SEARCH

SEARCH BY CITATION

Keywords:

  • HIV infections;
  • antiretroviral therapy;
  • national health programs;
  • Tanzania;
  • mortality;
  • lost to follow-up
  • infections par le VIH;
  • traitement antirétroviral;
  • programmes nationaux de santé;
  • Tanzanie;
  • mortalité;
  • perte au suivi
  • infección por VIH;
  • terapia antirretroviral;
  • programa nacional de salud;
  • Tanzania;
  • Mortalidad;
  • pérdida durante el seguimiento

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. References

Objective  To analyse survival and retention rates of the Tanzanian care and treatment programme.

Methods  Routine patient-level data were available from 101 of 909 clinics. Kaplan–Meier probabilities of mortality and attrition after ART initiation were calculated. Mortality risks were corrected for biases from loss to follow-up using Egger’s nomogram. Smoothed hazard rates showed mortality and attrition peaks. Cox regression identified factors associated with death and attrition. Median CD4 counts were calculated at 6 month intervals.

Results  In 88,875 adults, 18% were lost to follow up 12 months after treatment initiation, and 36% after 36 months. Cumulative mortality reached 10% by 12 months (15% after correcting for loss to follow-up) and 14% by 36 months. Mortality and attrition rates both peaked within the first six months, and were higher among males, those under 45 kg and those with CD4 counts below 50 cells/μl at ART initiation. In the first year on ART, median CD4 count increased by 126 cells/μl, with similar changes in both sexes.

Conclusion  Earlier diagnoses through expanded HIV testing may reduce high mortality and attrition rates if combined with better patient tracing systems. Further research is needed to explore reasons for attrition.

Objectif:  Analyser les taux de survie et de rétention dans le programme tanzanien de soins et traitement.

Méthodes:  Les données de routine des patients étaient disponibles dans 101 des 909 cliniques. Les probabilités de Kaplan-Meier pour la mortalité et de désistement après l’initiation de l’ART ont été calculées. Les risques de mortalité ont été corrigés pour les biais liés à la perte au suivi à l’aide du nomogramme Egger. Les taux de risque ajustés montrent des pics de mortalité et de désistement. La régression de Cox a identifié les facteurs associés à la mort et au désistement. Les taux médian de CD4 ont été calculés par intervalles de 6 mois.

Résultats:  Chez 88.875 adultes, 18% ont été perdus au suivi 12 mois après l’initiation du traitement et 36%, 36 mois après. La mortalité cumulée a atteint 10% en 12 mois (15% après correction pour les pertes au suivi) et 14% en 36 mois. Les taux de mortalité et le désistement ont atteint un pic endéans les six premiers mois et étaient plus élevés chez les hommes, ceux de moins de 45 kg et ceux dont la numération des CD4 était inférieure à 50 cellules/μl lors de l’initiation de l’ART. Dans la première année sous ART, le nombre médian de CD4 a augmenté de 126 cellules/μl, avec des changements semblables pour les deux sexes.

Conclusion:  Un diagnostic plus précoce grâce à l’extension des tests VIH pourrait réduire les taux élevés de mortalité et de désistement s’il est combiné avec un meilleur système de traçage des patients. Des recherches supplémentaires sont nécessaires pour explorer les raisons du désistement.

Objetivo:  Analizar las tasas de supervivencia y retención dentro del programa Nacional de Cuidados y Tratamiento del VIH de Tanzania.

Métodos:  Se contó con datos de rutina de pacientes de 101 de 909 clínicas. Mediante la prueba de Kaplan-Meier se calcularon las probabilidades de mortalidad y desgaste después de la iniciación del TAR. Se corrigió el sesgo de la pérdida durante el seguimiento en el riesgo de mortalidad utilizando el nomograma de Egger. Las tasas de riesgo mostraban los picos de mortalidad y desgaste. La regresión de Cox identificó factores asociados con la muerte y desgaste. Los conteos medios de CD4 se calcularon en intervalos de 6 meses.

Resultado:  De 88,875 adultos, un 18% se perdió durante el seguimiento tras 12 meses después de iniciado el tratamiento, y un 36% tras 36 meses. La mortalidad acumulativa alcanzó un 10% a los 12 meses (15% tras corregir la pérdida durante el seguimiento) y 14% a los 36 meses. Las tasas de mortalidad y desgaste aumentaron ambas dentro de los seis primeros meses, y eran mayores entre hombres, aquellos con menos de 45kg y aquellos con conteos de CD4 por debajo de 50 células/μl al momento de iniciar TAR. Durante el primer año recibiendo TAR, el conteo medio de CD4 aumentó en 126 células/μl, con cambios similares en ambos sexos.

