The Protective Effectiveness of Natural Rotavirus Infection in an American Indian Population

Transcription

1 1562 The Protective Effectiveness of Natural Rotavirus Infection in an American Indian Population Lawrence H. Moulton, Mary A. Staat, Mathuram Santosham, and Richard L. Ward Department of International Health, Department of Biostatistics, and Center for American Indian and Alaskan Native Health, Johns Hopkins University School of Hygiene and Public Health, Baltimore, Maryland; Division of Infectious Diseases, Children s Hospital Medical Center, Cincinnati, Ohio The degree of protection conferred by natural rotavirus infection was estimated through analyses of data gathered as part of a 2-year rotavirus vaccine study of 1185 Native American infants. In 292 placebo recipients with complete serum sample sets, rotavirus IgA antibody indicative of infection before 2 months of age was associated with a 58% decrease in symptomatic infections throughout the trial. In all 391 placebo recipients, the preventive effectiveness of an initial symptomatic infection was 72% overall and 94% within 6 months following the infection. In contrast to studies conducted at other sites in the United States, serotype G3 was the predominant serotype associated with gastrointestinal episodes (80%). The effectiveness of an initial serotype G3 episode with respect to preventing subsequent serotype G3 episodes was 91%. Rotavirus is the leading cause of severe diarrhea morbidity in infants and children throughout the world [1, 2]. Although prevention of rotavirus gastroenteritis has been deemed to be of high priority by international agencies and the Institute of Medicine [3], recent studies of tetravalent rhesus rotavirus vaccine have demonstrated overall efficacy against rotavirus disease of only 50% (with a range of 24% 66%) in children!2 years of age [4 10]. The protection conferred by a natural infection may be regarded as a target for vaccine development efforts. Most studies of rotavirus diarrhea have been unable to address the issue of natural immunity because of the relatively small numbers of episodes or follow-up time in the study group. Studies that have been able to furnish estimates generally have found high protective effectiveness of an initial rotavirus infection or symptomatic episode [11 15], but this has not always been the case in populations with high exposure [16]. Although rotavirus antibody has not been definitively established as the mediator of protection against rotavirus disease, it has been reported that anti-rotavirus antibody both results from and generally cor- Received 2 April 1998; revised 10 July Written informed consent was obtained from parents or guardians of all study participants. The study protocol was approved by the Johns Hopkins University Committee on Human Research, the Indian Health Service, and the tribal councils and health boards of the respective tribes. The opinions expressed by the authors are not necessarily those of the Indian Health Service. Financial support: NIH (AI to L.H.M.). The data were collected under a grant from Wyeth-Ayerst Research. Reprints or correspondence: Dr. Lawrence H. Moulton, Dept. of International Health, 615 N. Wolfe St., Baltimore, MD (lmoulton@ jhsph.edu). The Journal of Infectious Diseases 1998;178: by the Infectious Diseases Society of America. All rights reserved /98/ $02.00 relates with protection against further rotavirus infections [17 20]. A rotavirus vaccine efficacy trial was conducted in Native American communities in the southwestern United States in to evaluate the safety and efficacy of a monovalent rhesus rotavirus vaccine (RRV-S1) and a tetravalent rhesus rotavirus vaccine (RRV-TV) [6]. The relatively large number of rotavirus gastroenteritis episodes in the placebo group enabled us to control for age and seasonality in evaluating the role of rotavirus infections in preventing further symptomatic and asymptomatic rotavirus infections. During the trial, RRV-TV vaccine was 50% efficacious in preventing all rotavirus gastroenteritis episodes and 53% efficacious for serotype G3 episodes. In this population, unlike other US populations, serotype G3 predominated, affording us the opportunity to assess the protective effect of an initial serotype G3 infection. Subjects and Methods Study participants. Native American infants aged 2 6 weeks residing on the Gila River Indian Reservation, the Navajo Nation, the San Carlos Apache Indian Reservation, and the White Mountain Apache Indian Reservation were enrolled in a three-arm double-masked trial to evaluate the efficacy of RRV-TV ( pfu/ dose) and RRV-S1 ( pfu/dose) vaccines [6]. Of the 1185 infants enrolled between January and October 1992, there were 391 in the placebo arm, 398 in the RRV-S1 arm, and 396 in the RRV- TV arm. Disease surveillance continued until 31 January Overall, 91% of enrolled infants completed the first year of surveillance, and 77% reenrolled and completed the second year of surveillance. Analysis populations. For the analyses of asymptomatic rotavirus infections, we include the subset of participants for whom results of 3 serum samples were available for determination of

