Follow-up of post-discharge growth and mortality after treatment for SAM in Malawi

Location:  Malawi

What we know: There are limited data on long term outcomes following discharge from SAM treatment; what exists is largely pre-HIV. Current CMAM programmes rely on simple and short-term follow up programme indicators to measure treatment success.

What this article adds: A prospective cohort study in Malawi followed up mortality and growth outcomes in 1,024 patients over one year after their treatment for SAM in an inpatient (pre-CMAM) setting. Overall mortality was found to be high in both the short and medium term. Forty two per cent of admitted children died; 25% of these deaths occurred more than 90 days post admission. Of successful discharges, 77% were alive 1 year later. Excess mortality was associated with HIV infection; underlying disability (6%) also contributed. Long term survivors showed good weight-for height catch-up growth, some weight-for-age catch-up but minimal height-for-age catch-up, compared to sibling controls. The findings endorse the CMAM model of care.  Further investigation of longer term, non-mortality outcomes of SAM and the impact of early SAM treatment on stunting are needed. 

There are limited data on long term outcomes following discharge from severe acute malnutrition (SAM) treatment. The few studies that have looked at this are mostly old, pre- HIV and even pre- current definitions of SAM. This represents an important knowledge gap. The primary aim in a recently published study was to follow up patients over a year after their treatment for SAM to describe the longer term mortality and growth outcomes. A secondary aim was to identify key risk factors for good/poor long term outcomes, particularly those which might be amenable to change or which would help identify high risk individuals for more targeted future support.

This was a prospective cohort study, arising from a randomised controlled trial - the ‘ProNut study’ - which had enrolled malnourished children in a trial examining the efficacy of a prebiotic/probiotic mixture added to ready-to-use therapeutic food (RUTF)during SAM treatment.

Method

The study setting was the MOYO nutrition ward, Queen Elizabeth Hospital, Malawi. This serves a catchment population of both urban and rural Blantyre district. At the time, there were no community programmes for SAM active in the district; this meant no community mobilisation, no proactive identification of SAM cases and no sub-classification into ‘complicated’ and ‘uncomplicated’ SAM. Children diagnosed with SAM (either at local health centres or in the paediatric admissions area of Queen’s hospital) were all referred for inpatient care. Whilst anti-retroviral (ARV) treatment ARVs were always available (at Queen’s Hospital since 2004), guidance at the time was for children newly diagnosed with HIV to start ARVs if meeting stage 3 or 4 (WHO classification) which included failure to respond to nutritional therapy. As a result, many would not have been considered for ARVs until 1–2 months after presenting with SAM, and those with good nutritional recovery were not eligible unless another staging criterion was met. From October 2006, the research team were able to obtain CD4 counts which also informed the decision to start ARVs. Co-trimoxazole prophylaxis was provided from admission and advised to be continued indefinitely.

Cohort subjects were children admitted to MOYO nutrition ward for inpatient care from the 12th July 2006 to 9th March 2007 (ProNut recruitment period), which included both dry (July to November) and rainy seasons (December to March). Whilst not every carer consented to their child’s enrolment into ProNUT, almost all answered a detailed baseline questionnaire. Admission criteria to MOYO followed Malawi 2007 national guidelines, which defined SAM as weight-for-height <70% median (NCHS growth references) and/or mid upper arm circumference (MUAC) <110 mm and/or oedematous malnutrition. Because of the inpatient-only set-up with children self-presenting to care, it is likely that most would have been towards the ‘complicated’ end of the SAM spectrum. However, since the patient assessment at the time did not include an appetite test, an unknown (but likely small) proportion would have had ‘uncomplicated’ SAM - which in today’s treatment programmes would be treated at home.

Children who were successfully cured (defined as two consecutive visits at above 80% weight-for-height, no oedema, clinically stable) in MOYO’s combined inpatient/outpatient programme were asked to return for a ward-based review on the 1 year anniversary of their cure date. Non-cures and those failing to return for assessment were followed-up at their home by a mobile team who were usually able to determine outcome information at first field visit.  