Conclusión:  Los diagnósticos tempranos mediante la prueba ampliada de VIH puede reducir unas altas tasas de mortalidad y desgaste si se combina con mejores sistemas de seguimiento de los pacientes. Se requieren más estudios para explorar razones para el desgaste.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. References

HIV care and treatment programmes have scaled up rapidly over the past 5 years in sub-Saharan Africa, with large increases in numbers initiating treatment, and substantial mortality declines in the HIV infected population (UNAIDS 2010; WHO 2010). Estimated ART coverage in the region increased from 28% to 37% from December 2008 to December 2009 (WHO 2010).

However, increases in ART uptake have been accompanied by rising attrition in national programmes (Brinkhof et al. 2008; Cornell et al. 2010; WHO 2010). Meta-analyses of sub-Saharan African ART programmes have yielded cumulative retention figures of 75–80% after 12 months on treatment and 67–77% at 24 months (Fox & Rosen 2010; Tassie et al. 2010). High attrition leads to underestimates of mortality on ART as patients lost to follow-up experience higher mortality than those remaining in programmes (Brinkhof et al. 2009b). Rigorous monitoring systems are needed to ensure accurate estimates of programme performance indicators such as attrition, mortality and treatment outcomes.

Country-level analyses of routinely collected patient data in national programmes are important for evaluating programme performance over time. Despite numerous studies of ART patient outcomes in sub-Saharan Africa, few countries have presented national-level data from multiple sites incorporated into their ART programmes. Cohort analyses of patients in national programmes are often restricted to sub-regions (Boulle et al. 2008), or small sample sizes (Bussmann et al. 2008; Lowrance et al. 2009; Cornell et al. 2010). This paper is the first to report detailed national estimates of retention, mortality and treatment outcomes and their determinants in the Tanzanian ART programme, based on routinely reported individual-level data from 88 875 patients in 101 clinics over 5 years, with the aim of informing government efforts to improve programme performance and patient outcomes.

By 2007, it was estimated that 1.4 million people were living with HIV in Tanzania, corresponding to an adult prevalence of 5.7% (TACAIDS et al. 2008). The National AIDS Control Programme (NACP) is responsible for prevention, care and treatment services, and programme monitoring and evaluation. The Government started providing free ART for eligible patients (defined by a CD4 count <200 cells/μl or WHO stage 4) at the end of 2004, and rapidly decentralised treatment to regional and district hospitals in 2005, and health centres from 2007. By the end of 2009, Tanzania had approved 909 facilities to provide care and treatment services, enabling more than 300 000 patients to access ART.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. References

Data collection

Clinics routinely send electronic or paper reports to the National Care and Treatment database at NACP. Some facilities provide anonymised patient-level data covering socio-demographic and clinical characteristics, while others simply provide clinic-level reports of numbers in care. In 2009, 101 of the 909 clinics contributed electronic records of patient-level data which form the basis of this analysis. The remaining 808 clinics only produced clinic-level reports (based on paper documentation) on numbers in care and on treatment, with no individual patient characteristics. Since this analysis focuses on patient outcomes, only the data from the 101 clinics providing patient-level data could be used. The implications of this restricted sample are considered in the discussion. In February 2010, patient-level datasets from the 101 clinics, located in 37 districts in 10 regions, were cleaned, appended and merged into a dataset containing a single record for each patient. Unique national identifiers were used to follow transfers of patients between clinics.

Data analysis

NACP/Ministry of Health organised a data analysis workshop in February 2010, conducted by staff from the National Institute for Medical Research in Mwanza and the London School of Hygiene and Tropical Medicine (NACP 2010). Participants included members from the Ministry of Health, and ART programme implementing partner organisations. Analyses were performed using Stata 11 (StataCorp 2009).

Patients were included in the analysis if they had ever started ART and were 15 years or older at ART initiation. The patient’s final outcome was defined as their last known status, classified as: in follow up, dead, opted out from the programme, transferred to a clinic without electronic records, or lost to follow up (if more than 3 months elapsed since their last appointment and no other outcome was recorded). Attrition was defined as patients who were lost to follow-up, opted out or transferred to a non-reporting clinic, but excluded those reported to have died. Cross-checks were made for patients transferring between clinics, to ensure that ART start corresponded to the earliest reported date. Baseline clinical characteristics included WHO disease stage, CD4 count, weight and body mass index (BMI), measured at the closest time to ART start, from 3 months prior to 2 weeks following ART initiation.

