Advertisement
  • Loading metrics

Distributional change of women’s adult height in low- and middle-income countries over the past half century: An observational study using cross-sectional survey data

  • Jewel Gausman,

    Roles Conceptualization, Methodology, Visualization, Writing – original draft, Writing – review & editing

    Affiliations Department of Social and Behavioral Sciences, Harvard T. H. Chan School of Public Health, Boston, Massachusetts, United States of America, Women and Health Initiative, Department of Global Health and Population, Harvard T. H. Chan School of Public Health, Boston, Massachusetts, United States of America

  • Ivan Meija Guevara,

    Roles Conceptualization, Data curation, Formal analysis, Methodology, Writing – review & editing

    Affiliations Department of Biology, Stanford University, Stanford, California, United States of America, Stanford Center for Population Health Sciences, Stanford University School of Medicine, Stanford, California, United States of America, Department of Demography, University of California at Berkeley, Berkeley, California, United States of America

  • S. V. Subramanian,

    Roles Conceptualization, Methodology, Supervision, Writing – review & editing

    Affiliations Department of Social and Behavioral Sciences, Harvard T. H. Chan School of Public Health, Boston, Massachusetts, United States of America, Harvard Center for Population and Development Studies, Harvard University, Cambridge, Massachusetts, United States of America

  • Fahad Razak

    Roles Conceptualization, Methodology, Supervision, Writing – review & editing

    fahad.razak@mail.utoronto.ca

    Affiliations St. Michael’s Hospital, Toronto, Ontario, Canada, Institute of Health Policy, Management and Evaluation, University of Toronto, Toronto, Ontario, Canada, Li Ka Shing Knowledge Institute, University of Toronto, Toronto, Ontario, Canada

Distributional change of women’s adult height in low- and middle-income countries over the past half century: An observational study using cross-sectional survey data

  • Jewel Gausman, 
  • Ivan Meija Guevara, 
  • S. V. Subramanian, 
  • Fahad Razak
PLOS
x

Correction

4 Jun 2018: Gausman J, Mejía-Guevara I, Subramanian SV, Razak F (2018) Correction: Distributional change of women’s adult height in low- and middle-income countries over the past half century: An observational study using cross-sectional survey data. PLOS Medicine 15(6): e1002588. https://doi.org/10.1371/journal.pmed.1002588 View correction

Abstract

Background

Adult height reflects childhood circumstances and is associated with health, longevity, and maternal–fetal outcomes. Mean height is an important population metric, and declines in height have occurred in several low- and middle-income countries, especially in Africa, over the last several decades. This study examines changes at the population level in the distribution of height over time across a broad range of low- and middle-income countries during the past half century.

Methods and findings

The study population comprised 1,122,845 women aged 25–49 years from 59 countries with women’s height measures available from four 10-year birth cohorts from 1950 to 1989 using data from the Demographic and Health Surveys (DHS) collected between 1993 and 2013. Multilevel regression models were used to examine the association between (1) mean height and standard deviation (SD) of height (a population-level measure of inequality) and (2) median height and the 5th and 95th percentiles of height. Mean-difference plots were used to conduct a graphical analysis of shifts in the distribution within countries over time. Overall, 26 countries experienced a significant increase, 26 experienced no significant change, and 7 experienced a significant decline in mean height between the first and last birth cohorts. Rwanda experienced the greatest loss in height (−1.4 cm, 95% CI: −1.84 cm, −0.96 cm) while Colombia experienced the greatest gain in height (2.6 cm, 95% CI: 2.36 cm, 2.84 cm). Between 1950 and 1989, 24 out of 59 countries experienced a significant change in the SD of women’s height, with increased SD in 7 countries—all of which are located in sub-Saharan Africa. The distribution of women’s height has not stayed constant across successive birth cohorts, and regression models suggest there is no evidence of a significant relationship between mean height and the SD of height (β = 0.015 cm, 95% CI: −0.032 cm, 0.061 cm), while there is evidence for a positive association between median height and the 5th percentile (β = 0.915 cm, 95% CI: 0.820 cm, 1.002 cm) and 95th percentile (β = 0.995 cm, 95% CI: 0.925 cm, 1.066 cm) of height. Benin experienced the largest relative expansion in the distribution of height. In Benin, the ratio of variance between the latest and earliest cohort is estimated as 1.5 (95% CI: 1.4, 1.6), while Lesotho and Uganda experienced the greatest relative contraction of the distribution, with the ratio of variance between the latest and earliest cohort estimated as 0.8 (95% CI: 0.7, 0.9) in both countries. Limitations of the study include the representativeness of DHS surveys over time, age-related height loss, and consistency in the measurement of height between surveys.

Conclusions

The findings of this study indicate that the population-level distribution of women’s height does not stay constant in relation to mean changes. Because using mean height as a summary population measure does not capture broader distributional changes, overreliance on the mean may lead investigators to underestimate disparities in the distribution of environmental and nutritional determinants of health.

Author summary

Why was this study done?

  • An adult’s height may provide important insight into childhood health circumstances.
  • Height is also related to longevity, mortality from a variety of noncommunicable diseases, maternal mortality, and quality of life.
  • Mean height has decreased in some low- and middle-income countries and increased in others, but it is unclear if everyone in the population experiences similar changes in height.
  • This paper examines the relationship between changes in the mean of women’s height and the distribution of women’s height across a population.

What did the researchers do and find?

  • Data from 59 low- and middle-income countries (n = 1,122,845 women, aged 25–49 years) was obtained from the Demographic and Health Surveys, nationally representative surveys conducted at regular intervals around the world.
  • Mean-difference plots were used to visually examine changes in mean women’s height in relation to the distribution of height across countries.
  • Statistical analyses were conducted to examine the global associations between mean height and the standard deviation of height, as well as median height and the 95th and 5th percentile.
  • Seven countries experienced a decline in mean height (all located in sub-Saharan Africa), and 25 countries experienced a change in the standard deviation of height.
  • There was an inconsistent relationship between mean height and its standard deviation.

