The effect of nutritional status on immune capacity and immune responses in preschool children in a rural community in India*

The effect of nutritional status on immune capacity and immune responses in preschool children in a rural community in India* A. A. KIELMANN,1 I. S. UBEROI,2 R. K. CHANDRA,3 & V. L. MEHRA4 Cell-mediated immune response (CMI) and several aspects of humoral immune status and response were measured and related to nutritional status in preschool children in north India. CMI was measured by means of postvaccinal (BCG) tuberculin sensitivity and leucocytic blast cell transformation. Humoral immune response was measured by means of tetanus antibody production following vaccination with diphtheria-pertussis-tetanus vaccine. Immunoglobulins A, G, and M and complement (C3) were also determined. CMI, serum IgA, and C3 were found to be directly correlated with weight-for-age status. In the developing world, malnutrition and infec- tion account for the majority of deaths among pre- school children. The two have been shown to interact synergistically in that infectious illness brings about a deterioration in nutritional status and under- nutrition worsens infection. Several mechanisms seem to be responsible for this interaction. The most important are a decrease in immune capacity associ- ated with undernutrition and catabolism caused by a febrile illness (1-3). These mechanisms have been verified mostly in children in hospital with extreme forms of undernutrition or in adult volunteers, popu- lations unrepresentative of " rural " child com- munities. The principal objective of this investigation was to determine whether the high prevalence of low body weight in non-hospitalized preschool children in a rural community is associated with a decrease * This study was part of a WHO-sponsored collabora- tive investigation into the effects of malnutrition on immunity. Part of the analysis was supported by a Johns Hopkins University general research grant (GRS 30-25). 1 Assistant Professor, Department of International Health, School of Hygiene and Public Health, The Johns Hopkins University, 615 North Wolfe Street, Baltimore, MD 21205, USA. 'Director, Salvation Army Hospital, Dhariwal, Near Amritsar, Punjab, India. ' Professor of Pediatric Research, Memorial University of Newfoundland, and Janeway Child Health Centre, St. John's Newfoundland, Canada. Formerly at the All-India Institute of Medical Sciences, New Delhi, India. ' Department of Biochemistry, All-India Institute of Medical Sciences, New Delhi, India. in immune capacity in general and specifically in immune response to tetanus immunization. METHODS The study was carried out in nine villages of the former Narangwal Rural Health Research Centre in Ludhiana District, Punjab, India. In these villages, all children up to 3 years of age routinely received food supplements and/or infection control (curative health care and preventive services including diphtheria- pertussis-tetanus (DPT) and BCG vaccinations). The effects of these services on indices of child health were monitored by means of longitudinal surveys.a Cell-mediated immune capacity was determined by measuring tuberculin sensitivity (5 tuberculin units of PPD-S) 45, 235, and 735 days after BCG vacci- nation and by means of leucocytic blast cell trans- formation with phytohaemagglutinin (4). Humoral immune response was determined by measuring a From 1961 to May 1974 the Department of Inter- national Health, Johns Hopkins University, maintained a Rural Health Research Centre (RHRC) in Narangwal village, Ludhiana district, Punjab, North India. A series of research projects focused on crucial health questions important for rural communities and field work was carried out in 25 villages surrounding the RHRC. One of these projects, the Narangwal Nutrition Study, compared the relative effect of malnutrition and infection on child health and development. The present study was carried out within the Nutrition Project. A more detailed description of the Narangwal Projects can be found in: Johns Hopkins Uni- versity, Department of International Health. The Narangwal Population Study: integrated health and family planning services. Baltimore, Maryland, 1975 (mimeographed). 3543 - 477 - BULL. WORLD HEALTH ORGAN., Vol. 54, 1976 478 A. A. KIELMANN ET AL. serum immunoglobulins A, G, and M (5) as well as tetanus antibody levels (by indirect haemaggluti- nation) 45 and 90 days after DPT inoculation. Non- specific humoral immunity was measured by means of serum complement (C3) levels (5). For the determination of tuberculin sensitivity after BCG inoculation, an effort was made to enroll all children newly receiving BCG vaccine over the 2-year time span. Additional selection criteria were: absence of a history of tuberculosis, for the subjects as well as for the other members of the household; no previous BCG vaccination; age less than 36 months; and willingness of the parents to participate in the study. All children were ambulant and appeared " normally active" by local standards. At the beginning of the study, the population con- sisted of 223 children. For subsequent analyses of the effects of weight- for-age on tuberculin sensitivity, two different para- meters were measured: 1. The increase in tuberculin sensitivity from day zero (increment of induration) to provide a more sensitive indicator of the effects of nutrition on tuber- culin sensitivity." 