Why did it help or not help Why did it help or not help? Utilizing qualitative data to understand the results of adherence trials: the China CATS Study and the Uganda WiseMama Study Lora Sabin, DrPH, MA Associate Professor BUSPH Dept of Global Health February 13, 2018

Discussion goals Background: The CATS Study in China (2012-2014) ART adherence in the care cascade Previous work on adherence The CATS Study in China (2012-2014) Why did it work? The Uganda WiseMama Study (2015-2017) What happened? Summary and discussion High retention in HIV care and ART adherence are critical for success BOTH women’s health and elimination of mother-to-child transmission (MTCT) of HIV) There are some substantial obstacles, however – HIV+ pregnant and postpartum women (PPPW) face particular challenges related to BOTH retention in HIV care AND adherence

Background There are some substantial obstacles, however – High retention in HIV care and ART adherence are critical for success BOTH women’s health and elimination of mother-to-child transmission (MTCT) of HIV) There are some substantial obstacles, however – HIV+ pregnant and postpartum women (PPPW) face particular challenges related to BOTH retention in HIV care AND adherence Background

Background: ART retention and adherence With highly effective antiretroviral therapy (ART), we now have the tools to eliminate HIV/AIDS High ART adherence is critical to achieving the last pillar in the HIV care cascade High retention in HIV care and ART adherence are critical for success BOTH women’s health and elimination of mother-to-child transmission (MTCT) of HIV) There are some substantial obstacles, however – HIV+ pregnant and postpartum women (PPPW) face particular challenges related to BOTH retention in HIV care AND adherence

HIV treatment in China China: new infections still increasing Current estimate: 780,000 PLWHA Border epidemics still growing most rapidly Rapid scale-up of ART In 2002 – China implements national free HIV treatment By Sept 2011, ≈109,000 on ART By March 2014, ≈287,000 on ART Among Chinese patients on ART Non-adherence appears common Non-adherence contributes to drug resistance Interventions to improve adherence urgently needed Sources: China MoH (2011), China CDC (2013) Add something on adherence in China? Most studies to date have used SR, not terribly reliable Levels of drug resistance suggest that adherence is a factor in some populations

Previous China research and electronic drug monitoring (EDM) feedback Our previous work in Dali EDM-informed counseling significantly improved ART adherence and CD4 counts Conclusion: EDM-guided adherence support works, but is limited – it doesn’t provide real-time behavioral feedback Real-time monitoring (via Wisepill) Web-linked medication container that sends electronic signal to central server at each opening Allows reminders to be sent at specific times Patient experience with Wisepill Wisepill feasible/acceptable in Uganda (2010) Wisepill feasible/acceptable in China (2013)

So we hypothesized… Could real time reminders (via Wisepill) combined with data-informed counseling improve ART adherence? Could we combine this new technology that allows real time monitoring and therefore real time reminders WITH the kind of counseling we did with AFL?

CATS team & acknowledgements BU CGHD Lora Sabin (PhD) Mary Bachman DeSilva (ScD) Allen Gifford (MD) Christopher Gill (MD) Taryn Vian (PhD) Ariel Falconer (MPH) FHI 360/China Zhong Li (MS) Cheng Feng (PhD) (former China Country Director) Xie Wubin (MPH) Guangxi Provincial CDC ART Clinic (Nanning) Lan Guanghua (MD) All clinic staff members Ditan Hospital Xu Keyi (MD) Harvard University/Mass General Hospital David Bangsberg (MD) Jessica Haberer (MD) We gratefully acknowledge support from the National Institute for Drug Abuse. We appreciate help from: Shoshana Kahana, Richard Denisco, Bram Brooks, Mark Harrold, Evan Hecht, & Katherine Semrau. We thank most sincerely all the individuals in Nanning who participated in this research.