Survival time was censored at the date the data were exported for analysis for patients still in follow-up, or at the date of the last appointment for patients who opted out, transferred or were lost to follow-up. Kaplan–Meier survival probabilities were calculated up to 4.5 years after initiating ART, stratified by baseline characteristics.

Alternate analyses were conducted to allow for the unknown vital status of patients lost to follow-up. First, reported deaths were used to obtain a minimum estimate of mortality, assuming that all deaths had been recorded. A second analysis provided a maximum mortality estimate by assuming all patients lost to follow-up had died. A third estimate used the nomogram developed by Egger et al. (2011) to correct for excess mortality among those lost to follow-up. This adjustment is based on the relationship between mortality of patients in follow-up, mortality of those lost to follow-up (derived from tracing studies), and the proportion lost to follow-up.

Smoothed hazard rates showed peak periods of mortality and attrition. Cox proportional hazards regression identified baseline factors associated with mortality and attrition. To investigate CD4 changes over time, for each six month period following treatment initiation, the lowest CD4 count per patient was identified. Within those periods median CD4 values were displayed graphically by characteristics of interest.

Ethical approval was sought from NACP, who compiled the data and collaborated in all the analyses. Patients were only identified by study number to preserve anonymity, with other personal identifiers removed from the dataset before analysis. The database was password protected, and only accessible to researchers who had signed data sharing agreements with NACP.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. References

Patient characteristics

Of all patients registered in the 101 clinics included in the analysis, 88 875 (66% women) had initiated ART by February 2010. Of these patients, 88 817 had follow-up data [median 9.5 months, IQR: 3.4–21.6 months], accruing 115 223 person-years of observation. The male to female ratio and age structure were similar across all calendar years, with males older than females at ART start. In 2005, the median CD4 count at ART initiation was 115 [IQR: 41–195] among men and 121 [IQR: 51–197] among women. This steadily increased each year to reach 122 [IQR: 47–205] among men and 144 [IQR: 64–223] among women by 2009. Additional baseline characteristics are shown in Table 1.

Table 1.   Baseline characteristics of 88 875 patients who had started ART by December 2009
Category2005 or before2006200720082009
n%n%n%n%n%
Total70758.012 23913.818 34320.624 92528.026 29329.6
Sex
 Male248935.2431835.3607133.1833633.4865532.9
 Female458664.8792164.712 27266.916 58966.617 63867.1
Age (years)
 15–29120317.0226218.5366720.0517720.8544720.7
 30–39298042.1522442.7787943.010 54642.310 94541.7
 40–49200928.4321226.2464725.3615124.7661025.1
 ≥5087712.4154012.6214911.7304512.2328112.5
 Missing60.110.010.060.0100.0
 Median (IQR)37 (32–44)37 (31–44)36 (31–43)36 (30–44)36 (30–44)
Facility level
 Hosp457964.7855969.912 91570.417 28569.318 00868.5
 Health C.3074.43352.73822.15362.2263310.0
 Dispensary217430.7332227.2501627.3708028.4560721.3
 Missing150.2230.2300.2240.1450.2
Facility type
 Private187926.6334727.3491826.8618224.8724327.5
 Government519673.4889272.613 42573.218 74375.219 05072.4
BMI
 <1964710.5137112.3251715.0402117.5392116.2
 19–2472011.7121010.9209012.4328514.3355514.7
 25+3225.24554.17744.612535.512495.2
 Missing448072.6810572.711 40367.914 34762.615 49464.0
Weight (kg)
 <45135119.1300724.6458525.0595823.9607923.1
 45–54208029.4427434.9653835.6931337.4939735.7
 55+324045.8484239.6706838.6941537.810 35339.4
 Missing4045.71160.91520.82390.94641.8
WHO stage
 15688.08977.313857.5268910.8339412.9
 292413.1197216.1314117.1482719.4567621.6
 3209929.7429635.1774142.210 50542.110 19038.7
 496513.6188615.5309416.9347914.0296811.3
 Missing251935.6318826.0298216.3342513.7406515.5
CD4 cell count
 <50148321.0252520.6343918.7460518.5421816.0
 50–199289240.9476938.8755341.2985139.5914934.8
 ≥200128418.1205216.8328017.9500620.1547420.8
 Missing141620.0289323.6407122.2546321.9745228.3
 Median (IQR)119 (47–196)110 (45–190)121 (52–195)126 (54–203)137 (57–217)

Among the 29 869 men and 59 006 women initiating ART, 3580 (12.0%) men and 4746 (8.0%) women were reported to have died; 6198 (20.7%) men and 11 740 (19.9%) women were lost to follow up; whilst 3155 (10.6%) men and 6618 (11.2%) women transferred to other clinics by the end of follow-up. Under 0.5% of men and women had opted out. The remaining 16 793 (56.2%) men and 35 615 (60.4%) women were still in follow-up.

