|
|
| |||||||||||||||||||||||||||||||
Background Persistent diarrhea is a prominent feature of the acquired immunodeficiency syndrome in adults, but its cause and its effect on children with human immunodeficiency virus (HIV) infection are largely unknown, particularly in Africa.
Methods We studied a birth cohort of 429 infants born to HIV-positive or HIV-negative mothers in Zaire to determine the incidence of acute, recurrent (
Results Of the 238 infants whose mothers were HIV-positive, 53 were infected, 139 were uninfected, and the HIV status of 46 could not be determined. As compared with uninfected infants, infected infants had higher incidence rates for acute diarrhea (170 vs. 100 episodes per 100 child-years, P = 0.003), recurrent diarrhea (21 vs. 11, P = 0.12), and persistent diarrhea (19 vs. 4, P<0.003). Persistent diarrhea developed in 11 HIV-infected infants; all but 1 died. It also developed in 19 uninfected infants; all but 1 survived. The prevalence of stool pathogens was similar in the two groups. In a multivariate model, persistent diarrhea in an infant was independently associated with symptomatic HIV type 1 infection in the mother (relative hazard, 1.5; P = 0.08). The incidence of persistent diarrhea in the uninfected infants of seropositive mothers was nearly double that in the uninfected infants of seronegative mothers (4.9 vs. 2.7 episodes per 100 child-years), and the risk increased if the mother died (relative hazard, 10.4). Significant growth impairment and severe immunosuppression occurred in the six to eight weeks before the onset of persistent diarrhea.
Conclusions In Zaire, infants with HIV infection have an 11-fold increased risk of death from diarrhea, largely persistent diarrhea, which is often preceded by recurrent episodes of acute diarrhea, malnutrition, or immunosuppression. Illness and death of the mother increase that risk, even among her uninfected infants.
2 episodes), and persistent ( 14 days) diarrhea; outcome; and risk factors.
The risk factors for diarrhea and its natural history in infants with HIV-1 infection remain largely unknown. This report describes the results of a 21-month study that prospectively evaluated the incidence, clinical characteristics, and mortality of acute, recurrent, and persistent diarrhea in 429 infants with and without HIV-1 infection in Kinshasa, Zaire.
Methods
Study Site and Patients
The study was conducted at Mama Yemo Hospital, a 2000-bed public facility serving patients of lower socioeconomic status in Kinshasa. From October 1989 to March 1990, 4656 consecutive women attending the hospital's antenatal clinic were screened for HIV-1 antibodies6. All seropositive women and a control group of seronegative mothers matched for age and parity were recruited for longitudinal follow-up. Four hundred twenty-nine infants were enrolled at birth, including 238 from 232 HIV-1-seropositive women and 191 from 188 HIV-1-seronegative women.
The infants (and their mothers) were seen monthly from birth to 12 months of age and every 2 months thereafter by a team of study physicians in a separate clinic. Physical examinations, anthropometric measurements, and a standardized study questionnaire were completed at each visit. All medical care, medications, and transportation to the clinic were provided by the study. Universal immunization was provided with bacille Calmette-Guerin and oral poliovirus vaccine at birth, diphtheria toxoid-tetanus toxoid-pertussis vaccine at two, four, and six months, and Edmonston-Zagreb measles vaccine at six months. T-cell subgroups were measured at three-month intervals in most infants.
Mothers who missed a regularly scheduled visit were sought at home within one week of the missed visit. Mothers were strongly encouraged to bring their children to the clinic for all episodes of illness. When clinically necessary, the infants were hospitalized at Mama Yemo Hospital and followed by the study research nurse.
Study Design
To determine the incidence rate of diarrhea and its outcome, all reported episodes were followed until resolution or death. Detailed investigations to evaluate behavioral and environmental risk factors associated with acute, recurrent, and persistent diarrhea were initiated on May 22, 1990, when the cohort had reached a mean age of six months. All the infants who were still alive on that date were included and were followed for an additional 16 months or until death. Infants with acute diarrhea were seen by a study team (two physicians and a nurse) to collect information on clinical status, nutrition, and maternal and environmental factors. Follow-up information was collected with a standard questionnaire. Once each week, the fifth study infant who appeared for a routine visit was chosen as a healthy control and similarly evaluated.