CATS site: Nanning, Guangxi Autonomous Region CATS study site

Study objectives Primary Objectives Secondary Objectives To generate efficacy data of real-time feedback on adherence Secondary Objectives To generate efficacy data of real-time feedback on CD4 count, HIV viral load From protocol: We will have a minimum of 80% power to detect down to a 25 percentage-point difference in proportions achieving ≥95% adherence between arms at post-intervention (75% power post follow-up), and a minimum of 80% power to detect down to a difference of 80 cells/µl in mean change in CD4 count at post-intervention and post follow-up, with a two-sided alpha of p=0.05.

Methods

CATS study design (‘real-time feedback’ intervention) Month 3: Randomization of Enrolled Patients Adherence Monitoring Comparison: Usual Care PLUS (No reminders/ Adherence feedback) Intervention: Reminders & Adherence Feedback No Reminders/ feedback Month 0: Enrollment in Study Eligible patients Pre-intervention Period (3 months) Intervention (6 months) Intervention design and timeline Month 9: End of Active Intervention Post-Intervention Follow-up Period Randomized controlled trial Month 12: End of Follow-up Period

Enrollment and randomization Enrollment: 120 patients recruited Dec 2012-April 2013 Included: individuals 18 years of age and above, currently on ART or about to begin ART, deemed at risk for poor adherence own a cell phone Randomization: block stratified method After 3 months, patients stratified by ‘high’ or ‘low’ adherence, based on average adherence over entire 3 months ≥95% = ‘high adherence’(optimal) <95% = ‘low adherence’(suboptimal) Ensured balanced allocation to intervention/comparison arms We were concerned that at time of randomization, we might inadvertently randomize more high adherers to one group or the other. To avoid this and therefore bias, we used a stratified block randomized process. We calculated each patients months 1-5 adherence and put them into 2 groups – high v low adherers (greater than/less than 95% adherent) and then made sure we had equal randomization to Int/Control in each group.

What happened in intervention arm? 1. SMS reminder to cell phone if device unopened within 30 minutes of dose time Patients chose one of 10 possible reminders; examples: Carry on, carry on! Be healthy, have a happy family. :） 2. Wisepill data used in counseling sessions At monthly clinic visits, Wisepill report given to patient Patients <95% adherence given counseling using report What happened in comparison arm? No reminder messages Wisepill report NOT shared with patient

Wisepill report

Study endpoints Impact on adherence (primary endpoint) % ≥95% adherent post-intervention (M 9) Mean adherence in Month 9 Adherence measure (‘on time’ measure): # doses taken +/- 1 hour of scheduled time _________________________________________________________________________ # prescribed doses Methods: Technical failures as % expected openings (battery failures, etc.) Measures of adherence Main one is the one we found to be most predictive of CD4 and VL in Dali, incorporating both missed and off-time doses A second one is the simple proportion taken. Impact on clinical markers CD4 (cells/µl) mean change: M3 to M9 Undetectable Viral load (UDVL) (RT PCR: <50 copies/ml): % UDVL in M9

Results

Patients’ characteristics at randomization Intervention Comparison Characteristic N (%) or Mean (SD) (N=62) (N=57) p-value Gender (male) 41 (66.1) 35 (61.4) NS Age (years) 36.5 (10.7) 38.8 (9.9) Married 24 (38.7) 38 (66.7) ** Education level Primary only 14 (22.6) 13 (22.8) Middle/secondary school 34 (54.8) Beyond Secondary School 9 (15.8) Currently employed (yes) 35 (56.5) 31 (54.4) Monthly income (yuan) (n=64) 2593 (2456) 3333 (5950) Sample characteristics: Total of 119 were randomized - No stat sig difference in the 2 groups other than married variable. The comparison group was more likely to be married *p<0.05; ** p<0.01

Impact of the intervention Comparison of mean monthly adherence: pre-intervention vs. final intervention month Interpretation – by both EDM measures, patients in the intervention group had higher adherence in Month 9 compared to those in the control group. *p<0.05 **p<0.01 At Month 3, no significant differences between intervention and comparison arms. At Month 9, large increase in adherence in intervention arm, regardless of measure; no significant increase in comparison arm.

