Effectiveness of the Common Elements Treatment Approach for mental and behavioural health outcomes among women struggling to remain adherent to HIV treatment and who have experienced intimate partner violence in South Africa: secondary outcomes from a randomised controlled trial

Abstract

Background

Rates of intimate partner violence (IPV) and HIV in South Africa are among the highest globally. IPV is associated with a range of adverse mental health and HIV outcomes. The Common Elements Treatment Approach (CETA) is a transdiagnostic, evidence-based intervention delivered by lay providers.

Objective

To compare the effectiveness of CETA to active attention control in reducing IPV, depression, post-traumatic stress disorder (PTSD) and substance use among women at risk of poor HIV outcomes who have experienced IPV.

Methods

Women living with HIV with an unsuppressed viral load or at risk for poor adherence and experienced past 12-month IPV were recruited from Johannesburg-area clinics and randomised 1:1 to CETA or control (SMS HIV appointment reminders plus safety checks and planning). The primary trial outcome was HIV retention and viral suppression, under review elsewhere. This paper reports secondary outcomes, evaluated at 3 and 12 months: IPV, depression, PTSD and substance use.

Findings

Participants were enrolled between 11 November 2021 and 19 July 2023 and randomised to CETA (N=202) or control (N=197). In the intent to treat analysis, the Cohen’s d treatment effect for depression at 3 months was 0.24 (difference in mean change −3.1; 95% CI −6.1 to 0.1) and 0.48 at 12 months (−6.2; 95% CI −9.5 to –2.8). The PTSD treatment effect was 0.39 at 3 (−0.3; 95% CI −0.5 to –0.1) and 0.47 at 12 months (−0.3; 95% CI −0.5 to –0.2). Effect sizes were larger in a subgroup of participants with the top 50% of baseline symptom scores (depression: d=0.50, d=0.74; PTSD: d=0.58, d=0.94, at 3 and 12 months, respectively). There were no statistically significant differences in change for substance use or IPV. At baseline, only 12% of participants had past 3-month substance use and 32% had past 3-month or ongoing experiences of IPV, which made these outcomes challenging to evaluate.

Conclusions

CETA was effective for reducing depression and PTSD, including among high severity participants and at an extended follow-up. Future studies with increased power for substance use and IPV outcomes are warranted.

Clinical implications

CETA is a recommended treatment for depression and PTSD among this population.

Trial registration number

NCT04242992.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Intimate partner violence (IPV) and related mental health problems are common in South Africa and can lead to poor HIV outcomes, such as low retention in care and viral non-suppression. There is a lack of evidence-based mental healthcare options for women living with HIV who have experienced IPV.

WHAT THIS STUDY ADDS

• Among women living with HIV and past-year IPV experiences, we found that Common Elements Treatment Approach (CETA) was an effective treatment for depression and post-traumatic stress disorder compared with a control condition.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• CETA is recommended to treat common mental health problems among women with HIV and experiences of IPV.

Introduction

In South Africa, almost eight million people are living with HIV (PWH).¹ To achieve epidemic control, South Africa aims to achieve the UNAIDS 95-95-95 targets by 2030 whereby 95% of PWH know their HIV status, 95% of people who know their status are in care and receiving antiretroviral therapy (ART), and 95% of people on ART are virally suppressed.²

Certain populations remain well below the 95-95-95 targets, including women who have experienced intimate partner violence (IPV). A recent pooled analysis of 57 surveys in sub-Saharan Africa found that women who experienced IPV in the past year were less likely to be virally suppressed than those who had not experienced IPV.³ This is a particular concern in South Africa, where up to 50% of women experience IPV in their lifetime.⁴ Experiencing IPV is associated with increased risk for mental and behavioural health problems, including depression, post-traumatic stress disorder (PTSD) and substance use,⁵ which are themselves associated with poor HIV outcomes.⁶ Addressing mental health and IPV is critical to achieving the 95-95-95 targets in South Africa. However, there is a substantial mental health treatment gap in South Africa similar to most low-income and middle-income countries (LMICs): 75%–92% of individuals who need mental healthcare do not receive treatment.⁷

The Common Elements Treatment Approach (CETA) is an evidence-based, transdiagnostic intervention comprised of cognitive–behavioural therapy (CBT) components.⁸ CETA was designed to address a variety of commonly co-occurring mental and behavioural health problems. CETA can be delivered by lay counsellors at lower cost than traditional psychological treatments, thereby addressing a critical implementation barrier in LMIC. In diverse LMIC contexts and populations, CETA has been effective at treating a range of mental health problems in randomised controlled trials (RCTs).9,13 In a trial of women experiencing IPV perpetrated by a current male partner with unhealthy alcohol use, CETA reduced IPV substantially when both women and men received CETA.¹⁴ CETA has the potential to improve HIV treatment outcomes by addressing underlying barriers to retention and adherence (eg, IPV and other co-occurring mental health conditions). However, CETA has not been tested for mental health outcomes within the context of HIV and IPV in South Africa, and prior trials were also limited by short (ie, <6 months) follow-up.