Acute diarrhea was defined as a change in the infant's normal stool pattern as reported by the mother, characterized by at least one day of increased frequency, liquidity, blood, or mucus. Episodes were treated with fluids according to the recommendations of the World Health Organization7. Antimicrobial agents were prescribed for dysentery, if cultures grew shigella, salmonella, or campylobacter species, or if trophozoites of Entamoeba histolytica were detected. Mothers were instructed to return with their infants if the diarrhea worsened at any time or continued for 14 days.
Episodes lasting at least 14 days were considered to be persistent diarrhea. Diarrhea after an interval of three diarrhea-free days was arbitrarily considered to be a new episode. Infants with persistent diarrhea were seen every two weeks for the duration of the episode, and a standard questionnaire was completed. Treatment was directed toward maintaining hydration and nutrient intake.
Laboratory Methods
Antepartum serum samples from the women were tested for antibodies to HIV-1 by enzyme-linked immunosorbent assay (Ortho Diagnostics, Raritan, N.J.) and confirmed by Western blotting (Dupont, Geneva). Seropositivity was defined as a reactive enzyme-linked immunosorbent assay and at least two of the following bands on Western blotting: p24, gp41, and gp120/1608. Blood was obtained from the infants at two days of age and every three months thereafter. For the first two specimens, HIV was sought by subjecting approximately 1 µg of DNA to polymerase chain reaction with the SK 38/39 primer pair, which is specific for the nucleotide sequence of p24 core protein, as described elsewhere9. For HIV culture, 2 x 106 to 4 x 106 peripheral-blood mononuclear cells were cocultured with 4 x 106 to 8 x 106 fresh donor cells. Both cell populations were stimulated with phytohemagglutinin, and cultures were maintained in RPMI medium containing 10 percent fetal-calf serum and interleukin-2. Coculture supernatants were tested for p24 antigen every 3 days for 21 days (Coulter Immunology, Hialeah, Fla.).
Stool Analysis
Fresh stool was transported to the microbiology laboratory within two hours of collection, and classic bacterial and protozoal enteric pathogens were sought by standard methods. In addition, 10 colonies of Escherichia coli were picked from plates of MacConkey agar, suspended in sterile low-fat milk (2 percent fat), stored at -20 °C, and transported frozen to Boston to determine whether any of seven E. coli virulence factors were present. The frozen specimen was sampled with a sterile loop, cultured in Luria Bertani broth overnight at 37 °C, and spotted onto Luria Bertani agar. Plasmids containing genes associated with E. coli virulence10 (including heat-labile toxin, heat-stable toxin, enteropathogenic E. coli adherence factor, E. coli attaching and effacing factor, the invasion plasmid antigen of enteroinvasive E. coli, and DNA associated with the enterohemorrhagic 60-kd virulence plasmid of E. coli and with enteroaggregative E. coli) were obtained from Dr. J.B. Kaper (University of Maryland, Baltimore). After appropriate restriction-endonuclease digestion, probe DNA was purified and labeled with 32P by random priming for colony blot hybridization under stringent conditions11. Phenotypic and genotypic characteristics were highly correlated among a subsample of 50 probe-positive isolates, and any hybridizing colony was therefore considered to be a potential pathogen. In addition, the presence of fecal blood was determined by Hemoccult II (SmithKline Diagnostics, San Jose, Calif.), the presence of leukocytes by microscopical analysis of preparations stained with methylene blue, and the presence of fat by Sudan III staining.
