Abstract
Background
Helicobacter pylori infection affects about half of the world’s population and is usually acquired in childhood. The infection has been associated with chronic gastritis, peptic ulcer, and stomach cancer in adulthood. Little is known, however, about its consequences on child health. We examined the effect of H. pylori infection on growth among school-age children in the Colombian Andes by comparing growth velocity in the presence and absence of H. pylori infection.
Methods
Children who were 4–8 years old in 2004 were followed up in a community where infected children received anti-H. pylori treatment (n = 165) and a comparison community (n = 161) for a mean of 2.5 years. Anthropometry measurements were made every 3 months and H. pylori status ascertained by urea breath test every 6 months. Growth velocities (cm/month) were compared across person-time with and without infection, using mixed models for repeated measures.
Results
In the untreated community, 83% were H. pylori-positive at baseline and 89% were -positive at study end. The corresponding prevalences were 74% and 46%, respectively, in the treated community. Growth velocity in the pretreatment interval was 0.44 (standard deviation [SD] = 0.13) cm/month. Models that adjusted for age, sex, and height estimated that H. pylori-positive children grew on average 0.022 cm/month (95% confidence interval = 0.008 to 0.035) slower than H. pylori-negative children, a result that was not appreciably altered by adjustment for socioenvironmental covariates.
Conclusions
This study suggests that chronic H. pylori infection is accompanied by slowed growth in school-age Andean children.
Helicobacter pylori infection causes chronic inflammation of the stomach, which increases the risk of peptic ulcer disease and stomach cancer. Usually acquired in childhood, the infection often persists long-term without specific symptoms. It is common in young children in developing countries,1 where it has been observed to occur in association with malnutrition and slowed growth.2,3 Chronic infections can interfere with micronutrient absorption, appetite, metabolism, and related factors, and thereby diminish children’s growth.4 The causal nature of observed associations between H. pylori infection and deficient growth remains controversial, however, because human growth is also determined by other factors that may be independently associated with H. pylori infection, such as diet, socioeconomic status, and coinfections.5,6 Height deficits in children from the developing world are related more strongly to poverty and other environmental influences than to genetic influences on body size.1,7–9 A recent cohort study showed that new H. pylori infections appeared to reduce growth velocity in preschool children.10 The aim of the present analysis was to examine the hypothesis that persistence of H. pylori infection leads to reduced growth velocity in school-age children, using data from a study designed to investigate short-term consequences of H. pylori infection in children.
METHODS
Study Population and Design
The study was conducted in 2 rural communities, each comprising a pair of proximal villages, in the Andean region of Nariño, Colombia, where H. pylori infection is nearly universal by adolescence.11 Our goal was to prospectively observe the occurrence of selected health outcomes in children with and without H. pylori infection. We selected 2 comparable populations of children with high H. pylori prevalence, offered treatment aimed at eliminating the infection in one of the populations, and followed up both to compare growth in the presence and absence of this infection. Although the design includes an intervention for the purpose of observing the desired exposure contrast, it is essentially observational in nature, given that H. pylori status, rather than treatment, was the exposure of relevance to the research objective, and exposure was not allocated randomly to individual children; it was not practical to treat some children and not others within 1-school villages, and H. pylori treatment success depends heavily on factors that are not randomly distributed, such as adherence to complex treatment regimens and prevalence of antibiotic-resistant strains.12–15 We obtained written informed parental consent for all children and those with sufficient maturity signed assent forms, in accordance with the standards of the ethics review boards that approved this study: the Vanderbilt University review board and the research ethics committee of the Universidad del Valle.
We offered treatment to eliminate H. pylori infection in the Nariño-Genoy community, located 20 kilometers northwest of Pasto, the capital of the state of Nariño. We selected this community to receive treatment because an earlier treatment intervention had been carried out there. The community where treatment was not offered, La Laguna-Cabrera, is located at about the same distance from Pasto to the southeast. The 2 communities are similar in size, demography, and socioeconomic status, although there were subtle differences in baseline characteristics (Table 1).
TABLE 1.
Baseline Characteristics by Community Among 326 School-age Andean Children
| La Laguna-Cabrera (n = 161) | Nariño-Genoy (n = 165) | |
|---|---|---|
| No. (%) | No. (%) | |
| Child | ||
| Sex | ||
| Boys | 85 (53) | 84 (51) |
| Girls | 76 (47) | 81 (49) |
| Age (years) | ||
| 3.6–4 | 37 (23) | 30 (18) |
| 5 | 35 (22) | 31 (19) |
| 6 | 37 (23) | 45 (27) |
| 7 | 40 (25) | 40 (24) |
| 8 | 12 (7) | 19 (12) |
| Goes to school | ||
| No | 41 (25) | 39 (24) |
| Yes | 120 (75) | 126 (76) |
| No. siblings | ||
| 0 | 31 (19) | 39 (24) |
| 1 | 38 (24) | 56 (34) |
| 2 | 45 (28) | 38 (23) |
| 3–6 | 47 (29) | 32 (19) |
| Mother | ||
| Age (years) | ||