Adherence over time, stratified by baseline adherence (optimal v Adherence over time, stratified by baseline adherence (optimal v. suboptimal)

Proportion of subjects achieving adherence ≥ 95% Effect of Real Time Feedback on rates of optimal adherence in Month 9 RR 1.52 (1.16-2.00)** RR 1.68 (1.29-2.19)*** RR 2.34 (1.20-4.58)** Interpretation – this shows proportion at or above 95%, intervention vs. comparison group. For all, quite significant and same for low and high adherers Using on-time adherence measure

Insights from qualitative data Why did it work? Insights from qualitative data n=20 IDIs (intervention subjects, post-intervention) Main findings: 5 themes Salutary "supervisory effect" Subjects described a positive “supervisory” effect of EDM use Specific words included “self-imposed pressure,” “invisible pressure,” “a supervisor,” and “like a teacher overseeing.” Subjects knew WPC sent data; they liked ‘being watched’ WPC served as a cue to action, to take medication I think the bottle is equivalent to an invisible pressure to remind myself of taking medicine on time. It seems I won’t forget when I see it. Now, I basically begin to think that it’s half an hour for me to take my medicine at 9:30, and whatever I’m doing, I will have this in mind. –Male, 20, baseline optimal the motivating influence of objective feedback; the role of the reports in promoting accountability; text messages as simple reminders or to establish routines; and the usefulness of counseling in overcoming barriers.

Insights from qualitative data Why did it work? Insights from qualitative data Reports increased awareness and motivation All subjects liked the adherence reports; they were a helpful reference for maintaining or improving adherence Subjects said reports provided a visual depiction of their punctuality -- easy to read and use to see late or missed doses. They liked positive reports; got worried with poor reports Reports heightened their awareness of recent medication-taking behavior, including changes in such behavior: Now I see 96.6 [percent] for this month and it used to be over 98. I would certainly be aware of the wrong time taking my medicine in this month. I can’t make the same mistake the next month. It just reminds me to have this awareness. –Female, 38, baseline optimal the motivating influence of objective feedback; the role of the reports in promoting accountability; text messages as simple reminders or to establish routines; and the usefulness of counseling in overcoming barriers.

Insights from qualitative data Why did it work? Insights from qualitative data Reports promoted accountability Subjects revealed a desire to please the clinicians, to “follow the doctor’s orders” Subjects viewed their relationship with the doctor as contractual; being adherent was a “promise” to the doctors They saw themselves as trustworthy; wanted to keep promise The pillbox can prove that I have been taking my doses on time, and the clinicians would know that I am adherent…. to let them know I have been taking my doses on time, and I am following their instruction. –Male, 32, baseline optimal I take my promise seriously and I’m a man of my word. When I break my promise, the first thing I do is to blame myself, and next, I’d be concerned about whether other people are angry and think that I’m not trustworthy. –Male, 33, baseline suboptimal the motivating influence of objective feedback; the role of the reports in promoting accountability; text messages as simple reminders or to establish routines; and the usefulness of counseling in overcoming barriers.

Insights from qualitative data Why did it work? Insights from qualitative data Texts worked as simple reports/established routines Subjects viewed texts as a useful reminder for forgotten doses Views of the text messages differed by adherence at baseline: most optimal adherers said they were adherent without reminders, but texts were helpful in the case of unplanned events, unusual business, or when a dose was forgotten (rare) Among suboptimal adherers at baseline, subjects viewed texts as helpful in establishing a routine to promote on-time adherence and alerting them to missed doses (often) ...definitely useful. For example, it would remind you if you forget to take your medicine. The feeling is that I’d think about whether I had forgotten to take my medicine. I’d recall whether I had forgotten and then take it if I had. –Male, 35, baseline suboptimal)