Probability of dying after ART initiation

All 88 817 patients were included in the following survival analyses and Cox regressions. Sample sizes for analyses restricted to particular calendar years or baseline characteristics (which exclude missing values) are detailed in Table 1.

The minimum estimate (all deaths recorded) of the cumulative mortality risk was 7.5% (95% CI: 7.3–7.7%) 6 months after ART initiation, 9.5% (9.3–9.7) at 12 months, and 12.0% (11.8–12.3) at 24 months. The maximum estimate (assuming all patients lost to follow-up had died) was 18.0% (17.8–18.3), 25.7% (25.4–26.0) and 36.2% (35.8–36.6) at 6, 12 and 24 months, respectively. The steeper increase in mortality over time using the maximum estimate is due to steep rises in loss to follow-up over time from treatment initiation, increasing from 17.9% at 12 months to 27.4% at 24 months. Probability of dying was consistently higher in men than women, with one-year risks reaching 12.2% in men vs. 8.2% in women for the minimum estimate, and 29.1% in men vs. 24.0% in women for the maximum estimate.

Correcting the minimum mortality estimates for excess mortality in those lost to follow-up using Egger’s nomogram yielded an overall correction factor of 1.57 for mortality risk a year after treatment initiation, giving a 12-month cumulative probability of death of 15%. This shows a substantial underestimation of mortality due to bias from loss to follow-up. Separate analysis for each year of initiation gave correction factors varying from 1.4 to 1.7.

Mortality risks 12 months after treatment initiation according to the three variants are shown in Figure 1, by year of initiation. As the programme matured, mortality risks based on recorded deaths decreased: for patients starting in 2005 the probability of dying by 12 months was 11.5%, and this fell to 7.9% for those starting in 2009. In contrast, one-year mortality risks assuming all patients lost to follow-up were dead increased from 21.3% for patients starting in 2005 to 26.6% for those starting in 2009. This rise is explained by increasing loss to follow-up with each calendar year, from 11.5% one year after ART initiation in 2005 to 21.0% in 2009.

image

Figure 1.  One-year mortality risks by year of ART initiation, according to the three mortality scenarios.

Download figure to PowerPoint

Despite increases in loss to follow-up over time, the bias introduced by excess mortality in the lost to follow-up did not increase as the programme matured, as shown by the relatively stable difference between the corrected (medium variant) and minimum mortality estimates over time (Figure 1).

Figure 2 shows the minimum and maximum estimates of mortality risks since ART initiation, stratified by baseline characteristics. Mortality risks increased with baseline WHO stage, and decreased with increasing baseline CD4 count and BMI. The minimum estimate of mortality risk at 12 months was 19.1% (18.4–19.8) for patients in WHO stage 4 vs. 3.1% (2.6–3.6) for those in stage 1, and 16.5% (15.9–17.1) for patients with baseline CD4 counts below 50 cells/μl vs. 5.6% (5.3–6.0) for patients with CD4 counts above 200 cells/μl.

image

Figure 2.  Cumulative mortality risk over time since ART initiation, by baseline clinical characteristics, under minimum and maximum mortality assumptions. Column 1: if all deaths were recorded. Column 2: if all lost to follow-up were dead.

Download figure to PowerPoint

Mortality and attrition rates among ART patients

Changes in (recorded) mortality rates and attrition rates (including patients who transferred, opted out or were lost to follow-up) over time since ART initiation are shown in Figure 3. The crude mortality rate over the total follow-up period was 7.3 deaths/100 person-years (95% CI: 7.2–7.5). The peak mortality rate occurred within the first few months after treatment initiation, reaching over 10 deaths per 100 person-years. Rates then decreased with time, falling below 2% per person-year after 3 years on treatment. Attrition rates were comparatively higher than mortality rates, at around 20–30 per 100 person-years. The apparent steep decline in attrition rates in year 4 may be a selection (frailty) effect: as the data were censored in 2009, the only people still exposed to risk of attrition after 4 years were those who started treatment in 2004/2005 at the very start of the programme, who may have been selected for their higher adherence.

image

Figure 3.  Smoothed hazards of death and attrition following ART initiation.