Objective

We conducted an RCT of CETA among women in Johannesburg, South Africa who were at risk of poor retention and/or viral non-suppression and reported IPV in the past year. The primary outcome of the trial was a combined retention/viral suppression, and those results are under review. This paper reports on the effectiveness of CETA at reducing depression, PTSD, substance use and IPV, which were secondary trial outcomes, compared with an active control.

Methods

Setting

This was an individual, parallel RCT testing the effectiveness of CETA compared with an active attention control. The study was conducted at two large public sector HIV clinics in Johannesburg, South Africa. Women were recruited from the clinic after initially being referred by clinic staff who believed they might meet the inclusion criteria. Study staff provided a brief overview of the study and then conducted an initial abbreviated consent process for screening to confirm eligibility. The abbreviated consent process explained questions participants would be asked, including questions on IPV.

Participants

Women were eligible for the trial if they met the following:

• >18 years old.

• Living with HIV and on ART.

• At least one of the following:

  • Virally unsuppressed (>50 copies/mL).

  • Defaulted from treatment since their last viral load.

  • Missed or late (>14 days) clinic appointment in the past year.

• Any past 12-month experience of IPV as reported on the Severity of Violence Against Women Scale (SVAWS).¹⁵

• Literate and able to speak and read English, Zulu or SeSotho.

We excluded women who: (1) were currently psychotic or on an unstable psychiatric regimen; (2) had a suicide attempt/ideation with intent and plan and/or self-harm in the past month or (3) were enrolled in any other HIV treatment intervention study.

The IPV inclusion criterion was evaluated using a self-administered tablet-based version of the SVAWS physical/sexual violence subscale. Eligibility was defined as any IPV occurring in the past 12 months at the time of assessment. The SVAWS physical/sexual violence subscale consists of 27 items asking about the frequency of violence perpetrated by their male partner. All women who were screened were given resources for hotlines and local programmes supporting women experiencing IPV. Women who screened as eligible (eg, consented to the abbreviated process and indicated any past 12-month IPV) then completed a full consent process that provided a detailed overview of the study including: what the intervention and control arms were, their chances of receiving the intervention, follow-up procedures, safety monitoring, risks and benefits, and permission to allow the study team access to their medical records for HIV-treatment outcomes at 12 and 24 months.

Interventions

Common Elements Treatment Approach

Participants randomised to the intervention arm received CETA, a modular, transdiagnostic, flexible therapy approach implemented by lay counsellors. CETA is composed of nine evidence-based CBT elements to treat a variety of common problems, including violence, substance use, depression, anxiety and trauma-related symptoms.⁸

12 female lay counsellors were trained via a 10-day in-person training, and then participated in supervision groups, led by a local supervisor. Groups of 5–6 counsellors met weekly with a supervisor for 2 hours to review cases, present problems and decide which treatment elements should be used for each participant. Each supervisor had a weekly call with a CETA trainer to ensure treatment fidelity, provide close oversight of high-risk safety cases and address any implementation difficulties. To be hired as a lay counsellor, individuals had to have a minimum of a high school education and prior experience as a community health worker or related similar experience. These individuals were hired by the South African study team.

Those randomised to CETA were assigned a counsellor who provided 6–12 CETA sessions, aiming for these to be done on a weekly basis for 60 min each. Women were asked if they wanted to invite a male partner to participate in CETA. Only 15 male partners were invited and participated in CETA with a separate counsellor. As CETA is a modular, flexible treatment approach, the elements implemented, dosing and number of sessions were dependent on the client’s clinical presentation. On average, a participant receives 6–12 sessions to complete treatment. Due to COVID-19, which was widespread in South Africa during this trial, as well as challenges women reported in getting to the clinic (ie, distance and time to the clinic), we allowed for CETA to be delivered telephonically, although few participants chose this option (n=23). Participants were reimbursed ZAR50 (~US$3) for transportation costs incurred to attend CETA sessions.

Control

The active control condition included weekly SMS for 12 weeks to participants who provided reminders about their HIV care appointments and asked questions on their safety (ie, suicidal and homicidal intent and IPV) over the past week. In addition to weekly SMS messages, participants received usual care referrals for any mental health problems.

Safety for both arms

Given the potential high-risk nature of a study population experiencing IPV, participants in both arms received safety monitoring, including assessment and planning. At enrolment, counsellors met with each woman to assess their IPV risk and develop a safety plan as needed. Counsellors checked in weekly with all participants for safety assessments and asked: (1) Are you thinking of killing yourself? (2) Do you have a plan to kill yourself? (3) Do you have the means to kill yourself? and (4) Are you at risk of serious injury or death from interpersonal violence? For control participants, this was completed via SMS. For CETA participants, this was completed during the weekly sessions (either in person or telephonically). If a session was missed, participants were telephoned to complete the safety questions and additional planning, as needed. Additional precautions were taken to ensure client safety when completing these questions via SMS or phone, including ensuring privacy during telephone contact and/or personalising and coding SMS messages. For example, study staff may text participants ‘Remember to walk on Wednesday’ or ‘We look forward to seeing you on Wednesday’ to ensure privacy. If a participant responded yes to any of the questions, further safety assessment and immediate intensive safety planning was completed either telephonically or in person by a trained CETA counsellor. CETA trainers were contacted within 24 hours for all safety cases to ensure that local safety protocols were followed.

