Access to prophylactic cotrimoxazole and to timely initiation of antiretroviral therapy can tremendously improve the survival of HIV infected children 12345. HIV testing is the "gateway" or first step to introducing infected children to these interventions6. Unfortunately, the coverage of HIV testing still remains low in most resource constrained countries, particularly for children, and existing pediatric clinical algorithms have low sensitivity for detecting HIV, particularly among the severely malnourished 78. Thus, most HIV-infected children remain undiagnosed, even in households with an adult already on antiretroviral treatment (ART)9.

Reasons for low coverage of pediatric testing are varied but include high postnatal dropout from Prevention of Mother-to-Child Transmission (PMTCT) programmes, and relatively few testing facilities, especially in rural areas 1011. In 2005, Malawi had an estimated 91,000 HIV infected children (≤ 14 years) 12. Among them, the cumulative number on ART by March 2006 was 2718, representing only 3.0% of all infected children and 5.8% of patients on ART in Malawi to date 12.

The HIV epidemic is contributing to increased child mortality and severe malnutrition throughout Africa 1314. In Malawi and elsewhere where HIV is highly prevalent, studies suggest that 30% or more of all children admitted to inpatient nutrition rehabilitation units are HIV-infected 13151617. A prior study in Malawi demonstrated that many HIV-infected children with severe acute malnutrition (SAM) can achieve an adequate Weight-For-Height (WFH) with appropriate therapeutic feeding, although recovery times are significantly longer and mortality is much higher compared to HIV-uninfected children 18. Unfortunately in Africa, the low coverage of both HIV testing and therapeutic feeding and the multiple factors preventing caregivers bringing children to hospital and remaining with them for extended periods 71920 means that the majority of all HIV-infected African children do not receive any nutritional care. Thus, it is important to find innovative ways of identifying HIV-infected children, especially those from remote areas, to allow them to benefit from cotrimoxizole prophylaxis, nutritional care, and ARV treatment in developing countries.

Community-based Therapeutic Care (CTC) is an approach for managing SAM in children < 5 years that provides care close to where people live and focuses on early detection of SAM through community mobilization 21. Children with appetite and without complications are managed directly in the community in an Outpatient Therapeutic Programme (OTP) and provided with routine medical care and nutrient-dense, pathogen resistant Ready-to-Use Therapeutic Foods (RUTF) for nutritional rehabilitation 22 on a weekly basis. Children with generalized severe oedema, anorexia and medical complications are treated at inpatient facilities according to national protocols, until they are stabilized and appetite has returned and they are able to rejoin the OTP 2324. Children are discharged from the OTP when they achieve WFH ≥ 80%. Previous evaluations show that CTC achieves high coverage and has lower default and mortality rates compared to traditional therapeutic feeding programs 2526. Throughout Africa, home-based care programs are providing support for HIV-affected households. We undertook this study to identify appropriate ways of diagnosing HIV among malnourished children and to assess the feasibility and outcome of treating severely malnourished HIV-infected children in a community based program. This paper presents the uptake of HIV testing when integrated within an ongoing CTC programme, the comparative outcomes of HIV-infected and uninfected malnourished children enrolled in the programme and the diagnostic value of anthropometric and other admission characteristics for identifying children at risk of HIV where diagnostic testing is not available.

The study was carried out from December 2002 to May 2005 in the Dowa district of central Malawi, where antenatal and adult HIV prevalence were estimated to be 9.8% and 6.4%, respectively 2728 Voluntary HIV Counselling and Testing (VCT) was offered to caregivers and children who were enrolled or had recently graduated from a CTC programme run by the Ministry of Health and the non-governmental organization Concern Worldwide. Clinical records were reviewed to assess our primary outcomes: HIV testing uptake and test results and indices of nutritional recovery in HIV-infected and uninfected children. Two groups of caregivers were invited to participate: 1) those who were discharged from CTC prior to VCT introduction (retrospective cohort), and 2) those who entered CTC after testing was introduced (prospective cohort).

