Association of Maternal Short Stature with Stunting in Mexican Children: Common Genes Vs Common Environment

Association of maternal short stature with stunting in Mexican children: common genes vs common environment

S HernaÂndez-DõÂaz1*, KE Peterson1, S Dixit2, B HernaÂndez1, S Parra3, S Barquera4, J SepuÂlveda3 and JA Rivera3

1Harvard School of Public Health, Boston, MA, USA; 2Department of Human Nutrition and Dietetics, University of Illinois at Chicago, Chicago, IL, USA; 3Instituto Nacional de Salud PuÂblica, Col. Santa MarõÂa AhuacatitlaÂn, Cuernavaca, Morelos, Mexico; and 4Tufts University School of Nutrition Science and Policy, Medford, MA, USA

Objective: To evaluate the association between stunting in children and maternal short stature, controlling for potential environmental confounders. Design: 1988 Mexico National Nutrition Survey. Setting: Mexico Subjects: The ®nal sample size was 4663 pairs of children ( < 5 y) and their mothers (12 ± 49 y) from a total of 13 236 surveyed houses. Main outcome measures: Stunting (height-for-age Z-scores < 7 2). Results: The prevalence of stunting in children was 19%, and 10% of the mothers exhibited short stature ( < 145 cm). In the crude analysis, mothers with short stature were signi®cantly more likely to have stunted children (odds ratio (OR) 4.0; 95% con®dence interval (CI) 3.2 ± 4.8; P-value < 0.001). In a multiple logistic regression model the OR for child stunting was reduced, but remained signi®cant OR 2.0; 95% CI 1.6 ± 2.6; P-value < 0.001) after adjustment for region, urban=rural residence, socio-economic status, household size, child age and presence of infection in the past 14 d, and maternal age, body mass index (BMI), and educational level. Adjusted ORs varied between regions (Mexico City, OR 3.9; North Mexico, OR 3.1; Central Mexico, OR 2.0; South Mexico, OR 1.6. Comparison of crude vs adjusted estimates pointed to regional differences in the proportion of association between maternal and child short statures explained by environmental determinants. Conclusions: Maternal stature, re¯ecting her potential height and early environment, appeared to contribute to child height independently of the shared risk factors that could affect stature. Nonetheless, we could explain much of the association between stunting in children and maternal short stature by environmental factors, and part of the residual variability may be due to unmeasured determinants. Regional differences pointed to a predominance of environmental factors in explaining child stunting in poorer regions. Descriptors: children; short stature; stunting; Mexico; survey

Introduction maternal height provides information about growth potential in the child (Tanner et al, 1970). Nonetheless, the Stature is a heritable morphologic characteristic which can correlation between maternal and child stature may re¯ect be considered the phenotypic expression of genetic endow- not only shared genetic background, but also common ment and childhood environment (Tanner, 1960; Jepson environmental determinants that affect the mother during et al, 1994; Rivera et al, 1995; Martorell et al, 1988). To early childhood and maturation and subsequently the the degree that mothers express their genetic potential, growth of her offspring (Kramer, 1987). In developing countries, environmental factors that *Correspondence: S HernaÂndez-DõÂaz, Department of Maternal and Child impair the ful®llment of individual genetic potential for Health, Harvard School of Public Health, 677 Huntington Ave, Boston, MA 02115, USA. growth have important implications for understanding Guarantor: S HernaÂndez-DõÂaz. intergenerational patterns of malnutrition (Scrimshaw, Contributors: SH-D was lead author responsible for formulating the 1995). Availability of food, water supply and sanitation, research questions and analytic strategy, conducting the analysis, and and access to medical services determine physical and writing the manuscript; KEP worked in close collabortion with the ®rst behavioral conditions that in¯uence child growth through author in de®ning the hypothesis, interpreting the results and writing the manuscript; SD worked with the ®rst author from the early stages of the the proximate determinants of inadequate dietary intake project, contributing to the hypothesis de®nition and to the analysis plan; and infections (Pinstrup-Andersen et al, 1993; McGuire & BH participated substantially in the interpretation of the results in the Popkin, 1990). We could de®ne a vicious cycle of mal- preparation of the manuscript; SP provided input in writing and through nutrition across generations as follows (UNICEF, 1998). A group discussion at the analysis and interpretation stages; SB contributed substantially to the interpretation of ®ndings in the context of the literature mother may not express her genetic potential if she experi- on nutrition transition and intergenerational patterns; JS was responsible enced an adverse environment during infancy. This can be for the design, conduct and overall supervision of research emanating from re¯ected in her adult size and her metabolic status (Martor- the 1988 Mexican National Nutritional Survey, in coordination with JAR; ell et al, 1994, 1998; Martorell, 1996; Rao, 1984). Women JAR contributed substantially to the design and analytic strategy as well as with short stature will be at higher risk of having low birth interpretation of ®ndings in the context of the literature on nutrition transition and intergenerational patterns. weight (LBW) infants as a function of at least three factors: Received 21 July 1998; revised 25 May 1999; accepted 22 June 1999 genetic heritage, physical mechanisms such as sub-optimal Maternal short stature and stunting in children S HernaÂndez-DõÂaz et al 939 development of her pregnancy-related anatomical systems with favorable economic conditions, where stature in both (Kramer, 1987; Kramer et al, 1990; Gonzalez-Cossio et al, mother and child may better re¯ect genetic potential 1998; World Health Organization, 1995b; Ramakrishnan et (Tanner et al, 1970, Martorell et al, 1977). al, 1999), and metabolic mechanisms like maternal circu- lating glucose levels (Lowy, 1994; Barker, 1994). Infants with LBW in turn are at risk for poor growth during Methods childhood (World Health Organization, 1992; Gayle et al, 1987; Binkin et al, 1988) and may become small mothers In 1988 the Secretary of Health of Mexico carried out the (Falkner