The impact of rickets on growth and morbidity during recovery among children with complicated severe acute malnutrition in Kenya: A cohort study

Abstract The effects of rickets on children recovery from severe acute malnutrition (SAM) are unknown. Rickets may affect both growth and susceptibility to infectious diseases. We investigated the associations of clinically diagnosed rickets with life‐threatening events and anthropometric recovery during 1 year following inpatient treatment for complicated SAM. This was a secondary analysis of clinical trial data among non‐human immunodeficiency virus‐infected Kenyan children with complicated SAM (2–59 months) followed for 1 year posthospital discharge (ClinicalTrials.gov ID NCT00934492). The outcomes were mortality, hospital readmissions, and growth during 12 months. The main exposure was clinically diagnosed rickets at baseline. Of 1,778 children recruited, 230 (12.9%, 95% CI [11.4, 14 .6]) had clinical signs of rickets at baseline. Enrolment at an urban site, height‐for‐age and head circumference‐for‐age z scores were associated with rickets. Rickets at study enrolment was associated with increased mortality (adjusted Hazard Ratio [aHR] 1.61, 95% CI [1.14, 2.27]), any readmission (aHR 1.37, 95% CI [1.09, 1.72]), readmission for severe pneumonia (aHR 1.37, 95% CI [1.05, 1.79]), but not readmission with diarrhoea (aHR 1.05, 95% CI [0.73, 1.51]). Rickets was associated with increased height gain (centimetres), adjusted regression coefficient 0.19 (95% CI [0.10, 0.28]), but not changes in head circumference, mid‐upper arm circumference, or weight. Rickets was common among children with SAM at urban sites and associated with increased risks of severe pneumonia and death. Increased height gain may have resulted from vitamin D and calcium treatment. Future work should explore possibility of other concurrent micronutrient deficiencies and optimal treatment of rickets in this high‐risk population.

Children with complicated severe acute malnutrition (SAM) remain susceptible to serious infections during recovery (Berkley et al., 2016), and to our knowledge, no study has evaluated the effect of the presence of rickets on their recovery. In this study, we aimed to investigate the associations of clinically diagnosed rickets at study enrolment with anthropometric recovery and life-threatening events during 1 year following inpatient treatment for complicated SAM.  (Berkley et al., 2016). In the parent trial, there was no overall effect of daily co-trimoxazole prophylaxis on mortality or growth (Berkley et al., 2016). Inpatient management and referral of SAM cases to communitybased therapeutic feeding programmes upon discharge followed the Kenya national recommendations and World Health Organization (WHO) guidelines (WHO, 2005). Children with signs of rickets were treated with vitamin D (150,000 IU STAT intramuscular injection for children younger than 6 months and 300,000 IU for children older than 6 months) and calcium supplements (Purecal/Calcimax® 50-75 mg/kg per day for 3 months). The caregivers were advised to expose the children to sunlight for at least 30 min per day. Children receiving rickets treatment were reviewed after 2 weeks to ensure treatment was continuing, as well as during the planned monthly study followup visits for the first 6 months, then bimonthly up to Month 12. If rickets was unresolved or there was minimal improvement after 3 months, the doses were repeated.

| Study participants
All children admitted in the paediatric wards of the four hospitals during the study period were screened for SAM, and those eligible and consented were enrolled after completing the stabilization phase of complicated SAM inpatient treatment (WHO, 2013). Children were eligible for the trial if they had a negative HIV rapid-antibody test, were aged 2 to 59 months, and were hospitalized with complicated SAM defined as mid-upper arm circumference (MUAC) < 11.5 cm for children 6 to 59 months old or MUAC < 11.0 cm for children 2 to 6 months or oedema at any age (Berkley et al., 2016).