Download figure to PowerPoint

Table 2 presents crude and adjusted mortality rate ratios and attrition rate ratios by baseline characteristics. Males have significantly higher mortality rates, even after adjusting for baseline characteristics (P < 0.001). Mortality rates are higher for older age groups, and those initiating ART at lower weights, more advanced WHO stage, and lower CD4 counts. Those with missing CD4 counts display significantly higher mortality than those with counts between 50 and 200 cells/μl.

Table 2.   Hazard ratios for death and attrition, by baseline characteristics of ART patients for total follow-up period
Baseline characteristicHazard ratios for death (model 1)Hazard ratios for attrition* (model 2)
Crude (95% CI)Adjusted (95% CI) Crude (95% CI)Adjusted (95% CI)
  1. n for model 1 = 84 837; n for model 2 = 84 232.

  2. Fit of model 1 to data: LR χ2 = 2720.37, P < 0.001; fit of model 2 to data: LR χ2 = 770.41, P < 0.001.

  3. *Excluding those known to have died. Includes those lost to follow-up because they opted out, transferred out and provided no further information, and lost to follow-up for unknown reasons.

  4. Adjusted for all other factors included in the table.

Sex
 Male1.001.001.001.00
 Female0.65 (0.62–0.68)0.59 (0.56–0.62)0.95 (0.93–0.97)0.88 (0.86–0.91)
Age (years)
 15–290.98 (0.92–1.04)0.94 (0.88–1.00)1.20 (1.16–1.24)1.19 (1.15–1.24)
 30–391.001.001.001.00
 40–491.12 (1.06–1.18)1.07 (1.01–1.13)0.91 (0.89–0.94)0.91 (0.88–0.94)
 ≥501.36 (1.27–1.45)1.26 (1.18–1.36)1.01 (0.97–1.04)0.96 (0.92–1.00)
Weight (kg)
 <453.17 (3.00–3.35)2.99 (2.81–3.19)1.37 (1.33–1.41)1.32 (1.27–1.36)
 45–541.60 (1.51–1.69)1.56 (1.47–1.66)1.16 (1.13–1.19)1.14 (1.11–1.18)
 55+1.001.001.001.00
WHO stage
 10.35 (0.30–0.41)0.46 (0.39–0.54)0.94 (0.89–0.98)0.94 (0.89–0.98)
 20.80 (0.74–0.86)0.88 (0.82–0.95)0.99 (0.96–1.03)0.99 (0.96–1.03)
 31.001.001.001.00
 42.54 (2.41–2.67)2.09 (1.98–2.20)1.13 (1.10–1.17)1.13 (1.10–1.17)
CD4 cell count
 <502.19 (2.08–2.67)1.78 (1.68–1.90)1.16 (1.12–1.20)1.09 (1.05–1.13)
 50–1991.001.001.001.00
 ≥2000.74 (0.69–0.79)0.74 (0.68–0.80)1.15 (1.11–1.18)1.18 (1.14–1.22)
 Missing1.39 (1.31–1.47)1.20 (1.13–1.28)1.23 (1.19–1.27)1.19 (1.15–1.23)

Attrition rates were significantly higher for men, those with lower weight, more advanced disease stage, and CD4 counts <50 cells/μl or missing, showing a similar pattern to mortality rate ratios. However, patients with CD4 counts above 200 cells/μl also displayed high attrition rates compared to those with CD4 counts between 50 and 200 cells/μl, suggesting some patients who feel healthy may discontinue treatment. Unlike mortality, attrition was inversely associated with age, with rates being highest amongst young patients.

Improvements in CD4 counts

In the first year of treatment, median six-monthly CD4 counts increased by 126 cells/μl, and improvements continued over the 5 years, although in later years the rate of increase declined. CD4 counts were significantly higher in females than males at baseline (t = −18.58, P < 0.001), with a median of 136 cells/μl for women vs. 114 cells/μl for men, and improved more rapidly in women than in men (Figure 4). CD4 increases were similar across all age groups and health facilities (hospitals, health centres and dispensaries), and did not differ by baseline WHO stage. Patients who started ART with a lower CD4 count showed more rapid improvements in the first year than those with higher baseline counts (Figure 4). However, after five years on ART, median CD4 counts were similar for all patients regardless of initial count, converging to around 450 cells/μl.

image

Figure 4.  Improvements in median CD4 counts by various characteristics.

