Like other countries in sub-Saharan Africa, Malawi has high HIV infection rates. Since July 2004, antiretroviral therapy (ART) has been available free of charge to Malawian HIV patients through selected government, mission, and private health facilities. Adults with World Health Organization (WHO) clinical stage III or IV disease or with a CD4 count <200 cells/mm3 are eligible. First-line treatment is a generic combination of stavudine + lamivudine + nevirapine (d4T+3TC+NVP) (marketed as Triomune; Cipla Ltd, Mumbai, India), with 1 tablet taken twice daily. One of the key principles of ART is that doses are not missed. Nonadherence is the primary predictor of virologic treatment failure, and levels of adherence greater than 95% have been suggested as necessary to prevent the development of viral resistance.1-3 Before the wide-scale introduction of ART in Africa, concerns were raised that adherence may be low in poorly educated patients living in conditions of extreme poverty.4
The Malawi national HIV treatment guidelines recommend that adherence levels should be ≥95%. Before starting treatment, patients receive education about the importance of good adherence during group and individual counseling sessions. The only measure of adherence that is routinely used is “pill counting”; patients attend the clinic approximately every 4 weeks, and the number of d4T+3TC+NVP tablets remaining in their bottles is counted (by the patient or by the practitioner). Adherence is estimated by calculating the difference between the number of tablets that have been taken (using the pill count [PC]) and the number of tablets that should have been taken since the last clinic visit.
In this study, we measured adherence to d4T+3TC+NVP over a 4-week period using PCs and a short adherence questionnaire and compared these results with the adherence determined using a medication event monitoring system (MEMS; AARDEX Ltd., Zug, Switzerland). These are electronic pill bottle caps that record the time and date whenever the bottle is opened. Our hypothesis was that adherence measured by pill counting and the MEMS would be the same.
The study took place at the Queen Elizabeth Hospital in Blantyre, the only public hospital in Malawi's largest city. Patients were recruited from the ART clinic between December 2005 and April 2006. ART was provided free of charge, and there were no fees for attending the clinic. We recruited HIV-positive patients aged 16 years and older who had been on d4T+3TC+NVP for at least 3 months, self-administered their medication, and were mentally competent to give consent. Patients were not selected for the study; all patients who attended the clinic on the days when the study team was present and who fitted the entry criteria were approached. Potential recruits to the study were told that the aim was to compare different methods of measuring adherence. On recruitment, each subject was given his or her medication for the next month in a bottle closed with a MEMS cap (MC). The pill bottles with a MC looked similar but not identical to the usual d4T+3TC+NVP bottles, and the patients were not informed about the function of the MC. Ethics approval was obtained from the College of Medicine Research Ethics Committee, University of Malawi.
After approximately 4 weeks, the patients returned to the clinic as usual and were seen by the study nurse, who collected the MCs. A PC was taken of the number of tablets remaining, and the adherence questionnaire was completed by the patient. The questionnaire contained the following question: “Did you miss a tablet yesterday, last week, or last month, or have you ever missed a tablet?” In addition, the subject was asked whether he or she kept the Triomune in the pill bottle provided or in another container. The study nurse was not part of the regular ART clinic staff. The patients were told that the results of the questionnaires were confidential and would not be shared with their ART clinician.
The MC data were read using a MEMS communicator and processed using AARDEX PowerView 2.3.3 software. The MC adherence was calculated as a percentage; the number of bottle opening events divided by the prescribed number of doses multiplied by 100. On any day, if the number of openings exceeded the number of prescribed doses, the data were adjusted so that the number of openings equalled the number of prescribed doses. In addition, opening events repeated within 4 hours were treated as a single event. PC data were interpreted as follows: on recruitment, each patient was given 60 tablets of d4T+3TC+NVP. The observed (O) number of doses taken was calculated as 60 minus the number of tablets returned to the clinic at the next appointment. The expected (E) number of tablets taken was equal to the number of prescribed doses that should have been taken since the previous clinic appointment. The percentage PC adherence was equal to O/E multiplied by 100. The maximum PC adherence permitted was 100% for those patients who returned fewer tablets than expected. The data were analyzed using SPSS version 11 (SPSS Inc., Chicago, IL). The adherences measured by the MC or PC were not normally distributed; thus, nonparametric tests were predominantly used for comparisons.