2. The change in the size of induration between days 735 and 235 and between days 235 and 45 to be used also to determine the effects of change in nutritional status on tuberculin sensitivity. The children were allocated to three weight-for-age groups, namely, those at or above the median of all Narangwal study children, those below the median but above the tenth percentile, and those below the tenth percentile. These groups corresponded approxi- mately to 80% or higher, 65% to 79%, and less than 65% of the Harvard median (18), respectively. Selection of children for the determination of serum immunoglobulins, Cs, and tetanus antibody and for the estimation of leucocytic blast cell trans- formation was based primarily on the parents' willingness to have their children participate as well as on weight-for-age. Blood samples were taken from 55 children; arm, head, and chest circumfer- ence, height, and triceps skinfold thickness in relation to age were also determined for these children. The reference employed for weight, skinfold thickness, and head and chest circumferences were the respec- a Using only the absolute size of induration, two children with prevaccinal sensitivities of zero and 5 mm to PPD-S and identical postvaccinal sensitivities of 6 mm each would contribute equally, namely 6 mm, to the postvaccinal indu- ration mean. By using the change in sensitivity, the first contributes 6 mm and the second 1 mm. tive Harvard median curves. For arm circumference, the Wolanski median (6) was used. Vaccines and tuberculins Standardized batches of BCG, DPT, PPD-S, and PPD-B were provided by the World Health Organ- ization. The BCG vaccine was freeze-dried and originated from the Statens Seruminstitut, Copen- hagen, Denmark. The DPT vaccine contained ten binding units of tetanus toxoid and 1.8 mg of A1203 per 0.5 cm3; it was prepared at the Mecnikov Research Institute, Moscow, USSR. The PPD-S contained five tuberculin units and the PPD-B 0.1 mg per 0.1 ml; both originated from the Center for Disease Control, Atlanta, GA, USA. RESULTS Postvaccinal tuberculin (PPD-S) sensitivity All 223 children received BCG on day zero and those that reacted (indurations >6 mm) to either PPD-S or PPD-B on day zero +3 were excluded from the examination of postvaccinal tuberculin sensitivity. The remaining 204 children were then tested for tuberculin sensitivity 45, 235, and 735 days after BCG vaccination. Although both PPD-S and PPD-B were used, the analysis was restricted to post- vaccinal sensitivity to PPD-S. Children below 65% of the Harvard weight median had, for the first two intervals, the lowest proportion of positive increments. For all three intervals, the mean size of the positive increment was lowest in the underweight group. On day 235, the proportion of children with positive increments increased linearly with ascending weight-for-age. Positive increments were shown by 61 Y. of the children below 65% of the wveight median, by 72% of those between 65% and 79%, and by 84% of those at or above 80% of the Harvard weight median (Table 1). In order to determine whether a change in nutri- tional status between any of the three postvaccinal follow-up dates was accompanied by a change in tuberculin sensitivity, changes in induration size be- tween days 235 and 45 and between days 735 and 235 were related to changes in nutritional status.b For the first interval (day 235 to day 45), the pro- portion of children showing an increase in tuberculin sensitivity decreased linearly from the group that b The criterion for change in weight-for-age was arbi- trarily set at one-half standard deviation (6.2 percentage points) of the weight-for-age distribution of all 223 children enrolled in the study. EFFECT OF NUTRITIONAL STATUS ON IMMUNITY 479 Table 1. Incremental change in size of induration fol- lowing administration of 5 TU of PPD-S, according to nutritional status and time after BCG vaccination Increment of induration Interval Nutritional No. Positivestatus a No oiie Mean Standard No. % size deviation Day45-day0 <65% 51 27 53 3.9 5.12 656%-79% 83 61 73 5.5 4.68 > 80 % 60 39 65 5.0 5.55 All 194 127 65 5.0 5.05 Day 235- < 65% 44 27 61 3.6 4.54 day o b 65%-79% 74 53 72 5.1 5.09 > 80% 77 65 84 5.5 4.84 All 195 145 74 5.0 4.90 Day735-dayO <65% 16 14 88 6.0 5.83 65%-79% 38 32 84 7.6 5.12 > 80 % 22 16 73 7.6 7.24 All 76 62 82 7.4 5.70 a Expressed as a percentage of the Harvard weight median. b The test for a linear trend in the proportions of positive incre- ments with ascending nutritional status was highly significant (P = 0.004) for the day 235-day 0 interval. had an improved nutritional status (weight-for-age) to the one whose nutritional status had decreased (Table 2). This trend was further confirmed by simple linear regression. For the day 735-day 235 interval no such association could be found. Leucocytic blast cell transformation following stimu- lation with phytohaemagglutinin Radioactive thymidine uptake, measured as counts/ min in a liquid scintillation counter, were related to nutritional status through a multiple linear regression model which included terms representing differences among the four assay batches. The partial F-value for the curvilinear nutritional components from the fitted multiple regression equation was F 2, 38 = 3.54, giving a probability of a chance occurrence of less than 5 %. Following standardization of all counts to account for batch differences, means and standard deviations of the square roots of counts were determined for the three nutritional groups (Table 3). An increase in means was observed with rising nutritional status. When terms representing age, caste, and prior mor- bidity were added to the model, neither nutritional status, nor age, nor caste, nor morbidity was by itself significantly related to radioactive thymidine uptake when adjusted for each other factor. Nutri- tional status was still positively related to thymidine uptake, however. Since nutritional status, age, caste, and morbidity are highly interdependent,a a con- siderably larger study would be required to decide which of these factors independently influence or are causally related to T cell function. Humoral immune status and response Specific antibody response to tetanus immunization. There was no significant difference in antibody titre a We had found earlier that most underweight children were of low caste, that with increasing age weight-for-age levels decreased, and that previous infectious disease was related to low weight-for-age. Table 2. Effect of a change in nutritional status of at least one half of one standard deviation a on size of induration, comparing day 235 with day 45 Change in size of induration Nutritional No status Percentage Percentage Percentage mm Standard Student b-- negative zero positiveD m deviation " t Probai lity I mproved 17 41 6 53 +1.53 7.69 No change 56 40 20 39 +0.52 5.18 1.83 0.10 Deteriorated 1 1 64 18 18 -3.09 5.61 All 84 43 17 39 +0.25 5.90 a One standard deviation in the distribution of the nutritional status of the 223 children, expressed as a percentage of the Harvard weight-for-age median, was 13.3 %. b The observed linear trend in the percentage of children showing an increase in the size of induration with change in nutritional status was significant (P = 0.07). 480 A. A. KIELMANN ET AL. Table 3. Square root of thymidine uptake count by nutritional status Nutritional No. Square root of count status a NoMen SMean SD > 80% 15 118 46 65%-79% 18 116 35 < 65% 1 3 87 39 a Expressed as a percentage of the Harvard weight median. levels 45 or 90 days after DPT inoculation between children of differing weight-for-age levels. When the relationship between weight-for-age and increase in titre levels (from day zero to either day 45 or day 90) were examined by means of simple linear regression models, slopes for both the 45- and the 90-day titres were positive; statistical significance was, however, not reached. Immunoglobulin A, G, andM levels. The levels of all three immunoglobulins were found to be higher than those of a " normal " North American control population of identical age distribution: IgM levels were 41 %, IgG levels 16%, and IgA levels 14% higher in the study children than in the control population. Children with weight-for-age below 65% of the Harvard reference standard had IgA levels identical to those in the control population (Table 4). The IgA level and the IgA ratio a were found to be positively associated with a number of anthropo- metric indices (weight-, height-, head circumference-, chest circumference-, and arm circumference-for-age), with actual arm circumference, and with skinfold thickness. The IgA level and the IgA ratio were also positively correlated with the IgM level and the IgM ratio. The IgM level, the IgM ratio, the IgG level, and the IgG ratio showed no consistant associations with any of the anthropometric or biochemical indices of nutrition. The IgM level, the IgM ratio, the IgG level, and the IgG ratio were positively correlated with each other; the IgA level, the IgA ratio, the IgG level, and the IgG ratio showed no such