Download figure to PowerPoint

These CD4 trajectories do not take into account the upward bias in CD4 trends due to the elimination over time of patients who die. In Figure 4, we separated the trajectories of those who died, who showed smaller improvements in CD4 counts after ART initiation and declining counts prior to death. However, the exclusion of patients who died had a negligible effect on the median CD4 counts of the remaining patients, since they accounted for a small fraction of the total patient population. Those lost to follow-up showed improvements similar to patients in follow-up or transferring clinic, suggesting that the majority of patients lost to follow-up were not experiencing CD4 declines prior to attrition. This is not unexpected, given that death is thought to explain only around 40% of loss to follow-up (Brinkhof et al. 2009b).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. References

Routinely collected data from Tanzania’s national ART programme were used in a large-scale cohort analysis of mortality, attrition and CD4 response over time on ART, shedding light on the performance of the Tanzanian programme and the effect of the rapid scale-up on patient outcomes. Restricting the analysis to the 101 clinics with patient-level electronic data may have biased the estimates of mortality and attrition for the Tanzanian programme, since these clinics are generally larger, better equipped facilities, situated in more urban areas: of the 909 ART clinics 28% are hospitals, 51% health centres and 21% dispensaries, whereas the respective proportions contributing electronic data are 71%, 21% and 8% (NACP 2010). Recently established small, decentralised clinics are less likely to have computer equipment.

Patients from computerised clinics may differ systematically from those in other clinics, though it is difficult to determine in which direction this might bias estimates: while small decentralised clinics might be less well equipped than urban clinics, they can provide a more intimate follow up, and large hospitals may provide better services but attract patients with more advanced infection who have worse outcomes (TRAC 2008; Mutevedzi et al. 2010). In our study, no differences were identified in outcomes by type of facility, suggesting that restricting the analysis to the 101 clinics did not bias the estimates. As more clinics start collecting electronic data, it will become possible to investigate differences in patient outcomes between clinics included in this analysis and newly computerised clinics.

The 101 clinics are drawn from only 10 of the 23 regions in Tanzania, and may not be representative of the programme as a whole. However, the 10 regions span rural and urban areas across the country, and represent a range of HIV prevalence levels (from 1.8% to 15.7%) similar to the range for all 23 regions (1.5–15.7%) (TACAIDS et al. 2008). CTC programme coverage in the 10 regions ranged from 13% to 29%, compared to 21% nationally (11–42% over all 23 regions) (NACP 2010). These figures suggest the 10 regions were fairly representative of Tanzania in terms of HIV prevalence and coverage of care and treatment services.

Using routine data has the advantage of providing a realistic assessment of the performance of the ART programme in ordinary government clinics, in contrast to using data from research or trial clinics, where the majority of data on patient outcomes has been reported. Although these best-practice clinics often perform well, these successes cannot easily be translated to lower-resource clinics during programme scale-up.

Mortality risks in the Tanzanian programme are broadly comparable to those found in other African programmes (Egger et al. 2011). Mortality decline in the maturing programme is likely due to a combination of better programme delivery, improved treatment options, and more timely ART initiation: patients started ART at increasingly high CD4 counts between 2005 and 2009. Mortality rates peak in the first few months after ART initiation, then decline steeply, as commonly found in other programmes in low-income countries (Braitstein et al. 2006; Van Cutsem et al. 2011).

Probabilities of attrition in the Tanzanian programme are also similar to those found in other sub-Saharan African programmes (Fox and Rosen 2010; Tassie et al. 2010). Unlike mortality, attrition rates were more sustained over time from ART initiation, suggesting that loss to follow-up in later years is caused by factors other than death, which hinder clinic attendance or adherence to treatment. The finding that loss to follow-up increases over time as the programme is scaled up is corroborated by results from a cross-national study (Brinkhof et al. 2008), and may be due to increased difficulties of tracking a rapidly growing number of patients in expanding programmes. This calls for improved monitoring systems informed by research on the causes of attrition, in order to keep track of patient outcomes and evaluate programme performance. Various interventions have been shown to improve retention, including reducing costs to patients, using community outreach teams, building stronger links between health services and the community, and avoiding drug stock-outs (Forster et al. 2008; Harries et al. 2010). Patient tracing schemes, while effective in reducing loss to follow-up, are more costly (Rosen & Ketlhapile 2010; Tweya et al. 2010).

