Deworming and micronutrient status by community open defecation prevalence: An observational study using nationally representative data from India, 2016–2018

Background Micronutrient deficiencies are widespread in India. Soil-transmitted helminth (STH) infections are acquired by interaction with soil and water contaminated by human feces and lead to blood loss and poor micronutrient absorption. The current recommendation for control of STH-related morbidity is targeted deworming, yet little is known about the effectiveness of deworming on micronutrient status in varying sanitation contexts. Ranging between 1% and 40% prevalence across Indian states, open defecation (OD) remains high despite India’s investments at elimination by promoting community-wide sanitation. This variation provides an opportunity to study the relationship between deworming, micronutrient status, and OD at-scale. Methods and findings Cross-sectional datasets that were representative for India were obtained the Comprehensive National Nutrition Survey in 2016 to 2018 (n = 105,060 individuals aged 1 to 19 years). Consumption of deworming medication was described by age and community OD level. Logistic regression models were used to examine the relationship between deworming, cluster OD, and their interactions, with anemia and micronutrient deficiencies (iron, zinc, vitamin A, folate, and vitamin B12), controlling for age, sex, wealth, diet, and seasonality. These regression models further allowed us to identify a minimum OD rate after which deworming becomes ineffective. In sensitivity analyses, the association between deworming and deficiencies were tested in subsamples of communities classified into 3 OD levels based on statistical tertiles: OD free (0% of households in the community practicing OD), moderate OD (>0% and <30%), or high OD (at least 30%). Average deworming coverage and OD prevalence in the sample were 43.4% [IQR 26.0, 59.0] and 19.1% [IQR 0, 28.5], respectively. Controlling for other determinants of nutritional status, adolescents living in communities with higher OD levels had lower coverage of deworming and higher prevalence of anemia, zinc, vitamin A, and B12 deficiencies. Compared to those who were not dewormed, dewormed children and adolescents had lower odds of anemia (adjusted odds ratio 0.72, (95% CI [0.67, 0.78], p < 0.001) and deficiencies of iron 0.78, (95% CI [0.74, 0.82], p < 0.001) and folate 0.69, (95% CI [0.64,0.74], p<0.001)) in OD free communities. These protective effects remained significant for anemia but diminished for other micronutrient deficiencies in communities with moderate or high OD. Analysis of community OD indicated a threshold range of 30% to 60%, above which targeted deworming was no longer significantly associated with lower anemia, iron, and folate deficiency. The primary limitations of the study included potential for omitted variables bias and inability to capture longitudinal effects. Conclusions Moderate to high rates of OD significantly modify the association between deworming and micronutrient status in India. Public health policy could involve sequencing interventions, with focus on improving deworming coverage in communities that have achieved minimum thresholds of OD and re- triggering sanitation interventions in high OD communities prior to deworming days, ensuring high coverage for both. The efficacy of micronutrient supplementation as a complementary strategy to improve nutritional outcomes alongside deworming and OD elimination in this age group needs further study.

I was asked for a methodological report and I interpret that to include all aspects of the design and conduct of the study.The version I received did not have page numbers.My numbering has the abstract starting on page 1

Points of detail
Page 4 Is junk food a well-defined category of food?
Page 4 Does that mean that December to February is treated as the reference category?
Page 5 Since categorising an essentially continuous variable wastes information (Altman and Royston, 2006;Royston et al., 2006) and leads to models which are often implausible as they predict the effect remaining flat within categories and then jumping to a new value at the category boundary I wonder why the authors only present the model with continuous OD as a sensitivity analysis.To my mind it is clearly superior and has the additional benefit of detecting possible thresholds.
Page 5, Table 1 I do not understand what the meaning of averaging the wealth categories is.Strictly speaking the (four) quintiles are values on the principal component but it looks as though this is just the mean of {1, 2, 3, 4, 5} Page 7 and Figure 2 If the authors are relying on the confidence intervals to support their statement about vitamin A then I think more care is needed.Just because the confidence intervals for for the three scenarios do not overlap does not mean that they are not different in the statistical significance sense of different (Sedgwick, 2012).Vitamin A is monotonic with OD exposure as well.
Page 8 There seems to me to be more going on in Figure 3 than the text describes.The results for folate seem quite surprising as they indicate that in locations of high OD deworming is harmful.
Page 8 I do not use Stata so citing the Margins command is not very helpful to m or to other non-Stata people.Is it possible to describe what it does in words?This is not the case for some of the others.
Page 9 Does the phrase in moderate and high OD contexts' imply that there was an effect in the low OD context?

Figure S1
This is rather more dramatic than it should be since the y-axis zero has been suppressed.It is still fairly dramatic though as it reveals a halving of deworming over the range of OD.

Points of more substance Interaction
The authors have used separate regressions and have relied on eye-balling differences between them (pages 5, 7, and 9 are examples).I wonder why they have gone to such lengths to avoid fitting a model with interaction term(s).In fact the word interaction when it does occur in the text is used in other contexts.An interaction would enable us to see the change in (log) odds associated with a unit change in OD.

Who uses deworming?
The quite strong relationship between OD and deworming seems to me quite important.This is shown in Figure S1.This implies to me that the children who receive deworming are different in different areas.There must be a variety of social, economic and cultural reasons why people defecate in the open which could, in principle, be associated with propensity to deworm your child.How does this affect the public health message which the authors wish us to draw?

Summary
A few points for clarification and a couple of things which puzzle me.
Michael Dewey