T-Cell Analysis
We used automated routine hematologic analysis (Coulter Electronics, Luton, United Kingdom) and lymphocyte phenotyping with fluoresceinated monoclonal anti-CD4 and anti-CD8 antibodies and a fluorescence-activated cell analyzer (Becton Dickinson, San Jose, Calif.) to calculate the absolute CD4 and CD8 counts, the percentage of CD4 and CD8 cells, and the CD4:CD8 ratio.
Statistical Analysis
Confidence intervals for the incidence rates of diarrhea were computed on the assumption that episodes were independent and followed a Poisson distribution. The significance of differences between groups was calculated by the method of Kirkwood,12 and relative hazards were calculated by the method of Rothman and Boice13. Seroreverting infants of infected mothers and seronegative infants of uninfected mothers were analyzed together as "uninfected," except as otherwise indicated. Infants who died or were lost to follow-up before their infection status was determined were excluded from analysis. Symptomatic HIV disease in seropositive mothers was defined as the presence of chronic diarrhea or fever or the loss of more than 10 percent of previous weight.
Incidence rates of acute and recurrent diarrhea were calculated on the basis of events occurring during the final 16 months of the study; incidence rates and relative hazards of persistent diarrhea were calculated over the entire 21 months of the study. The uncorrected chi-square and Fisher's exact tests were used to measure strengths of association for categorical variables; Student's t-test was used for continuous variables. The threshold of significance was 5 percent, and all P values were two-tailed.
Kaplan-Meier product-limit estimates were computed with BMDP 1L and 2L software packages14,15. The Cox-Mantel log-rank statistic was used to compare the equality of two survival distributions. The Cox proportional-hazards model was used to determine independent factors associated with the time to an episode of persistent diarrhea or death due to persistent diarrhea. Intervals were calculated from birth until the beginning of an acute diarrheal episode that became persistent. When persistent diarrhea recurred, only the initial episode was considered in the survival analysis.
To evaluate the risk of persistent diarrhea associated with weight gain and the CD4:CD8 ratio, which changed over time, each episode of persistent diarrhea was randomly matched with an episode of acute diarrhea in a control infant matched for age (within one month) and HIV status. For this comparison, an age-adjusted range of values for the CD4 and CD8 cell counts was calculated from 1645 measurements made during routine well-baby visits in the full cohort of 330 uninfected infants and aggregated over each of the four six-month periods in the first two years of life.
Results
A total of 429 infants were prospectively followed from birth to a mean (±SD) age of 16.9 ±6.4 months. During that time 72 infants (17 percent) died; 87 (20 percent) were lost to follow-up, only 2 of whom were confirmed to be HIV-positive; and 270 (63 percent) were followed for the entire study. Overall, 53 infants had proved HIV-1 infection (group 1), 139 were uninfected offspring of infected mothers (group 2), and 191 were born to uninfected mothers (group 3). Infection status could not be established in 46 infants born to infected mothers (group 4). Twenty-eight infants in group 1 (53 percent), 2 in group 2 (1.4 percent), and 15 in group 3 (8 percent) died during the study. Twenty-seven infants in group 4 (59 percent) also died, most likely reflecting a high prevalence of unconfirmed HIV-1 infection among these infants.
Diarrhea was the most common cause of death in the first year (34 percent). The mortality rate from diarrhea was 12 per 1000 live births among the HIV-negative infants and 132 per 1000 live births among the infected infants (relative hazard, 10.9) (Table 1). Acute diarrhea, recurrent diarrhea, and especially persistent diarrhea were more frequent in infected infants than in seronegative infants (Table 2).
|
|
Acute Diarrhea
Sixty-five percent of the infants had at least one episode of acute diarrhea during the 16 months of intensive study. The incidence in HIV-infected children was 170 episodes per 100 child-years, as compared with 100 per 100 child-years in the uninfected infants (Table 2) (P<0.003). Although the overall rate of acute diarrhea in uninfected infants of seropositive mothers was similar to that in the infants of seronegative mothers, the incidence among the infants of symptomatic mothers was higher than the incidence among the infants of asymptomatic HIV-positive mothers and the infants of HIV-negative mothers (94 vs. 74 episodes per 100 child-years, P = 0.06). Acute diarrhea also occurred more often among the 12 uninfected infants whose infected mothers died during the study (148 episodes per 100 child-years) than among the 383 infants whose mothers remained alive (74 episodes per 100 child-years, P<0.03).