| 20–29 | 58 (36) | 73 (44) |
| 30–34 | 50 (31) | 53 (32) |
| ≥ 35 | 53 (33) | 35 (21) |
| Missing | 0 (0) | 4 (2) |
| Education (years) | ||
| 0–4 (incomplete primary) | 74 (46) | 72 (44) |
| 5 (complete primary) | 50 (31) | 46 (28) |
| 6–10 (incomplete secondary) | 15 (9) | 19 (12) |
| 11–17 (complete secondary or higher) | 22 (14) | 25 (15) |
| Missing | 0 (0) | 3 (2) |
| Occupation | ||
| Agriculture | 8 (5) | 0 (0) |
| Handicraft worker | 0 (0) | 14 (8) |
| Housewife | 106 (66) | 87 (53) |
| Sales | 20 (12) | 19 (12) |
| Service worker | 23 (14) | 34 (21) |
| Othera | 4 (2) | 7 (4) |
| Missing | 0 (0) | 4 (2) |
| Father | ||
| Lives with the child | ||
| No | 33 (21) | 48 (29) |
| Yes | 128 (80) | 117 (71) |
| Age (years) | ||
| 20–29 | 38 (24) | 38 (23) |
| 30–34 | 40 (25) | 50 (30) |
| ≥35 | 83 (52) | 70 (42) |
| Missing | 0 (0) | 7 (4) |
| Education (years) | ||
| 0–5 | 57 (35) | 47 (28) |
| 6–11 | 66 (41) | 68 (41) |
| 12–17 | 21 (13) | 17 (10) |
| Other | 17 (11) | 29 (18) |
| Missing | 0 (0) | 4 (2) |
| Occupation | ||
| Agriculture | 87 (54) | 18 (11) |
| Construction | 23 (14) | 45 (27) |
| Sales | 7 (4) | 10 (6) |
| Service worker | 5 (3) | 23 (14) |
| Transportation | 18 (11) | 28 (17) |
| Unemployed | 1 (1) | 1 (1) |
| Otherb | 16 (10) | 35 (21) |
| Missing | 4 (2) | 5 (3) |
| Residential stability | ||
| Home ownership | ||
| Own | 90 (56) | 88 (53) |
| Rent | 15 (9) | 37 (22) |
| Lives with relatives | 52 (32) | 36 (22) |
| Other | 4 (2) | 4 (2) |
| Child has lived in another community | ||
| No | 151 (93) | 153 (91) |
| Yes | 11 (7) | 16 (9) |
| No. times child changed residence | ||
| 0 | 98 (61) | 85 (52) |
| 1 | 40 (25) | 48 (29) |
| 2–6 | 22 (14) | 32 (19) |
| Missing | 1 (1) | 0 (0) |
| Housing conditions | ||
| Type of floor in house | ||
| Dirt | 45 (28) | 17 (10) |
| Cement | 84 (52) | 118 (72) |
| Otherc | 32 (20) | 30 (18) |
| Type of toilet facility | ||
| None | 2 (1) | 3 (2) |
| Latrine | 67 (42) | 7 (4) |
| Flush toilet | 37 (23) | 140 (85) |
| Toilet without running water | 53 (33) | 12 (7) |
| Other | 2 (1) | 3 (2) |
| Household crowding | ||
| Child shares bed | ||
| No | 40 (25) | 59 (36) |
| Yes | 121 (75) | 106 (64) |
| Household crowding (people/room) | ||
| 0.25–0.99 | 31 (19) | 66 (40) |
| 1–1.99 | 77 (48) | 75 (45) |
| 2–6 | 49 (30) | 23 (14) |
| Missing | 4 (2) | 1 (1) |
| Water sources | ||
| Drinking water source | ||
| Municipal water system, tap in house | 83 (52) | 120 (73) |
| Municipal water system, tap in patio | 76 (47) | 38 (23) |
| Other | 2 (1) | 7 (4) |
| Frequency of purifying child’s drinking water | ||
| Always | 103 (64) | 121 (73) |
| Sometimes/usually | 19 (12) | 20 (12) |
| Never/not usually | 38 (24) | 24 (15) |
| Missing | 1 (1) | 0 (0) |
| Health insurance information | ||
| Has health insurance | ||
| No | 2 (1) | 23 (14) |
| Yes | 159 (99) | 142 (86) |
Clerk, health worker, student, and textile worker.
Mining, wood and pulp, handicraft worker, craft and trades, textile and clothing, food services, skilled agricultural and fishery worker, law and justice, student, teacher, administrator.
Ceramic tile, wood, and linoleum.
An earlier study offered H. pylori treatment to preschool children and family members in Nariño-Genoy; the participating children and a comparison group in La Laguna-Cabrera were monitored for H. pylori status, height, and weight during 1999–2003. The effectiveness of treatment increased with age among the preschoolers, but was poor overall; just 43% of treated children tested negative at 3 months post-treatment. In 2004, the present study recruited 4–8-year-old children from the earlier study, along with other age-eligible children (eFigure 1, http://links.lww.com/EDE/A439). At baseline, children from both communities were tested for H. pylori infection and measured for height and weight. Study staff interviewed the child’s primary caretaker, using a structured questionnaire to record information on demographic, household composition, general socioeconomic, and hygiene-related factors. Follow-up began at 6 months after study enrollment and occurred at 3-month intervals thereafter. At each follow-up visit, the children’s height and weight were measured in duplicate, using the procedure recommended in Lohman’s anthropometric standardization reference manual16; agreement between the 2 measurements was monitored for quality control, and the average of the 2 measurements was used. Fresh stool specimens were collected from the children for ova and parasite examination. Every 6 months, children were tested for H. pylori status with the 13C-urea breath test.17 This analysis uses data from 2.5 years of follow-up.
We chose height-based growth velocity as the outcome of interest because growth defined in terms of height is considered a sensitive indicator of overall nutritional status throughout childhood, given that poor nutrition results in slowed growth and stunting.18 Measures that incorporate weight, such as weight-for-height or BMI, on the other hand, reflect nutritional problems at high and low values, cannot be interpreted independently of height, and are not a sensitive measure for identifying mild-to-moderately impaired nutritional status.18
In Nariño-Genoy, children who tested H. pylori positive at baseline received therapy to eliminate H. pylori infection by directly observed treatment with lansoprazole (30 mg/kg/d for children <30 kg, 60 mg/kg/d for children ≥30 kg), amoxicillin (50 mg/kg/d), metronidazole (30 mg/kg/d), and bismuth (262 mg/d) for 14 days. Study staff administered the medications twice daily either at the health care center or the child’s home. Treated children were given an additional breath test around 3 months post-treatment. Those whose treatment failed were retreated and retested during 2 additional treatment rounds.