et al, 1994). However, the chain of poor potential National Nutrition Survey (ENN) as a part of the larger growth ful®llment may be ameliorated, particularly if the National Health System Surveys. The multi-stage, strati®ed problem is identi®ed at an early stage and proper actions sampling design used in this cross-sectional survey ensured are implemented. In individual children, provision of nutri- that the results were representative of the four regions of ents and control of infections may permit partial post-natal the country: Mexico City and North, Central and South catch-up growth in LBW infants (Strauss & Dietz, 1997; Mexico. The sample for this study consisted of 7426 Villar et al, 1984; Golden, 1988; Martorell et al, 1994). At children less than 5 y of age and 19 278 women between the population level, secular trend data indicates that 12 and 49 y of age from a total of 13 236 surveyed houses, increased access to food supplies and decreased burden of with a non-response rate of 19.9%. Detailed information on infectious diseases result in gains in phenotypic stature that sampling design, calculation of sample size, data collection may be observed over generations (Hauspie et al, 1996; and data entry phases of the survey has been given else- Ulijaszek, 1994). Improvements in socio-economic status where (Rivera et al, 1995, SepuÂlveda-Amor et al, 1990). from one generation to another would, however, allow the Mothers and children were matched, yielding 5083 pairs. child only to partially ful®ll his genetic potential, since We excluded 420 pregnant women who had statistically small maternal size may have blunted post-natal growth signi®cant differences in weight, a control variable in these through intra-uterine restriction and endocrine imbalance. analyses, compared with non-pregnant women. In order to Therefore, several generations of an improved environment exclude aberrant values for children's height or weight, during childhood may be necessary to discern the real observations of height-for-age and weight-for-height Z- in¯uences on stature mediated through maternal genetic scores out of the 1 ± 99% interval were studied. The ®nal factors. sample size was 4663 pairs, corresponding to 3448 women. The epidemiological transition is a theory described ®rst Anthropometric measurements were obtained by trained by Omran, characterized by socio-economic changes assoc- ®eld workers following a standardized protocol (Habicht, iated with a shift in mortality and morbidity (Omran, 1971, 1974). Length (children younger than 24 months) and 1983). In the developing countries of Latin America, this height (children 24 ± 59 months), in centimeters, were transition is characterized by a slow decline in the pre- converted to Z-scores for age and gender of the valence of undernutrition and infectious diseases, coexist- NCHS=WHO reference population (Hamill et al, 1979) to ing with an increasing prevalence of obesity and other permit comparisons and to follow international guidelines chronic diseases (Bobadilla et al, 1993; Omran, 1983). for analysis of anthropometric data (World Health Organi- Greater, more rapid economic development in higher zation, 1995a). Children with a height-for-age Z-score social classes may result in disparities in health status value lower than 7 2 were classi®ed as stunted (World across socio-economic strata. Such disparity has been Health Organization, 1983, 1995a), a condition used as our termed epidemiologic polarization (ACC=SCN, 1993). In binary response variable. Maternal height, in centimeters, the late 1980s Mexico under went heterogeneous economic was our main exposure variable; heights less than 145 cm development across regions in different stages of the were classi®ed as short stature, in accordance with previous epidemiologic transition (SepuÂlveda-Amor et al, 1990). international reports (ACC=SCN, 1992). Maternal body Previous reports summarizing the nutrition situation in mass index (BMI), weight (kg)=height m2, was calculated Mexico identi®ed patterns that vary among different geo- and classi®ed as underweight, normal or overweight. graphical areas: a prevalence of obesity in the North, Underweight was de®ned as less than the 15th percentile underweight in Central Mexico and short stature in the and overweight as greater or equal than the 85th South (Rivera et al, 1995; SepuÂlveda-Amor et al, 1990; percentile of age-speci®c reference data for women (Must HernaÂndez et al, 1996). Regional differences in nutritional et al, 1991 a, b). parameters were explained largely by differences in socio- Maternal age at birth of the studied child was categorized economic factors (Rivera et al, 1995). To the extent that into three groups: 12 ± 19, 20 ± 34 and 35 ± 49 y. The prevalent outcomes re¯ect epidemiological polarization, mother's education was classi®ed into four levels: no educa- the correlation of stature within mother ± child pairs may tion, at least some elementary school, completed high have also been in¯uenced to varying degrees among school, and education beyond high school. Child age was regions. classi®ed as follows: under 6 months, 6 ± 11 months, 12 ± 23 In this report, our primary aim is to describe the months, 24 ± 35 months, and 36 ± 59 months. Mothers association of child stunting with maternal short stature, reported the children's age at introduction of foods and both crude and adjusted for post-natal environmental liquids other than breast milk, and duration of breast- determinants of child stature. We hypothesized that part feeding. This information, however, was obtained only of the association between maternal and child short statures for children under 12 months, so we did not use it in our can be explained by socio-economic determinants. By analysis. Diarrhea, respiratory infections and other infec- means of a regional strati®cation of the analysis, we also tions were recorded as dichotomous variables, indicating attempt to provide evidence of variability in the relevance their presence in the last 14 d. of environmental risk factors for poor growth across Environmental factors included as covariates are as regions. Higher association may be expected in regions follows. Regions: Mexico City, North, Central and South Maternal short stature and stunting in children S HernaÂndez-DõÂaz et al 940 Mexico; rural=urban: families were classi®ed as living in a Results rural municipality if more than half of the population of the Table 