Measurement and outcomes

Mental health and IPV outcomes were evaluated actively through ACASI, which allowed participants to read questions on a laptop screen and simultaneously hear the questions being read to them while wearing headphones. Text and audio were available in the language of the participant’s choice. ACASI can reduce social desirability bias,¹⁶ which we believed was important given the sensitive nature of the outcomes. The assessment was administered at baseline, and 3 and 12 months postbaseline. Participants received ZAR150 (~US$8) on completion of each assessment. Because our primary outcome was retention based, we made no attempt to retain patients in the study nor ensure they attended a follow-up assessment beyond standard outreach to complete the CETA intervention and follow-up assessments as any effort to do so would undermine the primary outcome (eg, whether CETA improves retention and viral suppression). The measures in this study have been used in several studies that have been conducted among women in sub-Saharan Africa.^{14 17 18}

The outcomes assessment included:

• SVAWS,¹⁵ a 46-item questionnaire evaluating the frequency and severity of IPV. The scale consisted of three subscales: threatened violence (19 items), physical violence (21 items) and sexual violence (6 items). Like the prior trial of CETA’s effectiveness on IPV in Zambia,¹⁴ we analysed the threatened violence subscale and a combined physical/sexual violence subscale as two discrete outcomes. At the 3-month and 12-month follow-up assessments, the reference period for the scale was the past 3 months, whereas the baseline assessment used a reference period of the past 12 months. Continuous subscale scores with higher scores indicating greater severity of IPV were used as outcomes (threatened scale possible range: 19–76; physical/sexual IPV scale possible range: 27–108).

• Centre for Epidemiological Studies-Depression Scale (CES-D),¹⁹ a 20-item scale of past week depression symptoms. A continuous score (possible range 0–60) with higher scores indicating greater severity of depression was used as the outcome.

• Harvard Trauma Questionnaire,²⁰ a 39-item scale of past week PTSD symptoms. A continuous score (possible range 1–4) with higher scores indicating greater severity of PTSD was used as the outcome.

• Alcohol, Smoking and Substance Involvement Screening Test,²¹ which evaluates past 3-month substance use for a range of substance types. A binary variable of any past 3 months was used as the outcome for analysis.

Randomisation

After the baseline assessment, participants were randomised 1:1 to CETA or control using blinded block randomisation through a computer accessed, randomisation code (generated by the US-based investigative team) accessed via a tablet by study staff. Participants were immediately notified of the randomisation result. As the intervention required specific counselling, neither counsellors nor participants were blinded to their study arm, but data capture and analysis was blinded with unblinding only after data analysis was completed on 3-month and 12-month mental health outcomes (and after all data, including HIV data, were collected for the primary 12-month outcome).

Sample size

Based on findings from prior HIV studies²² and assuming 80% power with a 2-sided α=0.05, we determined that 182 participants total (91/treatment arm) were required to detect a 20 percentage point increase in the proportion suppressed and retained (eg, primary HIV outcomes) between the CETA and control arms. The sample size was then increased to 400 total (200/treatment arm) to allow for sufficient power to evaluate mediators and moderators. Based on power calculations from previous CETA trials where the primary outcomes were mental health-based,¹⁴ we believed this sample size was sufficient to estimate at least a medium effect size (Cohen’s d=0.5) for the mental health and IPV outcomes.

Statistical analysis

Primary analyses were by intention-to-treat (ITT) and included all enrolled participants who were not disenrolled after randomisation (online supplemental figure 1). We used multiple imputation (20 sets of imputations) with chained equations to assign values to missing data. We used linear mixed effects regression models to assess the effect of CETA compared with the control group for continuous outcomes: IPV (threatened and physical/sexual), depression and PTSD. We used a generalised linear mixed effect model for the binary outcome of any substance use. Treatment group, time and an interaction term between treatment group and time were included as fixed effects with a random effect for participant ID to account for repeat measures of the same participant over time. Robust SEs were estimated for each outcome using the vce(robust) option in Stata.²³ For all continuous outcomes, we calculated Cohen’s d effect sizes.²⁴ All analyses were conducted in Stata V.16.²³

We conducted the following secondary analyses: (1) a per-protocol analysis where we excluded participants who did not complete CETA and (2) a subgroup analysis of participants in the top 50% of baseline symptoms for the IPV, depression and PTSD outcomes regardless of treatment arm to explore the effectiveness of CETA among participants with the highest severity problems. We also conducted a sensitivity analysis using complete case (ie, non-imputed) data to evaluate the robustness of our assumptions with regard to missing data.

.

Results

Participants were enrolled from 11 November 2021 to 19 July 2023, with 202 randomised to the CETA Arm and 197 randomised to the control arm (online supplemental figure 1). Of the 202 randomised to the CETA Arm, 144 (71.3%) completed CETA and averaged 7.6 sessions (SD 5.5) over an average of 12 (SD 7.0) weeks (online supplemental figure 1). The overall mean age was 40.6 years (SD: 8.8). Most participants were of South African nationality, unemployed but looking for work, reported never being married and experiencing high levels of food insecurity (table 1). Although all participants reported experiencing IPV in the past year (as an inclusion criterion), only 31.5% reported past 3 months or ongoing IPV (n=126) and 36.6% (n=146) were living with the IPV perpetrator. Baseline characteristics were similar when stratified by treatment arm.