A total of 2,592 children under 5 years of age had participated in the CTC programme prior to the study. Of these, 809 (31.2%) resided outside the study area and were not eligible for the RC. Of the 1783 possible participants, 180 had moved, 69 had died (according to key informants), 113 could not be located due to an improper address and 1,421 were invited to participate [Figure 1]. The cause of death was ascertained for 27 (39%) of the 69 children that died with the main causes being malaria (n = 8), HIV (n = 7), malnutrition (n = 7), and pneumonia (n = 3).

Of the 1421 children invited to participated, 148 (10.4%) did not turn up. Amongst the returnees, the median time between discharge from the programme and the invitation to participate in the study was 15.6 months (IQR: 10.5–23.3), the average age at CTC admission was 30.0 (17.2) months (median: 26.9; IQR: 17.6 – 37.8) and at study enrolment was 47.2 months (median: 44.3; IQR: 34.4–57. 2). VCT uptake was 92.2% (1174/1273) for children and 58.4% (743/1273) for caregivers. The reasons for refusing the HIV test were: need to consult with husband in 15.2% (15/99) of cases, not psychologically ready in 13.1% (13/99) of cases, fear of the results in 8.1% (8/99), no authority for allowing testing in 7.1% (7/99), not needed because the child is healthy in 7.1%(7/99), already tested in 5.1% (5/99) and others (believes already infected or fear of aggravating child anaemia or not willing to cause pain) in 2.0% of cases (3/99). Thirty-eight caregivers (38.4% of refusals) did not disclose the reasons for refusal.

735 children were eligible for the PC. The average age at admission was 26.5 (13.7) months (median: 23.0; IQR: 16.9–34.1). VCT uptake was 97% (714/735) for children and 64.1% (471/735) for parents in this cohort. The reasons for refusal were not collected by study design.

There was no difference in socio-demographic and clinical characteristics at admission between tested and not tested children in either the RC or the PC (data not shown).

Other baseline characteristics of children in the RC and PC groups are shown in Table 2. Family history of tuberculosis (TB) was more common RC (p < 0.01). Only 22 PC children (3.1%) and 29 RC children (2.5%) were HIV-infected (p = 0.45). HIV prevalence was similar amongst parents in both cohorts. Amongst caregivers tested, 5.4% (58/1081) of mothers and 2.9% (3/133) of fathers were HIV-infected giving an overall prevalence amongst caregivers of 5.0% (61/1214).

In both RC and PC cohorts, there was no significant difference in the sex ratio, age distribution and in frequency of family history of tuberculosis among HIV-infected and HIV-negative children (Table 3). HIV-infected children were more likely to be orphaned and to come from a household with at least one HIV proxy indicator (Table 3). Nutritional status at admission differed between HIV-infected and HIV-uninfected children (Table 4). HIV-infected children were more likely to be admitted with MUAC < 110 mm and less likely to have oedema than uninfected children in both cohorts. Oedematous malnutrition, however, was the most frequent admission characteristic in both infected and uninfected children. An equal proportion of HIV-positive (8/22 = 36.4%) and HIV-negative children (249/692 = 36.0%) required inpatient stabilisation prior to referral to OTP (p = 0.971). About one-third of all HIV-positive children were orphaned (at least one parent dead) compared with < 10% of HIV-negative children (p < 0.001 in both cohorts).

Nutritional recovery also varied by HIV status (Table 4). In the PC, 13 HIV-infected children (59.1%) and 523 HIV-uninfected (83.4%) achieved discharge WFH (p = 0.003). Average rate of weight gain was 4.7 g/kg/day in uninfected children and 2.8 g/kg/d in HIV-positive PC children. Median recovery time to discharge was 42 days for uninfected children and 56 days for HIV-positive children. Five HIV-positive children (22.7%) and 89 HIV-negative children (14.2%) defaulted from the programme (p < 0.001). Four HIV-positive children died during the CTC treatment (18.2%); mortality amongst uninfected children was 1.8% (p < 0.001). Estimated daily weight gain was lower in the RC children, in part due to the high prevalence of oedema at admission and use of admission weight to estimate this parameter.