What do these findings mean?

  • Changes observed in the distribution of height indicate that transitions in living standards and nutrition have not been experienced equally across many populations. A focus on mean women’s height does not capture these changes.
  • Future research on height as a population measure of well-being should include assessment of changes in the distribution as well as the mean.

Introduction

An individual’s maximum height is both heritable and heavily influenced by childhood environmental exposures [1,2]. Genetics play a limited role in the marginal changes observed in the heights of populations over time, while environmental factors, such as illness and nutritional deprivation, are thought to be the primary determinants [38]. Adverse circumstances during periods of rapid growth, such as those occurring in utero [9,10] and during childhood and adolescence [1115], have been associated with decreased adult height. Height is also associated with future health and well-being. Studies have found that increased height is negatively associated with mortality from a variety of noncommunicable diseases such as cardiac disease, stroke, and some types of cancer [1621]; adiposity and type 2 diabetes [22]; suicide [23]; and all-cause mortality [24]. Additionally, health-related quality of life [25] and age-related declines in cognitive function [26] are associated with adult height. In India, shorter maternal height is associated with increased child mortality, stunting, and wasting [27]. Short maternal stature has also been associated with increased perinatal mortality [2830], cesarean delivery [31], and maternal morbidity and mortality [32,33].

Given the sensitivity of height to nutritional and disease exposures, examining changes in adult height over time may provide insight into the underlying health status of a population. Two recent studies have found that mean height is declining in several countries in sub-Saharan Africa (SSA), while it has generally increased across the European, Eastern Mediterranean, South-East Asian, and the Western Pacific regions [3436]. Another study found that mean height has stagnated among men in some countries in Central Asia, the Middle East, and North Africa, while women have continued to grow taller [36]. In addition, mean height has served as a crucial measure of overall human health in a large body of economic work, for example Robert Fogel’s research on economic history and Angus Deaton’s research on consumption, poverty, and welfare [3,7,8].

Global trends in height [3,3441], and many population-level health metrics [4245], are frequently examined by analyzing changes in the mean without considering other properties, such as variance. Additionally, progress towards achievement of the Millennium Development Goals in nutrition is often discussed in terms of changes in mean height within populations [46,47]. The use of mean levels makes an important assumption that as the average level in the population changes, the dispersion around the mean remains constant, making the mean a good proxy for population-level changes [2]. Recent research has found that this assumption does not apply to body weight, where rising mean BMI is accompanied by a widening of the distribution, with gains concentrated among the highest percentiles of the BMI distribution and relatively little change among individuals who are underweight [48,49]. This has direct equity implications for use of the mean to measure population health, since underweight individuals would be expected to benefit from weight gain, whereas overweight individuals would benefit from weight loss.

In this paper, we examine whether relying on the mean as a summary measure of height may obscure significant changes in the distribution of height over time both within and across populations. We describe changes in mean height across birth cohorts, and explore whether there is an association between change in mean height and its distribution, by systematically examining the mean and standard deviation (SD) of women’s height both within and across a broad range of low- and middle-income countries (LMICs) over the past half century. We also examine the change observed in the left and right tails of the height distribution to understand distributional change over time.

Methods

We include the prospective analysis plan used to guide our analysis as S1 Analysis Plan. This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (S1 STROBE Checklist) [50]. The study is based on an anonymous, publicly available dataset with no personally identifiable information. Ethical approval for the Demographic and Health Surveys (DHS) was provided centrally by the ORC Macro Institutional Review Board and by individual review boards within each participating country.

Data sources

We used data from DHS surveys, which are nationally representative, cross-sectional household surveys that have been conducted regularly in more than 85 countries since 1984 [51,52]. DHS surveys are comparable over time and across countries, as DHS employs standardized interviewer training, measurement tools, data collection techniques, and questionnaires [53]. A multistage stratified cluster design that samples households within a selection of primary sampling units ensures that sampled households are nationally representative [54]. The final sample size typically ranges from 5,000 to 30,000 households [52].

The primary objective of DHS is to collect data on basic demographic and health indicators [54]. Anthropometric measurement for a specified subsample of women aged 15–49 years was initially included in DHS surveys in the early 1990s; thus, some surveys include women’s height for women born as early as 1940 [55]. In order to ensure accuracy in height measurements, height is measured by 2 well-trained individuals who have completed a standardized training process, and is measured to the nearest millimeter using a measuring board with a headpiece [56]. The method used for measuring height in DHS surveys was first introduced in 1987 and has remained consistent since its introduction [57]. Additionally, anthropometric data reported by DHS undergo regular quality assessment to ensure both their validity and reliability over time [58].

Study population and sample

The study population comprises 1,122,845 women aged 25–49 years from 59 countries with women’s height measures available from survey waves spanning from 1993 to 2013. Women aged less than 25 years were excluded from the sample in order to ensure that full height attainment had been achieved in order to limit potential bias [35]. Women with missing height values (n = 18,802), those excluded from the one-third or 50% subsample of women selected for anthropometric assessment (n = 246,276), and those who had implausible values recorded for height (less than 100 cm or greater than 200 cm) (n = 714) were also excluded from the study [35].

Outcome and exposure

Height (in centimeters) was specified as a continuous outcome. Using methods described previously, the SD and the 5th and 95th percentiles were used as outcomes in statistical modeling because they provide information on how the distribution of height changes between and within countries across birth cohorts [48,49].

The primary exposure of interest was birth cohort. Women in each country were divided into 4 birth cohorts: 1950–1959, 1960–1969, 1970–1979, and 1980–1989. Several steps were taken to limit bias within birth cohorts. While we had initially planned to include women born before 1950 from all countries in our analysis, we made 2 data-driven changes from our original analysis plan after assembling the data. As so few women in the sample were born before 1950 (n = 3,194 across all 59 countries), we excluded them from the analysis because the sample sizes of birth cohorts born before 1950 were too small within each country to offer meaningful comparisons. Similarly, birth cohorts that contained fewer than 500 women were also excluded from the analysis because they were found to be too small for robust analyses of distributional changes and introduced instability into the models, especially around the tails (5th and 95th percentile) of the distribution (n = 5,641).