1 displays health and socio-demographic character- municipality lived in communities with fewer than 15 000 istics of the mothers and children in the analytic sample. inhabitants, or an urban municipality when more than half The prevalence of short stature was 18.8% in children and of the population lived in communities with greater than 9.9% in women. Strati®ed by region, the prevalence of 15 000 inhabitants (SepuÂlveda-Amor et al, 1990); socio- stunting ranged from around 11% in the North and in economic status (SES): a principal components analysis of Mexico City to 33.7% in the South, mirrored by differences household characteristics was performed, including type of in prevalence of maternal short stature. The largest socio- ¯ooring, source of household water, type of drainage, economic disparities occurred between Mexico City, with sewage disposal and possession of a radio, television, 4.8% of the families in the lowest SES tercile and 52.6% in refrigerator, telephone and car. A composite factor was the highest SES tercile, and the South, with 58.8% of the created to represent `socio-economic level of the house- families living in the lowest and 18.2% living in the highest hold', which explained 67.5% of the total variance. Factor SES tercile. The prevalence of women with no education scores were calculated for each mother=child pair. These was 22.7% in the South and 3.8% in Mexico City. Strati®ed scores were divided into terciles representing low, middle by maternal stature, the prevalence of stunting was 16% in and high SES (Long-Dunlap et al, 1995), and were con- children with normal height mothers and 42.9% in children sidered both as a continuous and as a categorical variable. with short mothers. The proportion of families in each tercile was not exactly Table 2 displays crude and adjusted odds ratios for our 33% due to ties in the scores. Number of household main variables of interest. The associations changed with members: three groups were delineated ( < 4, 4 ± 8, > 8). adjustment for child factors (model 1), child and mother Preliminary analysis involved basic descriptive statis- factors (model 2), and child, mother and socio-economic tics and examination of prevalence rates of stunting within factors (model 3). The association between maternal and each covariate category using contingency tables. Logistic child short stature decreased, as we controlled for other regression was utilized to assess the association of cov- covariates in successive models, from a crude OR of 4.0 ariates, individually as well as jointly, with the dichot- (95% CI: 3.2 ± 4.8) to an adjusted OR of 2.0 (95% CI: 1.6 ± omous response variable (SAS, 1995). In the ®rst stage, 2.6). Odds of stunting increased in the South, and with we ®tted models to assess the associations between a child's age, lower maternal education, rural residence, low child's stunting and the main variables of interest, mater- socio-economic status, in crowded households, and with nal height (short stature vs normal) and region (with infections other than respiratory and diarrhea in the last 2 Mexico City as the reference). In the next stage, we weeks. estimated adjusted odds ratios (OR) by adding to the We also ®tted the ®nal model including maternal height model variables that were signi®cant at the univariate as a continuous variable (data not shown). Stunting in level, using a lenient cutoff point of P 0.10, or that children was associated with the height of the mother, needed to be included in order to control for confounding, with an OR of 0.92 (95% CI: 0.91 ± 0.94) corresponding assessed by more than 10% change on the main parameter to a 1 cm increase in mother's height, after adjusting for the estimates. We also explored whether an individual cov- other possible risk factors for stunting. This implies that a ariate could be best modeled as categorical or continuous 10 cm increase in mother's height would correspond to an (linear, polynomial, or log transformed), and conducted OR of 0.4. That is, the odds of encountering a stunted child analyses that included maternal height as a continuous of a woman who is 160 cm tall, for example, would be less variable. During the model building, BMI and SES were than half as great as the odds of ®nding a stunted child of a kept as continuous variables. We collapsed categories 150 cm tall mother. from other variables since there were no differences in In the models presented up to this point there were no stunting prevalence (education until high school and interaction terms between regions and maternal height (as a education beyond high school, children aged between 12 binary variable). Upon their inclusion we found that the and 36 months, and Mexico City and North regions). The magnitude of the association between maternal and child gender of the child was not included since it was not stature differed across regions. The association was stron- signi®cant and its inclusion=exclusion did not change ger in the North ± Mexico City combined category as other parameters. After modeling main effects, we tested compared with the Central (P-value 0.078), or South for effect modi®cation by including interaction terms (P-value 0.002) regions. Four separate models were between region and maternal height. also constructed, one for each region. The OR between The interpretation of the results from this cross-sectio- maternal and child short stature was higher in Mexico City nal study may be dif®cult due to the fact that nutritional (OR 3.9; 95% CI 2.2 ± 7.1) and the North region status is a dynamic phenomenon, especially in children (OR 3.1; 95% CI 1.4 ± 6.9) and lower in the South under 2 y of age. Therefore we repeated analyses focusing (OR 1.6; 95% CI 1.1 ± 2.2) and in the Central on children over 36 months of age, whose growth would (OR 2.0; 95% CI 1.1 ± 3.5) regions (Table 3 and have been already delayed. Another important character- Figure 1). Moreover, adjustment for environmental covari- istic of this particular study is that observations may not ates changed crude ORs less in Mexico City and the North be independent, given that each mother might be matched as compared with the South and Central regions. to more than one child (on average, two children). Since When we repeated the analysis focusing on children repeated measures analysis accommodates this type of over 36 months of age, neither maternal nor regional dataset with clustered data, we used generalized estimated associations with child stunting differed from the results equations (GEE; Liang & Zeger, 1986) to estimate robust presented. As a concluding analytic step, we ®tted ®nal 95% con®dence intervals (CI) and compared the results models using repeated measures techniques. The results with the simple logistic regression results presented were very similar, and the inferences that could be drawn below. Maternal short stature and stunting in children S HernaÂndez-DõÂaz et al 941 Table 1 Characteristics of mother ± child pairs by region, 1988 Mexico National Nutrition Survey

