Coffee intake has been associated with several adverse nutritional consequences in both animal models and in human studies. These include effects on birthweight (Golding 1995, Muñoz et al. 1986 and 1988, Narod et al. 1991), calcium metabolism (Hasling et al. 1992, Nolen 1981), thiamin status (Vimokesant et al. 1982), iron and zinc absorption (Hallberg and Rossander 1982, Morck et al. 1983, Pecoud et al. 1975) and metabolism (Muñoz et al. 1986 and 1988). Although caffeine is one component that may be responsible for some of these effects, many other components in coffee can alter physiological functions. For example, the influence of coffee on iron absorption is likely due to the high concentration of polyphenols (tannins) and other substances such as chlorogenic acid (Brune et al. 1989).

Most previous research on the nutritional impact of coffee has been conducted among adults, because the intake of coffee is highest in this age group. In Central America, however, coffee is also a common beverage for children (Flores 1976, Ministerio de Salud Pública 1981, Valverde et al. 1975). In Guatemala, for example, it is one of the first liquids given to infants, beginning as early as 2 mo of age (Quan de Serrano et al. 1991). By contrast, in North America, coffee is not considered an appropriate beverage for children, in part because it is believed to impair growth. This hypothesis, however, has never been adequately tested. Two preliminary reports suggested that coffee-drinking children in Zaire (Kuvibidila et al. 1992) and Guatemala (Schroeder et al. 1994) differed in anthropometric indices from their peers who did not drink coffee. Because of potentially confounding variables, however, it is difficult to draw causal inferences from observational studies.

To determine whether there is a causal association between coffee consumption and nutritional status, an experimental design is necessary. It is not ethical to assign at random children who do not consume coffee to be given coffee or not, but it is possible to eliminate coffee from the diet of those who already consume it. The objectives of the present study were to determine the effects of discontinuation of coffee intake among Guatemalan toddlers 12-24 mo of age on the following outcomes: 1) growth, 2) morbidity, 3) iron status and other hematological indices, and 4) behavioral outcomes, including sleep patterns and cognitive development. This paper presents the results for growth and morbidity; the other outcomes will be reported elsewhere.

SUBJECTS AND METHODS

The selection criteria for the study were that the children be 12-24 mo of age, coffee consumers for at least 2 mo, with an intake of more than 90 mL/d, and likely to be iron deficient at base line. Children were considered likely to be iron deficient if hemoglobin (Hb) was 115 g/L, hematocrit (Hct) was 0.35, or zinc protoporphyrin-heme ratio (ZPP/H) was >80 µmol/mol heme. These criteria for Hb and Hct are adjusted upward relative to the standard cutoffs of 110 g/L and 0.33, respectively, to correct for the altitude of Guatemala City (1300-1500 m) (Dirren et al. 1994).

Children were categorized into those who were initially frankly anemic (A) (defined herein as Hb < 105 g/L) or those who were either mildly anemic or not anemic [HB  105 g/L; herein labeled as "nonanemic" (NA)]. After this stratification, children were randomly assigned to the coffee or substitute group. Anemic children were provided with iron supplements (15 mg Fe/d as ferrous sulfate in Fer-N-Sol, Mead Johnson, Evansville, IN) for 2-3 mo because it was not considered ethical to leave this group untreated during the 5-mo study. The status of the "nonanemic" children was not considered severe enough to warrant immediate iron supplementation.

Recruitment was done by door-to-door canvassing in selected low-income neighborhoods of Guatemala City. At the time of recruitment, children were screened for eligibility, a blood sample was taken and anthropometric measures were completed. Eligible subjects were stratified into anemic and "nonanemic" groups, and random assignment was performed in blocks of 20, using a table of random numbers. Subjects were visited in the home shortly after recruitment to obtain demographic data, evaluate dietary intake using a food-frequency questionnaire and assess base-line sleep patterns. At this visit, subjects in the intervention group were given their first 2-wk supply of the coffee substitute and instructed in its use. Every 2 wk, all subjects were visited in the home to assess the child's intake of fluids (including use of the substitute, if assigned to that group), morbidity since the previous visit and sleep patterns. At the completion of the 5-mo intervention period, the final blood sample was collected, anthropometric and food frequency assessments were repeated, child behavioral development was assessed and an "exit" interview was conducted with mothers to determine their reaction to the intervention. Socioeconomic and demographic characteristics of the subjects will be reported elsewhere.

Because some brands of coffee in Guatemala contain substantial amounts of noncoffee fillers such as roasted grains, samples of each brand were analyzed for caffeine content as a marker of the "strength" of the coffee. Caffeine content was measured by HPLC. The results were used to calculate another index of coffee dose based on caffeine concentration.

Table 1. Characteristics of subjects, by initial anemia status and intervention group [View Table]

In categorizing the morbidity data, diarrhea was defined as more than 3 liquid or semi-liquid stools per day, and respiratory illness was defined as presence of any respiratory symptoms (cough, runny nose), with or without accompanying fever. Data are expressed as the percentage of days ill during the 5-mo intervention period.