A total of 434 stools were examined for blood, leukocytes, and fat, including 353 stools from infants with acute diarrhea, 20 stools from infants with persistent diarrhea, and 61 control stools. Gross blood was detected in only three acute diarrheal samples (0.8 percent) and in no controls. Leukocytes were seen in 8.4 percent of the acute diarrheal stools and 3.2 percent of the control stools (P = 0.11), and they were not more common in the HIV-1-infected infants than in the uninfected infants with diarrhea. Stainable fat was found with similar frequency in the HIV-positive and HIV-negative infants with diarrhea (27 and 20 percent, respectively; P = 0.24).
Possible pathogens detected in stool are shown in Table 3. There was no difference in the prevalence of any organism between seropositive and seronegative infants with diarrhea, or between seronegative infants with and without diarrhea. When any organism or any probe-positive E. coli was considered to be a possible pathogen, similar percentages of seropositive and seronegative infants with diarrhea and seronegative control infants without diarrhea had pathogens (62, 62, and 65 percent, respectively).
|
Recurrent Diarrhea
Two or more episodes of acute diarrhea, occurring in clusters of up to eight episodes per child, were observed in 126 infants (29 percent) during the 16 months of detailed observation. The median interval between episodes was 68 days. The incidence of recurrent episodes of acute diarrhea was 1.6 to 2.3 times higher in seropositive infants than in seronegative infants (Table 2). As compared with infants who had only one episode of acute diarrhea, those who had two or more episodes had a relative risk of persistent diarrhea of 4.4, increasing to 5.1 with three or more episodes and to 7.0 with four or more (overall P<0.001).
Persistent Diarrhea
Persistent diarrhea occurred 34 times in 31 of the 429 infants. Persistent diarrhea developed in 11 HIV-infected infants; all but 1 died. It also developed in 19 uninfected infants; all but 1 survived. Twelve episodes occurred in infected infants, 21 in uninfected infants, and 1 in an infant whose HIV status remained uncertain. Three infants (one of them infected) had two bouts of persistent diarrhea each. The mean age at the onset of persistent diarrhea was 9.3 months, and nearly 80 percent of the episodes occurred before 12 months of age. The incidence of persistent diarrhea was nearly five times higher in the infected infants than in the uninfected infants (P<0.003) (Table 2).
Infants with three or more episodes of acute diarrhea, whether infected or uninfected, were significantly more likely to have persistent diarrhea than those with two episodes or fewer (Figure 1) (P<0.001 by Cox-Mantel log-rank statistic). This effect was independent of HIV status in a Cox proportional-hazards regression model. Despite their greater tendency to have persistent diarrhea, HIV-1-infected infants with multiple episodes of acute diarrhea did not have an elevated risk of death from diarrhea if persistent diarrhea did not develop (data not shown).
|
|
Mortality from Diarrhea
Of the 45 deaths in children whose HIV status was known, 11 (24 percent) were due to persistent diarrhea, and 10 of the 11 (91 percent) occurred in infected infants. This corresponds to a relative risk of death of 6.1 in all infants if an episode of persistent diarrhea occurred during the observation period (P = 0.03). The median interval from the beginning of an episode of persistent diarrhea to death was 31 days, with a range of 14 to 54. Of the 28 deaths among the 53 infected infants, 14 were due to diarrhea, and 10 of those were due to persistent episodes. This corresponds to a cause-specific mortality rate of 113 per 1000 live births. Only 1 of 21 episodes of persistent diarrhea among 19 uninfected infants resulted in death (P = 0.001). Thus, in this cohort an infected infant with persistent diarrhea had an 83 percent likelihood of dying from it, as compared with a 4.7 percent likelihood in an uninfected infant (relative hazard from the Cox model, 82; P<0.001).