Statistical Analysis
The outcome measure, growth velocity in cm/month, was defined as the difference between 2 consecutive height measurements divided by the time elapsed. For each child, a maximum of 9 measurements of growth velocity were available from a maximum of 10 height measurements, with an average of 3 months between them; each child’s growth velocity value for each follow-up interval is based on the actual number of days between that child’s consecutive follow-up visits. The exposure of interest, H. pylori status, was treated as time-dependent. Because H. pylori prevalence increases with age in the study population and growth velocity is age-dependent, estimates were adjusted for age as a time-dependent covariate. Visit number, used to mark aging (time elapsed since baseline), was not completely collinear with age, because the visits did not occur precisely on schedule. We assessed socio-environmental factors to control confounding by poverty-related determinants of H. pylori prevalence and growth. These include presence of intestinal parasites, modeled as time-dependent, and factors that did not change over time (ascertained at baseline): sex, community of residence, mother’s and father’s education, number of siblings, time at current address, residential stability, type of housing, number of people in the house, number of rooms in the house, and household density (people/rooms). Individual treatment status was not included because our earlier work consistently showed it to have little effect on growth independent of H. pylori status.
Growth velocity in cm/month was estimated as the dependent variable using a mixed model (with fixed and random effects) for repeated measures, with each child starting at his/her own intercept, and with changes in growth velocity over time specific to the child’s baseline age. In constructing the model, a value (eg, the mean) must be specified for each covariate. The coefficient for each independent variable reflects the change in growth velocity for the corresponding exposure contrast, given the specified set of covariate values. The model accounts for intraperson correlation due to repeated measures, differing numbers of measurements across subjects, and missing data due to skipped visits, as the number of observations per subject and the length of intervals between measurements were allowed to vary. Random slopes were assessed for the time-dependent variables. The small fraction of children with missing values for height at a given visit had that visit information excluded from analysis. No imputation was made for missing values because estimates for each child used data from other visits. All data were analyzed using S-PLUS (version 10; Insightful Corp., Seattle, WA).
Unconditional models were constructed to determine the total variation in growth velocity, attributable to differences between and within participants. These models also indicated whether there was substantial variation in initial status and rate of change; if so, further analysis explored the effects of H. pylori infection and covariates. The effect of each covariate was examined separately due to sample size constraints. For continuous variables, care was taken to detect departures from linearity in dose-response to determine whether a transformation was needed. The set of selected covariates in the multivariable model was based on the smallest log likelihood (Akaike’s Information Criterion).19
RESULTS
Baseline characteristics of children by community are shown in Table 1. A total of 326 children were enrolled with initial anthropometry measurements (325 at baseline and 1 at visit 1); 303 (93%) had anthropometry measurements around 2.5 years later (2 years plus 186 days, on average), at the ninth visit (eFig. 1, http://links.lww.com/EDE/A439). The average number of follow-up visits was 8.7. A high degree of residential stability in the cohort and rigorous follow-up methods facilitated a high level of participation and minimized missing data. Breath test results were available for 96% (314/326) of the children at baseline and the proportion tested at each scheduled follow-up ranged from 92% to 98% (eFigure 2, http://links.lww.com/EDE/A439). The proportion with anthropometry measurements ranged from 93% to 98% each scheduled measurement. Baseline visits occurred between 17 August and 26 October, 2004 (average = 18 September), and the last visit used for this analysis occurred between 21 February and 24 April, 2007 (average = 26 March).
At baseline, 78.7% of the 314 children tested were H. pylori positive. There were no substantial differences in H. pylori status by sex, but children with positive tests were notably older on average than those with negative results (mean ages of 6.4 [SD = 1.25] years and 5.9 [1.26] years, respectively). Nariño-Genoy had lower prevalence of H. pylori infection than La Laguna-Cabrera (74% vs. 83%, χ2 P = 0.053), presumably due to the treatment administered there 5 years earlier. Mean height at baseline was similar in boys and girls, slightly lower in children with parasites, higher in H. pylori-positive children (due to their older mean age), and higher in Nariño-Genoy (Table 2). Of the enrolled children, 67% had participated in the earlier study. Before the initial treatment in 1999 (when the children were 6 to 48 months old), H. pylori prevalence was 43% in Nariño-Genoy children and 45% in La La-guna-Cabrera children. At that time, mean height was 79.3 (95% CI = 77.9 to 80.8) cm in Nariño-Genoy children and 76.9 (75.5 to 78.4) cm in La Laguna-Cabrera children, a smaller difference than that observed in 2004.
TABLE 2.
Average Height at Baseline by Selected Factors Among 326 School-age Andean Children
| Height (cm)
|
||||
|---|---|---|---|---|
| No. | Mean (cm) | SE | (95% CI) | |
| Sex | ||||
| Boys | 169 | 109.0 | 0.63 | (107.8–110.3) |
| Girls | 156 | 109.1 | 0.65 | (107.8–110.4) |
| H. pylori | ||||
| Negative | 67 | 107.9 | 1.00 | (105.9–109.9) |
| Positive | 247 | 109.5 | 0.52 | (108.5–110.5) |
| Community | ||||
| Nariño-Genoy | 165 | 111.3 | 0.62 | (110.1–112.5) |
| La Laguna-Cabrera | 160 | 106.8 | 0.61 | (105.6–108.0) |
| Community (adjusted)a | ||||
| Nariño-Genoy | 165 | 110.8 | 0.34 | (110.1–111.4) |
| La Laguna-Cabrera | 160 | 107.4 | 0.36 | (106.7–108.1) |
| Parasitesb | ||||
| No | 85 | 110.6 | 0.88 | (108.8–112.3) |
| Yes | 229 | 108.9 | 0.54 | (107.8–109.9) |
| Helminthes | ||||
| No | 272 | 109.4 | 0.49 | (108.4–110.4) |
| Yes | 42 | 108.9 | 1.25 | (106.4–111.3) |
| Protozoa | ||||
| No | 95 | 110.4 | 0.83 | (108.8–112.0) |
| Yes | 219 | 108.8 | 0.54 | (107.8–109.9) |
Adjusted for age, sex, and H. pylori status. Covariate values in model used for adjustment (mean at baseline): age in months = 75.7343; sex = 1.49; H. pylori status = 0.79.