One hundred patients were approached, and 88 were enrolled into the study. The reasons for refusal were not recorded. Of the 88 patients who completed the study, 4 indicated that they kept their d4T+3TC+NVP tablets in a different pill bottle from that given to them at the clinic, and 4 of the MCs were found to be faulty after they had been used, with no data retrievable from them. Therefore, 80 patients had MC data available for analysis. The remaining MCs seemed to function properly, recording the initial opening on recruitment when the tablets were added and the opening at the clinic for the PC after the month in the study.
The mean age of the subjects was 38.6 years, and 46 (57.5%) were female. The median duration on ART was 13.2 months (range: 3 to 63 months), and the mean time in the study was 28 days (range: 23 to 31 days). Figure 1A shows the MC adherence for the 80 subjects: 57.5% (n = 46) had adherence rates of ≥95%, and 16.2% (n = 13) had MC adherence rates of <80%. Overall, the mean MC adherence rate was 88.1% and the median was 97.5% (range: 8.9% to 100%). Mean MC adherence was not significantly different between women and men: 92.7% versus 81.9% (P = 0.19). There was no difference in mean time on treatment between those with MC adherence ≥95% and those with MC adherence <95% (17.0 vs. 19.1 months; P = 0.39).
The mean PC adherence was 98.6%, and the median PC adherence was 100% (range: 87.5% to 100%). PC adherence was significantly higher than MC adherence (P < 0.001). Figure 1B shows a plot of the MC adherence and PC adherence for all 80 of the subjects in the study. There is no clear relation between the adherence measured by PC and MC: 4 patients had MC adherence <20% but PC adherence of 100%. In the adherence questionnaire, all participants said they had taken their tablets the previous day, 2 said they had missed a tablet in the previous week, and only 4 admitted missing 1 tablet or more in the previous month. These 4 patients had MC adherence rates of ≥95%. Only 23 patients admitted ever having missed tablets since starting on ART. The mean MC adherence of these patients was 92.9% compared with 86.2% in the 57 patients who denied ever missing a tablet (P = 0.67).
The association between ART adherence and viral suppression, HIV resistance, and progression to AIDS and death is well established.1,5,6 A target of ≥95% adherence is often quoted as necessary for durable suppression of HIV viral load,3 although more recent data suggest that this level of adherence may not be so crucial for nonnucleoside reverse transcriptase inhibitor (NNRTI)-containing regimens.7 Reports from other countries in sub-Saharan Africa have generally reported adherence rates equal to or better than those from more developed countries. A recent meta-analysis comparing adherence rates from 31 studies from North America with those of 27 studies from sub-Saharan Africa concluded that adherence rates were significantly higher in Africa.8 For most of these African studies, patient self-reporting was the only measure of adherence used. The HIV community has been heartened by the high rates of adherence reported in the African studies.
Our study is the first to compare the MC with standard adherence measures in a government ART clinic in Africa. We could only find 1 other African study that has used MCs, and that was in a fee-paying clinic.9 We believe that the study population was representative of the ART clinic patient population; we were not selective, and none of the recruited patients were involved in any other clinical trials at the clinic, which might have influenced their adherence. Although it would have been ideal to validate measures of adherence with viral loads and CD4 cell count estimation, the “gold standard” measures of adherence, they are not currently available in this setting.