Analysis

For global models, we pooled data across birth cohorts and countries. A second dataset was also created that contained the percentile values for the height distribution in each country for each birth cohort. Stata version 14 [59] was used to conduct the statistical analysis. Graphics were produced using the statistical package R [60].

Modeling of distributional parameters

Multilevel regression was used to account for the nested structure of the data, as the outcome measures of interest (SD, 5th percentile, and 95th percentile) are nested within cohort i at level 1 which is nested in country j at level 2.

For example, the general statistical model at level 1 for the SD of height is specified as where Yij is the SD of height in cohort i in country j, β0j is the overall SD of height in each country, β1 is the slope of the relationship between mean height and SD of height, heightij is the mean height in cohort i in country j, and eij is the residual for cohort i in country j.

The model at level 2 is represented as where β0 is the grand mean SD of height across all countries and u0j is the residual of the mean SD of height for country j from the overall SD of height. Both models are combined to obtain Using data from all countries and cohorts, multilevel regression models were constructed to examine the association between (1) mean height and SD of height and (2) median height and the 5th and 95th percentiles of height.

Additional tests were conducted to determine whether country-level changes in the SD were significant across birth cohorts. The ratio of the variance in height between the latest cohort versus the earliest cohort was calculated, and 95% CIs were estimated. For each country, an F-test for equality of variance was performed to test the change in variance (SD2) between the earliest and most recent birth cohort [61]. Additionally, a t test was performed to formally evaluate whether the change in mean was significant between these same cohorts in each country.

Graphical analyses

As quantitative approaches to examining changes in distributions remain limited, we also employed a graphical approach. Mean-difference (MD) plots enable a visual comparison of 2 distributions by plotting the differences between corresponding quantiles from the 2 distributions on the y-axis against the means of the same quantiles on the x-axis [62,63]. On these plots, a point’s location in relation to 0 on the y-axis illustrates the direction and magnitude of the change in value of that point in one distribution compared to its value in the other [63]. In other words, if the 2 distributions being compared are equal, the resulting plot is a horizontal line at the value 0. An upward sloping line indicates increasing dispersion in the distribution while a downward sloping line suggests contraction.

In our analysis, we created a MD plot for each country using data from the earliest and latest available birth cohorts to visually illustrate the shifts in the distribution of height. Points that fall below the horizontal line drawn at 0 indicate losses in height at that place in the distribution, while points that fall above the line indicate gains in height at that place in the distribution.

Results

The final analytic sample comprised 857,053 women. Table 1 provides basic statistics on the population included in the study according to country and birth cohort, as well as change in mean height and SD of height for each country. While 26 countries exhibited statistically significant gains in mean height across birth cohorts, 7 countries experienced statistically significant declines, all of which are in SSA. Rwanda experienced the largest decline in mean height (−1.4 cm, 95% CI: −1.8, 1.0), from 158.0 cm among the cohort born in 1950–1959 to 156.6 cm among the cohort born in 1980–1989, while Colombia experienced the greatest gain in height (2.6 cm, 95% CI: 2.4, 2.8) from 143.8 cm among the cohort born in the years 1950–1959 to 143.6 cm among the cohort born in the years 1980–1989. Sierra Leone consistently had the largest spread in the height distribution, with a SD that ranged from 11.6 to 11.8 cm depending on the birth cohort. Benin experienced the largest relative expansion in the distribution of height, from a SD of 6.4 cm among the cohort born in the years 1950–1959 to 7.9 cm among the cohort born in the years 1980–1989; thus, the ratio of the SD among the latest versus the earliest birth cohorts was 1.52 (95% CI: 1.4, 1.6), while Uganda and Lesotho experienced the largest relative contraction in SD, with a ratio comparing the latest versus the earliest cohorts estimated to be 0.8 (95% CI: 0.7, 0.9) in both countries.

thumbnail
Table 1. Women’s height across countries by birth cohort (women aged 25–49 years).

https://doi.org/10.1371/journal.pmed.1002568.t001

Across countries, changes in the SD of height in relation to changes in mean height are complex and varied. Fig 1 summarizes the results presented in Table 1 regarding the observed change in mean and SD. Overall, 34 of the 59 countries included in the study experienced no significant change in the variance of height between the first and the last birth cohort. Sixteen of these countries experienced increases in mean height, 3 countries experienced decreases in mean height, and 15 countries experienced no significant change in either mean height or the SD of height.

thumbnail
Fig 1. Summary of changes in mean and variance of women’s height from first to last birth cohort by country.

https://doi.org/10.1371/journal.pmed.1002568.g001

Nearly approximately 40% of the countries included in the study (25 out of 59) experienced a significant change in the SD of height between the first and last birth cohorts (9 experienced increased dispersion and 16 experienced decreased dispersion). Of the 9 countries where variance increased, 4 exhibited an increase, 1 exhibited a decrease, and 4 exhibited no change in mean height, and all but 2 are in SSA. There was more geographic diversity in the countries in which the SD of height decreased, including countries in South America, Africa, Eastern Europe, and South Asia. Six of these countries experienced increases in mean height, 3 experienced decreases in mean height, and 7 experienced no significant change in mean height.