The main study outcome was to describe longer term post-treatment survival. This was defined as the child being seen or reliably reported to be alive at least 1 year after the original discharge date from MOYO. Deaths were subdivided into short term (whilst still an inpatient on MOYO); medium term (within the first 90 days of admission but after inpatient care); and longer term (defined as death >90 days following programme admission). Ninety days was chosen to be consistent with the upper limits of time normally spent in therapeutic feeding; three weeks of inpatient care and 10 weeks in the ‘outpatient therapeutic programme’ (OTP). We also assessed long term growth as a secondary outcome. Weight-for-height z score (WHZ), weight-for-age z score (WAZ), and height-for-age z-score (HAZ) were all assessed and compared with sibling controls. Sibling controls were identified at home visit and were defined as any child born to the same mother and living in the same household. We measured all siblings present at the time of follow-up visit but due to limited resources, were not able to return for a second visit if any live siblings were away from the household at the original visit. Additional variables were assessed at original admission to MOYO and are available with complete data collection sheets. They were clinical history, clinical signs assessed by a doctor or study clinical officer, family and socioeconomic status, HIV status (positive/negative/unknown) and HIV clinical staging according to WHO criteria.

Results

Over the eight month study period, a total of 1,024 children contributed to 1,187 malnutrition admission episodes. Final outcome information was found on 889/1024 (87%). Of the 135 (13%) with no 1-year outcome, 45/135 (33%) could not be traced at the address given; 42 (31%) did not give an address; 31 (23%) had missing notes; 7 (5%) lived too far away for the outreach team to visit and 10 (7%) had an outcome but at less than a year.

A total of 462/1024 (45%) children were known to be still alive at a year or more after discharge from treatment. Long term survival was greatest amongst those who had been successfully cured following initial treatment. Of 471 discharged from the OTP as nutritionally cured, 365 (77%) were still alive at a year or more after their first admission. Of the 427/1024 (42%) who died, most did so early on in programme; 238/427 (56%) of deaths were during initial inpatient treatment. Most deaths, 274/427 (64%), were among children with underlying HIV infection. Differently expressed, 274/445 (62%) of known seropositive children died while a total of 77/459 (17%) of known seronegative children died.

Because admissions at the time were based on NCHS growth references, NCHS z-scores were used for subsequent calculations. To enable future researchers to compare their data, the study authors also present admission anthropometry calculated using WHO growth standards. Key observations are that: children who died were younger than those still alive at 1 year; kwashiorkor (oedematous malnutrition) was the dominant type of SAM but risk of death was lower than in non-oedematous SAM; children who were more malnourished (lower z-scores) at baseline were more likely to die; HIV was prevalent and HIV-associated mortality was high, especially with advanced disease (as indicated by low CD4 count); disability - mostly neurodisability such as cerebral palsy - was common and deaths among disabled children were high. Another observation was the large number of children who, according to their personal health record (health ‘‘passport’’) or carer, were seen in an outpatient clinic in the six months prior to SAM admission. In addition, many of these were seen for symptoms potentially consistent with SAM (e.g. ‘‘swelling’’; ‘‘severely thin’’ or similar noted in health passport).

Risk factors for mortality were further explored using multivariable Cox Regression. Adjusting for age, oedema and HIV, low baseline MUAC, WHZ and WAZ are all strongly associated with death. Low HAZ is of borderline significance. Children with oedematous malnutrition were significantly less likely to die than those without. Though the youngest children below 12 months of age were significantly more likely to die, there was no clear age-related risk gradient among older children. HIV stands out as having the greatest adverse impact, an adjusted hazard ratio of 4.03(95% CI 3.08 to 5.25). Deaths among children whose HIV sero-status was unknown were especially high. Analysis also looked at clinical and social factors which might plausibly explain mortality. What is notable here is that after adjusting for baseline anthropometry, very few factors remain as independent predictors of death. Disability stands out as strongly and significantly associated with death in HIV negative patients, and only just outside the p< 0.05 threshold for HIV positive patients. Severe anaemia (PCV <10) was also associated with increased risk of death, but in HIV positive patients only.

Finally, the research looked at the final anthropometry of the children who had received treatment on the MOYO ward, compared to sibling controls. Median birth order of the MOYO children was second born. Over 90% of siblings were reported as never having had SAM themselves. Several points are of note. First is the complete catch-up of WHZ. From an initial programme discharge mean WHZ of -1.96 (SD 1.5), there was a 1.92 (95% CI 1.76 to 2.08) z-score catch-up, bringing the surviving group to a WHZ of 0.04 (SD1.0), comparable to sibling controls. WAZ also improved by 1.66 (95% CI 1.50 to 1.82) z-scores to a ‘long-term’ final value of -1.77 (SD1.1). This was, however, significantly below that of sibling controls; mean difference in z-scores was 0.55 (95% CI -0.71 to -0.38, p<0.01). HAZ also improved, albeit by only 0.37 (95% CI 0.21 to 0.53, p<0.01) z-scores, to a ‘long-term’ final value of -2.97 (SD1.3). This was again significantly lower than that of the sibling controls: mean difference -1.13 (95% CI -1.34 to -0.93, p<0.01). Median age of all the siblings who were measured was 61 months and median age of the ex-MOYO children who had a sibling measured was 41 months.