Table 1. Baseline demographics of individuals enrolled in a randomised trial of the effect of CETA on mental health outcomes in Johannesburg, South Africa, among women with HIV on ART who have experienced violence and challenges with adherence by treatment arm (N=399).

Demographic characteristic	CETA arm (n=202)	Control arm (n=197)	Total (n=399)
Age, mean (SD)	40.4 (8.3)	40.9 (9.3)	40.6 (8.8)
Race
Black African	196 (97.0)	192 (97.5)	388 (97.2)
Coloured	2 (1.0)	5 (2.5)	7 (1.8)
Missing	4 (2.0)	0	4 (1.0)
Education
Grade 7 or less	12 (6.0)	7 (3.6)	19 (4.8)
Grade 8–10	30 (14.9)	30 (15.2)	60 (15.0)
Grade 11–12	105 (52.0)	101 (51.3)	206 (51.6)
Completed grade school	34 (16.8)	45 (22.8)	79 (19.8)
Certificate/diploma/postsecondary	12 (5.9)	13 (6.6)	25 (6.3)
Degree	4 (2.0)	1 (0.5)	5 (1.3)
Unknown/missing	5 (2.5)	0	5 (1.3)
Currently in school	14 (6.9)	17 (8.6)	31 (7.8)
Employment
Unemployed and looking for work	141 (69.8)	120 (60.9)	261 (65.4)
Unemployed and not looking for work	5 (2.5)	3 (1.5)	8 (2.0)
Self-employed (part-time or full-time)	12 (6.0)	21 (10.7)	33 (8.3)
Employed (part-time or full-time)	38 (18.8)	50 (25.4)	88 (22.0)
Other/missing	6 (3.0)	3 (1.5)	9 (2.3)
Currently living with perpetrator of IPV	75 (37.1)	71 (36.0)	146 (36.6)
Marital status
Never married	147 (72.8)	142 (72.1)	289 (72.4)
Married	35 (17.3)	30 (15.2)	65 (16.3)
Divorced/separated	13 (6.5)	14 (7.2)	27 (6.8)
Widowed	4 (2.0)	11 (5.6)	15 (3.8)
Unknown/missing	3 (1.5)	0	3 (0.8)
Have a current partner	178 (88.1)	181 (91.9)	359 (90.0)
Have children	176 (87.1)	172 (87.3)	348 (87.7)
Suicidal ideation	92 (45.5)	73 (37.1)	165 (41.4)
Homicidal ideation	47 (23.3)	43 (21.9)	90 (22.3)
SVAWS Threatened Score, mean (SD) (range)	38.9 (13.7) (20–73)	37.6 (12.4) (19–68)	38.2 (13.1) (19–73)
SVAWS Physical Score, mean (SD) (range)	61.6 (23.2) (34–125)	59.6 (20.3) (34–107)	60.6 (21.8) (34–125)
CES-D Score, mean (SD) (range)	31.8 (13.0) (5–53)	29.2 (12.7) (3–54)	30.5 (12.9) (21–45)
HTQ Score, mean (SD) (range)	2.7 (0.7) (2.1–3.3)	2.5 (0.7) (2.0–3.1)	2.6 (0.7) (2.1–3.2)
Any past 3-month illicit substance use	28 (13.9)	19 (9.6)	47 (11.8)
Missed a visit within the past year	122 (60.4)	119 (60.4)	241 (60.4)
Late to a visit within the past year	69 (34.2)	54 (27.4)	123 (30.8)
Defaulted on treatment within the past year	114 (56.4)	113 (57.4)	227 (56.9)
Years on ART, mean (SD)	9.0 (5.7)	8.5 (5.7)	8.7 (5.7)
Viral load, copies/mL, median (IQR)	30 (20, 107.5)	39 (20, 139.5)	35 (20, 110)
0–50	125 (61.9)	122 (61.9)	247 (61.9)
51–500	44 (21.8)	38 (19.3)	82 (20.6)
501–1000	5 (2.5)	3 (1.5)	8 (2.0)
1001–5000	8 (4.0)	7 (3.6)	15 (3.8)
5001–50 000	6 (3.0)	8 (4.1)	14 (3.5)
>50 000	4 (2.0)	10 (5.1)	14 (3.6)
Missing	10 (5.0)	9 (4.6)	19 (4.8)

CETA= Common Elements Treatment Approach; Control Arm received 12 weeks of weekly SMS messaging; SVAWS = Severity of Violence Against Women Scale, CES-D = Center for Epidemiologic-Studies Depression Scale; HTQ = Harvard Trauma Questionnaire.

Based on the non-imputed dataset.

Total N=399 instead of 400 because there were individuals who were randomised and then disenrolled after randomisation. See online supplemental figure 1 for more details.

May sum to greater than the total sample size across these three markers of engagement with care as these are non-mutually exclusive categories.

ART, antiretroviral therapy; CES-D, Centre for Epidemiologic-Studies Depression Scale; CETA, Common Elements Treatment Approach; HTQ, Harvard Trauma Questionnaire; IPV, intimate partner violence; SVAWS, Severity of Violence Against Women Scale.