| Outcomes
The main outcomes were episodes of life-threatening events and anthropometry (MUAC, weight, length/height, and head circumference) during the course of the primary trial of 12 months. Life-threatening events were defined as all-cause mortality or all hospital readmissions.
We also examined readmissions with severe pneumonia or diarrhoea separately. Severe pneumonia was defined using WHO 2015 guidelines as presence of cough or breathing difficulty with either central cyanosis/ oxygen saturation < 90% or general danger signs (impaired consciousness/lethargy/inability to breastfeed or drink) or severe respiratory distress (chest indrawing/grunting; WHO, 2005). Diarrhoea was defined as at least three episodes of loose/watery stools in 24 hr requiring

Key messages
• Clinically defined rickets was present among 12.9% children with complicated severe acute malnutrition mostly from urban areas in Kenya.
• Rickets at study enrolment was associated with increased height gain, likely because of the calcium and vitamin D treatment provided.
• Despite treatment, children with rickets at study enrolment remained more susceptible to death and hospital readmission with severe pneumonia compared to children without rickets. hospitalization following WHO 2015 guidelines (WHO, 2005). Anthropometric recovery was defined by changes from the baseline values of absolute MUAC, and z scores for weight-for-height/length (WHZ), weight-for-age (WAZ), height/length-for-age (HAZ), and head circumference (HCZ) calculated using 2006 WHO growth references (WHO, 2011).

| Exposures
The main exposure of interest was clinically defined rickets at baseline, which was systematically collected as present or not on the trial baseline case report form. A clinical diagnosis of rickets was made by trained study clinicians at study enrolment by the presence of swelling of wrists and ankles, bowed legs, rachitic rosary, craniotabes, or features of rickets on wrist X-ray. Other baseline exposures examined were age, gender, randomization arm, recruitment site, weight-forheight, height-for-age, weight-for-age, head circumference-for-age z scores, oedema, and admission diagnosis.

| Data sources/measurements
At the time of recruitment to the trial, child demographics, anthropometry, and admission diagnosis were recorded on structured case report forms. During each scheduled visit, anthropometry was performed, and history of readmission to non-study hospital or treatment as outpatient was documented. Free walk-in clinics for unscheduled visits were provided at all the study hospitals where minor illness was treated as outpatient, and children with serious infections were admitted to study hospitals.
Readmission diagnosis and causes of deaths were documented by trained study clinicians directly from study hospital records. For any readmission to non-study hospitals, admission diagnosis was extracted from the hospital documents.
For community deaths, or those at non-study hospitals where deceased child medical records were not accessible, verbal autopsies were undertaken. Causes of all deaths were assigned by two paediatricians not involved in the trial using participant medical records or verbal autopsies.

| Ethics
The Kenya National Ethical Review Committee (SSC 1562) and Oxford Tropical Research Ethics Committee (reference number 18-09) approved the primary trial and this secondary analysis.

| Statistical methods
Children were stratified as either having diagnosis of clinical rickets or not at baseline. To identify factors associated with rickets at baseline, we used multivariable logistic regression analysis, retaining all the variables examined in the univariable model. For analysis of all-cause mortality, we used single event survival analysis with time at risk defined as the period from the date of enrolment in the study up to the date of death, lost to follow-up, withdrew, or completed study follow-up (Month 12). For readmissions to hospital, we performed multiple event survival analysis, allowing participants to contribute more than one event during the follow-up. Time at risk was period from time of enrolment to the trial to an episode of readmission or death or lost to follow-up or withdrew or completed study follow-ups.
Time to any life-threatening event (all-cause mortality, all readmissions, readmissions with severe pneumonia or diarrhoea) curves were fitted using the Kaplan-Meier method and log-rank test used to compare distribution of events between children with and without rickets at study enrolment. The effect of rickets on risk of the lifethreatening events was estimated using Cox proportional regression analysis and reported as hazard ratios (HR) and respective 95% confidence intervals. We performed univariable analysis with the main exposure (baseline rickets) as the only independent variable to yield the crude HR and added a priori risk factors (baseline age, randomization arm, and gender) plus factors associated with baseline rickets to obtain adjusted HR. To test for effect modification, we compared models with and without interaction terms using likelihood-ratio tests.
To examine effects of rickets on growth, we used generalized estimating equations (GEE) for regression models with exchangeable correlation structure and the respective changes in monthly anthropometry as the dependent variable. We constructed univariable GEE regression models with baseline rickets as the only independent variable to yield the crude regression coefficient. Then, in multivariable GEE regression models, we adjusted for potential confounders: age, gender, urban/rural site, and randomization arm.
For a sensitivity analysis, missing anthropometry from missed visits, or where follow-up was done at home and only MUAC was taken, were imputed using the "linear interpolation" approach and used in the GEE regression analysis. The method linearly interpolates a missing value by using values before and after the missing one (Engels & Diehr, 2003). This way, we only imputed anthropometry for missed follow-up; for example, if a child had anthropometry measurement at Visits 1 and 3, but missed anthropometry at Visit 2, Visits 1 and 3 measurements were used to impute value for Visit 2.
No formal sample size estimation was done for this secondary analysis because data from all the children enrolled in primary trial were analysed. The primary trial enrolled 1,778 children, which was adequate to detect one-third mortality reduction from 15% among the control group with a power of 90% assuming p < .05.
All statistical analysis was performed using STATA version 13.1 (STATACorp, College Station, TX, USA).
Children with rickets were younger than those without rickets: median (interquartile range) age 9 (6 to 13) and 11 (7 to 17) months, respectively (p = .002). Children with signs of rickets had lower MUAC, WHZ, and HAZ than those without but had larger HCZ and a lower prevalence of oedema (Table 1).

