Diagnosis of antiretroviral therapy failure in Malawi: poor performance of clinical and immunological WHO criteria


Corresponding Author M.C. Hosseinipour, UNC Project, 100 Mzimba Road, Private Bag A/104, Lilongwe, Malawi. Tel.: +265 1 755 056; Fax: +265 1 755 954; E-mail: mina_hosseinipour@med.unc.edu


Objectives  In antiretroviral therapy (ART) scale-up programmes in sub-Saharan Africa viral load monitoring is not recommended. We wanted to study the impact of only using clinical and immunological monitoring on the diagnosis of virological ART failure under routine circumstances.

Methods  Clinicians in two urban ART clinics in Malawi used clinical and immunological monitoring to identify adult patients for switching to second-line ART. If patients met clinical and/or immunological failure criteria of WHO guidelines and had a viral load <400 copies/ml there was misclassification of virological ART failure.

Results  Between January 2006 and July 2007, we identified 155 patients with WHO criteria for immunological and/or clinical failure. Virological ART failure had been misclassified in 66 (43%) patients. Misclassification was significantly higher in patients meeting clinical failure criteria (57%) than in those with immunological criteria (30%). On multivariate analysis, misclassification was associated with being on ART <2 years [OR = 7.42 (2.63, 20.95)] and CD4 > 200 cells/μl [OR = 5.03 (2.05, 12.34)]. Active tuberculosis and Kaposi’s sarcoma were the most common conditions causing misclassification of virological ART failure.

Conclusion  Misclassification of virological ART failure occurs frequently using WHO clinical and immunological criteria of ART failure for poor settings. A viral load test confirming virological ART failure is therefore advised to avoid unnecessary switching to second-line regimens.


Diagnostic de l’échec de l’ART au Malawi: mauvais performance des critères cliniques et immunologiques de l’OMS

Objectifs:  Dans la thérapie antirétrovirale (ART) le déploiement de programmes de suivi de la charge virale en Afrique sub-saharienne n’est pas recommandé. Nous avons voulu étudier l’impact de la seule utilisation du suivi clinique et immunologique sur le diagnostic de l’échec virologique ART dans les circonstances de routine.

Méthodes:  Les cliniciens dans deux cliniques ART urbains au Malawi ont utilisé le suivi clinique et immunologique afin d’identifier les patients adultes nécessitant le passage à un traitement ART de seconde ligne. Lorsque les patients répondaient aux critères d’échec clinique et/ou immunologique des directives de l’OMS et avaient une charge virale < 400 copies/ml, il en résultait une mauvaise classification de l’échec virologique ART.

Résultats:  Entre janvier 2006 et juillet 2007, nous avons identifié 155 patients avec des critères d’échec clinique et/ou immunologique de l’OMS. L’échec virologique ART a été mal classifié chez 66 (43%) patients. L’erreur de classification a été significativement plus élevée chez les patients répondant aux critères d’échec clinique (57%) que chez ceux avec des critères immunologiques (30%). En analyse multivariée, l’erreur de classification était associée au fait d’être sous ART depuis < 2 ans (OR = 7,42 (2,63 - 20,95)) et des CD4 > 200 cellules/μl (OR = 5,03 (2,05 - 12,34)). La tuberculose active et le sarcome de Kaposi étaient les conditions les plus courantes entraînant une mauvaise classification de l’échec virologique ART.

Conclusion:  une mauvaise classification de l’échec virologique ART se produit fréquemment avec l’utilisant les critères cliniques et immunologiques d’échec de l’ART de l’OMS dans les régions pauvres. Une mesure de la charge virale confirmant l’échec virologique ART est donc conseillée afin d’éviter des passages non nécessaires vers des traitements de seconde ligne.


Diagnóstico de fallo del TAR en Malawi: desempeño pobre de los criterios de la OMS clínicos e inmunológicos

Objetivos:  En los programas de aumento a escala de la terapia antirretroviral (TAR) en África sub-Sahariana, la monitorización de la carga viral no está recomendada. Queríamos estudiar el impacto de utilizar solamente la monitorización clínica e inmunológica en el diagnóstico del fallo virológico del TAR bajo circunstancias rutinarias.