Table 2 presents the ITT analysis for IPV, depression, PTSD and substance use. The Cohen’s d treatment effect of threatened IPV was 0.15 at 3 months (difference in mean change: −2.0; 95% CI −4.6 to 0.7) and 0.18 at 12 months (−2.4; 95% CI −5.2 to 0.4), suggesting almost no effect of CETA on threatened IPV. For physical/sexual IPV, we found a treatment effect of 0.08 at 3 months (−1.7; 95% CI −6.2 to 2.8) and 0.18 at 12 months (−3.9; 95% CI −8.5 to 0.7), again suggesting almost no effect at 3 months and a potentially small effect by CETA on physical/sexual IPV at 12 months. However, 38.2% (n=71) had not yet completed treatment at the time of the 3-month assessment and 8.6% (n=16) completed treatment on the same day as the 3-month assessment. Similarly, 27.4% (n=51) had not yet completed treatment at the time of the 12-month assessment and 1.1% (n=2) completed treatment on the same day as the 12-month assessment. The 27.4% of women who had not yet completed treatment at the time of the 12-month assessment were still attending CETA sessions, but inconsistently. During the time between sessions, these women were still receiving weekly safety check-ins.

Table 2. Estimated mean outcome differences and clinical effect sizes among the full study sample in a randomised trial of the effect of CETA on mental health outcomes in Johannesburg to South Africa among women with HIV on ART who have experienced violence and challenges with adherence (N=399).

	Mean estimates (95% CI)		CETA versus control
	CETA arm (n=202)	Control arm (n=197)	Mean difference in change (95% CI)	Cohen’s d
SVAWS threatening violence
Baseline	38.8 (36.8, 40.7)	37.6 (35.9, 39.3)	–	–
3 months	22.0 (20.8, 23.1)	22.7 (21.6, 23.8)	−2.0 (−4.6, 0.7)	0.15
12 months	21.4 (20.5 to 22.3)	22.5 (21.4 to 23.7)	−2.4 (−5.2, 0.4)	0.18
SVAWS physical and sexual violence
Baseline	61.4 (58.3, 64.7)	59.6 (56.8, 62.4)	–	–
3 months	37.2 (35.4, 39.0)	37.0 (35.4, 38.5)	−1.7 (−6.2, 2.8)	0.08
12 months	35.2 (34.0, 36.3)	37.2 (35.5, 38.9)	−3.9 (−8.5, 0.7)	0.18
CES-D
Baseline	31.8 (30.0, 33.6)	29.3 (27.6, 31.0)	–	–
3 months	15.6 (13.7, 17.4)	16.1 (14.2, 17.9)	−3.1 (−6.3, 0.1)	0.24
12 months	16.5 (14.5, 18.4)	20.0 (18.2, 21.9)	−6.2 (−9.5, –2.8)	0.48
HTQ
Baseline	2.7 (2.6, 2.8)	2.5 (2.4, 2.6)	–	–
3 months	1.7 (1.6, 1.8)	1.8 (1.7, 1.9)	−0.3 (−0.5, –0.1)	0.39
12 months	1.7 (1.6, 1.8)	1.9 (1.8, 2.0)	−0.3 (−0.5, –0.2)	0.47
Illegal substance use*
Baseline	0.3 (0.2, 0.4)	0.3 (0.2, 0.4)	–	–
3 months	0.1 (0.08, 0.2)	0.1 (0.09, 0.2)	0.8 (0.4, 1.4)	–
12 months	0.1 (0.06, 0.2)	0.2 (0.1, 0.3)	0.5 (0.2, 1.0)	–

CETA= Common Elements Treatment Approach; Control Arm received 12 weeks of weekly SMS messaging; SVAWS = Severity of Violence Against Women Scale, CES-D = Center for Epidemiologic-Studies Depression Scale; HTQ = Harvard Trauma Questionnaire.

Assessments were conducted 3 and 12 months after enrolment date. Therefore, some 3-month assessments took place while the CETA arm was still receiving treatment. For SVAWS, the follow-up assessments at 3 and 12 months used a reference period of the past 3 months, whereas the baseline assessment used a reference period of the past 12 months.

Mean estimates were based on predicted values from linear mixed effect models for continuous outcomes (SVAWS, CES-D, HTQ and alcohol use) and logistic mixed effect models for dichotomous outcomes (illegal substance use). All models included fixed effects for treatment (control=0, CETA=1), time (0=baseline, 1=3 months, 2=12 months) and an interaction term between group and time. The interaction term represents the mean difference in change from baseline to 3 and 12 months, respectively, between the treatment groups. Random effects included participant ID.

Cohen’s d effect size was calculated by dividing the mean difference in change by the baseline SD. The baseline SD is based on the non-imputed data.

Missing baseline, 3-month and 12-month data was imputed 20 times through multiple imputation with chained equations.

^*Illegal substance use is a dichotomous outcome; therefore, the risk and 95% CI at each timepoint is reported instead of the mean estimate and 95% CI, and the ratio of risk ratios and 95% CI is reported instead of the mean difference in change and 95% CI.