Figs 210 provide MD plots for all countries that experienced a significant change in the SD of height between the first and last birth cohort. Countries are grouped according to the direction of the change observed in terms of both mean height and the SD of height. Figs 24 include countries where variance was found to have increased, Figs 57 include countries where variance was found to have decreased, and Figs 810 include countries where we observed no change in the distribution between the first and last cohorts.

thumbnail
Fig 2. Mean-difference plots by country: Countries with increased variance and increased mean height (95% confidence interval shaded).

https://doi.org/10.1371/journal.pmed.1002568.g002

thumbnail
Fig 3. Mean-difference plots by country: Countries with increased variance and decreased mean height (95% confidence interval shaded).

https://doi.org/10.1371/journal.pmed.1002568.g003

thumbnail
Fig 4. Mean-difference plots by country: Countries with increased variance and no change in mean height (95% confidence interval shaded).

https://doi.org/10.1371/journal.pmed.1002568.g004

thumbnail
Fig 5. Mean-difference plots by country: Countries with decreased variance and increased mean height (95% confidence interval shaded).

https://doi.org/10.1371/journal.pmed.1002568.g005

thumbnail
Fig 6. Mean-difference plots by country: Countries with decreased variance and decreased mean height (95% confidence interval shaded).

DRC, Democratic Republic of the Congo.

https://doi.org/10.1371/journal.pmed.1002568.g006

thumbnail
Fig 7. Mean-difference plots by country: Countries with decreased variance and no change in mean height (95% confidence interval shaded).

https://doi.org/10.1371/journal.pmed.1002568.g007

thumbnail
Fig 8. Mean-difference plots by country: Countries with no change in variance and increased mean height (95% confidence interval shaded).

https://doi.org/10.1371/journal.pmed.1002568.g008

thumbnail
Fig 9. Mean-difference plots by country: Countries with no change in variance and decreased mean height (95% confidence interval shaded).

https://doi.org/10.1371/journal.pmed.1002568.g009

thumbnail
Fig 10. Mean-difference plots by country: Countries with no change in variance and no change in mean height (95% confidence interval shaded).

https://doi.org/10.1371/journal.pmed.1002568.g010

In countries with an increase in both the mean and the SD of height across birth cohorts, the gains in height appear to be primarily concentrated among the tallest segments of the distribution, with lesser gains observed among the shortest population segments (e.g., Benin). In countries where the mean increased but the SD decreased, the gains in height were greatest among the shortest segments of the population (e.g., Armenia).

Mali, Lesotho, Rwanda, and Democratic Republic of the Congo (DRC) all experienced decreases in mean height between the first and last birth cohorts; however, the distributional changes vary considerably. For Mali, the variance in height increased across birth cohorts. While all segments of the population lost height, the biggest losses occurred among the shortest segments. Conversely, for Lesotho, Rwanda, and DRC, the distribution contracted because the tallest population segments lost height. In Lesotho, despite the decreased mean, the shortest percentiles of the population gained height, in contrast to the large losses observed among the tallest percentiles.

Finally, among the 26 countries that exhibited no significant change in mean height, the distribution of height expanded in 4 countries and contracted in 7 countries. In the 4 countries where the distribution expanded, the shortest population segments lost height between the first and last birth cohorts while the tallest gained height (e.g., Cameroon). In the 7 countries that experienced a contraction in the distribution, gains were generally observed among the shortest population segments while the tallest either experienced losses in height or little change.

Multilevel regression models reveal that across all 59 countries, mean height did not have a significant relationship with SD (β = 0.015 cm, 95% CI: −0.032 cm, 0.061 cm). There was a significant relationship between median height and the tails of the distribution: for each 1-cm increase in median height, the 5th percentile of height increased by 0.915 cm (95% CI: 0.820 cm, 1.002 cm), and the 95th percentile of height by 0.995 cm (95% CI: 0.925 cm, 1.066 cm).

Discussion

Our study has 3 salient findings. First, consistent with other studies, we find that less than half of the 59 LMICs included in our study have experienced gains in mean women’s height, while the majority have experienced no significant change, or even a decline, in mean height [3436]. Second, we demonstrate that in many countries, the distribution of women’s height has not stayed constant across successive birth cohorts, and that variance does not have a consistent relationship with mean height changes. Multilevel models show that there is no evidence for an association between changes in mean height and SD in LMICs. Third, we find that a focus on the mean obscures distributional changes within countries.

The lack of significant change or declines in mean women’s height that we observed across countries may reflect inequitable progress globally in tackling some of the fundamental drivers of health disparities. Of the 26 countries that experienced increases in mean height, only 5 of them are in SSA while all 7 of the countries that experienced declines in mean height are in SSA. The pattern that we observed in attained height across birth cohorts among women is suggestive of either a stagnation or even deterioration in the environmental and nutritional circumstances that children in these countries faced through the 1990s [35] despite the fact that new medical technologies, political will, and stronger health systems have brought about substantial declines in infant mortality and acute childhood disease throughout SSA over the last several decades [64,65]. For example, the increased dispersion in women’s height observed in Cameroon may be a relic of the economic crisis and structural adjustment programs that the country experienced during the 1990s [66].

Our results also highlight the lack of a consistent relationship between changes in mean women’s height and its SD, which challenges the assumption that the mean is a useful summary measure of a population’s change in risk factors [48,49]. Reporting mean population change over time is a common way to track key public health indicators [36,4245], and the mean is frequently used to measure success in reaching health and development goals [67,68]. The findings of this study suggest that caution should be used when reporting only the mean as a summary measure, especially when improving health equity is a goal. Several studies have interpreted increases in mean as an indication of an improvement in health and development conditions [38,69,70]; however, this interpretation is only valid at a population level following the assumption that the variance stays constant around the mean. In our study, 26 countries exhibited an overall increase in mean height across birth cohorts, but our examination of the variance provides a much more nuanced picture.

In fact, the findings presented in this paper suggest that changes in the distribution of women’s height have occurred in all possible combinations of change in mean and variance (see Fig 1). The broad pattern observed suggests that in the countries where the distribution contracts, the shortest segments of the population generally do better than the tallest, whether that means more prominent gains or smaller losses in height. Conversely, where the variance increases, the tallest segments of the population generally fair better than the shortest. Fig 11 provides a smoothed representation of a graphical typology of the country-level changes that were observed.

thumbnail
Fig 11. Theoretical models of patterns of change in mean and variance.

