORIGINAL ARTICLE Evidence-based management of Mastalgia: A meta-analysis of randomised trials A. Srivastavaa,1, R.E. Mansela, N. Arvinda, K. Prasadb, A. Dharc,Ã, A. Chabraa aUniversity Department of Surgery, Cardiff University, UK bClinical Epidemiology Unit, All India Institute of Medical Sciences, Ansari Nagar, New Delhi 110029, India cDepartment of Surgery, All India Institute of Medical Sciences, Ansari Nagar, New Delhi 110029, India Received 23 February 2007; accepted 19 March 2007 KEYWORDS Summary Several agents have been utilised for therapy of mastalgia based on Danazol; data from small trials. No meta-analysis of trials on mastalgia exists. We have Bromocriptine; conducted a meta-analysis on trials on mastalgia published in the English language. Tamoxifen; Study was restricted to randomised controlled trials comparing Bromocriptine, Mastalgia; Danazol, Evening primrose oil (EPO) and Tamoxifen with placebo. The analysis was Meta-analysis carried out on the REVMAN statistical package. Weighted mean difference in the pain score in favour of Bromocriptine was À16.31(95% CI À26.35 to À6.27). Danazol produced a significant benefit with a mean pain score difference À20.23(95% CI À28.12 to À12.34). EPO did not offer any advantage over placebo in pain relief, mean pain score difference being À2.78 (95% CI À7.97 to 2.40). Tamoxifen achieved a relative risk (RR) of pain relief of 1.92 (95% CI 1.42–2.58). Tamoxifen is associated with least side effects and should be the drug of first choice. & 2007 Elsevier Ltd. All rights reserved. Introduction of cyclical and non-cyclical breast pain. This includes hormonal manipulation by Danazol, Bro- Mastalgia or breast pain is a very common symptom mocriptine, Tamoxifen and LH-RH analogue, goser- 1 among patients attending the breast clinic. A large elin. Non-hormonal agents used in mastalgia are number of agents have been tried in the treatment non-steroidal anti-inflammatory gels, iodides, plant derivatives like evening primrose oil (EPO) and Vitus Agnus Castus.2 There is considerable debate à Corresponding author. Tel.: +91 11 2658 8500x3206; about the choice of best agent for initial manage- fax: +91 11 2658 8641. ment of mastalgia. So far no meta-analysis has E-mail address: [email protected] (A. Dhar). 1Present address: Clinical Epidemiology Unit, All India Institute been described to evaluate the effects of these of Medical Sciences, Ansari Nagar, New Delhi, 110029, India. agent. We therefore performed a meta-analysis on 0960-9776/$ - see front matter & 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.breast.2007.03.003 ARTICLE IN PRESS 504 A. Srivastava et al. published randomised trials of 4 common agents independently first on the baseline characteristics used in the therapy. from a photocopied portion of the paper where the result section was masked. Similarly data on the Objective results were extracted from a photocopy where the baselines features were masked. Following extrac- This investigation was carried out to generate Level tion by 2 persons a consensus was arrived by I evidence regarding the effectiveness of Bromo- discussion wherever a discordance occurred. Since criptine, Danazol, EPO and Tamoxifen in the cyclical mastalgia is more common in most breast therapy of mastalgia. clinics, the data on only cyclical mastalgia was extracted from the published reports. In some papers the actual data on mean pain score was not Search strategy and inclusion criteria reported in numerical form. In such cases, the pain score value was derived from the graph. In cross- Articles on randomised trials on treatment of over trials the result on initial phase (i.e. before mastalgia were searched using the electronic data- cross-over) was used for analysis wherever this was bases, viz. Medline, Google and Embase. Articles reported. were also hand searched from text books of breast diseases, benign breast diseases and surgery. The Statistical analysis search was performed with the following key words: mastalgia, mastodynia, breast pain, benign breast The outcome of interest is reported as the mean disease, therapy & treatment. The search was pain score with the active drug and with placebo in confined to articles published in the English language. some studies while other studies have described The search was restricted to randomised clinical trials the number of cases achieving clinical response where an active drug was compared with placebo or (usually greater than 50% reduction in the mean another drug along with the placebo. We have pain score measured on a visual analogue scale or conducted this meta-analysis for four agents, which Cardiff breast pain scale). The pooled estimates of are commonly used in the therapy of breast standardised difference of mean pain score be- pain—Bromocriptine, Danazol, EPO