Métodos:  Los clínicos de dos clínicas urbanas de TAR en Malawi utilizaron monitorización clínica e inmunológica para identificar pacientes adultos para cambiar a segunda línea de tratamiento. Si los pacientes cumplían con los criterios clínicos o inmunológicos de fallo según las guías de la OMS y tenían una carga viral de <400 copias/ml había una clasificación errónea del fallo virológico del TAR.

Resultados:  Entre enero 2006 julio 2007, identificamos 155 pacientes con criterios de la OMS para fallo inmunológico o clínico. El fallo virológico del TAR ha sido mal clasificado en 66 (43%) pacientes. La clasificación errónea era significativamente mayor en pacientes que cumplían con los criterios clínicos de fallo (57%) que en aquellos con criterios inmunológicos (30%). En un análisis multivariado, el error en la classificacón estaba asociado con estar en TAR <2 años (OR=7.42 (2.63, 20.95)) y CD4 >200 cel/¼l (OR=5.03 (2.05, 12.34)). Una tuberculosis active y el sarcoma de Kaposi eran las condiciones más comunes responsables de la clasificación errónea de fallo virológico del TAR.

Conclusión:  La clasificación errónea de fallo virológico del TAR ocurre frecuentemente utilizando los criterios de fallo de TAR de la OMS, tanto clínicos como inmunológicos, para entornos pobres. Una prueba de carga viral que confirme es, por lo tanto recomendable, para evitar posibles cambios a tratamientos de segunda línea.


In the national scale-up programme of antiretroviral therapy (ART) nearly 150 000 Malawians had started free of charge ART by the end of 2007. The progress of individuals taking ART has been favourable. For instance, the 1-year survival on ART is 78% and 95% of the patients were reported to have good adherence (HIV Unit 2007). Even in well-organized ART programmes, some patients will develop ART failure, and this percentage is expected to increase as patients have longer duration on ART. As the number of individuals taking ART in sub-Saharan Africa expands rapidly, ART failure is an increasingly important public health problem.

Monitoring of ART in Malawi follows the public health approach [World Health Organization (WHO) 2006a] and is mostly done on clinical grounds. CD4 count testing is restricted to larger centres. Viral load tests are not required according to the WHO for monitoring ART in resource limited settings (WHO 2006a) and are currently rarely available in Malawi, because of high cost and the need for advanced technical expertise. ART guidelines for resource limited settings (Ministry of Health, Malawi 2006; WHO 2006a) define ART failure according to clinical criteria, based on the WHO staging system (WHO 2006b) or immunological criteria, based on declining or persistently low CD4 counts. In this setting, virological ART failure will generally be identified late. This may result in accumulation of resistance mutations (Hosseinipour et al. 2009) and hence a reduced chance of successful second line ART in an individual, and a higher risk of transmission of resistant HIV strains in the population (Vekemans et al. 2007). Additionally, clinical and immunological failure can be simulated by intercurrent diseases despite an undetectable viral load. Little is known about the impact of clinical and immunological monitoring to select patients for switching to second-line ART in poor settings. We conducted a study within the Malawi ART scale-up programme to explore the degree to which current WHO immunological and clinical criteria of ART failure lead to misclassification of virological ART failure in routine circumstances, and to determine factors associated with misclassification of virological ART failure.



The study took place in two urban ART centres in Malawi, a very poor country in sub-Saharan Africa that has an adult HIV prevalence of 14%. The Lighthouse, in the capital Lilongwe, is a non-governmental ART clinic located on the Kamuzu Central Hospital premises. The ART clinic of Queen Elizabeth Central Hospital (QECH) in Blantyre, the largest city of Malawi, is a government clinic that is coordinated by the College of Medicine, the only medical school in Malawi. Both clinics introduced free ART halfway through 2004 after they had been running a fee paying system since 2000. At the time of the study, approximately 4000 patients were alive on ART in the Lighthouse clinic and around 3500 in QECH. Although both sites had CD4 counting facilities, only the Lighthouse clinic used routine 6-monthly CD4 count monitoring. Laboratory tests for the ART programmes were performed at the University of North Carolina Project and the Johns Hopkins Project, research and care centres affiliated to the Lighthouse clinic and the College of Medicine respectively, and at the main laboratories of Kamuzu Central Hospital and QECH. Both sites had access to basic biochemistry, haematology, microbiology and radiology (including ultrasound) facilities.