Frequency (%), Frequency (%), Frequency (%), Frequency (%), Frequency (%), Variable Country North Central South Mexico City

Children stunting:a Yes 876 (18.8) 129 (10.9)* 220 (19.5)* 407 (33.7)* 120 (10.5)* Age (months): 0 ± 6 421 (9.0) 117 (9.9) 97 (8.6) 95 (7.9) 112 (9.8) 6 ± 12 485 (10.4) 129 (10.9) 111 (9.9) 139 (11.5) 106 (9.3) 12 ± 24 883 (18.9) 234 (19.7) 201 (17.8) 250 (20.7) 198 (17.3) 24 ± 36 941 (20.2) 221 (18.6) 244 (21.7) 245 (20.3) 231 (20.2) > 36 1933 (41.5) 485 (40.9) 474 (42.1) 478 (39.6) 496 (43.4) Gender: Male 2391 (51.3) 596 (50.3) 570 (50.6) 651 (53.9) 574 (50.2) Female 2272 (48.7) 590 (49.7) 557 (49.4) 556 (46.1) 569 (49.8) Diarrhea:b Yes 808 (17.5) 171 (14.6)* 189 (17.0)* 291 (24.1)* 157 (13.9)* Respiratory infection:b Yes 1528 (33.2) 345 (29.5)* 286 (25.8)* 458 (38.3)* 439 (38.7)* Other infections:b Yes 596 (13.0) 114 (9.8)* 141 (12.8)* 227 (19.0)* 114 (10.1)* Mother height: < 145 cm 343 (9.9) 31 (3.5)* 67 (8.2)* 184 (21.5)* 61 (6.9)* Weight:c Overweight 301 (8.7) 104 (11.6)* 66 (8.1)* 65 (7.6)* 66 (7.5)* Normal 2874 (83.4) 741 (82.5) 661 (81.3) 703 (82.1) 769 (87.3) Underweight 273 (7.9) 53 (5.9) 86 (10.6) 88 (10.3) 46 (5.2) Age (y): < 19 302 (8.7) 82 (9.1)* 60 (7.4)* 83 (9.7)* 77 (8.7)* 20 ± 34 2532 (73.4) 678 (75.5) 578 (71.1) 593 (69.3) 683 (77.5) 35 ± 49 614 (17.8) 138 (15.4) 175 (21.5) 180 (21.0) 121 (13.7) Education: None 350 (10.2) 35 (3.9)* 88 (10.9)* 194 (22.7)* 33 (3.8)* Elementary 1862 (54.2) 466 (52.2) 503 (62.1) 476 (55.7) 417 (47.4) High school 590 (17.2) 184 (20.6) 99 (12.2) 92 (10.8) 215 (24.4) Beyond high school 635 (18.5) 208 (23.3) 120 (14.8) 92 (10.8) 215 (24.4) Region: North 898 (26.0) Central 813 (23.6) South 856 (24.8) Mexico City 881 (25.6) Residence: Urban 2881 (83.8) 817 (91.2)* 581 (71.9)* 662 (77.3)* 821 (93.7)* Rural 555 (16.2) 79 (8.8) 227 (28.1) 194 (22.7) 55 (6.3) Socio-economic status: Low 971 (30.0) 219 (25.8)* 248 (33.3)* 463 (58.8)* 41 (4.8)* Middle 997 (30.8) 261 (30.8) 190 (25.5) 182 (23.1) 364 (42.6) High 1267 (39.2) 368 (43.4) 307 (41.2) 143 (18.2) 449 (52.6) No. of people in the house: 4 1221 (35.6) 359 (40.1)* 228 (28.2)* 223 (26.1)* 411 (47.1)* 5 ± 7 1580 (46.1) 407 (45.5) 373 (46.2) 417 (48.8) 383 (43.9) 8 630 (18.4) 129 (14.4) 207 (25.6) 215 (25.2) 79 (9.1)

Maximum sample sizes are 4663 for children and 3448 for mothers and household characteristics. The difference is due to the presence of more than one child per mother, and per household. aStunting was de®ned as height-for-age Z scores < 7 2. bIn the past 2 weeks. cUnderweight was de®ned as body mass index less than the 15th percentile and overweight as greater or equal than the 85th percentile of age-speci®c reference data. *Chi-Square P-value 0.001 in regional comparisons. from this GEE analysis were the same as the ones from the may explain part of the association between maternal and logistic regression. child height. A combination of shared environmental factors may modulate the constitutional genetic backgrounds, impairing the ful®llment of height potential in both mother Discussion and child. Previous publications have suggested that part of Our analysis was aimed at clarifying a potential association the parent ± child correlation in malnourished populations between maternal and child short stature in a country may be attributed to the environment (Martorell et al, 1977; undergoing an epidemiological transition. We found that Martorell & Habicht, 1986). Only with an improved envir- short stature in the mother was associated with a higher onment during childhood could we discern the real in¯u- prevalence of stunting in her children, although adjustment ences on stature mediated through genetic factors. for environmental conditions attenuated the association. Regions were de®ned considering anthropometric infor- From the results, we may infer that socio-economic deter- mation from previous studies, the political limits of each minants related to short stature in both mother and child state, demographic contiguity and the existing budget for Maternal short stature and stunting in children S HernaÂndez-DõÂaz et al 942 Table 2 Crude and adjusted OR for child stuntinga associated with child, maternal and environmental characteristics (1988 Mexico National Nutrition Survey)

Crude Model 1 Model 2 Model 3 OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) Variable (n 4663) (n 4663) (n 4648)a (n 4305)b

Mother's height: 145 cm 1 1 1 1 < 145 cm 3.95 (3.2 ± 4.8) 3.03 (2.4 ± 3.8) 2.42 (1.9 ± 3.0) 2.04 (1.59 ± 2.62) Region: North and Mexico City 1 1 1 1 Central 2.03 (1.7 ± 2.5) 1.97 (1.6 ± 2.4) 1.44 (1.2 ± 1.8) 1.03 (0.8 ± 1.3) South 4.25 (3.6 ± 5.1) 3.62 (3.0 ± 4.4) 2.39 (2.0 ± 2.9) 1.42 (1.1 ± 1.8) Children's age (months): 0±5 1 1 1 1 6 ± 11 5.58 (2.8 ± 11.1) 5.62 (2.8 ± 11.3) 6.12 (3.0 ± 12.4) 5.70 (2.8 ± 11.6) 12 ± 35 10.32 (5.5 ± 19.5) 11.23 (5.9 ± 21.4) 12.19 (6.4 ± 23.4) 11.28 (5.8 ± 21.8) 36 ± 59 12.18 (6.5 ± 23.0) 13.90 (7.3 ± 26.5) 15.19 (7.9 ± 29.2) 14.36 (7.4 ± 27.7) Mother's body mass index:c Mother's age (y): 0.94 (0.92 ± 0.95) Ð 0.94 (0.92 ± 0.96) 0.95 (0.93 ± 0.97) < 19 1.49 (1.2 ± 1.9) Ð 1.25 (0.9 ± 1.7) 1.32 (1.0 ± 1.8) 20 ± 34 1 1 1 35 ± 49 1.24 (1.0 ± 1.5) 0.91 (0.7 ± 1.1) 0.83 (0.7 ± 1.1) Mother's education: None 10.90 (8.4 ± 14.2) Ð 7.26 (5.4 ± 9.7) 2.96 (2.1 ± 4.2) Elementary 3.47 (2.8 ± 4.3) 3.04 (2.4 ± 3.8) 1.68 (1.3 ± 2.2) High school or beyond 1 1 1 Residence: Urban 1 Ð Ð 1 Rural 1.97 (1.7 ± 2.4) 1.25 (1.0 ± 1.6) Socio-economic status (per unit 0.42 (0.4 ± 0.5) Ð Ð 0.60 (0.5 ± 0.7) increase in SES score): People in household 41ÐÐ1 5 ± 7 1.84 (1.5 ± 2.2) 1.45 (1.2 ± 1.8) 8 3.89 (3.1 ± 4.8) 2.07 (1.6 ± 2.7) Other infections:d Yes 2.12 (1.8 ± 2.6) Ð Ð 1.31 (1.0 ± 1.7) No 1 1

aStunting was de®ned as height-for-age Z scores < 7 2. bReduction in sample size due to missing values in some of the variables included in the models. cPer unit change. dIn the past 2 weeks.