Nutritional status at follow-up varied according to HIV status. Approximately 15 months after discharge, 24 out of 28 (85.7%) HIV-infected RC children were not malnourished (WFH > 80% reference median and no bilateral pitting oedema). However, six of these children had a MUAC below 125 mm, including one with a MUAC < 110 mm, giving an overall malnutrition rate of 35.7% (10 out of 28) in HIV-infected children compared with a malnutrition rate of 2.0% (22/1094) in HIV-negative children (p = 0.001).

The predictive characteristics for variables included in clinical algorithms, for individual proxy indicators, and for the 3 algorithms presently used to diagnose paediatric HIV are shown in Table 5. All three of the clinical algorithms had positive likelihood ratio lower than 10 and negative likelihood ratio higher than 0.1 suggesting limited utility for ruling-out and ruling in paediatric HIV. The presence of one or more HIV proxy indicators in combination with MUAC < 110 mm was also a poor predictor of the presence of an HIV infection. In contrast, however, severely malnourished children with a MUAC > 110 mm and no proxy indicators were 10-fold less likely to be HIV-infected, suggesting that this combination may be useful for ruling out HIV where confirmatory paediatric testing is not widely available. The other criteria with high positive likelihood for predicting HIV (i.e., PLR > 10) are having a deceased father or living in a widow-headed household.

The study aims were to assess whether a CTC programme can be used as an entry point for HIV services, including HIV testing and treatment of malnutrition in HIV-positive children, and to compare outcomes of HIV-positive and HIV-negative children within the programme. More than half of the HIV-infected children in the PC (59.1%) recovered to a satisfactory nutritional status using CTC protocols, suggesting that SAM can be managed in the community for many HIV-infected children. Furthermore, about two-thirds of the infected children identified after discharge were still adequately nourished. This nutritional recovery occurred without use of antiretroviral therapy (ART), suggesting that severe malnutrition was primarily the result of semi-starvation among recovering study children. These findings are different from those normally observed in developed and middle income countries where severe malnutrition in HIV-positive children is typically caused by HIV-related metabolic disturbances, which do not improve without ART 42. Indeed, the poor response to nutritional intervention of cachexia, a form of malnutrition that is primarily due to chronic systemic inflammation is a well-known phenomenon 4344.

The 59.1% recovery rate for HIV-infected children observed in the PC arm of our study includes deaths in both the inpatient-based stabilisation phase and the outpatient-based recovery phase of care. This figure is similar to the 56% recovery rate reported in a study in southern Malawi where RUTF was used for Home Therapy (HT) in the recovery phase of care, after patients had been discharged from a hospital Nutrition Rehabilitation Unit (NRU) 18. As mortality amongst severely malnourished HIV-positive children is usually highest during the initial phase of treatment and may exceed 30% 45, the results from our study are encouraging.

It is possible that our improved recovery rates arise from the decentralised nature of the CTC model of care that is designed to remove barriers to access and promote early presentation before serious complications develop. However, the numbers of HIV subjects in our PC are small and these findings need to be confirmed by larger studies. The results of the present study are also encouraging when compared with the 3-month mortality rate of 42.9% among severely wasted (WFH < 70%) children started on ART, recently reported in Malawi 46. Mortality in this group was > 10-fold higher than among children starting on ART who were not acutely malnourished (WFH > 80%) 46.