Clinicians identified patients, aged 13 years or older, on the basis of clinical and/or immunological monitoring, for switching to second line ART. Patients were subsequently evaluated by ART failure study teams. After informed consent, a history and physical examination were done. Blood samples were taken for a confirmatory CD4 count (either FacsCount; Becton-Dickinson, Franklin Lakes, NJ, USA, or EPICS-MCL Pan-Leuco Gating method; Beckman Coulter, Brea, CA, USA) and HIV-1 viral load (Roche Amplicor®; Roche, Basel, Switzerland; detection level 400 copies/ml). Relevant clinical details and past laboratory results were extracted from ART clinic files. All CD4 counts identified in clinical files were retrieved and included regardless of the laboratory or CD4 quantification method employed. All case notes were reviewed by site principal investigators.

Antiretroviral therapy failure definitions

In WHO guidelines clinical failure is defined as the occurrence of a new or recurrent WHO stage 4 clinical event. We included cases of progressive Kaposi’s sarcoma as clinical failure when Kaposi’s sarcoma had worsened after initial stabilization or regression on ART. However, we did not include some WHO stage 3 conditions such as severe pneumonia or pulmonary tuberculosis as the WHO guidelines suggest in footnotes. WHO defines immunological failure as the confirmed decline of the CD4 count either to less than the pre-treatment value or to <50% of a peak value on ART, or a CD4 count persistently lower than 100 cells/ml. Patients need to have been on ART for at least 6 months with good adherence (WHO 2006a). In patients who fulfilled criteria of clinical and/or immunological failure of the WHO guidelines and whose viral load was <400 copies/ml, ART failure was rejected and misclassification of virological ART failure was present.

Statistical and ethical considerations

The study was approved by the Malawi National Health Sciences Research Committee and the University of North Carolina School of Medicine Committee for the protection of human subjects. Informed consent was obtained from all patients.

Student’s t-test was used to compare continuous variables. The chi-square and Fisher’s exact tests were used to compare categorical variables where appropriate. Associations between misclassification of virological ART failure and demographic, immunological and clinical characteristics of patients were assessed by binary logistic regression modelling. Covariates that were considered were sex, age >18 years, CD4 > 200 cells/μl, duration on ART, weight loss <10% of body weight, diagnosis of active tuberculosis and diagnosis of Kaposi’s sarcoma. sas version 9.1 (Cary, NC) was used for all statistical analyses.



Between January 2006 and July 2007, 155 patients were identified as clinical and/or immunological failures according to WHO criteria. The mean age was 39 years (range: 13–68); 50% were female; the mean CD4 was 152 cells/μl (range: 1–1099); the percentage with a CD4 count <200 cells/μl was 71% and the mean duration on ART was 30.6 months (range: 6.9–127.3). There were 116 patients with immunological failure criteria, 76 patients with clinical failure criteria and 37 with both. Sixty of the 155 patients were from the Blantyre site (38 with clinical failure, 41 with immunological failure and 19 with both) and 95 from the Lilongwe site (38 with clinical failure, 75 with immunological failure and 19 with both). Due to lack of advanced diagnostic facilities, most WHO stage 4 conditions were diagnosed presumptively. A total of 123 patients (79%) were on the standard first-line ART regimen, a generic formulation of lamivudine, stavudine and nevirapine. The remaining 32 patients were taking one of the alternative first-line regimens in which stavudine was replaced by zidovudine, nevirapine by efavirenz or both, because of side effects. At the start of triple-therapy ART, 153 (98%) patients had been naïve to antiretroviral drugs, and none had received single-tablet nevirapine or any other form of prophylaxis for prevention of mother-to-child HIV transmission.