The mean MC adherence rate was 88.1%, which was significantly less than the mean PC adherence rate of 98.2%. MCs may overestimate or underestimate adherence.10 The system cannot determine whether a tablet is actually taken each time the bottle is opened, and patients may remove more than a single dose at a time for later use. We did ask patients whether they used any other container for their tablets and excluded 4 patients for doing so. We cannot exclude the possibility that some patients removed extra tablets from their pill bottles for later use; this would lead to underestimation of MC adherence. MC data are a relatively objective measure of adherence and have been found to correlate well with viral suppression in developed countries.1,11
Pill counting is known to overestimate adherence compared with MC data,1,11,12 and this also seems to be the case in our study. Patients who have missed doses may discard their excess tablets at the end of the month so as to seem more compliant. Self-reported adherence using the questionnaire correlated poorly with the MC data; mean MC adherence was actually higher in those patients who admitted to missing tablets. Pill counting alone would have suggested that adherence was high in our study; 92.5% of the patients had ≥95% adherence. The MC data suggest that only 57.5% of the patients were achieving ≥95% adherence, however, and 16.2% of the patients had rates <80%. If these figures are representative of the population as a whole, this suggests a potential for less good long-term clinical outcomes and the development of resistance to antiretrovirals, particularly nevirapine. Follow-up in this study was limited to 1 month, and adherence is known to decline with time.9 In addition, just by being enrolled and monitored in the study, patients may have increased their adherence such that we have overestimated their chronic adherence behavior.
This study highlights deficiencies in using pill counting and self-reporting as the main measures of adherence in Malawi. Many clinicians are aware of these deficiencies, but there are no practical alternatives available at the current time. Pill counting and self-report also serve another important function in this setting by ensuring that the issue of adherence is raised at each consultation. Further studies are required in similar settings, in larger patient groups, and over longer periods. Such studies should include the validation of MC data by therapeutic drug monitoring, virologic outcome, and CD4 cell count changes. Improved methods are needed to detect nonadherence to ART in developing countries.
The authors thank the patients and staff at the ART clinic at the Queen Elizabeth Hospital in Blantyre, Malawi for participating in this study and Thokko Ganiza and Dr. Sarah White of the Malawi-Liverpool-Wellcome Trust Clinical Research Programme for their assistance with data entry.
David Bell, Joep van Oosterhout, and David Lalloo designed the study. The day-to-day running of the study and data acquisition were done by David Bell, Yamika Kapitao, Rosemary Sikwese, and Joep van Oosterhout. All authors were involved with the writing of this article and approved the final version for publication.
1. Arnsten JH, Demas PA, Farzadegan H, et al. Antiretroviral therapy adherence and viral suppression in HIV-infected drug users: comparison of self-report and electronic monitoring. Clin Infect Dis
2. Bangsberg DR, Charlebois ED, Grant RM, et al. High levels of adherence do not prevent accumulation of HIV drug resistance mutations. AIDS
3. Paterson DL, Swindells S, Mohr J, et al. Adherence to protease inhibitor therapy and outcomes in patients with HIV infection. Ann Intern Med
4. Harries AD, Nyangulu DS, Hargreaves NJ, et al. Preventing antiretroviral anarchy in sub-Saharan Africa. Lancet
5. Bangsberg DR, Perry S, Charlebois ED, et al. Non-adherence to highly active antiretroviral therapy predicts progression to AIDS. AIDS
6. Garcia DO, Knobel H, Carmona A, et al. Impact of adherence and highly active antiretroviral therapy on survival in HIV-infected patients. J Acquir Immune Defic Syndr
7. Bangsberg DR. Less than 95% adherence to nonnucleoside reverse-transcriptase inhibitor therapy can lead to viral suppression. Clin Infect Dis
8. Mills EJ, Nachega JB, Buchan I, et al. Adherence to antiretroviral therapy in sub-Saharan Africa and North America: a meta-analysis. JAMA
9. Oyugi JH, Byakika-Tusiime J, Charlebois ED, et al. Multiple validated measures of adherence indicate high levels of adherence to generic HIV antiretroviral therapy in a resource-limited setting. J Acquir Immune Defic Syndr
10. Bova CA, Fennie KP, Knafl GJ, et al. Use of electronic monitoring devices to measure antiretroviral adherence: practical considerations. AIDS Behav
11. Liu H, Golin CE, Miller LG, et al. A comparison study of multiple measures of adherence to HIV protease inhibitors. Ann Intern Med
12. Walsh JC, Mandalia S, Gazzard BG. Responses to a 1 month self-report on adherence to antiretroviral therapy are consistent with electronic data and virological treatment outcome. AIDS