As observed in other nutritional studies carried out in Malawi, the HIV-positive children in our cohorts recovered more slowly than the HIV-negative children 1845. The possible reasons for slower weight gain include reduced intake due to poor appetite, nutrient malabsorption, increased incidence of infections that were unresponsive to the broad-spectrum antibiotics used, and increased nutrient requirements due to HIV 47. Despite the possibility of reduced appetite especially at the beginning of treatment, we believe that HIV-positive children may need more RUTF than HIV-negative children to achieve similar growth rates and improvements in other nutritional indices. Increasing the amount of daily energy offered to HIV infected children may improve their weight gain and reduce the length of stay in the program, and further testing of this hypothesis is needed. Continued nutritional surveillance and supplementation after discharge may also help HIV-infected children to remain well-nourished. Reducing recovery time and subsequent length and cost of program participation will reduce default rates, which occurred, on average, after 56 days by families with all HIV-positive children but at 70 days for those who finally defaulted. Similarly, adapting CTC routine antibiotic treatment to the epidemiology of HIV-associated infections and inclusion of routine prophylactic cotrimoxazole for HIV-positive children, as currently recommended by WHO, may improve recovery in this group 148.

The low relapse rates following CTC is in contrast to NRU and other outpatient SAM treatment programmes where morbidity and mortality after discharge are high 495051. Although survival bias cannot be ruled out as an explanation for this finding, it is possible that the CTC design, which uses community mobilization and referral for early identification and treatment of SAM also improves long-term recovery compared to hospital-based treatment programmes. SAM is a progressive condition and prognosis is directly associated with the lead time to presentation and treatment. Initiating nutritional intervention as soon as SAM presents may be especially important for HIV-infected and exposed children.

The high VCT uptake for adults and children, the low HIV-prevalence amongst SAM children, and the low nutritional relapse rate amongst surviving HIV-positive and uninfected children over a year after discharge are all noteworthy in our study. The high VCT uptake is comparable to that observed recently in some NRUs in Malawi 16 but contrasts with anecdotal reports of reluctance to come forward for HIV testing offered by clinics and therapeutic feeding programmes. We believe that the "opt-out" approach used in this study, with HIV testing offered to everyone with the right to refuse rather than the standard "opt-in" approach where people have to specifically request HIV testing, contributed to the high uptake seen here 115253. We also believe that offering testing through a programme such as CTC that is well established in the community improves trust and reduces the fear of stigmatization. Caregivers were informed about the opportunity for HIV testing one week prior to travelling to the health centre and no substantial compensation was offered, ruling out the likelihood of coercion in the RC. The high uptake observed here suggests that CTC is a potentially innovative way to increase access to and coverage of HIV testing, particularly in rural areas 54. It is important to note however that CTC would have to be combined with other community based HIV testing and counselling approaches like the home based and mobile VCT in order to obtain good coverage; in prior studies we observed that only 16% of HIV-affected households has a malnourished child treated in the CTC programme in the past 18 months 95556.

There are a number of factors that are likely to have contributed to the low prevalence of HIV amongst severely malnourished children in this study. Chronic food insecurity, frequent common childhood illnesses, poor access to modern health care and suboptimal complementary feeding practices all cause SAM in the absence of HIV in Malawi 575859. The decentralised nature and the high coverage rates obtained by CTC programmes means that a higher proportion of admissions live in remote rural areas far from towns and main roads 60 in contrast to admissions in more centrally located urban hospital units. People from rural areas are all subsistence farmers, have poor income and low educational level and have no possibility for travelling within or outside the country. These factors are known to increase the prevalence of malnutrition and lower that of HIV 5761626364. Lastly, the low HIV prevalence might also be explained by the high mortality of HIV infected infants biasing our data to include those children who have survive infancy. Without effective treatment, it is estimated that over 50% of infants who acquired HIV infection through mother-to-child transmission will die by two years of age (compared to 8% of uninfected children) while in Malawi kwashiorkor, the most common form of SAM in children, peaks between 18 to 23 months of age 65666768. The relatively older average ages of children in both the PC (26.4 months) and RC cohorts (47.2 months), suggests the possibility of survival bias.

