An individually randomized controlled trial (RCT) .
Hay el Sellom in the southern suburbs of Beirut, Lebanon, an informal settlement of approximately 150,000 mainly Lebanese Shiites, with low levels of basic health care services, education and physical infrastructure .
A total of 33 schools, three large factories, 14 gynaecological clinics and eight satellite network providers facilitated recruitment over six weeks (1 April to 15 May 2009), administering a questionnaire to all potentially eligible women. This questionnaire comprised symptoms of vaginal discharge, general inclusion/exclusion criteria and (for those reporting vaginal discharge) items relating to common mental disorders (CMDs). To rule out RTIs, women reporting vaginal discharge were referred for laboratory tests and a pelvic examination conducted by female gynaecologists trained in the protocol. Swab specimens were transported and analyzed by a trained technician from the American University of Beirut laboratory. The following lab tests were used to rule out each of the five different RTIs of interest to this trial (coupled with their respective sensitivities and specificities): COBAS AMPLICOR Nisseria gonorrhea/Chlamydia trachomatis Test” (Roche Molecular Diagnostics) was used for qualitative in vitro detection of N. gonorrhea (sensitivity 96.4%, specificity 97.9%) and C. trachomatis (sensitivity 93.4%, specificity 96.7%). “Tv latex” and “Candida Latex” (Kalon Biological Ltd) were used for T. vaginalis (sensitivity 95%, specificity 99%) and candidiasis (sensitivity 80%, specificity 100%) detection. Nugent Score technique (sensitivity 87.5%, specificity 95%) was followed for the assessing of bacterial vaginosis [21–25].
Inclusion criteria were: currently married women aged 18-49; reporting symptoms of vaginal discharge; low to moderate scores on the Hopkins Symptom Checklist-25 (HSCL-25) (ranging between 2.1-3.3 for the depression sub-scale and 2.0-3.2 for the anxiety sub-scale); signed informed consent. Exclusion criteria were: pregnant or less than 8 weeks postpartum; post-menopause or hysterectomy; reporting treatment for severe mental illness; positive RTI test (bacterial vaginosis, trichomoniasis, fungal infection, chlamydia, and gonorrhoea).
Two different consent forms were used. The first was used during the recruitment phase to obtain permission to check for the inclusion/exclusion criteria including the pelvic examination and the lab tests. This consent form was read to the woman by an interviewer in front of a witness prior to the pelvic examination. The second consent was obtained by trained trial staff from women who were eligible to participate in the trial. It ensured that women understood the trial procedures and reasons behind their eligibility.
Randomization, concealment of allocation and blinding
Randomization was performed using a computer-generated allocation schedule, produced by an individual not involved in recruitment. The original intention was that the baseline measures and consent were to be obtained before allocation was determined remotely (by telephone) and then revealed to the woman. In the event, logistical constraints at the recruitment centres meant that remote allocation was not feasible and allocation was performed using the computerized system as planned, but in the field rather than remotely. An unintended consequence of this change in procedure was that the women were informed of their allocation into either the intervention or the control arm before they gave final written consent. As will be detailed in the analysis section, this departure from protocol necessitates particularly extensive comparison of characteristics of the (consenting) women at baseline, with suitable control for any differential selection across the trial arms. Blinding of allocation after randomisation was not feasible given the nature of the intervention.
This consisted of 12 group sessions implemented over a six-week period using local facilities, given only to the intervention arm of the trial. Each session was divided into two parts (delivered in an order determined by practical circumstances): a psychosocial component and a relaxation exercise component.
The psychosocial component lasted approximately 75 minutes and was delivered by five Masters level clinical psychologists, assisted by five social workers as co-moderators. These sessions were divided into directed and semi-structured social support discussion sessions, incorporating problem-solving skills building as well as venting. Once a group was formed its members remained in the same group throughout. Those delivering the intervention received two days of training from a senior clinical psychologist, and were provided with a manual describing each session in detail.