| Associations with rickets
Enrolment at an urban site, HAZ and HCZ were associated with rickets (Table 2). Breastfeeding, known tuberculosis at enrolment, and an index admission diagnosis of severe pneumonia were associated with signs of rickets in the univariable analysis; however, these associations attenuated in the multivariable analysis (Table 2).  Figure 1A. After adjusting for potential confounders, rickets at study enrolment was associated with mortality; adjusted HR 1.61 (95% CI [1.14, 2.27]); Table 3. There was no evidence of effect modification by age (p = .56), gender (p = .63), or site (p = .52).

| Rickets and growth
Clinical signs of rickets at study enrolment were associated with increased height growth; adjusted regression coefficient 0.19 (95% CI [0.10, 0.28]); Table 4 and Figure S1. However, signs of rickets had no effect on head circumference, MUAC, and weight growth (Table 4 and Figure S1).
Results were similar in a sensitivity analysis with imputation of missing anthropometric data, where signs of rickets were associated with increased height growth; adjusted regression coefficient 0.19 (95% CI [0.10, 0.28]), but not with head circumference, MUAC, or weight gain (Tables S3 and S4).  Note. CI = confidence interval. a Adjusted for gender, age, co-trimoxazole randomization arm, recruitment site, baseline absolute mid-upper arm circumference, oedema, and height-for-age z scores.

