Successful antiretroviral therapy (ART) for people infected with HIV/AIDS leads to reduced mortality, morbidity, and extended life expectancy,1,2 particularly for patients treated with higher CD4 status.2,3 In addition to individual health benefits for the treated patient, successful ART reduces viremia and decreases the potential for transmission of HIV to others.4 However, the medical and public health benefits of ART rely on long-term, continuous adherence to drug treatment.5 Defaulting from care has long been recognized as an important concern for ART programmes,6,7 but less well appreciated is the fact that a proportion of these patients only temporarily default and will later return to care, termed unstructured treatment interruptions.8 Such health care interruptions, once considered a potential way to reduce treatment costs and toxicity,9 have since been demonstrated in trials and cohorts to increase drug resistance, viremia, morbidity, and death.10,11–13 Health care interruptions seem to be common (approximately one-fifth of patients) across all regions of the world, but understanding how to reduce them still remains a work in progress.14
In this study, we aimed to determine patterns and predictors of unstructured treatment interruptions or loss to follow-up using data from a representative cohort from Uganda.
Data and Study Population
All patient data for this study are from the Mildmay Uganda observational cohort.15 Mildmay is a medical service organization providing ART and other health care services to the HIV-positive community in Uganda using a family-centered multidisciplinary approach to care. The organization opened in 1998 to provide specialized outpatient care for people living with HIV/AIDS and provide teaching and training centre for HIV/AIDS health care personnel. This study was restricted to patients who initiated ART between January 2004 and April 2011, were aged 14 years or older at baseline, and have at least a baseline CD4 cell count. Patient level data are collected for all patients and recorded into a central server. Details on our specific cohort are available elsewhere.15
Outcome Measures and Prediction Variables
Our primary outcomes were health care interruptions, defined as not accessing HIV clinical care for at least 12 months at Mildmay, and loss to follow-up, defined as not accessing care for greater than 12 months and not returning to care at Mildmay up to October 2011. Patients are expected to visit their physician at least once every 6 months to get access to care at Mildmay Uganda. The time to loss to follow-up was the difference between date at treatment initiation and date of last contact. Similarly, time to health care interruption stopped at the start of the period of 12, or more, months. Explanatory variables included demographic (age, sex, marital status, and education), behavioral (sexual activity, disclosure to partner, and partner tested), and clinical variables (CD4 cell count and hepatitis B). Baseline CD4 cell counts were obtained within 6 months before starting treatment. During the period of observation, according to Ugandan Ministry of Health guidelines, patients were considered eligible for treatment if they had a CD4 count ≤250 cells per cubic millimeter (this was changed to ≤350 cells/mm3 in 2012) or a relevant clinical indication.16 CD4 at time of health care interruption was equally measured within 6 months of departure. However, CD4 on return to health care services was restricted to 3 months of follow-up because of the rapid impact of ART on the reconstitution of the immune system.
We used the Fisher exact test and logistic regression to determine if there were any important differences between each pair of outcome groups with respect to demographic, behavioral, and clinical variables. We used Kaplan–Meier survival curves to demonstrate the proportion of health care interruption and loss to follow-up over time relative to explanatory variables of interest. For each Kaplan–Meier plot, a logrank test was to used assess whether associations were significant. Finally, the Cox proportional hazards regression was used to determine the independent predictors of each outcome. Model selection was approached from an information theory paradigm and accomplished by minimizing the Akaike information criterion among a collection of plausible models.17 First-level interactions were included in the model. Verification of proportional hazards testing the interactions between all variables and time and by visual inspection.18 All analyses were conducted in SAS version 9.3 (Cary, North Carolina, USA) and R version 2.15 (Vienna, Austria).
This study was approved by the institutional review boards of Mildmay Uganda, a Uganda National Council for Science and Technology approved board, and the University of Ottawa.