The improvements in median CD4 counts over time, and differences by baseline characteristics, are similar to those found in a multi-country cohort analysis of CD4 response to ART (Nash et al. 2008). Unsurprisingly, patients who subsequently died showed smaller improvements in CD4 counts compared to those who survived. In contrast, CD4 trajectories of patients lost to follow-up resembled more closely those of patients in follow-up than those of patients who died, indicating that death accounts for a relatively smaller proportion of loss to follow-up in patients with improving CD4 counts.

The higher mortality of men compared to women may be partly explained by men’s lower baseline CD4 counts. This highlights the need for interventions to ensure timely ART initiation in men. Similar associations between mortality and age, CD4 count, WHO stage and weight have been found in other sub-Saharan African programmes (Palombi et al. 2009; May et al. 2010). In particular, national and cross-national analyses in low-income countries have shown that low CD4 count at ART start is a major risk factor for mortality, further emphasising the importance of timely ART initiation (Brinkhof et al. 2009a; Johansson et al. 2010; May et al. 2010).

In light of this, the increase in CD4 counts at ART initiation over time is promising. However, the median baseline CD4 count in Tanzania, at 155 cells/μl in 2009, is still well below the WHO recommended threshold for ART initiation, which was recently increased from 200 to 350 cells/μl (WHO 2010). Patients in low-income countries continue to start ART at much lower CD4 counts than those in high-income countries, and differences persist even after 6 months on treatment, leading to higher mortality in patients from low-income countries (Braitstein et al. 2006). Ensuring timely ART initiation at higher CD4 counts, through expanded HIV testing and clinic decentralisation, could help reduce the high first-year mortality of patients, as well as improving retention (Chan et al. 2010). These measures, coupled with improved systems for monitoring patients lost to follow-up, would enable a more accurate assessment of Tanzania’s programme impact.