Insights from qualitative data Why did it work? Insights from qualitative data Counseling was useful in overcoming barriers All subjects who engaged in counseling were positive about it Optimal adherers had few/no counseling sessions but suboptimal adherers often engaged in counseling Suboptimal adherers discussed barriers with counselors and identified strategies to overcome them I learned more about counseling. I could forget if I was told just once, but through repeated counseling, I would take the information more seriously. The counselors gave me different examples, like a patient who could always take his ARVs on time or [one] who was not adherent, and the consequences. –Male, 22, baseline suboptimal The pill bottle makes some noises when it’s getting empty. They told me to put some cotton wool in it. –Male, 20, baseline optimal

Next step: The WiseMama Study High retention in HIV care and ART adherence are critical for success BOTH women’s health and elimination of mother-to-child transmission (MTCT) of HIV) There are some substantial obstacles, however – HIV+ pregnant and postpartum women (PPPW) face particular challenges related to BOTH retention in HIV care AND adherence

Background: ART retention and adherence for pregnant & postpartum women ART retention & high adherence: critical for pregnant and post-partum women (PPPW) For women’s health and elimination of mother-to-child transmission of HIV Problem: HIV+ PPPW face major challenges related to ART retention & adherence Uganda background: High burden country: 7.1% prevalence among adults; 8% in pregnant women Rapid scale-up of WHO’s Option B+ initiative to provide antiretroviral therapy (ART) to HIV+ pregnant women Over 95% of HIV+ pregnant women on ART by 2015 High retention in HIV care and ART adherence are critical for success BOTH women’s health and elimination of mother-to-child transmission (MTCT) of HIV) There are some substantial obstacles, however – HIV+ pregnant and postpartum women (PPPW) face particular challenges related to BOTH retention in HIV care AND adherence

Given positive findings from China and the need for ART retention and adherence support among PPPW, we wanted to know… Could triggered reminders (via WPC) combined with data-informed counseling improve ART retention and adherence in this vulnerable population?

The Uganda WiseMama Study Primary Objective To generate efficacy data of triggered reminders plus data-informed counseling (‘real-time feedback’) on ART retention and adherence among HIV-positive pregnant and postpartum women Today: adherence results only! From protocol: We will have a minimum of 80% power to detect down to a 25 percentage-point difference in proportions achieving ≥95% adherence between arms at post-intervention (75% power post follow-up), and a minimum of 80% power to detect down to a difference of 80 cells/µl in mean change in CD4 count at post-intervention and post follow-up, with a two-sided alpha of p=0.05.

WiseMama study design (‘real-time feedback’ intervention) Month 1: Randomization of Enrolled Patients Adherence Monitoring Comparison: Usual Care (No reminders/ Adherence feedback) No Reminders/ feedback Month 0: Enrollment in Study Eligible patients* Pre-intervention Period (1 month) Intervention ( 6 months) Intervention design and timeline  Month 7: End of Active Intervention (postpartum M3) Post-Intervention Follow-up Period (3 months) Intervention: Reminders & Adherence Feedback Randomized controlled trial Month 10: End of Follow-up Period Postpartum M6 *HIV+ pregnant women

Methods: Enrollment Eligibility: ART-naïve pregnant women >18 years, 12-26 weeks of gestation, attending 2 clinics where ANC/HIV integrated care provided: Entebbe Grade B Hospital Mityana District Hospital Once daily regimen: (Tenofovir, 3TC, Efavirenz)

Adherence Measures (ITT approach) Mean adherence Over entire intervention period Over pre-delivery and post-delivery periods In final 30 days of the intervention Proportion at selected thresholds (in each period): ≥95% adherent ≥80% adherent Adherence measure (‘on time’ measure): # doses taken +/- 2 hours of scheduled time _________________________________________________________________________ # prescribed doses