The relaxation and exercise component involved 30 minutes sessions run by physical trainers. The exercises were introduced gradually. They included teaching women how to engage in visual guided imagery exercises on their own, coupled with stretching and progressive muscle relaxation. Each session started with a summary of the previous session, a review of what was being practiced at home, followed by introducing an additional technique. A manual was also developed by a physical fitness specialist who supervised and trained the physical trainers over two days. A brief pamphlet was given to participants describing the components of each session.
The control group was a treat later group, which received the intervention after the study was completed. Those women in the control group were divided into groups and offered the same intervention protocol as in the trial, using the same facilitators and locale. During the trial controls received usual care but they were followed up by two clinical psychologists (as requested by the IRB), over the phone, every two weeks by administering the HSCL-25 -in order to ensure that their CMD status did not regress.
Data collection took place at baseline, 1.5 months and six months using face-to-face interviews conducted by trained interviewers. The primary outcome was MUVD at six months. A woman was considered to have MUVD if she reported a complaint of vaginal discharge as assessed by the question: “are you complaining currently from vaginal discharge?” and was concurrently not suffering from any RTIs, as confirmed by pelvic exam and laboratory tests. Further questions asked whether or not they were bothered by MUVD, to indicate the colour, odour, thickness, consistency and frequency of the discharge.
Secondary outcomes included common mental disorder (CMD) as assessed by the Hopkins Symptom Checklist 25 (HSCL-25). This is known to have good psychometric properties [26, 27] but was subjected to further field-testing in Lebanon prior to the trial. From these investigations, cut-off values of 2.10 and 2.00 on the HSCL-25 were determined for depression and anxiety respectively .
The third secondary outcome was somatisation, using the Scale for Assessment of Somatic Symptoms. The main purpose of the 1.5 months follow-up was to maintain contact with the women and is not presented further here.
Allowing for 10% attrition, 80-102 women in each group would provide 80-90% power to detect a difference of 20 percentage points (10% vs. 30%) in the primary outcome of reported MUVD at 6 months, using a 2-sided 5% significance level. There is no previous study to inform this effect size directly, but differences in CMD recovery rates of up to 40% between usual care and active interventions for CMD have been found [8, 9]. A 20 percentage point difference in MUVD was chosen because in our judgement such a difference would be clinically worthwhile and the effects on CMD suggest that such a difference is plausible.
Using SPSS version 16 and Stata version 11 the trial was analyzed and reported in accordance with CONSORT guidelines. Descriptive statistics were used to compare the two groups as randomized on 14 socio-demographic and clinical variables. Since (final) consent was obtained after randomization, those consenting and not consenting were compared on the basis of the same 14 characteristics.
The primary analysis involved calculating the difference in the percentages of women with MUVD at 6 months between the groups as randomized (including the 95% confidence interval (CI) and the number-needed-to-treat. For interpretation alongside secondary analyses the primary comparison was repeated using logistic regression to obtain the odds ratio, its 95% CI and p-value. The main secondary analyses involved multivariable logistic regression to adjust for baseline imbalances across the groups of (randomized and consented) women and those characteristics associated with consent. The regression analyses were then repeated using multiple regression models for the three secondary outcomes. All regression analyses were performed before and after imputing missing outcome data using multiple imputations by chained equations .
Additional secondary analyses for explanatory purposes involved investigating clustering effects of the intervention groups women were assigned to, using mixed effects regression for the primary outcome. Potential mediators were investigated by adding (separately) changes in HSCL-25 Anxiety and Depression sub-scales and the somatisation scale scores to the original regression models with the primary outcome. Adherence effects were investigated using instrumental variables regression with a linear term for the number of sessions attended. These were run both on the risk difference scale for the MUVD outcome and following a probit transformation. The results for both sets of models were very similar and only the former are reported. The final set of secondary analyses involved adding appropriate interaction terms to the regression models for the primary and secondary outcomes to investigate differential effects of the intervention according to age and baseline HSCL-25 scores, all as continuous variables.