and Tamoxifen. tween active drug and placebo has been calculated for those studies reporting the outcome as mean pain score. We calculated the pooled relative risk Exclusion criteria (RR) for studies reporting outcome as dichotomous data on a 2  2 table. The point estimates and 95% Trials without randomisation or without a placebo confidence limits have been computed for fixed or arm were excluded. Trials on agents other than random effect model as appropriate. The random above 4 agents were excluded. effect model is more appropriate where there is significant heterogeneity among the trials whereas a fixed effect model is employed in situation of no Method of quality assessment significant heterogeneity among the trial results. The meta-analysis has been performed by two The following criteria were employed to assess the investigator (KP and AS) on ‘‘REVMAN’’ Meta- quality of individual trials: method of randomisa- analysis statistical package (version 4.2.8) obtained tion, concealment in randomisation, blinding in from Cochrane Collaboration through the world assessment of response, and reporting of dropouts/ wide web. In some trials the standard deviation loss to follow-up. Two investigators (AS and NA) (SD) of the mean pain score was not given, assessed the quality of studies independently and therefore we have taken the SD from the trial coded each study as: excellent, good and poor. No which has used the same drug and assumed the study was excluded because of poor quality. same SD across the studies. A funnel plot to evaluate the publication bias was not drawn because of very small number of studies available. Data extraction The data on the important trial characteristics, Results patient population, dosage of drug with duration, response measure at 3 months and side-effects of We came across 34 studies on Bromocriptine with drugs were extracted by 2 investigators (AS and NA) 12 clinical trials. Of these 7 were RCTs and were ARTICLE IN PRESS Evidence-based management of Mastalgia 505 included in the meta-analysis. For Tamoxifen, 34 Kontostolis et al.14 (Table 3). Of these the data studies with 14 clinical trials (4 RCTs) were found. was presented as mean pain score in the first 4 Three of these 4 RCTs are included in our analysis. trials and are included in the meta-analysis. The excluded trial of Fentiman et al.3 compared 2 Kontostolis et al. reported binary outcome of different dosage schedules of Tamoxifen without responders and failures, without any pain score, any placebo arm. For EPO—total studies 20, hence their data is not included in the analysis. clinical trials 6, RCTs-3. All 3 trials were included There is insignificant heterogeneity amongst the in the analysis along with the trial of Goyal and trials (heterogeneity Chi-square ¼ 5.11, df ¼ 3, Mansel on Gamolenic acid. Regarding Danazol—total p ¼ 0.16) hence a fixed effect model is employed. studies 35, clinical trials 8 and RCTs-4. All 4 RCT The overall weighted mean score difference of were included in the meta-analysis. À20.23 is statistically significant (Z ¼ 5.02, The quality of RCTs was assessed on the above- po0.00001) demonstrating the effectiveness of mentioned criteria. It was judged to be excellent in Danazol in amelioration of breast pain. The forest 2 trials, good in 2 trials and poor in 2 trials of plot is shown in Fig. 3. Bromocriptine. Regarding Tamoxifen trials 2 trials were good and one trial was found to be of poor Evening primrose oil (EPO) quality. For trials on Danazol the quality was as follows—1 excellent, 1 good and 2 poor. Of 4 trials Three trials with outcome presented as mean pain of EPO, 1 was excellent, 1 good and 2 of poor score were found. These are the trials of Pashby quality. No trial was excluded from the meta- et al.,15 Blommers et al.16 and Preece et al.17 analysis on grounds of poor quality. (Table 4). The multicentre European trial of Goyal and Mansel18 has evaluated the role of Gamolenic Bromocriptine acid (GLA) in controlling mastalgia. Since GLA is considered to be the active ingredient of EPO, we Seven trials were analysed—those of Mansel et al.,4 have included this trial data in the meta-analysis of Mansel et al.,5 Hinton et al.,6 Nazli et al.,7 Parlati EPO trials. The results are presented in Fig. 4. The et al.,8 Durning and Sellwood9 and that of Blichert- overall weighted mean difference of À2.78 was not toft et al.10 The outcome in the first six trials is significant (test for overall effect Z ¼ 1.05; recorded as mean pain score, whereas the last p ¼ 0.29) indicating the ineffectiveness of EPO trial reported outcome only as dichotomous data compared to placebo. There is significant hetero- (Table 1). Hence meta-analysis has been carried out geneity amongst the trials (heterogeneity Chi- on the first six trials using mean pain score.