Background characteristics at randomization (n=133) Intervention Comparison Mean/% (SD) (n=69)   (n=64) p-value Age (years) 25.6 (6.8) 25.2 (4.6) 0.73 Gestation age (weeks) 20.4 (5.0) 21.9 (4.2) 0.06 Married 71.0 (45.7) 76.6 (42.7) 0.47 Education level completed Primary 44.9 (50.1) 39.1 (49.2) 0.50 Secondary 49.3 (50.4) 54.7 (50.2) 0.54 First pregnancy 24.6 (43.4) 31.3 (46.7) 0.40 Multiparous women, previous pregnancies 2.2 (1.4) 3.0 (2.1) 0.02 Someone else knew status at enrollment 43.5 (49.9) 40.6 (49.5) 0.74 Disclosed to husband/partner at enrollment 31.9 (46.9) 23.4 0.28 Mean adherence, pre-intervention period 78.5 (23.9) 75.9 (24.5) 0.53

Results: Mean adherence of intervention vs. control arms (n=133) Mean adherence was low and decreased over time No significant differences between intervention and comparison arms

Results: Intervention effect: 95% threshold Very few women reach the 95% threshold, especially in the post-delivery period. No significant differences between intervention and comparison groups.

Results: Intervention effect: 80% threshold Very few women were even 80% adherent, especially postpartum

Why no improvement in adherence? Mixed methods data indicated: the intervention did not address structural and interpersonal barriers Major barrier: Lack of disclosure to male partners: 24.8% of women had disclosed at enrollment Only 56% had disclosed by postpartum month 3 (n=100). Major reasons for not disclosing: At enrollment At PPM3 Fear husband will blame her for HIV 47% 32% Fear of divorce 68% 16% At enrollment At PPM3 Fear husband will blame her for HIV 47% 32% Fear of divorce 68% 16%

Qualitative analysis of post-intervention IDI and FGD data: Motivation to stay on ART was high during pregnancy: women wanted to prevent HIV transmission to their babies. Lack of disclosure to male partners prevented women from taking their medication openly and on time. After delivery women found it difficult to get to the clinic, especially if they did not disclose to partner. Transport to clinics was expensive, women had to ask partners for money to travel to the clinic; travel was time-consuming, and burdensome. Lack of food prevented women from taking their medications.

Qualitative results Needed to hide status and medication from partner …some men are so furious and rude so the women fear to tell them and the women have to take their medication in secrecy and fear that they might be caught. When we come to the hospital you hear women talking about how they have to hide their medication and always lie to where they are going because they are afraid of the husbands.. Whenever I don’t take my medication on time, it is because my spouse is around and so I have to wait for him to sleep. Then I take my medication. If he is away, I easily take my medication, but trouble comes when he comes back.

Qualitative results Economic dependence So, I wait for what my husband gives me to buy the daily food and the truth is that if he doesn’t give me money on certain days, I don’t take my drugs. I wait till the next day when I can get some posho (food), then I take my drugs. … if at all I tell him that I am HIV positive and he chases me, where will I get my daily bread or shelter? I can even not feel at home in my own household? So I decide to take my drugs secretly as he also goes on with his adultery. (Laughter)

Conclusions Real-time feedback did not improve ART retention or adherence among pregnant and postpartum women in two clinic sites in Uganda. This population experiences substantial challenges being retained in ART care and adhering to their ART medications, especially postpartum. Future studies should explore structural and other barriers and design interventions to address them. We plan to analyze quantitative data collected at multiple time points during the trial to contribute to this discussion.

WiseMama Study Team & Acknowledgements Boston University School of Public Health Lora Sabin (PhD) Lisa J Messersmith (PhD) Nafisa Halim (PhD) Mary Bachman DeSilva (ScD) Davidson H Hamer (MD) Allen L Gifford (MD) Rachael Bonawitz (MD) Anna Larson Williams (MPH) Mildmay Uganda Barbara Mukasa (MPH) Harriet Chemusto (MSc) Julia Gasuza (MPH) Philip Aroda (BA) Makerere University Seggane Musisi (MD) We gratefully acknowledge support from NIH/NIMH. We thank the late Esther Kawuma, Dickens Akena, Mary Odiit, Steve Purcell, Lyncy Ha, the clinicians at the two study clinics, and the women in Uganda who participated in WiseMama.

Thank you! Questions?