Discussion

In this SAM treatment programme, overall mortality was high in both the short and medium term, markedly above SPHERE targets of <10%. Long term survivors showed good weight-for height catch-up growth, some weight-for-age catch-up but minimal height-for-age catch-up, compared to sibling controls. These observations are rare in the SAM literature and therefore offer key lessons for policy makers and programme managers.

First is the need for more information on longer term outcomes so as to assess the true impact of SAM on child mortality. In this particular setting, routine reporting would have given a falsely optimistic impression of post-SAM mortality. Overall 427/1024 (42%) children died, but 105 (25%) of these deaths occurred >90 days after their original admission. They would have been missed by most nutrition programmes which only follow-up patients up to discharge from outpatient care. Other factors which can mislead routine reporting systems but which the study described include defaulter follow-up so as to ascribe correct survival status and correct capture of readmissions (of the 52 readmissions, 17 (33%) died) – if not done, denominator population is overestimated and % mortality thus underestimated.

The second key lesson is that patient clinical profile needs to be better accounted for, both when managing individual patients and when judging programme performance. Simple targets, such as SPHERE, cannot distinguish between poorly performing programmes and those caring for a clinically complex patient population where underlying factors are more significant than food insecurity. HIV infection was associated with much of the excess mortality and likely underestimated; even higher mortality was noted among children whose serostatus was unknown (too sick to test) but who were likely seropositive. Notable in FuSAM is the importance of underlying disability. Underlying disability affected 6% of children and accounted for significant excess mortality - a likely underestimate given that the study had no formal screening tool and therefore identified only the most obvious problems. At present, few programmes actively look for disability so this is a key area for future work.

The third message relates to stunting. Current treatments for SAM may not be sufficient to address stunting on their own. There was limited improvement in survivors’ height-for-age who were significantly stunted compared to sibling controls. This is even more notable given the siblings’ older median age: population prevalence of stunting increases over the first two years of life so that older cohorts might be expected to be more rather than less stunted. Whilst weight-for-age had improved, it also remained low, significantly below siblings’. The marked improvement of weight-for-height z-scores though in one sense welcome, likely reflects rapid weight gain combined with poor linear growth, a pattern which may be associated with later obesity and chronic disease. This needs further study.

FuSAM compares favourably to the few other studies looking at longer term SAM outcomes. Most of these other studies, however, took place pre-HIV.  FuSAM thus fills a key gap regarding HIV-era outcomes, especially in sub Saharan Africa where the problem is greatest.

The authors acknowledge a number of potential limitations of the study. For example, FuSAM was conducted when CMAM was not as widespread as it is today and was not available in Blantyre district where the work took place (though it did start shortly after, in mid-2008). Even though some risk factors for mortality (e.g. HIV, disability) are likely to be found in both inpatient and community settings, care is needed before generalising these findings to CMAM programmes. However, the results are consistent with CMAM’s key principle that earlier treatment is better to maximise survival.  

Resource limitations meant that it was not possible to follow-up regularly at fixed time-points after programme discharge, but instead relied on a single long term FuSAM visit. It is possible that some variables such as socioeconomic status or orphaning may have changed over time and hence have been missed.

Much more awareness of longer term outcomes of SAM is needed, with attention on important non-mortality outcomes such as growth, development, co-morbidities and pre-determinants for eventual adult non-communicable disease. Programmes should work more holistically, in closer collaboration with other child health services and form part of a seamless ‘continuum of care’.  Urgent future work is needed to elaborate exactly which interventions are most effective and most cost-effective for children with disability (and likely other) chronic underlying disease. Finally, the results support calls for, and revitalisation of, early identification and treatment of SAM in CMAM programming. Whilst other confirmatory research is urgently needed, it is very plausible that earlier treatment of SAM would favourably affect stunting and longer term post-SAM chronic disease – both key areas in current international malnutrition policy and programming. 

Footnotes