association. Following adjustment for sex, caste, age, and vil- lage group by means of multiple linear regression models, the correlation of the IgA level or IgA ratio with all examined anthropometric measurements or indices remained statistically significant (P< 0.05). The strongest association was between IgA ratio and weight-for-age, the weakest between IgA ratio and skinfold thickness. Table 5 shows the mean C3 levels for the study and reference populations in the three weight-for- age groups. The study children had significantly lower complement levels within all three groups than the reference population of identical age distri- bution. Children in the lowest weight-for-age group had less than 50% and those in the two higher a Observed immunoglobulin level divided by the immuno- globulin level of a North American reference population of identical age distribution. Table 4. Levels of immunoglobulin A by weight-for-age status Study population Reference population b Nutritional Standard Confidence Standard Confidence statusa No. Level error of rangeac No. Level error of range c(mg/I) the mean (mg/I) (mg/I) (hemg/I) (mg/I)(mg/I)(m/ > 80 % 19 520 60 460-580 19 430 60 370-490 65 %-79 % 19 490 50 440-540 19 440 60 380-500 < 65% 15 400 60 340-460 15 390 90 300-480 All 53 480 30 450-510 53 420 40 380-460 a Expressed as a percentage of the Harvard weight median. b The reference population was matched for age only. c Mean ± 1 standard error of the mean. EFFECT OF NUTRITIONAL STATUS ON IMMUNITY Table 5. Levels of complement (C3) by weight-for-age status Study population Reference population b Nutritional Standard C Standard status a No Level error of Confidence Level error of Confidence(m/Iohemen range c No. Lvl erro range c(mg/) th(e m/I) (mg/I) (mg/I) the mean (mg/l(mg/0 ~~~~~~~(mg/I) (gI > 80 % 19 910 70 840-980 19 1450 80 1370-1530 65 %-79 % 19 980 70 910-1050 19 1460 80 1380-1540 < 65 % 15 690 60 630-750 15 1500 100 1400-1600 All d 53 870 40 830-910 53 1470 50 1420-1520 a Expressed as a percentage of the Harvard weight median. b The reference population was matched for age only. c Mean ± 1 standard error of the mean. d The difference in C3 levels between the study and reference populations was significant (P < 0.001). nutritional groups between 60% and 70% of the complement levels of the reference population. C3 levels or ratios were also positively correlated with several other anthropometric indices (weight-, chest circumference-, and arm circumference-for- age). It is interesting that most morbidity variables showed a negative relationship with the C3 level or ratio, although none reached statistical significance. Following adjustment for age, caste, sex, cohort, and batch differences by means of multiple linear regression models, which also contained terms for the anthropometric variables under examination, C3 levels and C3 ratios continued to be significantly as- sociated with all anthropometric variables (P< 0.025). Visual examination of residual plots of IgA level, IgA ratio, C3 level, or C3 ratio against anthropo- metric indices, within the framework of the complete model, showed residuals distributed symmetrically around zero throughout the range of anthropometric measurements; this confirmed the linearity of the observed relationship within the examined range. DISCUSSION Following BCG vaccination, improved nutritional status was positively correlated with increments in the mean size of the induration induced by 5 tuber- culin units of PPD-S. This increase seems to be a result of the relative proportion of overall converters within the different nutritional groups rather than a measure of the real induration sizes, for we found no significant differences in the mean size of indu- ration for those with reactions larger than 5 mm or larger than 10 mm in the three nutritional groups (7), which suggests that sensitivity to tuberculin following BCG vaccination develops on an " all-or-nothings" basis. The finding that children with improved nutritional status showed an augmentation whereas those with a lower nutritional status experienced a decline in their sensitivity to tuberculin confirms the impression of other investigators (8) that postvaccinal tuberculin sensitivity cannot be used as an indicator the effective- ness of BCG vaccination. Twenty-four months after BCG vaccination, a decrease or improvement in weight-for-age was no longer related to changes in the increment of indu- ration. If anything, children whose weight-for-age had decreased from day 235 to day 735 showed a larger mean increase in increments than those whose weight-for-age had improved; this suggests that the initial induration response to BCG vaccination may be slower in poorly nourished children but that the long-term result may be similar. Another possible explanation may lie in the obser- vation that children that