| DISCUSSION
Clinical signs of rickets were common among Kenyan children who were hospitalized with complicated SAM and of prognostic importance for both mortality and for subsequent episodes of severe pneumonia despite treatment with calcium and vitamin D. Rickets occurred predominantly in urban sites and was associated with stunting. An association with increased head circumference is likely to have been due to the inclusion of craniotabes in the diagnostic criteria. However, the clinical diagnosis of rickets in the absence of systematic measurement of vitamin D, calcium, alkaline phosphatase, and their pathways is necessarily subjective. In the parent trial, the individual signs used to identify rickets were not collected; however, the young age of cases (9 months) indicates that the classical presentation of bowed legs in walking-age children is not the predominant syndrome, as was also demonstrated in an urban community setting in Nairobi (Jones et al., 2017).
Clinical signs of rickets are suggestive of calcium and/or vitamin D deficiency and have high sensitivity for these conditions (Thacher, Fischer, & Pettifor, 2002). However, they may also reflect a diet deficient in other micronutrients affecting both growth and immunity.
Evidence from other studies suggest that deficiency of vitamins A, D, and E can occur concurrently and are associated with risk of respiratory tract infections (Zhang et al., 2016). Children with clinical signs of rickets were more stunted and wasted; however, clinical rickets was associated with risk of mortality and serious infections requiring  Note. HAZ = height-for-age z score; HCAZ = head circumference-for-age z score; WAZ = weight-for-age z score, WHZ = weight-for-height z score; MUAC = mid-upper arm circumference. a Adjusted for gender, age, co-trimoxazole randomization arm, and urban/rural hospital. hospitalization even after adjusting for these anthropometrics measurements. Vitamin D plays a major role in modulating both innate and adaptive immunity (Rosen et al., 2016).
We found that linear growth, but not markers of wasting, was improved in children with signs of rickets at baseline, even after adjustment for age. This was largely accounted for by a smaller decline in height/length-for-age z score between baseline and Month 1 (Figure S1). The most likely explanation is that these children were treated with calcium and vitamin D additional to that contained in therapeutic milks and ready-to-use therapeutic food, and this promoted linear growth. Previous studies have reported that vitamin D supplementation increased height growth (Ganmaa et al., 2017). Another potential explanation is regression to the mean; however, the fact that the height/ length-for-age z scores of children with and without rickets do not converge ( Figure S1) strongly suggests otherwise.
Other studies, although not among children hospitalized with complicated SAM, have found evidence of association between vitamin D deficiency and severe pneumonia, gastrointestinal infection, and growth (Ganmaa et al., 2017;Haider et al., 2010;Muhe et al., 1997;Thornton et al., 2013). This partially concurs with our study findings; however, in our population, diarrhoea risk was not associated with clinical defined rickets.
There is inconclusive evidence from prior studies on the efficacy of vitamin D supplementation in reducing all-cause mortality, hospital admissions, or incidence of severe pneumonia and diarrhoea (Das, Singh, Panigrahi, & Naik, 2013;Yakoob et al., 2016). Our results suggest that the treatment for rickets that we gave did not result in reduction in risks of mortality or of readmission to that of the other children in the trial. We did not measure compliance with rickets treatment; however, there is the possibility that the current rickets treatment

| Strengths and limitations
The main strength of the study was its prospective design that attained a 95% follow-up for 1 year and the well documented life-threatening events.
The major limitation was the use of clinical signs to diagnose rickets.
Although clinical signs are sensitive in identification of rickets cases, they have poor specificity, and therefore, we could have missed some cases especially among children with oedema that could blur the clinical signs (Jones et al., 2017;Thacher et al., 2002), that is, many other children may have had low calcium or vitamin D, and this may have introduced bias.
Severe pneumonia and diarrhoea were defined using the WHO clinical signs; although these signs are sensitive, they have low specificity.
However, this reflects the diagnostic criteria commonly used in such poor-resource settings and do not preclude the fact that children diagnosed with clinical signs of rickets had higher risks of life-threatening infections requiring hospital admission or causing death.
This being a secondary data analysis of a clinical trial, active followups and the free walk-in clinics offered could have led to overidentification of serious infections leading to over-reporting of readmissions compared to children in the community. It is also possible, the nonsevere illness treated during the follow-up visits could have prevented them progressing to severe infection requiring readmission. Missing follow-up anthropometry could have introduced bias, but imputing missing data did not change our results significantly. Our results should be interpreted cautiously because it is not generalizable to children with uncomplicated SAM or complicated SAM infected with HIV.

| CONCLUSION
Rickets is common among children with complicated SAM, predominantly from urban sites, and is associated with increased risk of death and hospital admissions with severe pneumonia. Increased height growth may have been due to the calcium and vitamin D treatment provided. Future work should explore the possibility of other micronutrient deficiency and optimal treatment of rickets in this high-risk population.

ACKNOWLEDGMENTS
The authors wish to thank the participants and staff of co-trimoxazole prophylaxis trial. We thank the Wellcome Trust for personal fellowship awarded to J. A. B. The article is published with the permission of the Director of the Kenya Medical Research Institute.

CONFLICTS OF INTEREST
The authors declare that they have no conflicts of interest.

SUPPORTING INFORMATION
Additional Supporting Information may be found online in the supporting information tab for this article.