Data from 380 anthropometric studies, whose results have been published since 1960, were extracted from the literature and used as the basis for creating curves to represent the growth in stature and weight of populations in different parts of the world. Few truly representative national samples were found. No data were found from most parts of northern Asia whilst data from several countries in other regions were either very sparse or totally absent. A number of samples represent small tribal or other groups which are not typical of the modern populations of their countries. Other samples represent groups which are reasonably typical, if not strictly representative, of either the whole population of their country or important sub groups within it. The extent to which each of these samples should be taken into account when estimating the representative curve for its parent population was a matter of judgement as few publications give precise information on this point.
The curves attributed to each country were therefore approximations but this approximation was also necessary in order to achieve the objective of representing worldwide data by a relatively small number of curves.
The data for each major geographical area were examined country by country and a series of steps, described in the appropriate sections of the text, led to the grouping together of those countries whose data might reasonably be described by similar curves.
Twenty-two separate curves were found to be necessary to represent all the major variations between countries for stature and sixteen were required for weight. These curves are tabulated in table IV and illustrated in the appendix. The major part of most of these curves lay at or below the 50th centile for the United States. Some followed paths nearly parallel to lower U.S. centiles whilst others crossed the centiles.
The number of curves obtained was determined by the level of approximation adopted. If reliable data were available for every country and no approximation were allowed a larger number of curves would result. Alternatively, by greater simplification, the number of curves could be reduced. For example, in the case of stature (see table IV), curves MS 50 and FS 50 may be taken as representative of most populations of European origin but Africa and Latin America could not easily be represented by less than two curves for each sex, e.g. MS 50/5 and either MS 10 or MS 3 for males. It might be reasonable to use MS 5 as a substitute for these last two. Curves FS 20 and 70/10 would represent the statures of most female populations in either Africa or South America. The populations of south-east Asia seem to be quite well represented by curve 20/3 for both sexes. In the case of those populations for which there are no satisfactory data the best approximation would probably be obtained be selecting one of the above curves on the basis of ethnic and geographical background.
The great majority of male weight patterns can be represented by three curves. MW 50 and MW 60/30 describe most populations of European origin whilst MW 10 represents most African and Asian populations. Curve FW 50 represents the females of European ancestry in most parts of the world as well as some privileged groups from Africa and India. Most of the remaining African and Asian populations are reasonably well described by curve FW 20. Populations for which no data exist are most likely to fit one of the above curves, according to sex and origin.
The curves discussed above represent only mean stature and weight according to age. The variation in either parameter within a country is seldom known with accuracy as few samples have been sufficiently representative to give this information. The published standard deviations (Table I) may be judged in the light of the information on the nature and size of samples which is given in the annotated bibliography. Where there are no satisfactory local data, it may sometimes be justifiable to assume that a given mean stature, at whatever age it occurs in the curve appropriate to that country, would be associated with the same standard deviation as applies to that mean stature in the U.S. standards. However, the text draws attention to some cases in which the assumption would not be justified and there may be many others. As much knowledge as possible about the structure of the population should be taken into account before the assumption is made. The standard deviations related to given mean statures are shown in table V whilst table VI allows a similar argument to be adopted in defining the position of the 10th and 90th centiles for weight.
The relationship between stature and weight is not well documented. The differences between regions are quite small but the variation between samples within regions is often great. This variation may be studied in more detail by reference to table I.
The study has revealed considerable variation between populations in both stature and weight. It is not clear how much of this variation is due to adverse environmental factors. Privileged groups in some populations follow growth curves close to the U.S. 50th centile whilst their less privileged compatriots are smaller. However, in view of our very limited knowledge of the genetic structure of these populations, we cannot assume that the mean for the whole population would necessarily reach our 50th centile curves if nutrition were improved. We have even less basis for making this assumption for those countries where even privileged members of the population are small in stature and weight. Thus in general, we can only study populations as they are and any attempt to predict what they might be like in other circumstances is hazardous.