Contents - Previous - Next

This is the old United Nations University website. Visit the new site at


Secular trends in adults


The comparison of adults' heights measured at two different times (1974 1 year and 1988) shows a decrease in height during this interval, beginning between the ages of 41 and 50 years in men and between 31 and 40 years for women (table 1). As seen in figure 1 (see FIG. 1. Mean change in the stature of adults between measurements in 1974 and 1988, by sex and age group), the mean reduction over 10 years increased with age but in a non-linear fashion: in men, for instance, the reduction was approximately 0.5 cm per decade between the ages of 41 and 60 but was about three times larger during the next two decades (- 1.38 between 51 and 60 years and—1.60 between 61 and 70). As expected, linear regression analysis of height change on age provided a poor fit of the data, and the usual linearizing transformations (logarithmic and exponential) did not improve the fit.

TABLE 1. Unadjusted and age-adjusted heights of adults (centimetres) by age group and sex

Age in 1988 (years) N Height Change

per decade

Age-adjusted height
1974 1 yr 1988-89 Method 1a Method 2b
<41 26 160.71 (5.23) 160.76 (5.42) +0.05 (0.69) 160.76 (5.42) 161.60 (5.42)
41-50 79 160.75 (5.79) 160.04 (5.48) -0.49 (0.81) 160.48 (5.48) 161.68 (5.49)
51-60 71 159.32 (5.51) 158.53 (5.47) -0.55 (0.90) 159.52 (5.47) 161.10 (5.42)
61-70 37 161.43 (5.27) 159.44 (5.31) -1.38 (0.89) 161.81 (5.31) 162.93 (5.29)
71 11 159.00 (6.31) 156.71 (5.58) -1.60 (1.29) 160.68 (5.58) 161.31 (5.59)
<41 110 149.49 (5.11) 148.27 (5.21) -0.85 (0.86) 149.12 (5.21) 149.00 (5.32)
41-50 102 148.78 (5.34) 147.28 (5.33) -1.04 (0.90) 149.17 (5.33) 149.02 (5.40)
51 62 148.55 (5.14) 146.30 (5.34) -1.55 (1.19) 149.74 (5 34) 149.06 (5.34)

a. Longitudinal method.
b. Cross-sectional method.

The age-specific estimates of height changes per decade were used cumulatively to correct the 1988 heights for the effects of ageing. The correction factors used for each age group and sex are presented in table 2. The age-adjusted heights were then regressed against year of birth to examine the presence of secular trends in the heights of men and women respectively. Figure 2 (see FIG. 2. Secular trends in the stature of adults, corrected for age-associated shrinkage by longitudinal and cross-sectional methods)presents the age-adjusted heights by birth cohort and clearly shows the absence of change over time for both sexes. The regression analysis confirmed that birth year was not a significant determinant of age-adjusted height in adults.

TABLE 2. Correction factors used to adjust the 1988-1989 height of adults by the longitudinal method

Age group (years) Age-corrected height
31-40 no adjustment
41-50 observed height + (-0.05 + 0.49) = 0.44
51-60 observed height + (-0.05 + 0.49 + 0.55) =0.99
61-70 observed height + (-0.05 + 0.49 + 0.55 + 1.38) = 2.37
>=71 observed height + (-0.05 + 0.49 + 0.55 + 1.38 + 1.60) =3.97
31-40 observed height + (0.85)
41-50 observed height + (0.85 + 1.04) = 1.89
>=51 observed height + (0.85 + 1.04 + 1.55) = 3.44


The cross-sectional method using subischial height [3] provided age-adjusted heights very similar to those obtained by the longitudinal method. The regression of age-adjusted heights on birth year also demonstrated the absence of secular trends between 1905 and 1959. The coefficients obtained for the ageing effect (b1 in equation 2) were -0.10 cm per year for men and -0.11 cm per year for women, and the coefficients for subischial height were 1.254 and 1.32 for men and women respectively.

Analysis of the change in women's subischial height over time revealed a constant and minimal decrease of approximately 0.05 cm per decade. The assumption that the shrinkage associated with ageing originated mainly from the upper portion of the body was thus verified in this sample.