The infants with HIV-1 infection tended to be more wasted at the beginning of an acute diarrheal episode that became persistent than at the beginning of an episode that did not persist. The mean (±SD) weight-for-height z score of infected infants was -1.5 ±1.2 at the beginning of an episode that persisted, as compared with -0.8 ±1.5 at the beginning of a self-limited acute episode (P = 0.15). Among 26 analyzable pairs, mean (±SE) weight gain was significantly less in the month before an acute episode of diarrhea that persisted than in the month before one that resolved (59 ±69 vs. 348 ±93 g, P = 0.008).
T-cell subgroups were analyzed within the two months before an episode of persistent diarrhea in 9 of 12 infected children. As compared with diarrhea-free controls matched for age and HIV status, these infants had significantly depressed CD4:CD8 ratios (mean [±SD] ratio, 0.54 ±0.29 vs. 0.93 ±0.03; P = 0.001). All but one infant with persistent diarrhea had CD4: CD8 ratios that were below the 5th percentile of age-adjusted Zairian normal values.
Analysis of feeding, environmental, and behavioral data did not identify additional risk factors for acute or persistent diarrhea (data available elsewhere*).
Discussion
Our study provides a quantitative estimate of the effect of HIV-1 infection on morbidity and mortality from diarrheal disease among infants in Zaire. Episodes of acute, recurrent, and persistent diarrhea occurred significantly more frequently among infected infants than among uninfected controls. These data are similar to those observed with different reporting techniques in a comparable population in Zaire,4 and they probably underestimate the rates in the general population, because the study infants had universal measles immunization and readily available medical care. Diarrhea was the leading cause of mortality in the present cohort, accounting for 36 percent of all deaths. When HIV-1 infection was also present, the mortality rate from diarrhea increased 11-fold. And, as is now understood for all populations with access to treatment for acute dehydration, persistent diarrhea was much deadlier than acute diarrhea. The high rate of mortality from persistent diarrhea was strikingly concentrated in the group with HIV-1 infection, for which the relative hazard of death due to persistent diarrhea was 82, as compared with the uninfected group.
We did not find specific associations between particular enteric pathogens and HIV infection, even using DNA probes to detect E. coli with virulence genes. The similar prevalence of these organisms in HIV-negative infants with diarrhea and without indicates the high prevalence of pathogens in the community and the susceptibility of all infants to the same pathogens. Because we have not yet looked for enteric viruses, we cannot exclude specific associations with HIV-1 in these infants.
Some of the increased risk of death among the infants with HIV-1 infection may be attributable to the effects of malnutrition. Failure to thrive is a well-known manifestation of HIV-1 infection in children. Among HIV-1-infected infants in this cohort, not only was the weight-for-height z score lower at the onset of episodes of acute diarrhea that persisted than at the onset of self-limited episodes, but there was also significantly greater growth impairment in the six to eight weeks before an episode of persistent diarrhea than in the period before an episode of acute diarrhea that did not persist. This is consistent with cross-sectional data from a similar cohort showing that diarrhea was associated with severe malnutrition and HIV-1 infection16. Poor nutrition is also related to the occurrence of both recurrent17 and persistent18,19 diarrhea in populations without HIV-1 infection, and it is a reported risk factor for death from persistent diarrhea20,21,22. In fact, among Bangladeshi infants, malnutrition associated with persistent diarrhea increased the risk of death 68-fold,18 an increase approaching that associated with HIV-1 in our study.
The risk of persistent diarrhea was also related to the recurrence of acute diarrhea independently of HIV status. This is reminiscent of observations in Bangladesh, where children with persistent diarrhea had a 46 percent higher incidence of acute diarrhea during the preceding 12 months than controls without persistent diarrhea23.