Misclassification of virological antiretroviral therapy failure

Misclassification of virological ART failure was present in 66 (43%) of the 155 patients. The prevalence of misclassification of virological ART failure was significantly lower in patients fulfilling immunological criteria than in those fulfilling clinical failure criteria; more than half of the clinical failures had an undetectable viral load (Table 1). Patients with misclassification of virological ART failure were on ART shorter (22 months vs. 37 months, P < 0.001) and had higher CD4 counts (236 vs. 91 cells/μl, P < 0.001) than those who had virological failure. If we used a viral load threshold of >10 000 copies/ml instead of >400 copies/ml, misclassification of ART failure occurred in 83 patients (53%; 95% CI: 45%–61%).

Table 1.   Misclassification of virological antiretroviral therapy failure in patients fulfilling WHO* criteria of immunological and/or clinical failure
  Misclassification of ART failure†
Alln% (95% CI)
  1. *WHO immunological failure: CD4 drop >50% or <pre-treatment CD4 or CD4 persistently <100 cells/ml. WHO clinical failure: new or recurrent stage IV condition (2).

  2. ART, antiretroviral therapy; CI, confidence interval; WHO, World Health Organization.

  3. †Fulfilling WHO criteria of immunological and/or clinical failure and having an HIV-1 viral load <400 copies/ml.

Clinical failures764356.6 (45.1, 68.1)
 New stage IV event311651.6 (33.0, 70.2)
 Recurrent/progressive stage IV301963.3 (45.0, 81.6)
 New and recurrent/progressive stage IV15853.3 (24.7, 81.9)
 Immunological failures1163530.2 (21.7, 38.7)
  Persistent CD4 < 100 cells/ml291344.8 (25.6, 64.1)
  CD4 drop >50%22313.6 (0, 29.2)
  CD4 drop to below pre-treatment241250.0 (28.4, 71.6)
Any combination of above41717.1 (7.2, 32.1)
Clinical and immunological failures371232.4 (18.4, 50.0)
All failures1556642.6 (34.7, 50.5)

Conditions causing misclassification of virological antiretroviral therapy failure

Kaposi’s sarcoma, wasting syndrome, extra-pulmonary tuberculosis and oesophageal candidiasis were the most frequent conditions leading to a diagnosis of clinical failure. Kaposi’s sarcoma and extra-pulmonary tuberculosis gave rise to the highest number of misclassifications of virological ART failure (Table 2). When we excluded all cases of Kaposi’s sarcoma and extra-pulmonary tuberculosis, the percentage of misclassification decreased from 43% (95% CI: 35%, 51%) to 32% (95% CI: 24%, 40%).

Table 2.   Conditions causing a diagnosis of clinical failure
WHO stage IV event New (% misclassification of ART failure)*Recurrent/ progressive (% misclassification of ART failure)
  1. There were 48 new WHO stage IV events in 46 patients. There were 45 patients with recurrent or progressive WHO stage IV events. Fifteen patients had both a new and a recurrent or progressive WHO stage IV event on presentation.

  2. ART, antiretroviral therapy; HSV, herpes simplex virus; WHO, World Health Organization.

  3. *HIV viral load <400 copies/ml.

Any48 (50%)46 (58%)
Kaposi’s sarcoma3 (33%)30 (77%)
Wasting syndrome13 (54%)4 (0%)
Extra-pulmonary tuberculosis13 (85%)4 (25%)
Oesophageal candidiasis12 (17%)4 (0%)
Cryptococcal meningitis4 (75%)2 (50%)
Pneumocystis pneumonia2 (0%)1 (0%)
Invasive cervical carcinoma01 (100%)
Mucocutaneous HSV, persistent1 (0%)0

Factors associated with misclassification of virological antiretroviral therapy failure

In the univariate analysis of co-variates of misclassification of virological ART failure, CD4 count above 200 cells/μl, duration of ART shorter than 2 years and diagnosis of Kaposi’s sarcoma were associated with misclassification of virological ART failure. After adjustment for sex, age >18 years, CD4 > 200 cells/μl, duration on ART, weight loss <10% of body weight, diagnosis of active tuberculosis and diagnosis of Kaposi’s sarcoma, only CD4 count >200 cells/μl and duration on ART shorter than 2 years remained independent predictors of misclassification of virological ART failure (Table 3). When we did the same analysis separately for patients with immunological failure and those with clinical failure, the results were similar with one exception, CD4 > 200 cells/μl no longer being significantly associated with misclassification of virological ART failure among patients presenting with immunological failure (data not shown).