(A) Decreased variance; (B) no change in variance; (C) increased variance.

https://doi.org/10.1371/journal.pmed.1002568.g011

Our results build on a recent study of 54 LMICs that suggests that height declined globally amongst the poorest wealth quintiles and increased amongst the richest wealth quintiles [35]. From a health equity standpoint, the increased dispersion that we see with regard to height may be suggestive of growing disparities in the distribution of environmental and nutritional inputs, especially in SSA. In Mali, the situation appears particularly bleak in that height has decreased across all population segments, but most markedly among the shortest segments.

Furthermore, our study points toward the possibility of growing disparities in exposures leading to adverse health outcomes, especially early-life episodes of noncommunicable diseases associated with loss of adult height. There remains a considerable research agenda in understanding the burden of these diseases among the most disadvantaged populations in LMICs, and little attention has been paid to identifying effective means of prevention and control among these population subgroups [71]. While there are strong arguments for population-based approaches that are meant to shift the entire distribution of risk factors downwards [2,72,73]—such as those that address inequities in access to economic opportunity, health care, education, and technology—the rising variance in some populations argues for the need for both population-wide and targeted interventions, such as nutrition programs targeting the children most at risk, to ensure both a rise in mean levels and a concomitant improvement in health equity [49].

Our study builds on a body of literature that examines height in relation to socioeconomic inequality [74]. Globally, a positive association has been found between mean height and a country’s Gini coefficient [74]. Additionally, the gap in potential versus attained height may be one way in which socioeconomic disparities are reproduced across generations. While several studies have found that adult height is likely to be at least partially determined by the economic conditions experienced during childhood [11,7577], an individual’s height may in turn influence later socioeconomic potential [7881] through school achievement, economic participation, and income generation [9,8284]. Strauss and Thomas use historical data from Vietnam to argue that changes in the distribution of height provide insight into the changes in the distribution of resources [84]. They argue that prior to 1955, improvements in the distribution of resources in Vietnam led to considerable gains in mean height over time, which were characterized by the shortest segments of the population gaining more height than the tallest segments. However, after 1955, the improvements in mean height plateaued, and the distribution of height in the north began to expand—a phenomenon that they argue may reflect the disproportionate effects of war on the poor in the north.

Rising socioeconomic disparities, however, may not fully explain distributional changes for biological measures such as height. A recent analysis of changes in the BMI distribution in the US demonstrated that dispersion was increasing overall and within socioeconomic groups and suggested that other factors, such as assortative mating, social norms, and genetic predisposition, may play a role [49]. It is possible that some of these factors may also influence changes in the distribution of height. Another potential explanation for the widening dispersion that we observed relates to the possibility of there being some selection bias as a result of the historically adverse conditions found in many LMICs, which may have eliminated shorter individuals from earlier birth cohorts, leaving only the taller individuals, who were more likely to survive [12]. Moreover, recent advances in medical technology, including many of the same interventions that have led to the dramatic reductions in child mortality that have been observed across SSA in recent decades, raise the possibility that the phenomenon observed in our study is the result of increased survival among the shortest population segments in more recent birth cohorts. If shorter people are more likely to survive in more recent birth cohorts than in older birth cohorts, this would be a desirable scenario that would cause the distribution of height to widen. Finally, changes in ethnic diversity resulting from increased migration, such as refugee flows in some countries, may account for some of the changes observed in the relationship between the height parameters examined in this study.

The findings of this study should be interpreted in light of several limitations. First, women tend to lose height in old age; however, given the age range of the women in the study, we do not expect this to have played a significant role as height loss is expected to be minimal during the age range included in this study [35,85]. Second, DHS birth cohorts are not necessarily meant to be representative across birth cohorts, and thus averages may not be consistent over time across birth cohorts. To ensure that this did not influence our results, we examined each cohort to ensure that its mean remained similar as it advanced through survey waves. We found no notable differences in attained height comparing cohorts across survey years in this manner; thus, we do not believe that this would have had an effect on the findings. Third, while DHS uses extensive quality control procedures to ensure consistency in height measurements across surveys and years, there may be some small variation due to measurement error, but we do not expect this to have substantially influenced the results. Finally, the data used in this study include only women; thus, the results may not be fully generalizable to entire populations.

Conclusions

Examining the height trends of populations using both the mean and the variance may enrich research on health equity [86], as adult height may serve as a link between early-life experiences and health in later life, including noncommunicable disease and longevity [36]. While increases in mean height may be indicative of progress in improving standards of living, such improvements have not occurred equitably in many countries. In several countries, despite marked improvements in mean women’s height overall, height among the shortest segments of the study population did not significantly change or even declined. In conclusion, the findings of this study indicate that the population distribution of women’s height does not necessarily stay constant in relation to the mean, and the overreliance on the mean as a summary measure may underestimate inequalities.

Supporting information

S1 Fig. Residual plots (predicted versus residual estimates) and quantile–quantile (QQ) plots for the multilevel regression models for 59 DHS countries.

https://doi.org/10.1371/journal.pmed.1002568.s003

(DOCX)

Acknowledgments

Data for this project were accessed using the Stanford Center for Population Health Sciences (PHS) Data Core. The PHS Data Core is supported by a National Institutes of Health (NIH) National Center for Advancing Translational Science Clinical and Translational Sciences Award (UL1 TR001085) and by internal Stanford University funding.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