Helminthes or protozoa detected in stool.
SE indicates standard error.
The first follow-up visit occurred 6 months postbaseline, before treatment was administered, permitting estimation of the pretreatment average growth velocity, which was 0.44 (SD = 0.13) cm/mo overall (Table 3). During the pretreatment interval, age-adjusted estimates showed little difference in growth velocity by sex, a 0.018 cm/month (95% CI = −0.52 to 0.17) difference favoring H. pylori-positive children and a 0.082 cm/month (0.062 to 0.12) difference favoring La Laguna-Cabrera children. Further adjustment for sex, height, and H. pylori status did not reduce the pretreatment growth velocity difference between the communities. Unadjusted mean height was higher in Nariño-Genoy throughout follow-up (Table 4).
TABLE 3.
Pretreatment Growth Velocitya by Selected Factors Among School-age Andean Children
| No. | Growth Velocity (cm/Month)
|
||||||
|---|---|---|---|---|---|---|---|
| Unadjusted
|
Age-adjusted
|
||||||
| Mean | SE | (95% CI) | Mean | SE | (95% CI) | ||
| Total | 318 | 0.437 | 0.010 | (0.414–0.421) | |||
| Sex | |||||||
| Boys | 163 | 0.436 | 0.010 | (0.416–0.455) | 0.434 | 0.010 | (0.415–0.452) |
| Girls | 154 | 0.439 | 0.010 | (0.419–0.459) | 0.441 | 0.010 | (0.422–0.461) |
| H. pylori status | |||||||
| Negative | 64 | 0.435 | 0.016 | (0.404–0.467) | 0.423 | 0.015 | (0.393–0.454) |
| Positive | 243 | 0.438 | 0.008 | (0.421–0.454) | 0.441 | 0.008 | (0.425–0.456) |
| Community | |||||||
| Nariño-Genoy | 160 | 0.394 | 0.009 | (0.375–0.412) | 0.397 | 0.009 | (0.379–0.415) |
| La Laguna-Cabrera | 158 | 0.482 | 0.010 | (0.463–0.501) | 0.479 | 0.009 | (0.461–0.497) |
| Community (adjusted)b | |||||||
| Nariño-Genoy | 160 | 0.396 | 0.009 | (0.377–0.414) | |||
| La Laguna-Cabrera | 158 | 0.482 | 0.010 | (0.463–0.502) | |||
Over the 6 months after baseline assessment.
Adjusted for age, sex, H. pylori status, and baseline height. Covariate values in model used for adjustment (mean at baseline): age in months = 75.7343; sex = 1.49; H. pylori status = 0.79; height in cm = 109.2915.
TABLE 4.
Mean Height Across Follow-up by Community Among 326 School-age Andean Children
| Visit | Nariño-Genoy
|
La Laguna-Cabrera
|
Total
|
|||
|---|---|---|---|---|---|---|
| Mean (cm) | (SD) | Mean (cm) | (SD) | Mean (cm) | (SD) | |
| 0a | 111.3 | (8.03) | 106.8 | (7.74) | 109.1 | (8.19) |
| 1 | 113.7 | (7.89) | 109.5 | (7.55) | 111.6 | (7.99) |
| 2 | 115.5 | (7.76) | 110.7 | (7.43) | 113.1 | (7.96) |
| 3 | 117.1 | (7.85) | 112.1 | (7.39) | 114.6 | (8.02) |
| 4 | 118.7 | (7.84) | 113.2 | (7.33) | 115.9 | (8.07) |
| 5 | 120.2 | (7.62) | 114.9 | (7.30) | 117.6 | (7.91) |
| 6 | 121.6 | (7.63) | 116.1 | (7.30) | 118.9 | (7.95) |
| 7 | 123.3 | (7.74) | 117.8 | (7.23) | 120.6 | (7.97) |
| 8 | 124.3 | (7.78) | 119.2 | (7.29) | 121.7 | (7.95) |
| 9 | 125.7 | (7.75) | 120.2 | (7.33) | 122.9 | (8.01) |
Baseline visit.
eFigure 2 (http://links.lww.com/EDE/A439) shows the numbers tested, treated, and retreated for H. pylori infection, along with positivity status, across follow-up. Of 120 children given initial treatment, 109 completed all doses. From the baseline H. pylori prevalence of 79% among all children, there was a declining trend through the fourth visit, with 56% positive, and then increasing to 67% at study end. In La Laguna-Cabrera, 83% were positive at baseline, with the prevalence increasing to 89% at study end. In Nariño-Genoy, 74% were positive at baseline, with the prevalence declining to 27% at the fourth visit and increasing to 46% at study end. As the cohort aged, unadjusted mean height became higher in H. pylori-negative children, starting at visit 3 (Table 5).
TABLE 5.