Table 3.   Predictors of misclassification of ART failure among patients fulfilling WHO criteria of immunological and/or clinical failure (n = 155)
PredictorTotalMisclassification of Art failureUnadjusted OR (95% CI)Adjusted OR (95% CI)P-value
nn (%)
  1. ART, antiretroviral therapy; CI, confidence interval; OR, odds ratio; WHO, World Health Organization.

  2. *Presence of Kaposi’s sarcoma, whether leading to clinical ART failure or not.

  3. †Active tuberculosis, pulmonary or extra-pulmonary.

Age >18 years14364 (44.7)4.05 (0.86, 19.15)3.57 (0.50, 25.55)0.21
Sex (male)7736 (46.8)1.40 (0.74, 2.66)1.58 (0.69, 3.59)0.28
CD4 > 200 cells/mm34430 (68.2)4.46 (2.11, 9.44)5.03 (2.05, 12.34)<0.001
Kaposi’s sarcoma*3726 (70.3)4.72 (2.11, 10.58)2.49 (0.96, 6.44)0.06
Tuberculosis†3316 (48.5)1.35 (0.63, 2.93)1.17 (0.43, 3.18)0.76
Weight loss <10%11754 (46.2)1.79 (0.82, 3.89)2.02 (0.76, 5.38)0.16
Duration on ART:
 <2 years6543 (66.2)7.82 (3.30, 18.52)7.42 (2.63, 20.95)<0.001
 2–3 years4013 (32.5)1.93 (0.74, 5.02)2.55 (0.85, 7.60)0.04
 >3 years5010 (20.0)1.01.0 

Misclassification would be reduced from 43% (95%CI: 35%, 51%) to 38% (95%CI: 30%, 46%) if we restricted immunological criteria by excluding patients with CD4 > 200 cells/μl and if we increased the minimum duration on ART from 6 months to 1 year for both clinical and immunological criteria. However, eight patients would not have been identified with virological ART failure if these modifications were made.


We have demonstrated that the diagnosis of virological ART failure in circumstances where HIV viral load tests are not available is problematic. Using current WHO clinical and immunological criteria for ART failure in resource limited settings leads to a high percentage of misclassification of virological ART failure. This has large public health impact in countries with ART scale-up programmes in sub-Saharan Africa. Had we not confirmed ART failure with a viral load test, a high percentage of our patients would have started second line ART unnecessarily. Second-line ART regimens are more expensive and complicated, have a higher pill burden and therefore may be less successful than the standard first-line regimens in resource limited settings.

Our results show that clinical monitoring was most prone to misclassification of virological ART failure. In sub-Saharan Africa the CD4 count at the start of ART is often very low, and it may take much longer than 6 months of adequate HIV suppression before the CD4 count reaches a level that protects against opportunistic infections. Secondly, despite adequate HIV suppression, immune reconstitution can remain incomplete (Florence et al. 2003), and certain WHO stage 4 conditions may present despite successful ART with adequate CD4 response (Ledergerber et al. 1999). Finally, insufficient diagnostic capacity in resource limited settings may lead to incorrect diagnosis of WHO stage 4 conditions. A study with a computer model of the effect of ART in resource limited settings predicted negligible benefits of virological monitoring over clinical monitoring with respect to long-term survival on ART (Phillips et al. 2008). This conclusion might need modification if the high percentage of misclassification of ART failure we and others (Basenero et al. 2007; Mee et al. 2008; Reynolds et al. 2009) observed using clinical monitoring and the negative effects associated with unnecessary switching to second line ART would be accounted for in the model. Although to a significantly lesser degree than clinical monitoring, immunological monitoring is also vulnerable to misclassification of virological ART failure in our setting. There are indications that common conditions such as malaria (Van Geertruyden et al. 2006), bilharzia (Kallestrup et al. 2005), tuberculosis (Day et al. 2004) and severe bacterial infections (Venet et al. 2008) can lead to temporary reductions of the CD4 count. Medications such as vincristine chemotherapy for Kaposi’s sarcoma, and high-dose corticosteroids may reduce CD4 counts. Irregular or infrequent CD4 monitoring can obscure chronological trends. The interpretation of CD4 count results can therefore be difficult in clinical practice.