References

  1. 1. Silventoinen K. Determinants of variation in adult body height. J Biosoc Sci. 2003;35(2):263–85. pmid:12664962
  2. 2. Rose GA, Khaw K-T, Marmot M. Rose’s strategy of preventive medicine: the complete original text. New York: Oxford University Press; 2008.
  3. 3. Deaton A. Height, health, and development. Proc Natl Acad Sci U S A. 2007;104(33):13232–7. pmid:17686991
  4. 4. Grasgruber P, Cacek J, Kalina T, Sebera M. The role of nutrition and genetics as key determinants of the positive height trend. Econ Hum Biol. 2014;15:81–100. pmid:25190282
  5. 5. Dubois L, Kyvik KO, Girard M, Tatone-Tokuda F, Pérusse D, Hjelmborg J, et al. Genetic and environmental contributions to weight, height, and BMI from birth to 19 years of age: an international study of over 12,000 twin pairs. PLoS ONE. 2012;7(2):e30153. pmid:22347368
  6. 6. Silventoinen K, Kaprio J, Lahelma E, Koskenvuo M. Relative effect of genetic and environmental factors on body height: differences across birth cohorts among Finnish men and women. Am J Public Health. 2000;90(4):627. pmid:10754982
  7. 7. Deaton A. The great escape: a review of Robert Fogel’s “The Escape from Hunger and Premature Death, 1700–2100”. J Econ Lit. 2006;44(1):106–14.
  8. 8. Fogel RW. The escape from hunger and premature death, 1700–2100: Europe, America, and the third world. Cambridge: Cambridge University Press; 2004.
  9. 9. Victora CG, Adair L, Fall C, Hallal PC, Martorell R, Richter L, et al. Maternal and child undernutrition: consequences for adult health and human capital. Lancet. 2008;371(9609):340–57. pmid:18206223
  10. 10. Haeffner L, Barbieri M, Rona R, Bettiol H, Silva A. The relative strength of weight and length at birth in contrast to social factors as determinants of height at 18 years in Brazil. Ann Hum Biol. 2002;29(6):627–40. pmid:12573079
  11. 11. Tucker-Seeley RD, Subramanian S. Childhood circumstances and height among older adults in the United States. Econ Hum Biol. 2011;9(2):194–202. pmid:21159565
  12. 12. Bozzoli C, Deaton A, Quintana-Domeque C. Adult height and childhood disease. Demography. 2009;46(4):647–69. pmid:20084823
  13. 13. Steckel RH. Stature and the standard of living. J Econ Lit. 1995;33(4):1903–40.
  14. 14. Perkins JM, Subramanian S, Smith GD, Özaltin E. Adult height, nutrition, and population health. Nutr Rev. 2016;74(3):149–65. pmid:26928678
  15. 15. Sørensen HT, Sabroe S, Rothman KJ, Gillman M, Steffensen FH, Fischer P, et al. Birth weight and length as predictors for adult height. Am J Epidemiol. 1999;149(8):726–9. pmid:10206622
  16. 16. Emerging Risk Factors Collaboration. Adult height and the risk of cause-specific death and vascular morbidity in 1 million people: individual participant meta-analysis. Int J Epidemiol. 2012;41(5):1419–33. pmid:22825588
  17. 17. Barker DJ, Osmond C, Golding J. Height and mortality in the counties of England and Wales. Ann Hum Biol. 1990;17(1):1–6. pmid:2317001
  18. 18. Watt G, Hart C, Hole D, Smith GD, Gillis C, Hawthorne V. Risk factors for cardiorespiratory and all cause mortality in men and women in urban Scotland: 15 year follow up. Scott Med J. 1995;40(4):108–12. pmid:8787109
  19. 19. Njølstad I, Arnesen E, Lund-Larsen PG. Body height, cardiovascular risk factors, and risk of stroke in middle-aged men and women. Circulation. 1996;94(11):2877–82. pmid:8941116
  20. 20. Green J, Cairns BJ, Casabonne D, Wright FL, Reeves G, Beral V, et al. Height and cancer incidence in the Million Women Study: prospective cohort, and meta-analysis of prospective studies of height and total cancer risk. Lancet Oncol. 2011;12(8):785–94. pmid:21782509
  21. 21. Zhang B, Shu X-O, Delahanty RJ, Zeng C, Michailidou K, Bolla MK, et al. Height and breast cancer risk: evidence from prospective studies and Mendelian randomization. J Natl Cancer Inst. 2015;107(11):djv219. pmid:26296642
  22. 22. Asao K, Kao WHL, Baptiste-Roberts K, Bandeen-Roche K, Erlinger TP, Brancati FL. Short stature and the risk of adiposity, insulin resistance, and type 2 diabetes in middle age—the Third National Health and Nutrition Examination Survey (NHANES III), 1988–1994. Diabetes Care. 2006;29(7):1632–7. pmid:16801590
  23. 23. Magnusson PKE, Gunnell D, Tynelius P, Davey Smith G, Rasmussen F. Strong inverse association between height and suicide in a large cohort of Swedish men: evidence of early life origins of suicidal behavior? Am J Psychiatry. 2005;162(7):1373–5. pmid:15994722
  24. 24. Jousilahti P, Tuomilehto J, Vartiainen E, Eriksson J, Puska P. Relation of adult height to cause-specific and total mortality: a prospective follow-up study of 31, 199 middle-aged men and women in Finland. Am J Epidemiol. 2000;151(11):1112–20. pmid:10873136
  25. 25. Christensen TL, Djurhuus CB, Clayton P, Christiansen JS. An evaluation of the relationship between adult height and health-related quality of life in the general UK population. Clin Endocrinol (Oxf). 2007;67(3):407–12. pmid:17573903
  26. 26. Case A, Paxson C. Height, health, and cognitive function at older ages. Am Econ Rev. 2008;98(2):463. pmid:25152537
  27. 27. Subramanian S, Ackerson LK, Smith GD, John NA. Association of maternal height with child mortality, anthropometric failure, and anemia in India. JAMA. 2009;301(16):1691–701. pmid:19383960
  28. 28. Lawn JE, Lee AC, Kinney M, Sibley L, Carlo WA, Paul VK, et al. Two million intrapartum‐related stillbirths and neonatal deaths: where, why, and what can be done? Int J Gynaecol Obstet. 2009;107(Suppl 1):S5–19.