Mean Height Across Follow-up by H. pylori Status Among 326 School-age Andean Children
| Visit |
H. pylori |
|||||
|---|---|---|---|---|---|---|
| Positive
|
Negative
|
|||||
| No. | Mean (cm) | (SD) | No. | Mean (cm) | (SD) | |
| 0 | 247 | 109.5 | (8.1) | 67 | 107.9 | (8.7) |
| 1 | 254 | 111.9 | (7.9) | 64 | 110.3 | (8.4) |
| 3 | 196 | 114.0 | (7.9) | 113 | 115.6 | (8.1) |
| 5 | 190 | 117.0 | (7.9) | 119 | 118.6 | (7.9) |
| 7 | 188 | 119.5 | (7.9) | 120 | 122.4 | (7.8) |
| 9 | 199 | 122.4 | (7.9) | 99 | 124.6 | (7.5) |
Multivariable Mixed Models for Repeated Measures of Growth Velocity
Growth velocity depends strongly on age and on height attained at earlier ages, therefore we took care to model these relationships optimally. The relationship between age and growth velocity showed a slowly declining trend, with a prepubertal slowdown around 9 years and a pubertal growth spurt after 10 years (evident only among girls). Figure 1 shows the increasing width of the growth curve CI after 10 years. The best fitting model for age was quadratic, closely mimicking the curve from 5 to 10 years, although a linear model also approximated the growth velocity values well. The model correctly shows that the average growth velocity was 0.49 (standard error [SE] = 0.0058) cm/month for a 6.3-year-old-child, and 0.44 (0.0058) cm/month for an 8.8-year-old child. Because of extreme variability in growth velocity among the 29 children over 10.5 years at the last visit, a model that incorporated the start of the female growth spurt had a worse fit, with real values farther from fitted values, than a simpler model. Thus, the final model for age over time was a growth curve identical to the one in the graph for children from 5–10 years.
FIGURE 1.
Growth velocity by age among 326 school-age Andean children followed for a mean of 2.5 years. Model included fixed effects (quadratic age, height, H. pylori status, number of siblings, father’s years of education) and random effects (age, H. pylori status). The dotted line represents the 95% confidence intervals. The black lines at the bottom represent data points, to show their density over time.
There was a notable difference in the growth curve by sex, a well-known phenomenon, with boys having a shallower curve as they approach 10 years (P value of the interaction coefficient = 0.045). At 8.8 years, growth velocity was 0.44 cm/month in girls and 0.41 cm/month in boys. Our initial multivariable model had a different starting value for each child, and separate slopes for sex and the square of age. To this model, we added height and H. pylori status at the end of each interval, as time-dependent variables. Independent of age and sex, the average change in growth velocity was small for each additional centimeter of initial height (0.007 cm/month [95% CI = 0.006 to 0.008]). Community of residence was added to the model, but was not retained due to its small and imprecisely estimated effect on growth velocity over time (0.0041 cm/month [SE = 0.0042]) and the fairly small (14%) change-in-effect its inclusion had on the coefficient for H. pylori status. Adjusted only for baseline age and sex, the community effect was quite strong (0.022 cm/month [SE = 0.0036]), but most of this effect corresponded to height differences; additional adjustment for height reduced the community effect to 0.0067 cm/month (SE = 0.0039). The time-dependent model with age, sex, and height minimizes confounding of the H. pylori effect by height, age at baseline, and aging (growth velocity can accelerate and decelerate in the same child, and is not constant, decreasing as children age and get taller). The independent effects of height and age on growth velocity become apparent in the time-dependent scenario, though they are not observed in the model for the pretreatment interval, which does not have a time-dependent component. Older and taller children have lower growth velocity, and at a given age, taller children still have lower growth velocity.
This model yields the instantaneous effect of H. pylori on growth velocity independent of height, age, and aging. During intervals of H. pylori positivity, children grew on average 0.022 cm/mo (95% CI = 0.008 to 0.035) slower than during periods of H. pylori negativity. H. pylori-negative children had higher growth velocity across age, sex, and time elapsed since baseline. The effect, however, is most notable for the next visit after a change in H. pylori status, and this holds up for any next visit at any point in time. Thus, the effect appears to be acute, and is also cumulative. This model lets us estimate cumulative growth gains or deficits for hypothetical contrasts; for example, children who were H. pylori-positive at baseline became negative after the initial treatment and stayed negative throughout follow-up accumulated an average gain of 0.66 cm (95% CI = 0.24 to 1.05) compared with children who stayed positive throughout follow-up, independent of age, sex, and height.
Because of noted baseline differences in the communities, we carefully assessed measured socioenvironmental factors as potential confounders. None of these changed the effect estimates for H. pylori status in the multivariable model by 10% or more (Table 6). Thus, the main model contained a random intercept for each child, age, quadratic age, sex, height, and H. pylori status. It would be of interest to examine stratum-specific estimates of growth velocity generated by the multivariable model for subgroups of children who were persistently negative or positive or whose status changed over time in either direction; however, breaking up the data in this manner compromises the validity and stability of the growth velocity estimates, because the model derives its power for valid estimation of longitudinal effects from the totality of the cohort observed over the entire follow-up duration. Given that H. pylori status changes over time, strata based on H. pylori status do not contain the same individuals throughout follow-up. Furthermore, the intraper-son variability in growth velocity is substantial, with as much as an order of magnitude difference among measurements in the same child, whereas the interperson variability is rather small across the cohort. Thus, the stability of the growth velocity estimates is substantially reduced in subgroups, and the same is true for comparing estimates across distinct periods of follow-up because the growth velocity variance increases as the follow-up duration diminishes. Stratum-specific comparisons of growth velocity estimates are particularly problematic, however, because substantially different distributions and interrelationships of key confounders may lead to lack of comparability in adjusted estimates. For purposes of model checking, Table 7 shows unadjusted estimates of growth velocity by H. pylori status in pre- and post-treatment follow-up intervals. Although these estimates cannot be used to validly estimate growth velocity differences between groups, we were able to confirm in the multivariable model the observation that the only notable increase in growth velocity occurs in children who were H. pylori-positive at baseline and cleared the infection after treatment. We did this by estimating in our multivariable model a coefficient for the contrast between persistent negativity and post-treatment clearance, resulting in a small difference of 0.10 cm/month, with a large standard error of 0.17 cm/month.
TABLE 6.