2 JID 1998;178 (December) Repeat Rotavirus Infection 1563 immune responses. This subset comprises the 292 placebo recipients from whom samples were obtained just before the first immunization (median age, 2 months) and the third immunization (median age, 5 months) and at 12 months of age. We restrict analyses of the protective effects of asymptomatic infections to this group to avoid selection biases and noncomparability across subgroups defined by what sets of serologic results were available. Otherwise, for example, we might be comparing a child who had evidence of infection in a given interval with a child who had no opportunity to have an infection detected in that interval because of lack of the follow-up serum sample. For the analyses of the protective effect of symptomatic rotavirus infections, we have included all 391 placebo recipients. Definitions of episodes and infections. Symptomatic rotavirus episodes were characterized by the isolation of rotavirus in infants with gastrointestinal illness (three or more watery or loose stools in a day or projectile vomiting). We required 30 days between defined episodes to ensure the distinction between repeat symptomatic episodes. Asymptomatic infections are defined by a 4- fold rise in rotavirus IgA between successive blood samples among those who did not have a symptomatic rotavirus infection between the samples. At the first sample, those with rotavirus IgA 20 U/ ml were considered to have had a prior rotavirus infection. This was considered conservative, since the limit of detection was 4 U/ ml. The term infection in isolation refers to both symptomatic rotavirus episodes and asymptomatic infections. Because the last blood specimen was collected at 12 months of age, only symptomatic rotavirus infections were detectable after that time. Surveillance and analysis of stools. Homes of study participants were visited weekly to interview the caretakers regarding their infants health and to collect stool specimens from those infants who had an episode of gastrointestinal illness. There was no systematic collection of stools from asymptomatic children. Specimens could be collected at any time between episode onset and 7 days after episode resolution. After initial storage at 20 C for 1 day, they were stored at 70 C until the next monthly shipment to the J. N. Gamble Institute of Medical Research (Cincinnati) for rotavirus detection [21] and serotyping [22]. Rotavirus IgA analysis. Serum rotavirus IgA was determined by ELISA as described previously [23], except that a lysate of RRV rather than was used to bind rotavirus IgA to the wells of the plates. Statistical analyses. Results are phrased in terms of the effectiveness (complement of the risk or rate ratio) of an infection or episode and its associated 95% confidence interval. We used Taylor series based confidence intervals for the first set of analyses of risk ratios [24], in which follow-up time in the analysis intervals was relatively constant across participants, with P values from twosided Fisher s exact tests [25]. In the second set of analyses, time-to-event Cox survival models were fit to analyze the intervals from enrollment to the start of the first rotavirus episode and from the first to the second episode. This is accomplished by taking as the outcome the time from enrollment to the second episode but including a time-dependent variable that is zero until the start of an initial rotavirus episode, at which time it jumps to unity [26 28]. The exponentiated coefficient of this timedependent variable gives the instantaneous relative risk of the occurrence of a second episode among those with a first episode compared with the risk in those who have not yet experienced an episode. Subjects were censored at the earliest of the following: a second rotavirus episode, loss to follow-up, or the end of the study (31 January 1994). Variations on this approach are described below with the relevant results. A close adjustment for age was incorporated using a three-knot restricted cubic spline [29]. Calendar time was used for the Cox model time line to account for seasonality of rotavirus exposure. Thus, participants enter the risk set at the time of year at which they were enrolled, and when an episode occurs, the status (zero, one, or two episodes) of each participant at that calendar time is ascertained and compared with the status of each other participant. Asymptomatic infections were ignored in this set of time-to-event analyses for two reasons: Their serotype was unknown, and they occurred very frequently throughout the observation period to almost all infants, so their timing relative to the symptomatic episodes was not known. Against this background level of asymptomatic episodes, the observed effectiveness associated with the symptomatic episodes might be interpretable as follows: It is the maximum effectiveness we could expect, in this population, of a vaccine delivered at the same point in time (in place of the episode). Results Protective effects of early asymptomatic or symptomatic infections. Table 1 contains the asymptomatic and symptomatic rotavirus infection and episode patterns of all 292 participants enrolled in the placebo arm with results from all 3 serum samples. At the time of enrollment (receipt of first placebo dose), 21% (62/292) had experienced a rotavirus infection, according to our definition (rotavirus IgA 20 U/mL). Of those enrolled by 8 weeks of age, 8% (8/116) had evidence of a prior rotavirus infection, compared with 30% (53/176) of those enrolled between 8 and 25 weeks of age. Including these initial infections, by the end of the study, 88% (256) of the participants experienced at least one rotavirus infection (only symptomatic infections were recorded after a child was 12 months of age), 20% (58) experienced at least two infections, and 2% (5) experienced three infections. These data are further summarized to yield the effectiveness analyses in table 2. The effectiveness of a preenrollment infection with respect to preventing rotavirus disease episodes during the entire trial was 58% (table 2). This increased to 70% when attention was restricted to episodes occurring only within the first year of life (not shown). The effectiveness of a single rotavirus infection (either asymptomatic or symptomatic) at any time before the third administration of placebo was 45% for prevention of a subsequent rotavirus gastroenteritis episode, while it was 100% for those who experienced two infections (table 2). This latter estimate, however, was based on only 10 early double infections. Protective effects, overall and serotype-specific, of symptomatic infections. There were 128 symptomatic rotavirus episodes in the entire placebo group of 391 infants. Their serotype distribution was as follows: 29 indeterminate, 12 G1, 8 G2, and