  29. 29. Özaltin E, Hill K, Subramanian S. Association of maternal stature with offspring mortality, underweight, and stunting in low-to middle-income countries. JAMA. 2010;303(15):1507–16. pmid:20407060
  30. 30. Kozuki N, Katz J, Lee AC, Vogel JP, Silveira MF, Sania A, et al. Short maternal stature increases risk of small-for-gestational-age and preterm births in low-and middle-income countries: individual participant data meta-analysis and population attributable fraction. J Nutr. 2015;145(11):2542–50. pmid:26423738
  31. 31. Sheiner E, Levy A, Katz M, Mazor M. Short stature—an independent risk factor for cesarean delivery. Am J Obstet Gynecol. 2003;189(6):S144.
  32. 32. Merchant KM, Villar J, Kestler E. Maternal height and newborn size relative to risk of intrapartum caesarean delivery and perinatal distress. BJOG. 2001;108(7):689–96. pmid:11467692
  33. 33. Tsu VD. Maternal height and age: risk factors for cephalopelvic disproportion in Zimbabwe. Int J Epidemiol. 1992;21(5):941–6. pmid:1468857
  34. 34. Akachi Y, Canning D. Health trends in sub-Saharan Africa: conflicting evidence from infant mortality rates and adult heights. Econ Hum Biol. 2010;8(2):273–88. pmid:20634153
  35. 35. Subramanian S, Özaltin E, Finlay JE. Height of nations: a socioeconomic analysis of cohort differences and patterns among women in 54 low-to middle-income countries. PLoS ONE. 2011;6(4):e18962. pmid:21533104
  36. 36. NCD Risk Factor Collaboration. A century of trends in adult human height. eLife. 2016;5:e13410. pmid:27458798
  37. 37. Baten J, Blum M. Growing tall but unequal: new findings and new background evidence on anthropometric welfare in 156 countries, 1810–1989. Econ Hist Dev Reg. 2012;27(Suppl 1):S66–85.
  38. 38. Cavelaars AEJM Kunst AE, Geurts JJM, Crialesi R, Grötvedt L, Helmert U, et al. Persistent variations in average height between countries and between socio-economic groups: an overview of 10 European countries. Ann Hum Biol. 2000;27(4):407–21. pmid:10942348
  39. 39. Akachi Y, Canning D. The height of women in sub-Saharan Africa: the role of health, nutrition, and income in childhood. Ann Hum Biol. 2007;34(4):397–410. pmid:17620149
  40. 40. Moradi A. The nutritional status of women in sub-Saharan Africa, 1950–1980. Oxford: National Research Council; 2006.
  41. 41. Crimmins EM, Finch CE. Infection, inflammation, height, and longevity. Proc Natl Acad Sci U S A. 2006;103(2):498–503. pmid:16387863
  42. 42. Danaei G, Finucane MM, Lu Y, Singh GM, Cowan MJ, Paciorek CJ, et al. National, regional, and global trends in fasting plasma glucose and diabetes prevalence since 1980: systematic analysis of health examination surveys and epidemiological studies with 370 country-years and 2·7 million participants. Lancet. 2011;378(9785):31–40. pmid:21705069
  43. 43. Farzadfar F, Finucane MM, Danaei G, Pelizzari PM, Cowan MJ, Paciorek CJ, et al. National, regional, and global trends in serum total cholesterol since 1980: systematic analysis of health examination surveys and epidemiological studies with 321 country-years and 3·0 million participants. Lancet. 2011;377(9765):578–86. pmid:21295847
  44. 44. Danaei G, Finucane MM, Lin JK, Singh GM, Paciorek CJ, Cowan MJ, et al. National, regional, and global trends in systolic blood pressure since 1980: systematic analysis of health examination surveys and epidemiological studies with 786 country-years and 5·4 million participants. Lancet. 2011;377(9765):568–77. pmid:21295844
  45. 45. Finucane MM, Stevens GA, Cowan MJ, Danaei G, Lin JK, Paciorek CJ, et al. National, regional, and global trends in body-mass index since 1980: systematic analysis of health examination surveys and epidemiological studies with 960 country-years and 9·1 million participants. Lancet. 2011;377(9765):557–67. pmid:21295846
  46. 46. Hasan MT, Magalhaes RJS, Williams GM, Mamun AA. Forecasting the progress towards the target of Millennium Development Goal 1C in children under 5 years of age in Bangladesh. Public Health Nutr. 2015;18(10):1728–36. pmid:25648950
  47. 47. Stevens GA, Finucane MM, Paciorek CJ, Flaxman SR, White RA, Donner AJ, et al. Trends in mild, moderate, and severe stunting and underweight, and progress towards MDG 1 in 141 developing countries: a systematic analysis of population representative data. Lancet. 2012;380(9844):824–34. pmid:22770478
  48. 48. Razak F, Corsi DJ, Subramanian S. Change in the body mass index distribution for women: analysis of surveys from 37 low-and middle-income countries. PLoS Med. 2013;10(1):e1001367. pmid:23335861
  49. 49. Krishna A, Razak F, Lebel A, Smith GD, Subramanian S. Trends in group inequalities and interindividual inequalities in BMI in the United States, 1993–2012. Am J Clin Nutr. 2015;101(3):598–605. pmid:25733645
  50. 50. von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. PLoS Med. 2007;4(10):e296. pmid:17941714
  51. 51. Rutstein SO. Guide to DHS statistics. Calverton (Maryland): ORC Macro; 2006.
  52. 52. Corsi DJ, Neuman M, Finlay JE, Subramanian S. Demographic and Health Surveys: a profile. Int J Epidemiol. 2012;41(6):1602–13. pmid:23148108
  53. 53. ORC Macro. Demographic and Health Survey interviewer’s manual. Calverton (Maryland): ORC Macro; 2006.
  54. 54. ICF International. Demographic and Health Survey sampling and household listing manual. Calverton (Maryland): Measure DHS; 2012.
  55. 55. Garenne M. Trends in nutritional status of adult women in sub-Saharan Africa. Calverton (Maryland): ICF Macro; 2011.