Estimates of the Effect of H. pylori Status on Growth Velocity (cm/Month) From Baseline Through the Last Follow-up Visit, Adjusted for Socioenvironmental Factors That Influence Growth
| Covariate Added to Main Modelb |
H. pylori Statusa |
Socioenvironmental Covariateb |
||
|---|---|---|---|---|
| Coefficientc Contrast in Growth Velocity (cm/Month) | SE | Coefficientc Contrast in Growth Velocity (cm/Month) | SE | |
| No additional covariatesc | −0.022 | 0.0076 | — | — |
| No. siblings (per sib) | −0.022 | 0.0076 | 0.0036 | 0.0025 |
| Mother’s education (per year) | −0.022 | 0.0076 | −0.0024 | 0.0012 |
| Father’s education (per year) | −0.021 | 0.0076 | −0.0036 | 0.0012 |
| Household density (people/rooms) | −0.023 | 0.0076 | 0.011 | 0.0098 |
| Community (Nariño-Genoy vs. La Laguna-Cabrera) | −0.019 | 0.0083 | 0.0041 | 0.0042 |
H. pylori status is time dependent, classified according to breath test result at each visit, contrasting positive and negative person-time; total numbers exposed vary by visit.
Due to sample size constraints, socioenvironmental covariates were added to the model one at a time.
All models included age, quadratic age, sex, H. pylori status, and height, with all but sex modeled as time-dependent variables.
TABLE 7.
Comparison of Pre- and Post-treatment Unadjusteda Growth Velocity Estimates by H. pylori Status Among School-age Andean Children
| H. pylori Status | No. | Unadjusted Growth Velocity (cm/Month)
|
|
|---|---|---|---|
| Mean | (95% CI) | ||
| Pretreatment interval (6-month interval from baseline to visit 1) | |||
| Negative at baseline | 67 | 0.435 | (0.403–0.467) |
| Positive at baseline | 247 | 0.438 | (0.404–0.467) |
| Post-treatment interval (6-month interval from visit 1 to visit 3) | |||
| Negative at baseline (n = 64 tested at visit 3) | |||
| Negative at visit 3 | 45 | 0.464 | (0.420–0.508) |
| Positive at visit 3 | 19 | 0.481 | (0.428–0.534) |
| Positive at baseline (n = 235 tested at visit 3) | |||
| Negative at visit 3 | 66 | 0.558 | (0.511–0.604) |
| Positive at visit 3 | 169 | 0.469 | (0.445–0.494) |
Unadjusted estimates are presented for purposes of data description but cannot be validly compared due to confounding from substantially different distributions of baseline age and height across groups.
We also checked the growth velocity model against a model for estimating mean height at time points throughout follow-up. This model included fixed effects for age (mean at baseline = 6.3 years), community, baseline H. pylori status, number of siblings (mean = 1.8), father’s years of education (mean = 5.3), and visit number, as well as random effects for age and H. pylori status. The slope for the change in height differed by both age and H. pylori status. Nariño-Genoy children were taller on average at baseline, and this contrast held up over time. The change from positive to negative H. pylori status corresponded to a height increase of 0.11 cm, compounded over time. For every sibling in the house, there was a height decrease of 0.5 cm (a fixed effect not compounded over time). For every year of father’s education, there was a height increase of 0.33 cm (also a fixed effect). Figure 2 shows mean height predicted by this model, stratified by H. pylori status. The model estimates are very close to the unadjusted stratified mean heights (Table 2). The stratified mean heights over follow-up show a clear and consistent height advantage among H. pylori-free children, beginning at the third follow-up visit.
FIGURE 2.
Model-predicted mean height across follow-up by H. pylori (Hp) status at each follow-up visit among school-age Andean children. The mean height model included fixed effects for age, visit number, community, baseline H. pylori status, number of siblings, and father’s education, and random effects for age and H. pylori status.
DISCUSSION
Our study of the effect of H. pylori infection on growth among school-age children in the Colombian Andes shows that H. pylori negativity was associated with increased growth velocity after considering various factors associated with reduced growth and H. pylori status. Our statistical model accounted for the different distributions of age and height in compared groups; the model showed an excellent fit with known growth patterns and with alternate methods of examining height gains in the study population; and it produced estimates that were stable across varied model specifications. Our longitudinal design with closely spaced assessments of height and H. pylori status permits distinguishing whether infection dynamics (eg, acquisition, elimination, and unchanging status) differentially affect growth velocity. The observed timing of changes in growth velocity immediately subsequent to changes in H. pylori status suggests a modest rebound in growth velocity right after the infection clears, with an irreversible height gain while the growth velocity effect fades. This pattern is consistent with a catch-up effect resulting from the removal of a factor that slows growth.
The present findings are consistent with observed decreases in growth velocity after the acquisition of a new H. pylori infection in a cohort of preschool children in Cali, Colombia.20 In the Cali cohort, the growth velocity differences between infected and uninfected children of the same age and sex disappeared after 1 year following infection onset, but infected children were shorter on average and no catch-up growth was observed while the infection persisted.20 Studies of the effect of H. pylori infection on childhood growth in diverse locations have reported conflicting results, perhaps due to differences in the distribution of factors that affect growth, or due to differences in study design, such as timing of measurements or statistical methods used to estimate effects on growth. A study of Scottish children aged 7–11 years showed that H. pylori-positive children grew less than uninfected children.21 In contrast, among 7–11-year-old Alaska Native children, minimal differences in growth were associated with H. pylori clearance.22 A study of Gambian infants showed that H. pylori infection was associated with reduced growth after controlling for local growth curves, season of birth, and level of diarrheal disease.23 Cross-sectional studies have shown conflicting evidence regarding the association between H. pylori infection and indicators of slowed growth.5,24–37
Human growth is complex and depends on factors that may also be associated with acquisition or persistence of H. pylori infection, such as diet, socioeconomic status, and coinfections.38 Thus, accurately estimating the effect of H. pylori infection on growth is challenging due to the number of potential confounders, many of which are inadequately controlled in studies.38 We examined several socioenvironmental factors as potential confounders, but did not identify any for which adjustment appreciably altered the association between H. pylori status and growth velocity. Although differences in pretreatment growth velocity in the comparison communities were not fully accounted for by adjusting for mismatches in baseline age, height, and H. pylori status, there was minimal evidence of an independent community effect in the model of effects on growth velocity over time.