Our estimated adult HIV prevalence of 5.0% is predictably lower than the antenatal HIV prevalence rate in the central region (9.8%) but similar to comparable adult prevalence rates of 6.4% and 4.1% for the region and for an adjacent district, respectively, as reported in the 2004 Malawi Demographic and Health Survey 6970. Our study suggests that targeting adult caretakers of malnourished children for HIV testing is feasible but additional outreach and counseling efforts may be needed to increase uptake.

Our analysis confirms that clinical algorithms designed to diagnose paediatric HIV are neither sufficiently sensitive nor specific in severely malnourished children and that blood tests are therefore required to confirm the diagnosis 71. In context where blood tests are not available, the combination of MUAC > 110 mm and absence of proxy indicators can be used to rule out the presence of HIV. These family history variables could be incorporated into future CTC protocols for confirmed or suspected paediatric HIV in setting without possibility of blood tests 72.

SAM is one of several criteria for initiating ART in HIV-positive children 73. The fact that some previously undiagnosed HIV-positive children recovered from malnutrition, and were still healthy and asymptomatic an average of 15 months after discharge from CTC without ART, suggests that the presence of malnutrition should not be the sole criteria for initiating ART in food insecure settings. One possibility is that initiation of ART could be reserved for children who do not respond to CTC or at least could be delayed until nutrition improvement to minimize antiretroviral side effects. Several studies have reported that HIV-infected children tend to develop marasmus rather than kwashiorkor and that CD4 count remains higher in HIV-infected children with kwashiorkor compared to those with marasmus 177475. Therefore, delayed initiation of ART could be considered for children with kwashiorkor. This approach could help to prevent unnecessary exposure to ARV drugs that have side effects and toxicity and to reduce the risk of developing resistance 76. Further research, probably in the form of randomized trials, is urgently needed to strengthen this evidence base before any change of practice is recommended.

Several limitations of our study deserve mention. This study was carried out in conjunction with an ongoing CTC programme and clinical records were reviewed in order to obtain data on nutritional recovery. Although programme procedures were standardized, we were unable to verify nutritional measurements for accuracy. As noted previously, the RC may be subject to survival bias, and therefore we have limited our use and interpretation of the RC data. The statistical power of these analyses is also limited by the small number of HIV-positive children in the study and by further reduction of the sample size due to missing data for age and nutritional status. Nevertheless, the data from both cohorts paint a consistent picture with regard to the potential positive impact of Community-based Therapeutic Care for managing SAM in HIV-positive and uninfected children in rural Africa.

The results of the present study demonstrate that CTC is a valuable entry point for HIV testing for severely malnourished children and that good recovery rates can be achieved in HIV-infected severely malnourished children admitted to the program. These results indicate that CTC can be used to improve the coverage of HIV services, especially in rural areas. The approach has several important advantages over traditional inpatient therapeutic care, including earlier intervention, greater coverage, and increased accessibility. All of these characteristics are particularly important for providing timely care to HIV-exposed and vulnerable children. Additional research on feeding protocols for HIV-infected children and on timing of ART initiation are needed to refine CTC protocols in settings where HIV is common.

The authors Paluku Bahwere, Kate Sadler, Saul Guerrero and Steve Collins work for Valid International, an organisation that has been engaged in the research and development of Community-based Therapeutic Care. Dr Steve Collins is also unpaid director of Valid Nutrition, a not-for-profit company established to research and manufacture ready-to-use therapeutic food in developing countries. There is no conflicting interest for all the other authors.

PB: concept and design of the research, data collection, data analysis and interpretation, drafting of the paper. EP: concept and design of the research, critical revision of paper content. MCJ: data collection, analysis of data and revision of the manusdcript. KS: concept and design of the research, data analysis and interpretation, revision of the draft of the paper. CHGT: concept and design of the research, critical revision of paper content. SG: design and concept of the research, data collection and revision of the manuscript. SC: concept and design of the research, data analysis and interpretation, revision of the draft of the paper. All authors read and approved the final manuscript.