ART, antiretroviral therapy; CES-D, Centre for Epidemiologic-Studies Depression Scale; CETA, Common Elements Treatment Approach; HTQ, Harvard Trauma Questionnaire; SVAWS, Severity of Violence Against Women Scale.

The treatment effect for depression was small at 0.24 at 3 months (−3.1; 95% CI −6.1 to 0.1). By 12 months, the treatment effect increased to 0.48 (-6.2; 95% CI −9.5 to –2.8) showing not just a sustained but a larger effect on depression. For PTSD, the treatment effect was 0.39 at 3 months (−0.3; 95% CI −0.5 to –0.1) and 0.47 at 12 months (−0.3; 95% CI −0.5 to –0.2) again showing a sustained, medium effect size by CETA on PTSD. For substance use, the ratio of risk ratios was 0.8 at 3 months (95% CI 0.4 to 1.4) and was 0.5 at 12 months (95% CI 0.2 to 1.0). This suggests that substance use was lower among participants who received CETA compared with control; however, these findings were not statistically significant. Our sensitivity analysis using the non-imputed results was similar (online supplemental table 1).

Table 3 presents the per-protocol analysis that excluded participants who did not complete CETA (n=56). In this subsample, the treatment effect of threatened IPV was small at 0.26 at 3 months (−3.3; 95% CI −6.2 to –0.4) and 0.23 at 12 months (−2.9; 95% CI −6.0 to 0.1), unlike the ITT results. Similarly, for physical/sexual IPV, we found an imprecise but small treatment effect at three (Cohen’s d: 0.15) and 12 (Cohen’s d: 0.19) months. Treatment effects for depression were greater when compared with the ITT results at both time points with an effect of 0.41 at 3 months (−5.3; 95% CI −8.7 to –1.9) and 0.52 at 12 months (−6.7; 95% CI –10.6 to –2.8). For PTSD, we saw a sustained, medium-sized effect of the intervention at 0.53 at 3 months (−0.4; 95% CI −0.6 to –0.2) and 0.54 at 12 months (−0.4; 95% CI −0.6 to –0.2). Lastly, for the outcome of any substance use at 3 months the ratio of risk ratios was 0.8 (95% CI 0.5 to 1.2) and at 12 months the ratio of risk ratios was 0.6 (95% CI 0.4 to 0.9).

Table 3. Per-protocol analysis 1: estimated mean outcome differences and clinical effect sizes among participants who completed CETA in a randomised trial of the effect of CETA on mental health outcomes in Johannesburg to South Africa among women with HIV on ART who have experienced violence and challenges with adherence (N=343).

	Mean estimates (95% CI)		CETA versus control
	CETA arm (n=146)	Control arm (n=197)	Mean difference in change (95% CI)	Cohen’s d
SVAWS threatening violence
Baseline	39.4 (37.2, 41.5)	37.6 (35.9, 39.3)	–	–
3 months	21.1 (20.1, 22.0)	22.6 (21.4, 23.7)	−3.3 (−6.2, –0.4)	0.26
12 months	21.3 (20.2, 22.4)	22.5 (21.3, 23.6)	−2.9 (−6.0, 0.1)	0.23
SVAWS physical and sexual violence
Baseline	61.7 (58.0, 65.4)	59.6 (56.8, 62.4)	–	–
3 months	35.9 (34.4, 37.3)	37.0 (35.4, 38.5)	−3.3 (−8.2, 1.7)	0.15
12 months	35.2 (33.8, 36.6)	37.2 (35.5, 38.9)	−4.1 (−9.0, 0.8)	0.19
CES-D
Baseline	33.1 (31.0, 35.2)	29.2 (27.6, 31.0)	–	–
3 months	14.7 (12.8, 16.6)	16.1 (14.4, 17.8)	−5.3 (−8.7, –1.9)	0.41
12 months	17.0 (14.8, 19.2)	19.9 (17.8, 22.0)	−6.7 (−10.6 to –2.8)	0.52
HTQ
Baseline	2.7 (2.6, 2.8)	2.5 (2.4, 2.6)	–	–
3 months	1.6 (1.5, 1.7)	1.8 (1.7, 1.9)	−0.4 (−0.6, –0.2)	0.53
12 months	1.7 (1.6, 1.8)	1.9 (1.8, 2.0)	−0.4 (−0.6, –0.2)	0.54
Illegal substance use
Baseline	0.2 (0.1, 0.4)	0.2 (0.1, 0.4)	–	–
3 months	0.1 (0.07, 0.2)	0.1 (0.09, 0.2)	0.8 (0.5, 1.2)	–
12 months	0.09 (0.05, 0.2)	0.2 (0.1, 0.3)	0.6 (0.4, 0.9)	–

CETA= Common Elements Treatment Approach; Control Arm received 12 weeks of weekly SMS messaging; SVAWS = Severity of Violence Against Women Scale, CES-D = Center for Epidemiologic-Studies Depression Scale; HTQ = Harvard Trauma Questionnaire.

Assessments were conducted 3 and 12 months after enrolment date. Therefore, some 3-month assessments took place while the CETA arm was still receiving treatment. For SVAWS, the follow-up assessments at 3 and 12 months used a reference period of the past 3 months, whereas the baseline assessment used a reference period of the past 12 months.