The existing body of evidence regarding the relationship between H. pylori infection and growth has come mainly from cross-sectional studies, which are limited by the inability to establish temporality and thus the direction of detected associations remains ambiguous. Our study has the advantage of prospective follow-up of changes in growth patterns subsequent to changes in H. pylori status over time. Furthermore, few studies have analyzed growth velocity, which is considered a more sensitive indicator of health status than weight or height attained at a specific age.18 This prospective study suggests that H. pylori infection is accompanied by slowed growth in school-age Andean children.
Acknowledgments
Supported by the National Cancer Institute (PO1CA028842).
We thank Alicia Rosero for data management, Luz Stella Garcia and Nora A. Ordoñez for laboratory support, and Tito Collazos for administrative support.
Footnotes
Supplemental digital content is available through direct URL citations in the HTML and PDF versions of this article (www.epidem.com).
References
- 1.Megraud F, Brassens-Rabbe MP, Denis F, Belbouri A, Hoa DQ. Seroepidemiology of Campylobacter pylori infection in various populations. J Clin Microbiol. 1989;27:1870–1873. doi: 10.1128/jcm.27.8.1870-1873.1989. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Sullivan PB, Thomas JE, Wight DG, et al. Helicobacter pylori in Gambian children with chronic diarrhoea and malnutrition. Arch Dis Child. 1990;65:189–191. doi: 10.1136/adc.65.2.189. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Holcombe C, Omotara BA, Eldridge J, Jones DM. H. pylori, the most common bacterial infection in Africa: a random serological study. Am J Gastroenterol. 1992;87:28–30. [PubMed] [Google Scholar]
- 4.Stephensen CB. Burden of infection on growth failure. J Nutr. 1999;129(suppl 2S):534–538. doi: 10.1093/jn/129.2.534S. [DOI] [PubMed] [Google Scholar]
- 5.Goodman KJ, Correa P, Tengana Aux HJ, DeLany JP, Collazos T. Nutritional factors and Helicobacter pylori infection in Colombian children. J Pediatr Gastroenterol Nutr. 1997;25:507–515. doi: 10.1097/00005176-199711000-00004. [DOI] [PubMed] [Google Scholar]
- 6.Sood MR, Joshi S, Akobeng AK, Mitchell J, Thomas AG. Growth in children with Helicobacter pylori infection and dyspepsia. Arch Dis Child. 2005;90:1025–1028. doi: 10.1136/adc.2004.066803. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Global childhood malnutrition. Lancet. 2006;367:1459. doi: 10.1016/S0140-6736(06)68620-0. [DOI] [PubMed] [Google Scholar]
- 8.Humphrey JH. Underweight malnutrition in infants in developing countries: an intractable problem. Arch Pediatr Adolesc Med. 2008;162:692–694. doi: 10.1001/archpedi.162.7.692. [DOI] [PubMed] [Google Scholar]
- 9.Olivieri F, Semproli S, Pettener D, Toselli S. Growth and malnutrition of rural Zimbabwean children (6–17 years of age) Am J Phys Anthropol. 2008;136:214–222. doi: 10.1002/ajpa.20797. [DOI] [PubMed] [Google Scholar]
- 10.Bravo L, Mera R, Reina J, et al. Impact of Helicobacter pylori infection on growth of children: a prospective cohort study. J Pediatr Gastroenterol Nutr. 2003;37:614–619. doi: 10.1097/00005176-200311000-00021. [DOI] [PubMed] [Google Scholar]
- 11.Goodman KJ, Correa P, Tengana Aux HJ, et al. Helicobacter pylori infection in the Colombian Andes: a population-based study of transmission pathways. Am J Epidemiol. 1996;144:290–299. doi: 10.1093/oxfordjournals.aje.a008924. [DOI] [PubMed] [Google Scholar]
- 12.Khurana R, Fischbach L, Chiba N, et al. Meta-analysis: Helicobacter pylori eradication treatment efficacy in children. Aliment Pharmacol Ther. 2007;25:523–536. doi: 10.1111/j.1365-2036.2006.03236.x. [DOI] [PubMed] [Google Scholar]
- 13.Cai W, Zhou L, Ren W, Deng L, Yu M. Variables influencing outcome of Helicobacter pylori eradication therapy in South China. Helicobacter. 2009;14:91–96. doi: 10.1111/j.1523-5378.2009.00718.x. [DOI] [PubMed] [Google Scholar]
- 14.Wolle K, Malfertheiner P. Treatment of Helicobacter pylori. Best Pract Res Clin Gastroenterol. 2007;21:315–324. doi: 10.1016/j.bpg.2006.11.001. [DOI] [PubMed] [Google Scholar]
- 15.Malfertheiner P, Peitz U, Treiber G. What constitutes failure for Helicobacter pylori eradication therapy? Can J Gastroenterol. 2003;17(suppl B):53B–57B. doi: 10.1155/2003/241508. [DOI] [PubMed] [Google Scholar]
- 16.Lohman TG, Roche AF, Martorell R. Anthropometric Standardization Reference Manual. Champaigne, IL: Human Kinetics Books; 1988. [Google Scholar]
- 17.Graham DY, Klein PD. What you should know about the methods, problems, interpretations, and uses of urea breath tests. Am J Gastroenterol. 1991;86:1118–1122. [PubMed] [Google Scholar]
- 18.WHO Working Group. Use and interpretation of anthropometric indicators of nutritional status. Bull World Health Organ. 1986;67:929–941. [PMC free article] [PubMed] [Google Scholar]
- 19.Singer JD, Willett JB. Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence. New York: Oxford University Press; 2003. [Google Scholar]
