Authored by: Dr. Robert Black, Chair, Department of International Health, Johns Hopkins University Link to Bio
Research on the impact of immunizations on stunting rates have suffered from an inability to isolate the impact of immunizations from the impact of a wide variety of other factors which may impact stunting rates including mother’s education, poverty rates etc.
This challenge stems from the fact that it would be unethical to conduct a randomized control trial (RCT) that provides vaccinations to some children and a placebo to others and then measure stunting rates, as the placebos would elevate childhood mortality rates.
Without RCT’s, we can assess the impact of vaccinations only through observation, and this research has thus far been somewhat inconclusive, with some small reduction of stunting detected in certain, but not all, geographies.
The ideal way to answer this question would be for growth in early childhood be an outcome of placebo-controlled vaccine efficacy trials. Such trials are designed to determine if there is an effect on an infectious disease outcome and have not included growth or development of stunting as an outcome. For example, the Cochrane Review of trials of vaccines for the prevention of rotavirus diarrhea, the most important cause of severe acute diarrhea in LMIC, did not include a growth outcome for any of the trials.1 Therefore, it is necessary to address this question using observational data, either from longitudinal surveillance or surveys, while dealing with the limitations of such data for causal inference.
A common type of published analysis is to examine the correlation between vaccine coverage and prevalence of stunting, defined as the child having a length/height-for-age more than two standard deviations lower than the median length/height-for-age of an international reference population. An example is an analysis of data from an Indonesian nutritional surveillance system.2 Partial- or non-receipt of three doses each of oral polio vaccine and DPT vaccine and measles vaccine was associated with increased stunting. However, there are several confounding challenges with this study; non-receipt of immunizations was associated with maternal age, education, income and distance to the health post. These factors may be associated with stunting independent of immunizations, and failure to control for them in the analysis means that there cannot be a causal inference regarding immunizations and stunting. When there are factors that are associated both with the exposure of interest, in this case immunizations, and with the outcome i.e. stunting this is referred to as confounding.
Other analyses have attempted to control for this possibility in various ways, often unsuccessfully, including a survey of several villages in Kenya.3 A logistic regression was used to assess a number of dietary and health service exposures, controlling for age, sex and a socio-economic status (SES), which was based on the agricultural resources of the household. In this analysis, children without up-to-date immunizations were more than twice as likely to be stunted than children with current immunizations. While their measure of SES was not associated with immunization status, other SES factors such as maternal education that were not included in the regression analysis, may have been, resulting in residual confounding.
Still other analyses have attempted to reduce the possible effect of confounding using more advanced statistical methods including an analysis of the effect of measles vaccination on child growth which used data from Demographic and Health Surveys from 65 LMIC.4 For causal inference, the authors used conditional logistic regression models with stunting at 12-59 months of age as the outcome to determine the effect of measles vaccination. The regression models used household and mother fixed effects that allowed comparison across siblings, controlling for observed and non-observed household confounders, and also controlled for a number of other characteristics of the mother and child, including vaccination with DPT vaccine as an indicator of the likelihood of getting any immunizations. Measles vaccination decreased the probability of being stunted by 11.5%. The analysis found a gender interaction with an effect on measles vaccination reducing stunting in boys but not girls.
An analysis of longitudinal survey data from Ethiopia, India and Vietnam also looked for an effect of measles vaccine on stunting.5 Regression analysis used propensity score matching to control for systematic differences between children who were vaccinated for measles or not. Children at 7-8 years of age who had been vaccinated for measles in infancy in India had a slightly higher height-for-age (0.13 z-score) than unvaccinated children; no significant effect was found in Ethiopia or Vietnam. At 11-12 years of age, there were no significant effects on stunting in any of the countries. Using the same survey data for India, the effect of receiving Hemophilus influenzae type b vaccine in early childhood on stunting was analyzed using a similar propensity score matching approach.6 A small (0.17 and 0.22 z-score) positive effect on height-for-age was found for 11-12 and 14-15 year olds.
Finally, another type of analysis has tried to explain the changes in the prevalence of stunting in seven Sub-Saharan African countries, considering the changes in distal and proximate determinants between two DHS surveys between 2005 and 2014.7 In five of the seven countries, the proportion of children fully immunized increased, coinciding with a decline in the prevalence of stunting, but immunization status was not included in the decomposition analysis so there is no statistical basis for its effect on stunting.
Looking at our Exemplar countries, we – again – find mixed data on the role of immunizations in stunting reduction. In each country a multivariable hierarchical (distal, intermediate and proximate levels) model was used to examine the association of a broad set of possible determinants, including DPT3 and measles vaccinations, on mean height-for-age z-score change over time (generally 2000-2016/17). Two countries had substantial increases in coverage of DPT3 and measles vaccine.
In Senegal, DPT3 improvement in coverage of these vaccines was associated with 9% of the explainable change in stunting in the decomposition analysis. In Ethiopia, in spite of substantial improvement in vaccine coverage, there was no meaningful effect on stunting. Two countries, Peru and Nepal, had relatively high vaccine coverage (71/72%) at the beginning of the time period and modest improvements; in neither country did change in vaccine coverage explain any of the change in stunting. In Kyrgyz Republic the initial vaccine coverage was almost universal and declined by a few percentage points, so as would be expected did not have any association with change in stunting prevalence.8
In conclusion, in LMIC immunization coverage with a set of routine childhood vaccines or with measles or Hemophilus influenzae type b vaccines may provide some protection from stunting, but the small effect on height is not found in all settings and may not persist to mid-childhood. The observational nature of the data available, even with the best statistical methods, has limitations for causal inference and results may still be influenced by residual confounding making small effects, even if statistically significant, questionable.
In the absence of rigorous vaccine-efficacy trials that measure childhood stunting or linear growth as an outcome, the current state of evidence is inconclusive.
References:
- 1. Soares-Weiser K, Bergman H, Henschke N, Pitan F, Cunliffe N. Vaccines for preventing rotavirus diarrhoea: vaccines in use. Cochrane Database Syst Rev 2019; 2019(10).
- 2. Semba RD, De Pee S, Berger SG, Martini E, Ricks MO, Bloem MW. Malnutrition and infectious disease morbidity among children missed by the childhood immunization program in Indonesia. Southeast Asian Journal of Tropical Medicine and Public Health 2007; 38(1): 120.
- 3. Bloss E, Wainaina F, Bailey RC. Prevalence and predictors of underweight, stunting, and wasting among children aged 5 and under in western Kenya. J Trop Pediatr 2004; 50(5): 260-70.
- 4. Bogler L, Jantos N, Bärnighausen T, Vollmer S. Estimating the effect of measles vaccination on child growth using 191 DHS from 65 low-and middle-income countries. Vaccine 2019; 37(35): 5073-88.
- 5. Nandi A, Shet A, Behrman JR, Black MM, Bloom DE, Laxminarayan R. Anthropometric, cognitive, and schooling benefits of measles vaccination: Longitudinal cohort analysis in Ethiopia, India, and Vietnam. Vaccine 2019; 37(31): 4336-43.
- 6. Nandi A, Deolalikar AB, Bloom DE, Laxminarayan R. Haemophilus influenzae type b vaccination and anthropometric, cognitive, and schooling outcomes among Indian children. Ann N Y Acad Sci 2019; 1449(1): 70-82.
- 7. Buisman LR, Van de Poel E, O'Donnell O, van Doorslaer EK. What explains the fall in child stunting in Sub-Saharan Africa? SSM-population health 2019; 8: 100384.
- 8. Results provided by N Akseer, Z Bhutta.