Head circumference

Figure 3 (see FIG. 3. Secular trends in the head circumference of adults) shows the absence of any secular trend in adults' head circumference (confirmed by the regression analysis of head circumference on birth year). This corroborates the absence of a secular trend in height found in this population.


Secular trend in the length of three-year-olds

A positive and statistically significant linear trend in the length of three-year-olds between 1968 and 1988 was observed for both treatment groups (table 3 and FIG. 4. Secular trends in the length of three-year-old children. by supplement type (The estimates and confidence-interval limits of the linear and quadratic contrasts used to test secular trends were as follows. Atole: linear, 1.85 0.61 [1.24, 2.46]; quadratic, -0.54 0.98 [-1.52, 0.44]. Fresco: linear, 2.71 0.61 [2.10, 3.32]; quadratic, 1.33 0.99 [0.34, 2.321].)). In the fresco group, the quadratic term was also statistically significant, which corresponds to the slight decrease in length observed between 1969 and 1974, followed by a positive linear trend thereafter. In addition, during the supplementation years (1969-1977) the length of children from the atole villages was consistently greater than that of children from the fresco villages, and this trend disappeared after the supplementation period. In fact, the differences were seen only during the supplementation years, and they were not statistically significant at baseline (1968) and in 1988. (The difference between treatments was tested by including treatment in a regression model of length on age and age squared. Treatment was not statistically significant in either 1968 or 1988.)

TABLE 3. Length of three-year-old Guatemalan children (centimetres) by treatment and year of data collection

  Atole Fresco
N Mean SD N Mean SD
1968a 196 85.14   179 84.70  
1969-70 82 85.53 3.69 78 84.82 4.50
1971-72 100 86.58 4.17 101 84.63 4.02*
1973-74 117 86.09 3.56 110 84.46 3.94*
1975-76 117 86.86 3.83 101 84.69 4.11*
1977 35 86.55 3.54 39 85.23 3.66
1988 344 87.62   363 87.99  

a The means in 1968 and 1988 were estimated by regressing the length of children 9-60 months old on age and age squared, using ordinary least squares. Separate regressions were done for each year (1968 and 1988) and treatment group (atole and fresco).
* Differences between the atole and fresco groups statistically significant at the p < 05 level.

Overall, the length of three-year-olds increased by 2.5 cm (atole) and 3.3 cm (fresco) in the 20-year period.


The present analysis of secular trends covers two different populations and time periods: adults born between 1905 and 1959 and three-year-old children born between 1965 and 1985. In the adults, no secular trend was observed for either age-adjusted height or head circumference, whereas for children's length a significant positive trend was seen in the four villages studied. There were no reasons to expect similar trends for adults and children, not only because the time periods covered were different but also because the children were measured during an intervention trial that was intended to affect their growth. Therefore, the remainder of the discussion proceeds separately for adults and children.

Secular trends in adult height and head circumference

The availability of two measurements of adults' height with an approximate interval of 14 years allowed us to estimate the age- and population specific effect of ageing on height reduction. The importance of using an age-specific correction factor was highlighted by the fact that reduction in stature was not linearly associated with age, particularly in men. Previous studies have documented a similar non-linear relationship [7, 10]. Despite this, in the absence of longitudinal information, most studies of secular trends use cross-sectional data and a correction factor that assumes a linear relationship between shrinkage and ageing.

Our study also illustrates the importance of using a population-specific correction factor, since both the magnitude and the timing of the age-associated shrinkage in our population differed from previous reports. In men, the magnitude of the cumulative ageing effect on height reduction was higher than that in previous longitudinal studies. We estimated a cumulative reduction of approximately 4 cm between 30 and 70 years of age, which was slightly higher than previously documented levels of 3.6 cm [7], 3 cm [10], and 2.8 cm [16]. In our sample of women, the height reduction occurred earlier (before the age of 40) and was of larger magnitude (3.44 cm between 20 and 50 years) than previously reported. Other studies have found the reduction starting only in the early fifties [10] and being approximately 1 cm by 50 years of age [7]. Differences in the timing and magnitude of the height reduction have been related to population differences in initial height, socio-economic factors, bone mineralization, and osteoporosis and vertebral fractures [8, 17, 18]. Low socio-economic status, poor bone mineral status, and a high frequency of osteoporosis and fractures and deformities caused by hard physical work are potential explanations for the higher and earlier stature reduction in our Guatemalan sample, particularly in women.