- 20.Mera RM, Correa P, Fontham EE, et al. Effects of a new Helicobacter pylori infection on height and weight in Colombian children. Ann Epidemiol. 2006;16:347–351. doi: 10.1016/j.annepidem.2005.08.002. [DOI] [PubMed] [Google Scholar]
- 21.Patel P, Mendall MA, Khulusi S, Northfield TC, Strachan DP. Helicobacter pylori infection in childhood: risk factors and effect on growth. BMJ. 1994;309:1119–1123. doi: 10.1136/bmj.309.6962.1119. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Chimonas MA, Baggett HC, Parkinson AJ, Muth PT, Dunaway E, Gessner BD. Asymptomatic Helicobacter pylori infection and iron deficiency are not associated with decreased growth among Alaska Native children aged 7–11 years. Helicobacter. 2006;11:159–167. doi: 10.1111/j.1523-5378.2006.00395.x. [DOI] [PubMed] [Google Scholar]
- 23.Thomas JE, Dale A, Bunn JE, et al. Early Helicobacter pylori colonisation: the association with growth faltering in The Gambia. Arch Dis Child. 2004;89:1149–1154. doi: 10.1136/adc.2002.015313. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Fialho AM, Braga AB, Queiroz DM, Rodrigues MN, Herbster ID, Braga LL. The association between Helicobacter pylori infection and height in children from an urban community in north-east Brazil. Ann Trop Paediatr. 2007;27:55–61. doi: 10.1179/146532807X170510. [DOI] [PubMed] [Google Scholar]
- 25.Mohammad MA, Hussein L, Coward A, Jackson SJ. Prevalence of Helicobacter pylori infection among Egyptian children: impact of social background and effect on growth. Public Health Nutr. 2008;11:230–236. doi: 10.1017/S1368980007000481. [DOI] [PubMed] [Google Scholar]
- 26.Braga AB, Fialho AM, Rodrigues MN, Queiroz DM, Rocha AM, Braga LL. Helicobacter pylori colonization among children up to 6 years: results of a community-based study from Northeastern Brazil. J Trop Pediatr. 2007;53:393–397. doi: 10.1093/tropej/fmm051. [DOI] [PubMed] [Google Scholar]
- 27.Quiñonez JM, Chew F, Torres O, Bégué RE. Nutritional status of Helicobacter pylori-infected children in Guatemala as compared with uninfected peers. Am J Trop Med Hyg. 1999;61:395–398. doi: 10.4269/ajtmh.1999.61.395. [DOI] [PubMed] [Google Scholar]
- 28.Castro-Rodriguez JA, Leo-Barua R, Penny M. Helicobacter pylori is not a determinant factor of persistent diarrhoea or malnutrition in Peruvian children. Trans R Soc Trop Med Hyg. 1999;93:537–539. doi: 10.1016/s0035-9203(99)90372-3. [DOI] [PubMed] [Google Scholar]
- 29.Kehrt R, Becker M, Brösicke H, Krüger N, Helge H. Prevalence of Helicobacter pylori infection in Nicaraguan children with persistent diarrhea, diagnosed by the 13C-urea breath test. J Pediatr Gastroenterol Nutr. 1997;25:84–88. doi: 10.1097/00005176-199707000-00014. [DOI] [PubMed] [Google Scholar]
- 30.Naficy A, Frenck RW, Abu-Elyazeed R, et al. Seroepidemiology of Helicobacter pylori infection in a population of Egyptian children. Int J Epidemiol. 2000;29:928–932. doi: 10.1093/ije/29.5.928. [DOI] [PubMed] [Google Scholar]
- 31.Suoglu OD, Gokce S, Saglam AT, Sokucu S, Saner G. Association of Helicobacter pylori infection with gastroduodenal disease, epidemiologic factors and iron-deficiency anemia in Turkish children undergoing endoscopy, and impact on growth. Pediatr Int. 2007;49:858–863. doi: 10.1111/j.1442-200X.2007.02444.x. [DOI] [PubMed] [Google Scholar]
- 32.Clemens J, Albert MJ, Rao M, et al. Sociodemographic, hygienic and nutritional correlates of Helicobacter pylori infection of young Bangladeshi children. Pediatr Infect Dis J. 1996;15:1113–1118. doi: 10.1097/00006454-199612000-00012. [DOI] [PubMed] [Google Scholar]
- 33.Perri F, Pastore M, Leandro G, et al. Helicobacter pylori infection and growth delay in older children. Arch Dis Child. 1997;77:46–49. doi: 10.1136/adc.77.1.46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Richter T, Richter T, List S, et al. Five- to 7-year-old children with Helicobacter pylori infection are smaller than Helicobacter-negative children: a cross-sectional population-based study of 3,315 children. J Pediatr Gastroenterol Nutr. 2001;33:472–475. doi: 10.1097/00005176-200110000-00010. [DOI] [PubMed] [Google Scholar]
- 35.Choe YH, Kim SK, Hong YC. Helicobacter pylori infection with iron deficiency anaemia and subnormal growth at puberty. Arch Dis Child. 2000;82:136–140. doi: 10.1136/adc.82.2.136. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Leandro Liberato SV, Hernandez GM, Torroba AL, et al. Helicobacter pylori infection in the child population in Spain: prevalence, related factors and influence on growth [in Spanish] An Pediatr (Barc) 2005;63:489–494. doi: 10.1016/s1695-4033(05)70247-2. [DOI] [PubMed] [Google Scholar]
- 37.Yang YJ, Sheu BS, Lee SC, Yang HB, Wu JJ. Children of Helicobacter pylori-infected dyspeptic mothers are predisposed to H. pylori acquisition with subsequent iron deficiency and growth retardation. Helicobacter. 2005;10:249 –255. doi: 10.1111/j.1523-5378.2005.00317.x. [DOI] [PubMed] [Google Scholar]
- 38.Goodman KJ, Joyce SL, Ismond KP. Extragastric diseases associated with Helicobacter pylori infection. Curr Gastroenterol Rep. 2006;8:458–464. doi: 10.1007/s11894-006-0035-3. [DOI] [PubMed] [Google Scholar]