Table 3 Crude and adjusteda association of child stuntingb with maternal short stature by region (South, North, Central and Mexico City); 1988 Mexico National Nutrition Survey

Maternal Crude OR Adjusted OR tature (95% CI) (95% CI)

Mexico DF (n 1101) 145 cm 1 1 < 145 cm 4.49 (2.7 ± 7.6) 3.94 (2.2 ± 7.1) North (n 1100) 145 cm 1 1 < 145 cm 3.31 (1.6 ± 6.8) 3.08 (1.4 ± 6.9) Central (n 1009) 145 cm 1 1 Figure 1 Association of child stunting with maternal short stature < 145 cm 2.83 (1.8 ± 4.5) 1.98 (1.1 ± 3.5) strati®ed by region (South, North, Central and Mexico City); 1988 South (n 1095) Mexico National Nutritional Survey. 145 cm 1 1 < 145 cm 2.47 (1.9 ± 3.3) 1.57 (1.1 ± 2.2) national average. The central region was similar to the northern region, but states exhibited greater economic a Adjusted by child age (0 ± 6, 6 ± 12, 12 ± 36, > 36), maternal age ( < 19, heterogeneity. Mexico City had the highest proportion of 20 ± 34, 35 ± 49), mother's body mass index, mother's education (none, elementary, high school or beyond), municipality (urban, rural), SES, population living in urban areas and greatest food avail- numbers of people in the household ( < 5, 5 ± 7, > 7), and infections other ability of all the regions. The southern region was mostly than diarrhea or respiratory infections during the last 14 days. rural, with the lowest PCI of all the regions, and a high bStunting was de®ned as height-for-age Z scores < 7 2. density of indigenous people. The National Population Council of Mexico reported that the South had higher the project (SepuÂlveda-Amor et al, 1990). The northern levels of socio-economic marginalization than the North region was predominantly industrial with high density of and Mexico City (CONAPO, 1995), but also was econom- urban population and per capita income (PCI) above the ically heterogeneous due to states that were oil producers Maternal short stature and stunting in children S HernaÂndez-DõÂaz et al 943 (like Tabasco) or had tourism income (like Quintana Roo) et al, 1980; Mosley & Chen, 1984; Kuh & Wadsworth, (Kunz and Cortina, 1986). 1989). The prevalence of stunting increased gradually over Two conclusions may be drawn from regional compar- the child's age, consistent with growth retardation follow- isons in our study, one from the measures of prevalence and ing exposure to adverse external conditions that exist in the other from the measures of association. Firstly, a greater spite of being born with high stature potential (Allen, 1994; prevalence of child short stature in the South may re¯ect a Beaton, 1993; Victoria, 1992). predominance of environmental conditions associated with We acknowledge some potential limitations to the study. child stunting, but also may re¯ect ethnic differences in More appropriate socio-economic indicators, such as growth potential, since this is the region that holds the income or food expenditure, as well as usual dietary largest proportion of the indigenous population (INN, intake or illness patterns and incidence, would have per- 1997). Our results suggest that regional differences in mitted greater control for confounding in the analysis. short stature are due mainly to economic factors, inasmuch These unmeasured factors could have explained part of as those differences were diluted once we adjusted by the association between maternal and child short stature socio-economic variables. That is, the relative importance remaining in our adjusted ®nal model. Hence, the ®nal of region on child short stature diminishes when we adjusted association is a compendium of genetic heritage, substract the socio-economic component from it. In fact, intra-uterine environment, and residual effects of common with a better control for environmental factors, regional external factors insuf®ciently accounted for in the analysis differences would probably disappear. So, as some authors (due to limitations inherent in national surveys that require have suggested, linear growth may even be independent of cross-sectional methods that minimize respondent burden ethnicity in developing regions (Habicht et al, 1974). Our and study cost). Similarly, given the cross-sectional survey results are consistent with other studies at an international design, we were able to account for current environmental level, in which about three-quarters of the variability conditions of the family, but lacked information on the SES among countries in the prevalence of stunting could be of the parents when they were children. However, control explained by socio-economic factors (Frongillo et al, for past SES is unnecessary for the purpose of this study, 1997). since the two pathways through which past SES may affect Secondly, we tried to document evidence of polariza- child short stature (maternal stature and current SES) are tion, looking for heterogeneous patterns of short stature considered in the analysis. Moreover, mobility among within mother ± child pairs across regions. The crude OR social classes and geographic regions, and economic for the association between mother and child short stature changes during the 1980s, will make maternal childhood doubled from South to North. Similarly, a strati®ed analysis SES a poor surrogate for the SES currently affecting her in Chile has found that maternal short stature is associated children. with a two-fold increase in risk of child short stature among Short stature in women has