RESULTS

Characteristics of children who completed the study are shown in Table 1. There were no significant differences between treatment groups (coffee vs. substitute) with respect to child age, sex, birthweight, initial anthropometric status, percentage still breast-fed, age when coffee was first introduced, coffee preparation methods, initial coffee or caffeine intake, or initial or final hematological indices. The average length-for-age Z-score at the beginning of the study was more than 2 SD below the NCHS median, indicating a high prevalence of stunting. The average age at which coffee was introduced was 7-8 mo. Very few of the children received milk with their coffee. Initial coffee intake averaged 160-200 mL/d [0.35-0.51 g/(kg·d) of dry coffee] and did not differ significantly among groups. Caffeine concentration of the various brands used ranged from 7 to 56 mg/g dry coffee. Caffeine intake at the beginning of the study averaged 55-107 mg/d [6.1-11.0 mg/(kg·d)], depending on the subgroup.

Table 2 shows the frequency of consumption of foods from various food groups before and after the intervention period. There were no significant differences between treatment groups in initial or final intakes except within the anemic subgroup, in which children in the coffee group had a higher initial intake of beans than children in the substitute group. Similarly, the change in frequency of consumption of each food type during the 5-mo study period was generally similar between treatment groups; the only exceptions were the change in bean consumption within the anemic subgroup (decreased in the coffee group, increased in the substitute group) and the change in bread consumption (increased in the coffee group, decreased in the substitute group).

Table 2. Food intake by children before and after the intervention period, by initial anemia status and intervention group [View Table]

Table 3. Weight and length gain of children during the 5-mo intervention period, by initial anemia status and intervention group [View Table]

The growth data were re-examined after excluding children whose initial coffee intake was <100 mL/d, and who therefore would not be expected to show much of a response to its discontinuation. The remaining 96 children represented 70% of the total sample. Among these 96 children, there were no significant differences between treatment groups in initial characteristics (socioeconomic, demographic, child age, sex, birthweight, Z-scores or coffee intake) or dietary intake. Within the anemic subgroup, weight gain was 46% higher in those given the substitute (n = 11) compared with children in the control group (n = 18) (P < 0.05). The difference was not significant in the "nonanemic" subgroup. For length gain, there was a marginally significant (P = 0.07) 22% increase in the substitute group compared with controls when both "nonanemic" and anemic subgroups were combined. The results were similar when other variables of coffee or caffeine "dose" (e.g., grams dry coffee or milligrams caffeine, either per day or per kilogram per day) were used to exclude children whose initial intake was relatively low. The criterion based on volume per day (>100 mL/d) was chosen because there is less potential error in the values, given the uncertainty in recall data for coffee preparation method.

Additional statistical analyses were performed to evaluate whether the growth results would differ if we excluded children who were considered noncompliers in either the substitute or coffee group (as defined above), or tested for interaction effects with child gender, initial anthropometric status or use of iron supplements. None of these analyses substantially altered the results described above.

Table 4. Morbidity of children during the 5-mo intervention period, by initial anemia status and intervention group [View Table]

To investigate the possible reasons for the opposite trends in the morbidity outcomes in the "nonanemic" vs. anemic subgroups, we completed additional analyses controlling for iron supplementation during the intervention period. It has been hypothesized that iron supplementation may increase the risk of certain infections (Murray et al. 1978 and 1980, Weinberg 1974). As mentioned in Subjects and Methods, the anemic children were supposed to receive iron supplements for 2-3 mo during the intervention; most (though not all) of their parents did in fact administer the supplements. Some children in the "nonanemic" group were also given iron supplements by their parents even though these were not provided by the study. When we included iron supplementation as a variable in the statistical models, it was positively associated with illness prevalence (P < 0.01); this effect was greater in the anemic subgroup than in the "nonanemic" children (significant interaction term, P = 0.05). Controlling for iron supplementation, illness prevalence was still significantly lower in the substitute group than in controls among the "nonanemic" children, but there was no longer even a marginal difference by treatment group among the anemic children (P > 0.10).

Elevated C-reactive protein values (>10 mg/L) are considered to be reflective of recent or current infection or inflammation and therefore may be an indirect indicator of illness prevalence. The percentage of children with elevated C-reactive protein levels was similar between groups prior to the intervention (25 vs. 22% in the coffee vs. substitute groups, respectively) but somewhat lower in the substitute group at the end of the intervention (24 vs. 14% in the coffee vs. substitute groups, respectively, ² = 2.42, P = 0.12). Similar results were found when the analysis was performed separately for the "nonanemic" and anemic subgroups.

The differences in morbidity between groups did not explain the effect of the intervention on growth. In other words, when the morbidity variables were included in regression models with weight or length gain as the outcomes, there was no change in the results shown in Table 3.