were underweight on day zero had significantly improved their arm circum- ference-for-age by day 735 (7). Arm circumference largely reflects muscle mass rather than subcuta- neous fat. It seems, therefore, that although neither weight-for-age nor skeletal growth caught up with those of " normal " children (as demonstrated by the difference in head circumference-for-age), the underweight children adapted to a low nutritional intake by reducing their linear growth. Once this compensation had been made, normal or quasi- 481 A. A. KIELMANN ET AL. normal muscular development and immune response serumed. Tetanus antibody production seems not to have been impaired in the underweight group, and this confirms earlier data on hospitalized malnourished children (9). As noted by G. Edsall (personal com- munication, 1975), this agrees with the results of a more extensive investigation in the Gambia where, as part of the overall WHO-sponsored effort, the same reagents were used and essentially the same research protocol was followed as in our study. Of the three immunglobulins examined, only IgA was directly related to a variety of anthropometric indices of nutrional status. This finding suggests that IgA production is dependent on the degree of infec- tion and nutritional status. In this study only serum IgA was measured. However, our earlier data on hospitalized malnourished children showed reduced secretory IgA and specific IgA antibodies to two viral antigens (10). IgA as a double molecule (dimer) in the external secretions may be largely responsible for the inactivation of respiratory and gastrointes- tinal antigens or may prevent their attachment to the mucosal epithelial cells-a critical prerequisite for pathogenicity and symptomatology. Deficient IgA production in the underweight child may thus partly explain the generally high levels of upper res- piratory infections and perhaps the diarrhoeal dis- ease so common in developing countries where nutritional levels are usually depressed. IgA has been shown to be an active antiviral agent in the respiratory tract (11). Our results contrast with those of other investi- gators in that, in our survey, malnourished children did not show higher IgM or IgA levels compared with well-nourished ones. IgG levels were higher in children below 80% of the Harvard weight standard than in those above, but the differences were not statistically significant. Several explanations may account for these dif- ferences. First, most other studies have dealt with severely malnourished, hospitalized children who, in addition to malnutrition, suffered from concomitant infections. Our study children were considered " nor- mally active" and, at the time of examination, did not suffer from clinically overt infectious disease. Second, many studies we examined compared im- munoglobulin levels in malnourished children of the developing world to immunoglobulin levels of a reference population in the USA or in Europe who suffer milder and overall considerably less infectious disease. Alvarado & Luthringer (12) suggested that high circulating IgA levels may result from seepage of secretory IgA into the systemic circulation because of damage to the intestinal mucosa in the severely malnourished child, rather than from an absolute increase in IgA production. This hypothesis seemed to be supported by the results from other surveys on " normal " child communities where IgA levels were in general found not to be raised (13, 14), but was refuted by the absence of the reaction of sera of malnourished children to antiserum to the secretory component of IgA (10). Recently, a group of investigators working in Colombia found statistically significant reductions in IgA concentrations in the tears of malnourished (Gomez grade II and III) children. They suggested that the reduced IgA activity might be due to reduced synthesis (15). In ours, as in another community survey (16) and in our hospital-based studies (17), C3 levels were found to be directly dependent on levels of nutrition as well as significantly lower than in European refer- ence populations. This may be a phenomenon both of reduced production owing to low nutritional status and of increased utilization caused by frequent and prolonged infection. Lowered production, in turn, may contribute to lowered resistance to invading organisms and increased severity of illness. Of special significance was the observation that selected indicators ofimmune capacity-postvaccinal tuberculin sensitivity, leucocytic blast cell trans- formation, and levels of serum IgA and C3-were significantly reduced in a large proportion of " nor- mal" children. Until recently, only the extremes of undernutrition-kwashiorkor and marasmus-were thought to decrease immune capacity to levels of clinical importance, hence the focusing of investi- gations on the severely malnourished, hospitalized child. On average 10%, and during the summer up to 20%, of preschool children in the area we studied were at or below a nutritional level shown to be associated with diminished immune capacity and normal children in this area lost up to 10% of their body weight following immunization with live agents (19); for these reasons, it is extremely import- ant to plan and phase child health services so as to maximize the beneficial and minimize the deleterious effects. 