Our study has several limitations. We cannot fully evaluate the WHO criteria of ART failure and draw conclusions on the sensitivity and specificity of immunological and clinical monitoring to detect virological ART failure because virological monitoring in the population on ART was not performed. Rather, we have evaluated the positive predictive value (PPV) of the WHO criteria (misclassification of ART failure is the inverse of the PPV). In a study from Uganda that focussed on WHO immunological criteria only (Reynolds et al. 2009), misclassification was higher than in our study (84%). In a South African study, if either immunological criteria or clinical criteria were present misclassification was also higher (79%). Our relatively favourable results may be explained by the comparatively high prevalence of virological ART failure in our study population.

The patient population in both ART clinics may not have been consistently screened for ART failure, because we relied on busy clinicians to identify patients under routine circumstances. Some patients fulfilling clinical or immunological failure criteria may therefore have been missed, but all patients who started second line ART during the study period passed through the study teams. Although we cannot exclude the possibility of selection bias, the strength of our study situation is that it reflects clinical practice in Malawi and other countries in the region closely, which supports the external validity of the results.

We took a commonly used viral load threshold of >400 copies/ml for virological failure, which is lower than the value of 10 000 copies recommended by WHO for resource limited settings (WHO 2006a). Had we used this higher threshold, the percentage of misclassification would have increased from 43% to 53%. If some of the patients with low but detectable viraemia had virological blips (Karlsson et al. 2004) rather than virological ART failure, misclassification of virological ART failure would even have been more prevalent than 43%, which would strengthen the main conclusion of our study.

The possibility of ascertainment bias must be considered. The immunological failure diagnoses were often based on only one or few CD4 results that were available per patient and some were from laboratories that did not have strict quality control measures in place. Definitive diagnoses of WHO stage 4 conditions could not be made frequently due to lack of sophisticated investigations, and presumptive clinical diagnoses may have contributed to poor prediction of virological ART failure. However, these limitations are common throughout sub-Saharan Africa and our diagnostic capacity is still relatively high compared with most other clinics in Malawi. The percentage of misclassification of virological ART failure we found is therefore likely to be representative of ART clinics in resource limited settings.

Our analysis suggests ways in which clinical and immunological failure criteria can be modified to restrict misclassification of virological ART failure. The first is extension of the minimum duration of ART. Studies from Malawi (Ferradini et al. 2006; Aberle-Grasse et al. 2007) and South Africa (Orrell et al. 2007) indicate that virological ART failure in the first year of ART is rare, suggesting that changing the minimum duration of ART from 6 months to 1 year would not delay ART failure diagnosis in many cases. The second is exclusion of those with a CD4 count of more than 200 cells/μl from immunological failure. Excluding all cases with tuberculosis and Kaposi’s sarcoma would further reduce misclassification. However, if we implemented these modifications the prevalence of misclassification of virological ART failure remained unacceptably high (24%) and a large number (n = 36; 40%) of truly failing patients would have been denied access to second line ART. A different approach to the diagnosis of ART failure in settings where viral load testing is not available is therefore urgently required. Diagnosis of ART failure based on a combination of treatment and adherence history details, clinical events, haemoglobin concentration and total lymphocyte or CD4 count (Colebunders et al. 2006) holds promise but was based on expert opinion and needs validation.

In conclusion, implementation of the current WHO criteria of immunological and clinical ART failure for resource limited settings leads to very high rates of misclassification of virological ART failure. Until better ART failure criteria have been formulated, we advise to use clinical and immunological monitoring to triage patients for a viral load test to confirm virological ART failure, in order to avoid frequent unnecessary initiation of second line ART regimens.


We are grateful for the funding of the study from the National AIDS Commission of Malawi. We would like to thank the staff of the ART clinic of QECH and the Lighthouse clinic for their help with data acquisition, the HIV Unit of the Ministry of Health for their advice and support, and Prof. Malcolm E.M. Molyneux and Prof. Joseph J. Eron for critically reviewing an earlier version of the manuscript. This study was presented at the 4th International AIDS Society Conference on HIV Pathogenesis, Treatment and Prevention; Sydney, Australia; 22–25 July, 2007 (Abstract WEAB101).