Table 5. Reaction to coffee substitute and attitudes and beliefs about coffee at the end of the intervention period, by intervention group [View Table]

DISCUSSION

The potential mechanisms for an inhibitory effect of coffee on child growth are unclear. In a cross-sectional, observational study, it is likely that children who drink more coffee consume less of the more nutritious beverages such as milk and juice, and this could account for differences in growth status. In this experimental study, however, we can rule out this confounding influence because the coffee substitute did not provide any additional nutrients, and there was no detectable change in intake of nutritive fluids as a result of the intervention. Furthermore, food intake patterns were generally similar between treatment groups both before and after the intervention. Thus, it appears that there is something about coffee itself that may inhibit growth. Caffeine is one obvious candidate, especially given that prenatal caffeine intake is known to affect birthweight (Golding 1995). In rats, birthweight is depressed by maternal coffee intake even though there is no effect on maternal food intake or gestational weight gain (Muñoz et al. 1986). The average caffeine intake of children in this study was about 9 mg/(kg·d) at base line, which is 5-18 times greater (per kg) than the estimated intake of 6- to 10-y-old children in the U.S. (Arbeit et al. 1988, Ellison et al. 1995). Caffeine is a central nervous system stimulant that can increase blood pressure, heart rate, lipolysis and metabolic rate (Berger 1988, Koot and Deurenberg 1995, Nehlig and Debry 1994). If this increase in energy expenditure is large enough and is not compensated by increased energy intake, it could lead to reduced growth. We do not have quantitative data on energy intake from the present study to evaluate this possibility.

Caffeine is not the only component in coffee that may alter growth because coffee contains dozens of compounds whose physiological effects are not well understood. For example, it is possible that coffee affects growth indirectly through effects on iron or zinc status. In this study, we did not observe any significant effect of discontinuing coffee on iron status (except in children taking iron supplements) (see Table 1 and Romero-Abal et al. 1996); thus, it is unlikely that this mechanism applies in this population. When use of iron supplements was included in the statistical models for the growth response, it did not change the significance of the intervention group effect in either the anemic or the "nonanemic" subgroups. There is some evidence that coffee may alter zinc metabolism (Muñoz et al. 1986), and because zinc is clearly a factor in growth stunting among children in developing countries (Allen 1994), this possibility deserves further exploration. We did not observe any effect of discontinuing coffee on plasma zinc levels (see Table 1), but it is questionable whether this indicator adequately reflects zinc status.

The morbidity differences found in this study are intriguing. Respiratory illness prevalence was significantly lower in children who discontinued coffee within the "nonanemic" subgroup (the majority of the sample), but there was a marginally significant difference in the opposite direction within the anemic subgroup. Controlling for iron supplementation eliminated the latter trend, but not the former. The results are consistent with the hypothesis that in the anemic subgroup, iron supplements increased the risk of morbidity, perhaps by making more iron available to pathogenic microorganisms (Brunser et al. 1993, Murray et al. 1978 and 1980, Weinberg 1974). Data presented elsewhere suggest that coffee decreased the amount of iron absorbed from supplements (Romero-Abal et al. 1996); if the above hypothesis is correct, this could explain why those in the coffee group (within the anemic subgroup) had less illness than those in the substitute group. In the "nonanemic" subgroup, one would predict that much less iron would be absorbed from supplements regardless of coffee intake (because these children were less iron deficient); under these circumstances, coffee intake led to greater illness. The trend towards a lower prevalence of elevated C-reactive protein levels in the substitute group (in both the anemic and "nonanemic" subgroups) after the intervention is consistent with a reduction in morbidity due to discontinuation of coffee. Furthermore, many of the mothers in the substitute group commented spontaneously that their children were healthier during the intervention. This was reflected in the percentage who felt that coffee had adverse effects on child health (67% in the substitute group vs. 47% in the coffee group). Further research is required to verify that there is an effect of coffee on illness rates, and if so, what mechanisms might be involved. We can conclude, however, that the morbidity differences observed in this study did not explain the effect of the intervention on growth.

It has often been suggested that changing dietary patterns is difficult, but we found that acceptance of the coffee substitute was quite high in this population. In the exit interview, most of the mothers in the substitute group said that they would continue to withhold coffee after the study ended and would not give coffee to a subsequent child. Considering that long-term behavioral change was not the objective of the intervention, and no mention was made to mothers about the potentially adverse effects of coffee, this degree of acceptance of the substitute is remarkable. Most of the mothers in the substitute group did not find it difficult to withhold coffee from their children. These findings imply that if a campaign to reduce coffee intake by toddlers in Guatemala were to be mounted, with clear guidelines regarding substitute beverages, it would probably be well accepted.

The final question is whether the results presented here are sufficient to justify a public health recommendation to avoid coffee consumption by toddlers in such populations. Although the difference in growth rate observed (in those with initial coffee intake >100 mL/d) was marginally significant, a 22% increase in length gain over a 5-mo interval could be biologically important. It should be noted that this difference was due solely to eliminating coffee. A greater effect might be seen if more nutritious substitutes for coffee such as milk or juice were used instead of the hot, colored sugar-water chosen for this study. The feasibility of promoting such alternatives requires investigation because they are more expensive and may be more subject to contamination if refrigeration is unavailable. Development of low cost, safe and nutritious alternatives to coffee for young children in Latin America should become a research priority.