In total, 6970 patients aged 14 years or older initiated ART between January 2004 and April 2012. The median age was 36 [interquartile range (IQR): 29–43) years and 2375 (34.1%) were men. Patients were followed for a median of 3.0 years (IQR: 2.0–4.25). During this observation period, 784 (11.2%) patients had a health care interruption of at least 1 year and 217 (3.1%) patients were lost to follow-up. Table 1 displays the characteristics between those retained in care, those with health care interruptions while on treatment, and those lost to follow-up. Interruptions to health care of 12 months or more were more likely among those initiating at a CD4 cell count of 250 cells per cubic millimeter or higher [odds ratio (OR): 1.29, 95% confidence interval (CI): 1.11–1.50) and less likely to have a higher education (OR: 0.76, 95% CI: 0.62–0.92). Adolescents were much more likely to be lost to follow-up (OR: 3.11, 95% CI: 2.23–4.34). In contrast, having a partner (OR: 0.22, 95% CI: 0.16–0.31) or being sexually active at baseline (OR: 0.40, 95% CI: 0.28–0.55) was protective.
Using a time-to-event analysis had very little effect on the strength of associations between predictor and outcome variables. Figure 1 shows that patients initiating with higher CD4 cell counts (≥250 cells/mm3) were more likely to go a full year without seeing a physician. These figures also show that those who have interruptions do so within the first 2 years of treatment as opposed to the steadier rate of loss to follow-up over a 7-year period observed in Figure 2. The results of the Cox proportional hazards regression model are summarized in Table 2. These suggest that both CD4 cell counts above 250 cells per cubic millimeter [adjusted hazard ratio (aHR): 1.23, 95% CI: 1.04 to 1.47) and higher education (aHR: 0.77, 95% CI: 0.64 to 0.93) are independently protective of health care interruptions.
Figure 2 demonstrates that CD4 has no significant relationship to loss to follow-up. Results of the Cox proportional hazards regression suggest that age is an important predictor of loss to follow-up. Adolescents are much more likely to be lost to follow-up (aHR: 1.94, 95% CI: 1.38 to 2.73) than those aged 20–49 years. On closer inspection, adolescent females were most likely to be lost to follow-up. Having some secondary education (aHR: 0.38, 95% CI: 0.23 to 0.64), having a partner (aHR: 0.28, 95% CI: 0.20 to 0.38), and being sexually active at baseline (aHR: 0.46, 95% CI: 0.33 to 0.64) were associated with longer retention to care. All tests for the proportionality assumption were not significant.
We examined CD4 changes before and after health care interruptions. These show a generally adequate CD4 cell count (median: 313, IQR: 190–478, n = 332) at time of departure. Determining the difference in CD4 cell counts before and after health care interruptions was limited by the availability of CD4 cell count measurements on return to care (within 6 months of return to care). Of the 53 (16.0%) patients interrupting care with CD4 cell counts below 200 cells per cubic millimeter, the vast majority returned with higher CD4 cell counts. The median change was 207 (IQR: 35–357). In contrast, the median change in CD4 cell counts among the 279 (84%) that left care with CD4 cell counts above 200 was a decrease of 130 (IQR: −23 to −286).
Our study found an important trend of both health care interruptions and loss to follow-up among a comparatively large proportion of our patient population. Most notably, patients who had health care interruptions were more likely to have higher baseline CD4 counts and patients who were lost to follow-up were more likely to be adolescents.
The relatively low level of loss to follow-up (3.1%) in our cohort likely reflects the fact that Mildmay Uganda works within a geographic catchment area that has a functional adherence support system and can trace patients if required.6 Other cohorts in Africa have reported loss to follow-up as high as 50%6; however, a proportion of these may in fact be interrupters, returning to care at a later date. One of the reasons for large proportions of patients being lost from care in early cohorts is possibly considering absence from clinics for a shorter duration, such as 3 months, as lost from care. Another reason is that patients had switched to newer and closer clinics (ie, self-transfers).19,20
In our study, we observed an important proportion (11%) of patients who temporarily discontinued care and then resumed care after at least 1-year long interruption. Although we do not fully understand the reason for health care interruptions, we found that patients at a higher CD4 status when they initiated ART were more likely to experience a health care interruption. This may support the notion that patients initiating ART without having experienced an illness event may require alternative approaches to adherence support. Other studies have suggested that patients subsequently resume care when their health fails again.21 We are now initiating qualitative work to examine this issue in further detail. Currently, Mildmay Uganda engages patients through group activities and social networks. Moreover, Mildmay personnel will occasionally do home visits. These methods are used to improve retention to care. From the very limited literature on interventions for health care interruptions, we can envision using short message service reminders and community support groups.22–24 These may be targeted at patients believed to be at higher risk for health care interruptions. Before such methods can be implemented, however, we first must determine the use and availability of any required technologies among our cohort.