Mean estimates were based on predicted values from linear mixed effect models for continuous outcomes (SVAWS, CES-D, HTQ and alcohol use) and logistic mixed effect models for dichotomous outcomes (illegal substance use). All models included fixed effects for treatment (control=0, CETA=1), time (0=baseline, 1=3 months, 2=12 months) and an interaction term between group and time. The interaction term represents the mean difference in change from baseline to 3 and 12 months, respectively, between the treatment groups. Random effects included participant ID.

Cohen’s d effect size was calculated by dividing the mean difference in change by the baseline SD. The baseline SD is based on the non-imputed data.

Missing baseline, 3-month and 12-month data was imputed 20 times through multiple imputation with chained equations.

Illegal substance use is a dichotomous outcome; therefore, the risk and 95% CI at each time point are reported instead of the mean estimate and 95% CI, and the ratio of risk ratios and 95% CI is reported instead of the mean difference in change and 95% CI.

ART, antiretroviral therapy; CES-D, Centre for Epidemiologic-Studies Depression Scale; CETA, Common Elements Treatment Approach; HTQ, Harvard Trauma Questionnaire; SVAWS, Severity of Violence Against Women Scale.

Table 4 presents the treatment effects after restricting analyses to participants in the top 50% of baseline symptoms for each outcome regardless of treatment arm. When compared with the ITT results, the effect of CETA on threatened IPV at 3 months was similar when restricted to the top 50% of baseline threatened IPV scores (ITT Cohen’s d: 0.15; Top 50% Cohen’s d: 0.19). However, by 12 months, this restricted analysis found a small treatment effect of 0.32 (−3.1; 95% CI −6.6 to 0.4) compared with the ITT’s effect of 0.18. Unlike the ITT analysis which found almost no effect for physical/sexual IPV at both time points, when limited to those in the top 50% of symptoms, a small treatment effect of 0.30 was observed at 12 months (−4.9; 95% CI −10.3 to 0.5). When limited to those in the top 50% of depressive symptoms, treatment effects were even greater at 3 months (Cohen’s d: 0.50) and 12 months (Cohen’s d: 0.74) months. Similarly, for PTSD, effects were even greater at both time points and suggested a large effect at 12 months (3-month Cohen’s d: 0.58; 12-month Cohen’s d: 0.94).

Table 4. Subgroup analysis: estimated mean outcome differences and clinical effect sizes among participants in the top 50% of baseline symptoms in a randomised trial of the effect of CETA on mental health outcomes in Johannesburg to South Africa to among women with HIV on ART who have experienced violence and challenges with adherence.

	Mean estimates (95% CI)		CETA vs control
	CETA arm	Control arm	Mean difference in change (95% CI)	Cohen’s d
SVAWS threatening violence (CETA arm n=107; control arm (n=97)
Baseline	49.2 (47.3, 51.1)	47.7 (45.9, 49.4)	–	–
3 months	23.2 (21.2, 25.2)	23.5 (21.6, 25.4)	−1.8 (−5.3, 1.7)	0.19
12 months	21.6 (20.1, 23.1)	23.2 (21.4, 25.0)	−3.1 (−6.6, 0.4)	0.32
SVAWS physical and sexual violence (CETA arm n=103; control arm n=98)
Baseline	80.0 (76.5, 83.4)	76.7 (74.0, 79.5)	–	–
3 months	38.8 (35.8, 41.7)	37.5 (35.2, 39.8)	−1.9 (−7.3, 3.4)	0.12
12 months	36.1 (33.8, 38.3)	37.8 (35.2, 40.3)	−4.9 (−10.3, 0.5)	0.30
CES-D (CETA arm n=115; control arm n=92)
Baseline	41.3 (40.1, 42.6)	40.1 (38.6, 41.7)	–	–
3 months	16.7 (14.2,19.3)	18.8 (16.0, 21.6)	−3.3 (−7.4, 0.8)	0.50
12 months	18.0 (15.4, 20.5)	21.7 (18.6, 24.8)	−4.9 (−9.4, –0.5)	0.74
HTQ (CETA arm n=111; control arm n=91)
Baseline	3.2 (3.1, 3.3)	3.1 (3.0, 3.2)	–	–
3 months	1.7 (1.6 to 1.8)	1.9 (1.8, 2.0)	−0.3 (−0.5, –0.05)	0.58
12 months	1.8 (1.7 to 1.9)	2.1 (2.0, 2.2)	−0.4 (−0.6, –0.2)	0.94

CETA= Common Elements Treatment Approach; Control Arm received 12 weeks of weekly SMS messaging; SVAWS = Severity of Violence Against Women Scale, CES-D = Center for Epidemiologic-Studies Depression Scale; HTQ = Harvard Trauma Questionnaire.

Assessments were conducted 3 and 12 months after the enrolment date. Therefore, some 3-month assessments took place while the CETA arm was still receiving treatment. For SVAWS, the follow-up assessments at 3 and 12 months used a reference period of the past 3 months, whereas the baseline assessment used a reference period of the past 12 months.