been de®ned previously as a low socio-economic levels, compared with a more than height of less than 145 cm (ACC=SCN, 1992), but the use four-fold increase among high socio-economic levels of arbitrary cutoff points to de®ne low and high maternal (Amigo et al, 1997). In a review of 24 studies, Mueller stature might lead to spurious conclusions. Because of that, found that parent ± child correlation for stature was higher we also ®tted a model with height as a continuous variable. in European than in non-European countries. One potential With a linear term for mother's height, we did not ®nd explanation was that some non-European samples were a statistically signi®cant interaction between maternal experiencing some degree of malnutrition, and the addi- short stature and regions, probably due to a nonlinear tional stress of nutritional de®ciency might increase the relationship. environmental component of stature variability (Mueller, The exclusion of pregnant women and women without 1976). The weaker association in poorer groups may be children may affect the representativeness of the sample if attributed to the overwhelming impact of environmental the nutritional status of these groups differed from the factors on growth, which leaves little space for maternal analytic sample. Nevertheless, we were interested in study- in¯uences. In Mexico, adverse conditions in the South may ing the particular associations described in our main be impairing ful®llment of height potential, obscuring hypothesis (that is association between maternal and child inherited similarities. stature) and not in describing nutritional status at the Furthermore, when we controlled for socio-economic national level, reported elsewhere (Rivera et al, 1995; factors, the association between maternal and child stature SepuÂlveda-Amor et al, 1990). One ®nal limitation of decreased more in poorer regions. Therefore, within that observational studies in general, and in cross-sectional smaller effect of maternal factors in the South, a greater studies in particular, is that causality can never be ascer- proportion still may be explained by common extrinsic, and tained. Although able to identify factors associated with not genetic, determinants. This ®nding is consistent with child stunting, we cannot assume that stunting is a con- previous publications, which found that in less developed sequence of the reported factors, mainly because we are areas much of the variation in height is the result of studying prevalence, and because unmeasured factors may environmental in¯uences on linear growth, especially be confounding the results. However, the relative contribu- those that affect growth in the ®rst few years of life tion of risk factors for child stunting gains strength insofar (Martorell et al, 1977; Martorell & Habicht, 1986). as the results are consistent with other studies. Moreover, In terms of other associations, we found that the variables the temporal order between maternal and child stature can representing socio-economic status showed the strong- be presumed, even when dealing with prevalent outcomes. est association with child stature, compared to the other In summary, we found that the association between covariates considered. The prevalence of stunting was maternal and child short stature was stronger in wealthy associated with lower SES levels, overcrowding in the regions, and that the contribution of socio-economic factors home and lower maternal education. Similar results have to this association was substantial, particularly in poorer been found in other studies (Islam et al, 1994; Cochrane regions. These ®ndings support the relevance of improving Maternal short stature and stunting in children S HernaÂndez-DõÂaz et al 944 external conditions during childhood as means to facilitate Gonzalez-Cossio T, Sanin LH, HernaÂndez Avila M, Rivera J & Hu H ful®llment of height potential. Poor growth re¯ects not only (1998): Length and weight at birth: the role of maternal nutrition. Salud PuÂb. Mex. 40, 119 ± 126. past malnutrition, but also risk of present or future meta- Haas J, Murdoch S, Rivera J, Martorell R. Early nutrition and later bolic problems. Some authors have proposed that syndrome physical work capacity (1996): Nutr. Rev. 54 (Suppl 2), S41 ± S48. X (an entity characterized by obesity, noninsulin-dependent Habicht JP (1974): Estandarizacion de metodos epidemiologicos cuanti- diabetes mellitus (NIDDM), dyslipidemias and cardiovas- tativos sobre el terreno. Bol Of Saint Panam. 76, 375 ± 385. Habicht JP, Martorell R, Yarbrough C, Malina RM & Klein R (1974): cular diseases (CVD)) should be renamed `the small baby Height and weight standards for preschool children. How relevant are syndrome' due to the association between LBW and the ethnic differences in growth potential? Lancet 1, 611 ± 614. development of these diseases in adult life (Barker et al, Hamill PV, Drizd TA, Johnson CL, Roche AF & Moore WM (1979): 1993; Dorner & Plagemann, 1994). Stunting has also been Physical growth: National Center for Health Statistics percentiles. Am. associated with diminished intellectual capacity and J. Clin. Nutr. 32, 607 ± 629. Hauspie R, Vercauteren M & Susanne C (1996): Secular changes in impaired work performance later in life, with the corre- growth. Horm. Res. 45, 8 ± 17. sponding adverse economic consequences (Allen, 1995; HernaÂndez B, Peterson K, Sobol A, Rivera