Our study found that adolescents had the greatest risk of loss to follow-up. This likely reflects some of the known additional risks for poor adherence in adolescents such as having left care due to boarding school restrictions on access, denial of illness, and unstable home situations. We hope future qualitative work will help discern the different motivations in this cohort. Although our study monitored patients for an average of 3 years of follow-up, it is possible, even likely, that as longer periods on treatment are experienced, additional patterns of attrition from care will occur. This reinforces the challenge of lifelong therapy and retention.
Limitations of this study include the fact that patients come from a single health service and may not be applicable to other AIDS service organizations or other countries. We collected only a limited number of patient level variables because these are routinely collected in the cohort. We did not examine the viral status of patients to determine if they developed drug resistance during their health care interruptions as neither resistance testing nor viral load monitoring are common in Uganda. We considered a treatment interruption as an absence from care of at least 12 months, which can be considered a conservative definition as others have used shorter periods, from 1 day to 6 months, to determine health care interruptions8,21; using a shorter definition would likely result in the identification of a higher proportion of treatment interruptions. Finally, we were only able to track patient care at Mildmay Uganda. Mildmay Uganda is a well-established facility near a major urban centre, Kampala, and may therefore not be representative of all AIDS providers in East Africa or elsewhere. Therefore, some patients may have temporarily sought care at other clinics outside of the catchment area. Given that only 16% of patients returned with equal or better CD4 cell counts, such patients represent a minority.
Previous research that examined reasons for health care interruptions in the African settings examined short-term interruptions and found that cost and pharmacy stock-outs explained interruptions.21 However, these reasons are unlikely to explain 1-year gaps in care, particularly because ART care is free at Mildmay and pharmacy stock-outs are rare in PEPFAR programme clinics. They typically implement a buffer period of supplies. Our study therefore has important implications for the delivery of successful ART. There is currently considerable debate about how to roll out ART for the combined effectiveness of patient survival and treatment as prevention.25,26 Some argue to initiate patients immediately after diagnosis of infection to reduce the risk of transmission. However, our study found that higher immune status at baseline was indicative of the risk of health care interruptions. If earlier initiation is encouraged, a renewed emphasis on treatment adherence and retention in care will be necessary. The Ugandan Ministry of Health is recommending the decentralization of HIV/AIDS care from the clinic to the community settings. Evidence indicates that this can be as effective as clinic-based care and reduces costs in resource-constrained settings.27 However, these improvements are relevant to more rural populations and not urban or periurban, such as Mildmay’s populations.
In conclusion, our study found an important proportion of all patients initiated on adult ART experienced a health care interruption of at least 1 year before resuming care. Understanding the reasons for health care interruptions or loss to follow-up now represent important targets to improve the health of patients and decrease their potential for HIV transmission.
1. Mills EJ, Bakanda C, Birungi J, et al.. Life expectancy of persons receiving combination antiretroviral therapy in low-income countries: a cohort analysis from Uganda. Ann Intern Med. 2011;155:209–216.
2. Mills EJ, Bakanda C, Birungi J, et al.. Mortality by baseline CD4 cell count among HIV patients initiating antiretroviral therapy: evidence from a large cohort in Uganda. AIDS. 2011;25:851–855.
3. Mills EJ, Bakanda C, Birungi J, et al.. The prognostic value of baseline CD4+ cell count beyond 6 months of antiretroviral therapy in HIV-positive patients in a resource-limited setting. AIDS. 2012;26:1425–1429.
4. Cohen MS, Chen YQ, McCauley M, et al.. Prevention of HIV-1 infection with early antiretroviral therapy. N Engl J Med. 2011;365:493–505.
5. Hogg RS, Heath K, Bangsberg D, et al.. Intermittent use of triple-combination therapy is predictive of mortality at baseline and after 1 year of follow-up. AIDS. 2002;16:1051–1058.
6. Rosen S, Fox MP, Gill CJ. Patient retention in antiretroviral therapy programs in sub-Saharan Africa: a systematic review. PLoS Med. 2007;4:e298.