Mean estimates were based on predicted values from linear mixed effect models for continuous outcomes (SVAWS, CES-D, HTQ and alcohol use) and logistic mixed effect models for dichotomous outcomes (Illegal Substance Use). All models included fixed effects for treatment (Control=0, CETA=1), time (0=baseline, 1=3 months, 2=12 months) and an interaction term between group and time. The interaction term represents the mean difference in change from baseline to 3 and 12 months, respectively, between the treatment groups. Random effects included participant ID.

Cohen’s d effect size was calculated by dividing the mean difference in change by the baseline SD. The baseline SD is based on the non-imputed data.

Missing baseline, 3-month and 12-month data was imputed 20 times through multiple imputation with chained equations.

Top 50% was based on taking any participant, regardless of treatment arm, who had a score greater than or equal to the median for that specific symptom.

ART, antiretroviral therapy; CES-D, Centre for Epidemiologic-Studies Depression Scale; CETA, Common Elements Treatment Approach; HTQ, Harvard Trauma Questionnaire; SVAWS, Severity of Violence Against Women Scale.

Discussion

In a prior trial of women experiencing ongoing IPV in Zambia by their current male partner, CETA was effective in reducing IPV and unhealthy alcohol use among the male partner, with the trial stopped early by its Data and Safety Monitoring Board due to effectiveness.¹⁴ In Thailand, Iraq, Colombia and Ukraine, CETA was effective in reducing depression and PTSD compared with wait list control groups at post-treatment assessments.9,12 In line with these prior trials, we found effects of CETA for depression and PTSD. The current study expands on those findings by having a stronger, active attention control condition and an extended follow-up time point at 12 months demonstrating long-term CETA impacts.

In contrast to the prior CETA trial on IPV, however, our ITT analysis found null or small effect sizes of CETA for threatened and physical/sexual IPV. There are several possible explanations for the smaller observed impact. First, in this trial, we included women who had experienced any violence in the past 12 months. In the Zambia trial, the inclusion criterion for IPV was past 3 months and almost all of the women in the Zambia trial were experiencing ongoing IPV from their current male partner. For many women enrolled in the present study, they either did not have a current partner (22%) or were not cohabiting with their partner (55%), and the IPV they experienced was not ongoing, which renders the SVAWS score less sensitive to change through an intervention like CETA. Enrolling participants with more recently experienced and/or ongoing IPV is recommended for future trials. Relatedly, in the Zambia trial, all women joined the study with their current male partners, who were the perpetrators of the IPV. In this trial, although we allowed male partners to be included, only 15 enrolled. A primary mechanism through which IPV was reduced in the Zambia trial was by a reduction in unhealthy alcohol use among male partners; in this study, CETA was largely only delivered to women, which precluded a key mechanism through which CETA was previously found to be effective. Taking these two trials together, the disparate findings suggest that engaging men in IPV reduction interventions is important.²⁵

We found notable effect sizes for depression and PTSD. In an exploratory moderator analysis, we observed two novel findings for these outcomes. First, CETA treatment effects were larger among participants with the most severe symptoms. CETA is designed to address a range of mental health problems, and a range of symptom severity.⁸ This distinguishes the intervention from others that are designed to treat lower threshold mental health problems and often have stricter exclusion criteria (ie, excluding individuals with the most severe problems).26,28 Second, this was the first time CETA treatment effects on depression and PTSD were evaluated at an extended follow-up—12 months postbaseline. Prior CETA trials and most global mental health trials have been limited by short follow-up. We found that CETA treatment effects were larger at 12 months than at 3 months for depression and PTSD. Reduced 3-month effects may be attributed to a majority (54.8%) of the treatment cases either not completing CETA treatment prior to or completing CETA on the same day as the assessment time point. Our findings could indicate that CETA is effective for depression and PTSD among both milder and more severe levels, with treatment effects largely sustained for a year. However, these findings should be interpreted with caution as these were secondary analyses and we were not specifically powered to detect effects for these subgroups.

Our findings should be interpreted within the context of several limitations. First, findings for our substance use outcome should be considered exploratory in nature given the low prevalence of substance use at baseline, as we were underpowered for this outcome. Second, this trial used self-reported assessment of all outcomes and is thus subject to measurement error, although the use of ACASI likely reduced the risk of social desirability. Furthermore, despite our use of screening questions and eligibility criteria requiring individuals to be literate and read and speak the language, there was the potential for individuals to not fully understand ACASI and/or provide inaccurate responses purposefully leading to additional measurement error.²⁹ Third, much of the IPV reported in this trial was considered historical in nature—many of the women enrolled in the study were no longer with the male partner who perpetrated the violence. Simultaneously, we did not require a male partner to enrol in CETA. Therefore, a combination of these factors limited our ability to detect clinical effects for the outcome of IPV.

Clinical implications

Findings from this trial support the use of CETA to address issues related to trauma and depression among women living with HIV who reported experiences of IPV in the last 12 months. This is especially needed given the limited availability of appropriate and affordable mental health treatment options in South Africa and in LMIC more broadly. Further research is needed to definitively evaluate CETA’s effectiveness for substance use. While treatment effects for IPV were small, evidence from this trial and the prior trials of CETA suggest that CETA is an effective approach to addressing IPV; however, it is important to engage and include male partners when possible.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.

Data availability statement

No data are available.