Re-thinking "non-response" to wasting treatment: Exploratory analysis from 14 studies

Abstract

Children who receive therapeutic feeding for wasting treatment but do not reach the anthropometric definitions of recovery (usually within 12-16 weeks) are categorised as 'non-responders' and considered as treatment failures. We conducted a pooled analysis to explore the growth trajectories of non-responders and the appropriateness of the definition of 'non-response'. We pooled 14 studies of children aged 6-59 months receiving treatment for wasting. We included children classified by their studies as recovered or as non-responders. Observing the pooled data of non-responders' mid-upper arm circumference (MUAC), weight, weight-for-age z-score, weight-for-height z-score and daily weight gain rate, we found that the first quartile differentiated those who did not grow at all versus those that demonstrated some growth. We therefore defined 'low growth non-responders' as < 25th percentile anthropometric gain between admission and exit using the non-responders' pooled study data, and 'high growth non-responders' as ≥ 25th percentile gain. We plotted the growth trajectories of MUAC-, weight- and height-related indices of the recovered, high growth and low growth non-responder groups over time using mixed effects generalised additive models. We compared age, sex and anthropometric characteristics of the three groups and explored predictors of non-response category using a multivariate multinomial logistic regression model. For all outcomes, the high growth non-responders started with a worse anthropometric status compared to those who recovered, but then tracked along a near-parallel growth trajectory. The low growth non-responders showed limited growth throughout treatment. High growth non-responders are better viewed as 'delayed responders' and may need to be kept longer under treatment to recover and reduce the risks from early discharge. Low growth non-responders are the true treatment failures and should be referred for further investigations as quickly as possible. In conclusion, non-responders are not a homogenous group; ~75% of them respond well to treatment and ~25% are treatment failures.

Fig 1. Flow chart.

¹For those children re-categorised, recovered refers to children who had at least one anthropometric criterion above the threshold of the WHO definition of acute malnutrition (MUAC ≥125mm or WHZ ≥−2) at the point of discharge from the nutritional programme, alongside the absence of any other criterion of severe wasting (i.e., MUAC≥115mm and WHZ≥-3 and no oedema), with a minimum length of stay of 12 weeks. ²For those children re-categorised, non-responder refers to children who had both MUAC and WHZ below the threshold of the WHO definition of acute malnutrition (MUAC <125mm and WHZ <−2) at the point of discharge from the nutritional programme, with a minimum length of stay of 12 weeks. ³Defaulters and unknown status with length of stay <12 weeks.

Fig 2. Panel of modelled adjusted weekly means of a) MUAC, b) weight, c) WHZ and d) WAZ over time comparing children classified as recovered with non-responders.

MUAC, mid-upper-arm circumference; WAZ, weight-for-age z-score; WHZ, weight-for-height z-score.

Fig 3. Panel of modelled adjusted weekly means of a) MUAC, b) weight, c) WHZ and d) WAZ over time using 25^th percentile of MUAC gain to define high growth non-responders (MUAC≥25^th percentile; ≥2mm) and low growth non-responders (MUAC <25^th percentile; <2mm).

MUAC, mid-upper-arm circumference; WAZ, weight-for-age z-score; WHZ, weight-for-height z-score.

Fig 4. Panel of modelled adjusted weekly means of a) MUAC, b) weight, c) WHZ and d) WAZ over time using 25^th percentile of MUAC gain to define high growth non-responders (MUAC≥25^th percentile; ≥2mm) and low growth non-responders (MUAC <25^th percentile; <2mm) in children with severe wasting¹at baseline.

MUAC, mid-upper-arm circumference; WAZ, weight-for-age z-score; WHZ, weight-for-height z-score. ¹Severe wasting defined as WHZ <-3 or MUAC <115mm.

Fig 5. Panel of modelled adjusted weekly means of a) MUAC, b) weight, c) WHZ and d) WAZ over time using 25^th percentile of MUAC gain to define high growth non-responders (MUAC≥25^th percentile; ≥2mm) and low growth non-responders (MUAC <25^th percentile; <2mm) in children with moderate wasting¹ at baseline.

MUAC, mid-upper-arm circumference; WAZ, weight-for-age z-score; WHZ, weight-for-height z-score. ¹ Moderate wasting defined as WHZ ≥ -3 and <-2, or MUAC ≥ 115mm and < 125mm.

Fig 6. Multivariate multinomial logistic regression using 25^th percentile of MUAC gain to define high growth non-responders (MUAC ≥25^th percentile;≥2mm) and low growth non-responders (MUAC <25^th percentile; <2mm).

Reference group are the recovered children. Odds Ratio >1 indicate increased odds of being in either the high growth non-responder or low growth non-responder group compared to being recovered. P values show the Wald test result after all variables in model adjusted for. MUAC, mid-upper-arm circumference; WHZ, weight for-height/length z-score; WAZ=weight-for-age z score. Morbidity includes diarrhoea, cough, fever, vomiting and/or rash, malaria at inclusion or during follow-up.

Fig 7. Area under the ROC of the multivariate multinomial logistic regression model corresponding to Fig 6.

Model accuracy = 87.7%. Area under the ROC=74.1%.