J, Sepulveda U & Lezana MA Pollitt et al, 1995; Haas et al, 1996; Martorell et al, (1996): Sobrepeso en mujeres de 12 a 49 anÄos y ninÄos menores de cinco 1992, 1996; Pelletier et al, 1995). Investments in proper anÄos en MeÂxico. Salud PuÂb. Mex. 38, 178 ± 188. INN (1997): Encuesta Nacional de AlimentacioÂn en el medio rural (ENAL). interventions to alleviate environmental determinants, fun- Mexico City: Instituto Nacional de la NutricioÂn `Salvador ZubriaÂn.' damentally in regions delayed in the epidemiologic transi- Islam AM, Rahman MM & Mahalanabis D (1994): Maternal and socio- tion, would diminish the prevalence of short stature and economic factors and the risk of severe malnutrition in a child: a case- functional consequences later in life. control study. Eur. J. Clin. Nutr. 48, 416 ± 424. Jepson A, Banya W, Hassan-King M, Sisay F, Bennett S & Whittle H. (1994): Twin children in The Gambia: evidence for genetic regulation Acknowledgements ÐThe authors gratefully acknowledge Arthur Sobol of physical characteristics in the presence of sub-optimal nutrition. Ann. Â and Drs Mauricio HernaÂndez-Avila, Kurt Long, Miguel A HernaÂn, Walter Trop. Pediatr. 14, 309 ± 313. Willett and David Wypi, for contributions to database management, Kramer M (1987): Determinants of low birth weight: methodological analysis and interpretation. S HernaÂndez-DõÂaz was a fellow of the Real assessment and meta-analysis. Bull. WHO 65 (5), 663 ± 737. Colegio Complutense, established in cooperation with Harvard University, Kramer M, Olivier M, McLean F, Dougherty G, Willis D & Usher R during the development of this article. (1990): Determinants of fetal growth and body proportionality. Pedia- trics 86, 18 ± 26. References Kuh D & Wadsworth M (1989): Parental height: childhood environment and subsequent adult height in a national birth cohort. Int. J. Epidemiol. ACC=SCN (1992): Second Report on the World Nutrition Situation, Vol I, 18, 663 ± 668. Chap 4, Women's Nutrition Status: ACC=SCN, pp 51 ± 57. Kunz I & Cortina M (1986): Regionalizacion socioeconomico-demo- ACC=SCN (1993): Second Report on the World Nutrition Situation, Vol II, gra®ca y de salud de la Republica Mexicana: un instrumento para la Country Trends, Methods and Statistics, Chap 1, Mexico: ACC=SCN, planeacion e investigacion en atencion primaria a la salud. Mexico pp 85 ± 90. City: Centro de Investigaciones en Salud Publica-Secretaria de Salud. Allen LH (1994): Nutritional in¯uences on linear growth: a general Liang Kung-Yee & Zeger Scott L (1986): Longitudinal data analysis using review. Eur. J. Clin. Nutr. 48 (Suppl), S75 ± S89. generalized linear models. Biometrika 73 (1), 13 ± 22. Allen LH (1995): Malnutrition and human function: a comparison of Long-Dunlap K, Rivera-Dommarco J, Rivera-Pasquel M, HernaÂndez- conclusions from the INCAP and Nutrition CRSP studies. J. Nutr. 125 Avila M & Lezana MA. (1995): Feeding patterns of Mexican infants (Suppl 4), S1119 ± S1126. recorded in the 1988 National Nutritional Survey. Salud PuÂb. Mex. 37 Amigo H, Bustos P & RadrigaÂan ME (1997): Is there a relationship (2), 120 ± 129. between the short stature of parents and their children? Rev. Med. Chile Lowy C (1994): Regulation of intrauterine growth: the role of maternal 125, 863 ± 868. health. Horm. Res. 42, 203 ± 206. Barker D (1994): Outcome of low birthweight. Horm. Res. 42, 215 ± 230. Martorell R (1996): The role of nutrition in economic development. Nutr. Barker DJP, Hanes CN, Fall CHD, Osmond C, Phipps K & Clark PMS Rev. 54 (Suppl. 4), S66 ± S71. (1993): Type 2 (non-insulin dependent) diabetes mellitus, hypertension Martorell R & Habicht J-P (1986): Growth in early childhood in developing and hyperlipidaemia (syndrome X): relation to reduced fetal growth. countries. In: Human Growth: a Comprehensive Treatise, 2nd ed, Vol. 3, Diabetologia 36, 62 ± 67. ed. F Falkner, & JM Tanner, 241 ± 262. New York: Plenum Press. Beaton G (1993): Which age groups should be targeted for supplementary Martorell R, Yarborough C, Letching A, Delgado H & Klein R (1977): feeding? In: Nutritional Issues in Food Aid. A Report Based on an Genetic ± environmental interactions in physical growth. Acta Paediatr. ACC=SCN Symposium, Discussion paper no. 12, Geneva, pp. 37 ± 57. Scand. 66, 579 ± 584. Binkin NJ, Yip R, Fleshood L & Trowbridge FL (1988): Birth weight and Martorell R, Mendoza F & Castillo R (1988): Poverty and Stature in childhood growth. Pediatrics 82, 828 ± 834. Children Linear Growth. Retardation in Less Developed Countries. Bobadilla JL, Frenk J, Lozano R, Frejka T & Stern C (1993): The Nutrition Workshop Series, Vol 14. epidemiologic transition and health priorities. In: DT Jamison, WH Martorell R, Rivera J & Kaplowitz H. (1992): Consequences of stunting in Mosley, AR Measham, & JL Bobadilla (eds.) Control of Disease in earlychildhoodforadultbodysizeinruralGuatemala.Ann.Nestle48,85 ± 92. Developing Countries, New York: Oxford Medical Press, pp. 51 ± 63. Martorell R, Kettel LK & Schroeder DG (1994): Reversibility of stunting: Cochrane S, O'Hare D & Leslie J. (1980): The Effects of Education on epidemiological ®ndings from children in developing countries. Eur. J. Health. World Bank Working Papers, no. 405. Washington, DC: World Clin. Nutr. 48 (Suppl 1), S45 ± S57. Bank. Martorell R, Ramakrishnan U, Schroeder DG & Ruel M (1996). Repro- CONAPO (1995): Indice de marginalidad en MeÂxico 1990. MeÂxico: ductive performance and