7. Fox MP, Rosen S. Patient retention in antiretroviral therapy programs up to three years on treatment in sub-Saharan Africa, 2007-2009: systematic review. Trop Med Int Health. 2010;15(suppl 1):1–15.
8. Moore DM, Zhang W, Yip B, et al.. Non-medically supervised treatment interruptions among participants in a universally accessible antiretroviral therapy programme. HIV Med. 2010;11:299–307.
9. Reynolds SJ, Kityo C, Hallahan CW, et al.. A randomized, controlled, trial of short cycle intermittent compared to continuous antiretroviral therapy for the treatment of HIV infection in Uganda. PLoS One. 2010;5:e10307.
10. Hammer SM, Eron JJ Jr, Reiss P, et al.. Antiretroviral treatment of adult HIV infection: 2008 recommendations of the International AIDS Society-USA panel. JAMA. 2008;300:555–570.
11. Oyugi JH, Byakika-Tusiime J, Ragland K, et al.. Treatment interruptions predict resistance in HIV-positive individuals purchasing fixed-dose combination antiretroviral therapy in Kampala, Uganda. AIDS. 2007;21:965–971.
12. Nachega JB, Hislop M, Dowdy DW, et al.. Adherence to nonnucleoside reverse transcriptase inhibitor-based HIV therapy and virologic outcomes. Ann Intern Med. 2007;146:564–573.
13. Nachega JB, Stein DM, Lehman DA, et al.. Adherence to antiretroviral therapy in HIV-infected adults in Soweto, South Africa. AIDS Res Hum Retroviruses. 2004;20:1053–1056.
14. Kranzer K, Ford N. Unstructured treatment interruption of antiretroviral therapy in clinical practice: a systematic review. Trop Med Int Health. 2011;16:1297–1313.
15. Funk A, Kanters S, Nansubuga M, et al.. Cohort profile: the MUg observational cohort. Int J Epidemiol. 2012;41:1594.
16. Ministry of Health. National Antiretroviral Treatment and Care Guidelines for Adults and Children. Kampala, Uganda: Earnest Publishers; 2010.
17. Akaike H. A new look at the statistical model identification. IEEE Trans Automatic Control 1974;19:716–723.
18. Allison PD. Survival Analysis Using the SAS System: A Practical Guide. Cary, NC: SAS Institute; 1995.
19. Geng EH, Nash D, Kambugu A, et al.. Retention in care among HIV-infected patients in resource-limited settings: emerging insights and new directions. Curr HIV/AIDS Rep. 2010;7:234–244.
20. Geng EH, Bangsberg DR, Musinguzi N, et al.. Understanding reasons for and outcomes of patients lost to follow-up in antiretroviral therapy programs in Africa through a sampling-based approach. J Acquir Immune Defic Syndr. 2010;53:405–411.
21. Kranzer K, Ford N. Unstructured treatment interruption of antiretroviral therapy in clinical practice: a systematic review. Trop Med Int Health. 2011;16:1297–1313.
22. Decroo T, Telfer B, Biot M, et al.. Distribution of antiretroviral treatment through self-forming groups of patients in Tete province, Mozambique. J Acquir Immune Defic Syndr. 2011;56:e39–e44.
23. Pop-Eleches C, Thirumurthy H, Habyarimana JP, et al.. Mobile phone technologies improve adherence to antiretroviral treatment in a resource-limited setting: a randomized controlled trial of text message reminders. AIDS. 2011;25:825–834.
24. Lester RT, Ritvo P, Mills EJ, et al.. Effects of a mobile phone short message service on antiretroviral treatment adherence in Kenya (WelTel Kenya1): a randomised trial. Lancet. 2010;376:1838–1845.
25. Cohen MS, Muessig KE, Smith MK, et al.. Antiviral agents and HIV prevention: controversies, conflicts, and consensus. AIDS. 2012;26:1585–1598.
26. Cohen MS, Dye C, Fraser C, et al.. HIV treatment as prevention: debate and commentary-will early infection compromise treatment-as-prevention strategies? PLoS Med. 2012;9:e1001232.
27. Jaffar S, Amuron B, Foster S, et al.. Rates of virological failure in patients treated in a home-based versus a facility-based HIV-care model in Jinja, southeast Uganda: a cluster-randomised equivalence trial. Lancet. 2009;374:2080–2089.