The comparison of our longitudinal method with the cross-sectional method commonly used to correct for age-associated height reduction provided surprisingly similar results: the age-adjusted means were very comparable, and both methods agreed in showing the absence of a secular trend in the four villages. These results are encouraging, suggesting that little or no bias is introduced by using the cross-sectional method based on the assumption of a linear relationship between stature loss and ageing, even though the assumption is incorrect. Considering the scarcity of available longitudinal data sets for the analysis of secular trends, it is useful to know that cross-sectional data sets can be used with some confidence.

Another assumption of the cross-sectional method that could be tested in the women of the present study was the absence of reduction in subischial height associated with ageing (absence of shrinkage of the lower portion of the body). Only a minimal decrease was observed, on the order of approximately 0.05 cm per decade, and it was independent of age. This difference could be attributable to a systematic measurement bias in the 1988 data collection, since a single anthropometrist took all the measurements then, whereas a team was used for the 1974 measurements. If this was the case, systematic bias could cause an overestimation of the shrinkage effect on total height of approximately 0.05 cm per decade for each age group, which would cumulate to an overestimate of 0.15 cm for the older subjects (51 years old or over). The magnitude of this effect is small and represents only about 4% of the estimated cumulative shrinkage experienced by women in this age group.

This analysis could not be done for men because of the unavailability of sitting height information for 1974. However, the slight increase in height (0.05 cm per decade) observed between 20 and 30 years of age suggests that systematic underestimation of height in 1988 did not occur. In this survey, men and women were measured by a different anthropometrist. It is thus possible that the individual assigned to women systematically underestimated height, whereas the one assigned to men did not. If such a bias existed, however, its overall impact would be negligible, as discussed above. Nevertheless, it would have been interesting to examine shrinkage of the lower part of the body in our sample of men since a previous study (of men only) did document that some shrinkage occurred between the ages of 25 and 82 years, and that the amount increased with age until the age of 75 [16].

The absence of secular trends in adult stature between 1905 and 1959 was as documented for other developing countries [4-6]. Although some improvements in global health and nutrition indicators occurred in Guatemala during this period, improvements in environmental sanitation, health care, and food security at the household level appear to have been insufficient to produce an increase in adult stature over time.

In summary, our results suggest that all of the apparent increase in adults' height between 1905 and 1959 was due to an ageing effect rather than a secular trend. The findings were similar whether a longitudinal or a cross-sectional approach was used to correct height for the effects of ageing. The absence of a secular trend in head circumference corroborates the results obtained for height.

Secular trend in the length of three-year-olds

For three-year-olds born between 1965 and 1985, a positive and significant linear trend in length was observed. Interestingly, no differences were seen between children from the atole and fresco villages before the intervention and 12 years after its termination, but clear differences in favour of the atole villages were found throughout the supplementation period. This suggests that the higher rate of length increase seen between 1969 and 1977 in the atole villages was caused by the intervention. Because of the absence of data between 1977 and 1988, however, it is impossible to quantify the rate at which the deceleration in the changes in length occurred in the atole villages after the supplementation ended.

Interpretation of the results on secular trends in children was complicated by the absence of data on control villages or on the region as a whole. Such information would be helpful to determine whether the trend seen in the fresco villages during the intervention period (specifically between 1973 and 1977) was due to the health services offered by INCAP in all villages or whether it reflected overall improvements in standards of living in the region. It would also facilitate the interpretation of the positive trend seen after the intervention trial. Although it is more likely that the trends observed during this period are attributable to overall improvements in schooling, infrastructure, and availability of services in the villages [19], a possible long-term effect of INCAP's presence cannot be ruled out. Quantification of this effect is impossible in the absence of information on other villages from the region.

Overall, the estimated increase in the length of three-year-olds over the 20-year period ranged between 2.5 and 3.3 cm. This improvement represents only approximately 27% of the total deficit in length (11 cm) seen in these children in 1968, compared with the fiftieth percentile of the reference standards [20]. Effects of this magnitude were obtained at the community level with a calorie supplementation of approximately 10% of the recommended daily intake [21] provided during the first three years of life.

Contents - Previous - Next