This analysis of routine data from the Tanzanian Programme highlighted several improvements in performance over time, such as mortality reduction and more timely treatment initiation at increasing CD4 counts. The programme should continue investing efforts in raising the mean CD4 count at initiation to the WHO recommended threshold through expanded HIV testing, paying particular attention to men, whose lower baseline CD4 counts contribute to their persistently higher mortality. The other major remaining challenge is improving retention. Given that the majority of loss to follow-up does not appear to be death-related, more research is needed to understand the causes of attrition, in order to develop interventions to keep patients on treatment and reintegrate those lost to follow-up (including those with high CD4 counts) back into the programme before their health deteriorates. As one of the causes of attrition is transferring to non-computerised clinics, improving reporting of transfers and record linking between clinics will be key to improving performance. These are priority issues to address in order to ensure that recent achievements in ART access and outcomes can be built on to pave the way towards universal treatment access in Tanzania.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. References
  • Boulle A, Bock P, Osler M, et al. (2008) Antiretroviral therapy and early mortality in South Africa. Bulletin of the World Health Organization 86(9), 678687.
  • Braitstein P, Brinkhof MW, Dabis F, et al. (2006) Mortality of HIV-1-infected patients in the first year of antiretroviral therapy: comparison between low-income and high-income countries. Lancet 367(9513), 817824.
  • Brinkhof MW, Boulle A, Weigel R, et al. (2009a) Mortality of HIV-infected patients starting antiretroviral therapy in sub-Saharan Africa: comparison with HIV-unrelated mortality. PLoS Medicine 6(4), e1000066.
  • Brinkhof MW, Dabis F, Myer L, et al. (2008) Early loss of HIV-infected patients on potent antiretroviral therapy programmes in lower-income countries. Bulletin of the World Health Organization 86(7), 559567.
  • Brinkhof MW, Pujades-Rodriguez M & Egger M (2009b) Mortality of patients lost to follow-up in antiretroviral treatment programmes in resource-limited settings: systematic review and meta-analysis. PLoS One 4(6), e5790.
  • Bussmann H, Wester CW, Ndwapi N, et al. (2008) Five-year outcomes of initial patients treated in Botswana’s National Antiretroviral Treatment Program. AIDS 22(17), 23032311.
  • Chan AK, Mateyu G, Jahn A, et al. (2010) Outcome assessment of decentralization of antiretroviral therapy provision in a rural district of Malawi using an integrated primary care model. Tropical Medicine & International Health 15, 9097.
  • Cornell M, Grimsrud A, Fairall L, et al. (2010) Temporal changes in programme outcomes among adult patients initiating antiretroviral therapy across South Africa, 2002–2007. AIDS 24(14), 22632270.
  • Egger M, Spycher BD, Sidle J, et al. (2011) Correcting mortality for loss to follow-up: a nomogram applied to antiretroviral treatment programmes in Sub-Saharan Africa. PLoS Medicine 8(1), e1000390.
  • Forster M, Bailey C, Brinkhof MW, et al. (2008) Electronic medical record systems, data quality and loss to follow-up: survey of antiretroviral therapy programmes in resource-limited settings. Bulletin of the World Health Organization 86(12), 939947.
  • Fox MP & Rosen S (2010) Patient retention in antiretroviral therapy programs up to three years on treatment in sub-Saharan Africa, 2007–2009: systematic review. Tropical Medicine & International Health 15(Suppl. 1), 115.
  • Harries AD, Zachariah R, Lawn SD & Rosen S (2010) Strategies to improve patient retention on antiretroviral therapy in sub-Saharan Africa. Tropical Medicine & International Health 15, 7075.
  • Johansson KA, Robberstad B & Norheim OF (2010) Further benefits by early start of HIV treatment in low income countries: survival estimates of early versus deferred antiretroviral therapy. AIDS Research and Therapy 7(1), 311.
  • Lowrance DW, Ndamage F, Kayirangwa E, et al. (2009) Adult clinical and immunologic outcomes of the national antiretroviral treatment program in Rwanda during 2004–2005. Journal of Acquired Immune Deficiency Syndromes 52(1), 4955.
  • May M, Boulle A, Phiri S, et al. (2010) Prognosis of patients with HIV-1 infection starting antiretroviral therapy in sub-Saharan Africa: a collaborative analysis of scale-up programmes. Lancet 376(9739), 449457.
  • Mutevedzi PC, Lessells RJ, Heller T, Barnighausen T, Cooke GS & Newell ML (2010) Scale-up of a decentralized HIV treatment programme in rural KwaZulu-Natal, South Africa: does rapid expansion affect patient outcomes? Bulletin of the World Health Organization 88(8), 593600.
  • NACP (2010) Report on the Care and Treatment of HIV in Tanzania 2010. National AIDS Control Programme, Dar es Salaam.
  • Nash D, Katyal M, Brinkhof MW, et al. (2008) Long-term immunologic response to antiretroviral therapy in low-income countries: a collaborative analysis of prospective studies. AIDS 22(17), 22912302.
  • Palombi L, Marazzi MC, Guidotti G, et al. (2009) Incidence and predictors of death, retention, and switch to second-line regimens in antiretroviral-treated patients in sub-Saharan African Sites with comprehensive monitoring availability. Clinical Infectious Diseases 48(1), 115122.
  • Rosen S & Ketlhapile M (2010) Cost of using a patient tracer to reduce loss to follow-up and ascertain patient status in a large antiretroviral therapy program in Johannesburg, South Africa. Tropical Medicine & International Health 15, 98104.
  • StataCorp (2009) STATA/IC 11.1. StataCorp, College Station, TX, USA.
  • TACAIDS, ZAC, National Bureau of Statistics, Office of the Chief Government Statistician, Macro International (2008). Tanzania HIV/AIDS and Malaria Indicator Survey 2007–08. TACAIDS, ZAC, NBS, OCGS, Macro International Inc., Dar es Salaam.
  • Tassie JM, Baijal P, Vitoria MA, Alisalad A, Crowley SP & Souteyrand Y (2010) Trends in retention on antiretroviral therapy in national programs in low-income and middle-income countries. Journal of Acquired Immune Deficiency Syndromes 54(4), 437441.
  • TRAC (2008) TRAC Report on the Evaluation of Clinical and Immunologic Outcomes from the National Antiretroviral Treatment Program in Rwanda, 2004–2005. Treatment and AIDS Research Centre, Republic of Rwanda.
  • Tweya H, Gareta D, Chagwera F, et al. (2010) Early active follow-up of patients on antiretroviral therapy (ART) who are lost to follow-up: the ‘Back-to-Care’ project in Lilongwe, Malawi. Tropical Medicine & International Health 15, 8289.
  • UNAIDS (2010) Global Report: UNAIDS Report on the Global AIDS Epidemic 2010. UNAIDS, Geneva.
  • Van Cutsem G, Ford N, Hildebrand K, et al. (2011) Correcting for mortality among patients lost to follow up on antiretroviral therapy in South Africa: a cohort analysis. PLoS One 6(2), e14684.
  • WHO (2010) Towards Universal Access: Scaling up Priority HIV/AIDS Interventions in the Health Sector: Progress Report 2010. World Health Organization, Geneva.