482 EFFECT OF NUTRITIONAL STATUS ON IMMUNITY 483 ACKNOWLEDGEMENTS A major portion of the field work was done by Ms Roselyn Daniels. Dr P. Talwar of the All-India Institute of Medical Sciences made it possible for us to perform the immunological assays. In Baltimore, Dr W. Blot and Mr G. Smith did most of the programming, and Dr H. Abbey, Dr G. W. Comstock, Dr G. G. Graham, Dr C. McCord, and Dr C. E. Taylor greatly assisted in the interpretation of the results. RtiSUME EFFET DE L'ETAT NUTRITIONNEL SUR LA CAPACITt IMMUNITAIRE ET LES IdPONSES IMMUNITAIRES CHEZ LES ENFANTS D AGE PRPSCOLAIRE DANS UNE COMMUNAUTI RURALE EN INDE Chez des enfants d'age prescolaire appartenant A une collectivit6 rurale du Punjab, Inde septentrionale, on a mesure divers aspects de l'immunit6 A mediation cellu- laire et A mediation humorale; on a cherch6 A etablir des relations entre ces resultats et le poids des enfants en fonction de leur Age. L'immunite A mediation cellu- laire a et6 determin6e par la mesure de l'hypersensibilite A 5 unit6s de tuberculine (PPD-S) 45, 235, et 735 jours apres vaccination par le BCG et par la mesure de la transformation blastique des leucocytes. En ce qui conceme la capacite immunitaire humorale, elle a ete mesuree par la production d'anticorps antitetaniques 45 et 90 jours apres inoculation de vaccin DT-Coq, de meme que par le dosage des immunoglobulines A, G et M ainsi que du compl6ment (C3). La sensibilite a la tuber- culine, traduite par l'augmentation de la taille de l'indu- ration A divers intervalles apres la vaccination BCG s'est rev616e en rapport direct avec le poids en fonction de l'age: une diminution ou une augmentation du poids en fonction de l'Age etait suivie par une diminution ou une augmentation de la reaction d'hypersensibilite. La trans- formation blastique des leucocytes n'a ete trouvee reduite que chez les enfants dont le poids en fonction de I'Age 6tait inferieur A 65% de la mediane des < normes * de Harvard. En ce qui conceme les parametres de l'immunite humorale, les IgA et le compl6ment (C3) seriques se sont reve16s en relation directe avec le poids en fonction de l'age. La production d'anticorps antitetaniques et d'IgG ainsi que d'IgM seriques s'est montree sans relation avec l'etat nutritionnel. L'ampleur du probleme de sante 1i6 A la nutrition est illustree par le fait suivant: dans la collectivite examinee, parmi les enfants d'Age prescolaire, la proportion de ceux dont le poids en fonction de l'age 6tait egal ou inferieur A 65% de la mediane des # normes * de Harvard etait de 10% en moyenne et pouvait atteindre 20% au cours de P'kte. A ce niveau ou en dessous, l'immunite cellulaire ainsi que les taux d'immunoglobuline A et de compl6- ment (Cs) presentaient une diminution evidente. REFERENCES 1. CHANDRA, R. K. In: Brent, L. & Holborow, J. ed. Progress in immunology IL Amsterdam, North Hol- land, 1974, pp. 355-358. 2. FAULK, W. P. ET AL. Tropical diseases bulletin, 72: 89 (1975). 3. CHANDRA, R. K. World review of nutrition and diet- etics, 25: 166 (1976). 4. NOWELL, P. C. Cancer research, 20: 462 (1960). 5. MANcINI, G. ET AL. Immunochemistry, 2: 235 (1965). 6. BURGESS, H. J. L. & BURGESS, A. P. Journal of tropical pediatrics and African child health, 15: 189 (1969). 7. KIELMANN, A. A. Immune responses and weight fluctuations following immunizations in ruralpreschool children. Doctoral dissertation, Johns Hopkins Uni- versity, 1976. 8. COMSTOCK, G. W. & EDWARDS, P. A. Scandinavian journal of respiratory diseases, 53: 207 (1972). 9. CHANDRA, R. K. Journal of pediatrics, 81: 1194 (1972). 10. CHANDRA, R. K. British medical journal, 2: 583 (1975). 11. BELLANTI, J. A. Pediatrics, 48: 715 (1971). 12. ALVARADO, T. & LUTHRINGER, D. G. Clinical pedi- atrics, 10: 174 (1971). 13. ANDERSON, R. I. & BUCK, A. A. Zeitschrift fiir Tropenmedizin und Parasitologie, 24: 447 (1973). 14. GOLUBJATKOV, R. & STEADMAN, M. American journal of epidemiology, 95: 542 (1972). 15. McMURRAY, D. N. ET AL. Federation proceedings, 35: 588 (1976). 16. NEUMANN, C. G. ET AL. American journal of clinical nutrition, 28: 89 (1975). 17. CHANDRA, R. K. Archives of disease in childhood, 50: 225 (1975). 18. STUART, H. C. & STEVENSON, S. S. In: Nelson, W. ed. Textbook of pediatrics, 7th ed. Philadelphia, W. B. Saunders, 1959. 19. KIELMANN, A. A. American journal of clinical nutri- tion, 30 (1977) (in press).