nutrition during childhood. Nutr. Rev. 54 Consejo Nacional de PoblacioÂn. (Suppl), S15 ± S21. Dorner G & Plagemann A (1994): Perinatal hyperinsulinism as possible Martorell R, Ramakrishnan U, Schroeder DG, Melgar P & Neufeld L predisposing factor for diabetes mellitus, obesity and enhanced cardi- (1998): Intrauterine growth retardation, body size, body composition, ovascular risk in later life. Horm. Metab. Res. 26, 213 ± 221. and physical performance in adolescence. Eur. J. Clin. Nutr. 52 (Suppl. Falkner F, Holzgreve W & Schloo RH (1994): Prenatal in¯uence on 1), S43 ± S53. postnatal growth: overview and pointers for needed research. Eur. J. McGuire JS & Popkin BM (1990). Helping Women Improve Nutrition in Clin. Nutr. 48 (Suppl), S15 ± S24. the Developing World: Beating the Zero Sum Game. Technical Paper Frongillo EA, de Onis M & Hanson KPM (1997): Socioeconomic and 114. Washington, DC: World Bank. demographic factors are associated with worldwide patterns of stunting Mosley WH & Chen LC (1984): An analytical framework for the study of and wasting. J. Nutr. 127, 2302 ± 2309. child survival in developing countries. In: Child Survival: Strategies for Gayle HD, Dibley MJ, Marks JS & Trowbridge FL (1987): Malnutrition in Research. Cambridge: Cambridge University Press. the ®rst two years of life. Am. J. Dis. Child. 141, 531 ± 534. Mueller WH (1976): Parent ± child correlations for stature and weight Golden MHN (1988): Is complete catch-up possible for stunted malnour- among school aged children: a review of 24 studies. Hum. Biol. 48, ished children? Eur. J. Clin. Nutr. 48 (Suppl. 1), S45 ± S57. 379 ± 397. Maternal short stature and stunting in children S HernaÂndez-DõÂaz et al 945 Must A, Dallal G & Dietz WH (1991): Reference data for obesity: 85th Smith DW, Truog W, Rogers JE, Greitzer LJ, Skinner AL, McCann JJ & and 95th percentiles of body mass index (wt=ht2) and triceps skinfold Harvey MA. (1976): Shifting linear growth during infancy: illustration thickness. Am. J. Clin. Nutr. 53, 839 ± 846. of genetic factors in growth from fetal life through infancy. J. Pediatr. Must A, Dallal G & Dietz WH (1991): Reference data for obesity: 85th 89, 225 ± 230. and 95th percentiles of body mass index (wt=ht2) and triceps skinfold Strauss RS & Dietz WH (1997): Effects of intrauterine growth retardation thickness: a correction. Am. J. Clin. Nutr. 53, 773. in premature infants on early childhood growth. J. Pediatr. 130, Omran AR (1971): The epidemiologic transition. A theory of the epide- 95 ± 102. miology of population change. Milbank Memorial Fund Q. 49, Tanner JM (1960): Genetics of the human growth. Symp. Soc. Study Hum. 509 ± 538. Biol. 3, 43. Omran AR (1983): The epidemiologic transition. A preliminary update. Tanner JM, Goldstain H & Whitehouse RH (1970): Standards for chil- J. Trop. Pediatr. 26, 305 ± 316. dren's height at ages 2 ± 9 years allowing for height of parents. Arch. Pelletier DL, Frongillo EA, Schroeder DG & Habicht JP (1995): The Dis. Child. 45, 755 ± 762. effects of malnutrition on child mortality in developing countries. Bull. Ulijaszek SJ (1994): Between population variation in pre-adolescent WHO 73, 443 ± 448. growth. Eur. J. Clin. Nutr. 48 (Suppl), S5 ± S14. Pinstrup-Andersen P, Burger S, Habicht J-P & Peterson KE (1993): Protein UNICEF (1998): State of the World Children Report. New York: Oxford energy malnutrition. In: DT Jamison & WH Mosley (eds.), Disease University Press. Control Priorities in Developing Countries. London: Oxford University Victoria CG (1992): The association between wasting and stunting: an Press. international perspective. J. Nutr. 122, 1105 ± 1110. Pollitt E, Gorman K, Engle PL, Rivera JA & Martorell R (1995): Nutrition Villar J, Smeriglio V, Martorell R, Brown CH & Klein RE (1984): in early life and the ful®llment of intellectual potential. J. Nutr. 125 Heterogenous growth and mental development in intrauterine growth- (Suppl 4) S111 ± S1118. retarded infants during the ®rst three years of life. Pediatrics 74, Ramakrishnan U, Martorell R, Schroeder DG & Flores R (1999): Role of 783 ± 791. integenerational effects on linear growth. J. Nutr. 129, 544S ± 549S. World Health Organization (1983): Medicion del Cambio del Estado Rao RH (1984): The role of undernutrition in the pathogenesis of diabetes Nutricional. Geneva: WHO. mellitus. Diabetes Care 7, 595 ± 601. World Health Organization (1992): Low Birth Weight: A Tabulation of Rivera J, Gonzalez-Cossio T, Flores M, HernaÂndez-Avila M, Lezana MA Available Information, Maternal Health ands Safe Motherhood Pro- & SepuÂlveda-Amor J (1995): De®cit de talla y emaciacion en menores gramme, WHO=MCH=92.2. Geneva: WHO. de cinco anÄos en distintas regiones y estratas en Mexico. Salud PuÂb. World Health Organization (1995a). Physical Status: the Use and Inter- Mex. 37, 95 ± 107. pretation of Anthropometry. WHO Technical Report Series no. 854. SAS (1995): SAS=STAT User's Guide, V610. NC: SAS Institute Inc. Geneva: WHO. Scrimshaw NS (1995): The new paradigm of public health nutrition. Am. J. World Health Organization (1995b). Maternal anthropometry and Public Health 85, 622 ± 624. pregnancy outcomes: a WHO collaborative project. Bull. WHO 73 SepuÂlveda-Amor J, Lezana MA, Tapia-Conyer R, Valdespino JL, Madri- (Suppl). gal H, Kumate J (1990): Estado nutricional de preescolares y mujeres en Mexico: resultados de una encuesta probabilistica nacional. Gac. Med. Mex. 126, 207 ± 225.