Repeated episodes of acute diarrhea often lead to increasingly severe protein-energy malnutrition, specific micronutrient deficiency (e.g., zinc or vitamin A deficiency), or a combination of the two, especially when nutrient intake is limited24. Independently of HIV-1 infection, malnutrition impairs tissue repair, reduces immunologic function, and may inhibit host defenses, ultimately compromising the ability to resolve acute diarrhea25. The infants with HIV-1 infection who had persistent diarrhea in this study were markedly immunosuppressed in the weeks before the episode, as measured by the CD4:CD8 ratio. We used the CD4:CD8 ratio for this analysis because the absolute numbers of CD4 and CD8 T cells varied greatly among the infants and, in individual infants, with age. Whether this increased immunosuppression was due entirely to the effects of HIV-1 infection itself or was in part due to the resulting malnutrition is not clear. In either event, the cyclic effects of diarrhea, malnutrition, and immune dysfunction can produce an accelerated downward course in these patients.
The clinical course was also adversely affected by the early introduction of milk formula or solid food, both of which are associated with a 25 to 200 percent increase in the duration of diarrhea in infants without HIV infection26. Together, these data suggest that nutritional or biologic factors associated with recurrent episodes of acute diarrhea condition the host for subsequent persistent diarrhea.
Public health measures to prevent recurrent diarrhea and blunt the associated malnutrition may have a substantial effect on early mortality in infants with HIV infection in Africa, although their effect on the natural history of the infection remains to be determined. Because all infected infants will probably succumb to AIDS, the greater concern shifts to the uninfected offspring of infected mothers. The data in this study show that maternal health is an important risk factor for diarrheal morbidity in infants. Although the frequency of diarrhea was similar in uninfected infants born to HIV-positive and HIV-negative mothers, the number of episodes increased when the mothers had clinical AIDS and was even greater if the mothers died during the observation period. The risk of persistent diarrhea also increased among infected infants whose mothers died and tended to increase if the mothers were symptomatic. These findings may be explicable in the light of data from Mexico City showing that the daytime absence of mothers was a significant risk factor for persistent diarrhea in their children. Thus, it is reasonable to propose that in our cohort diarrheal morbidity, and by implication mortality, was related to the ability of the mother to care for her infant and maintain infant hygiene and nutrition, especially during episodes of acute diarrhea needing increased attention. The implication of a maternal-care factor in the prevention of excess diarrheal episodes and mortality becomes extremely important given the estimated 5.5 million AIDS orphans there will be in Africa by the end of the decade2. Interventions designed to delay morbidity and mortality among adults attributable to HIV infection may make a major contribution to the long-term survival of the uninfected children of infected mothers. Unfortunately, most programs for children still fail to recognize this and to direct resources to parental health.
Supported by an International Cooperation in AIDS Research grant (PO1-AI-26698, to Dr. Keusch) from the National Institute of Allergy and Infectious Diseases.
We are indebted to Andy Vernon, Robin Ryder, Veronique Battner, Frieda Behets, and the team of the perinatal study and the oral rehydration center at Mama Yemo Hospital for assistance in the field; and to David Acheson, Arthur Donohue-Rolfe, Sabine DeBreuker, Lisa Lincicome, and Ann Kane at New England Medical Center for laboratory studies.
Source Information
From the Division of Geographic Medicine and Infectious Diseases, New England Medical Center, Boston (D.M.T., W.M.R., G.T.K.); the Centers for Disease Control and Prevention, Atlanta (M.E.S.); the International Cooperation in AIDS Research Unit, Projet SIDA, Kinshasa, Zaire (D.M.T., U.A., K.K., B.K., M.M., T.T., C.K.); the Department of Pediatrics, Mama Yemo Hospital, Kinshasa, Zaire (F.D.); and the National Institute of Allergy and Infectious Diseases, Bethesda, Md. (C.B.).
Address reprint requests to Dr. Keusch at the Division of Geographic Medicine and Infectious Diseases, New England Medical Center, Box 041, 750 Washington St., Boston, MA 02111.
References
| |||||||||||||||||||||||||||||||
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
Previous