  56. 56. Macro International. Anthropometry, anemia, and HIV testing field manual. Calverton (Maryland): Macro International; 2008.
  57. 57. Garrett DA, Sangha JK, Kothari MT, Boyle D. Field-friendly techniques for assessment of biomarkers of nutrition for development. Am J Clin Nutr. 2011;94(2):685S–90S. pmid:21677055
  58. 58. Assaf S, Kothari MT, Pullum TW. An assessment of the quality of DHS anthropometric data, 2005–2014. Calverton (Maryland): ICF International; 2015.
  59. 59. StataCorp. Stata Statistical Software. Release 14. College Station (Texas): StataCorp; 2015.
  60. 60. R Core Team. R: a language and environment for statistical computing. Vienna: Foundation for Statistical Computing; 2013.
  61. 61. Brown MB, Forsythe AB. Robust tests for the equality of variances. J Am Stat Assoc. 1974;69(346):364–7.
  62. 62. Cleveland WS. Visualizing data. Summit (New Jersey): Hobart Press; 1993.
  63. 63. Flegal KM, Troiano RP. Changes in the distribution of body mass index of adults and children in the US population. Int J Obes. 2000;24(7):807–18.
  64. 64. Wang H, Liddell CA, Coates MM, Mooney MD, Levitz CE, Schumacher AE, et al. Global, regional, and national levels of neonatal, infant, and under-5 mortality during 1990–2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet. 2014;384(9947):957–79. pmid:24797572
  65. 65. Rajaratnam JK, Marcus JR, Flaxman AD, Wang H, Levin-Rector A, Dwyer L, et al. Neonatal, postneonatal, childhood, and under-5 mortality for 187 countries, 1970–2010: a systematic analysis of progress towards Millennium Development Goal 4. Lancet. 2010;375(9730):1988–2008. pmid:20546887
  66. 66. Pongou R, Salomon JA, Ezzati M. Health impacts of macroeconomic crises and policies: determinants of variation in childhood malnutrition trends in Cameroon. Int J Epidemiol. 2006;35(3):648–56. pmid:16567342
  67. 67. Wang YC, Orleans CT, Gortmaker SL. Reaching the healthy people goals for reducing childhood obesity: closing the energy gap. Am J Prev Med. 2012;42(5):437–44. pmid:22516482
  68. 68. Waage J, Banerji R, Campbell O, Chirwa E, Collender G, Dieltiens V, et al. The Millennium Development Goals: a cross-sectoral analysis and principles for goal setting after 2015. Lancet. 2010;376(9745):991–1023. pmid:20833426
  69. 69. Stevens GA, Finucane MM, Paciorek CJ, Flaxman SR, White RA, Donner AJ, et al. Trends in mild, moderate, and severe stunting and underweight, and progress towards MDG 1 in 141 developing countries: a systematic analysis of population representative data. Lancet. 2012;380(9844):824–34. pmid:22770478
  70. 70. Mohsena M, Goto R, Mascie-Taylor CN. Maternal nutritional status (as measured by height, weight and BMI) in Bangladesh: trends and socio-economic association over the period 1996 to 2007. Public Health Nutr. 2016;19(08):1438–45.
  71. 71. Miranda JJ, Kinra S, Casas JP, Davey Smith G, Ebrahim S. Non-communicable diseases in low- and middle-income countries: context, determinants and health policy. Trop Med Int Health. 2008;13(10):1225–34. pmid:18937743
  72. 72. Rose G. Sick individuals and sick populations. Int J Epidemiol. 2001;30(3):427–32. pmid:11416056
  73. 73. Rose G. Strategy of prevention: lessons from cardiovascular disease. Br Med J (Clin Res Ed). 1981;282(6279):1847–51.
  74. 74. Moradi A, Baten J. Inequality in sub-Saharan Africa: new data and new insights from anthropometric estimates. World Dev. 2005;33(8):1233–65.
  75. 75. Peck MN, Lundberg O. Short stature as an effect of economic and social conditions in childhood. Soc Sci Med. 1995;41(5):733–8. pmid:7502105
  76. 76. Silventoinen K, Lahelma E, Rahkonen O. Social background, adult body-height and health. Int J Epidemiol. 1999;28(5):911–8. pmid:10597991
  77. 77. Li L, Manor O, Power C. Early environment and child-to-adult growth trajectories in the 1958 British birth cohort. Am J Clin Nutr. 2004;80(1):185–92. pmid:15213047
  78. 78. Barker DJ, Eriksson J, Forsén T, Osmond C. Infant growth and income 50 years later. Arch Dis Child. 2005;90(3):272–3. pmid:15723915
  79. 79. Carba DB, Tan VL, Adair LS. Early childhood length-for-age is associated with the work status of Filipino young adults. Econ Hum Biol. 2009;7(1):7–17. pmid:19261549
  80. 80. Thomas D, Frankenberg E. Health, nutrition and prosperity: a microeconomic perspective. Bull World Health Organ. 2002;80(2):106–13. pmid:11953788
  81. 81. Adair LS, Fall CH, Osmond C, Stein AD, Martorell R, Ramirez-Zea M, et al. Associations of linear growth and relative weight gain during early life with adult health and human capital in countries of low and middle income: findings from five birth cohort studies. Lancet. 2013;382(9891):525–34. pmid:23541370
  82. 82. Case A, Paxson C. Stature and status: height, ability, and labor market outcomes. Cambridge (Massachusetts): National Bureau of Economic Research; 2006.
  83. 83. Meyer HE, Selmer R. Income, educational level and body height. Ann Hum Biol. 1999;26(3):219–27. pmid:10355493
  84. 84. Strauss J, Thomas D. Health, nutrition, and economic development. J Econ Lit. 1998;36(2):766–817.
  85. 85. Cline MG, Meredith KE, Boyer JT, Burrows B. Decline of height with age in adults in a general population sample: estimating maximum height and distinguishing birth cohort effects from actual loss of stature with aging. Hum Biol. 1989;61(3):415–25. pmid:2807265
  86. 86. Gakidou EE, Murray CJ, Frenk J. Defining and measuring health inequality: an approach based on the distribution of health expectancy. Bull World Health Organ. 2000;78(1):42–54. pmid:10686732