3 1564 Moulton et al. JID 1998;178 (December) Table 1. Frequency of different rotavirus infection patterns based on age(s) at positive serologic assays and presence or absence of symptoms. Pre-2M 2M 5M 5M 12M 12M 24M No. of children % of children 0 0 A S A A A 0 A S A A S A 0 S A A A S A 0 S S S A 0 0 S S A SS A S A A A A A S A S A 0 A S A S S S 0 S S S SS S SS Total 62A 43A 103A S 68S 29S NOTE. There were 319 infections in 292 placebo recipients with complete data during study period. 2M, 5M, 12M, median age at time of 3 serum samples ( M months); 24M, median age at end of year 2 of follow-up. A, infection defined by rise in IgA only; S, rotavirus gastrointestinal illness episode documented by both stool analysis and symptomatology (SS, 2 such episodes); 0, no positive serology or stool specimen. Before first serum specimen (pre-2m), no data on symptoms were collected, and some infections may have been symptomatic. 79 G3. We fit a series of Cox models that analyze the time from enrollment to second symptomatic rotavirus episode, with the first such episode treated as a time-dependent covariate (table 3). The first model (A) evaluated the effectiveness of a symptomatic infection of any serotype for prevention of a subsequent symptomatic infection of any serotype. The estimated effectiveness was 72%. The analysis then was restricted to only serotype G3 episodes by censoring at the first nondeterminate or non serotype G3 episode, with the relevant covariate being the first serotype G3 episode (model B). The serotype-specific effectiveness for serotype G3 was 91%. Similarly, we fit serotypespecific models restricted to serotype G1 and G2 episodes, with estimated effectiveness of 544% ( P.08) and 100% ( P.06), respectively, but they were based on very few episodes. We also fit a model to estimate heterotypic protection: The effectiveness of an initial serotype G3 episode for prevention of serotype G1 or G2 episodes was 44% ( P.51). Again, small numbers render interpretation difficult and result in wide confidence intervals. For example, even a 50% heterotypic effectiveness would not be incompatible with the data. We fit models to estimate the degree of waning of effectiveness of an initial episode with respect to subsequent episodes. The first such model was similar to model A but with an additional interaction term denoting whether the second episode occurred within 6 months of the first episode or not. We found that within the first 6 months following an initial symptomatic rotavirus episode, there was 94% effectiveness associated with that initial episode, dropping to only 36% thereafter ( P.03 for interaction, i.e., difference between the two estimates). The second model was similar to model B, that is, for serotype G3 episodes only. It gave an estimate of 100% effectiveness within 6 months after the initial episode and 72% thereafter (P.15 for the difference). Discussion In this American Indian population, there appeared to be a high degree of exposure to circulating rotavirus, with evidence of 86% of the infants having acquired an infection by the end of the first year of the trial, and 34% having had a symptomatic rotavirus episode detected in the first year. Reinfection rates were also high; 26% of those with at least one infection had at least a second, and 10% of those with an infection in the first year had a symptomatic rotavirus infection in the second year, in spite of the sharply declining attack rates over time. These figures contrast sharply with those in the general US population. In one US multicenter study, only 16% of placebo recipients had a symptomatic rotavirus infection in the first year, and of the 40% who had any evidence of rotavirus infection during the first year, only 1% had a symptomatic infection during the second year [30]. In another US multicenter study, 25% of placebo recipients had a symptomatic rotavirus infection in the first season [7], and in a Finnish multicenter study, this figure was only 16% over two seasons. Moreover, unlike in the general US population, rotavirus infections remained endemic throughout the year in our study population. These Table 2. Effectiveness of rotavirus infections (asymptomatic or symptomatic) for prevention of further symptomatic rotavirus infections during study period. Age interval Previous infection No. with symptomatic infection during interval/total (%) Effectiveness (%) 95% confidence interval 2M 24M Pre-2M 10/62 (16) None 88/230 (38) M 24M One!5M 19/99 (19) None 68/193 (35) M 24M Two!5M 0/10 (0) None 68/183 (37) a.015 NOTE. Effectiveness (1 risk ratio ) 100%. 2M, 5M, 12M, median age at time of 3 serum samples ( M months); 24M, median age at end of year 2 of follow-up. a Calculated exactly under odds ratio approximation to risk ratio. P

4 JID 1998;178 (December) Repeat Rotavirus Infection 1565 Table 3. Time-to-episode Cox models for evaluating effectiveness of a given first rotavirus episode for prevention of a second episode. Model Effectiveness (%) 95% confidence interval P A. Time to any episode after first episode Overall: mo.: mo.: B. Time to serotype G3 episode after first serotype G3 episode Overall: mo.: mo.: NOTE. Effectiveness (1 risk ratio ) 100%. Results are given for entire study period (overall) and whether interval of risk for second episode is within 6 months ( 6 mo.) or after 6 months (16 mo.) following first episode., not evaluable because of insufficient numbers. differences could be due to any of a number of variables: factors related to lower socioeconomic status, genetic factors, or other environmental or behavioral factors. Our data analyses indicate there was a substantial protective effect associated with rotavirus infections before enrollment, from 58% to 70% depending on the length of observation, 2 years or 1, respectively. Of the placebo infants in our subset with 3 serum samples, 116 were enrolled before 8 weeks of age; of these, 8% were positive for IgA compared with 30% of the 176 enrolled at 8 weeks. Although this suggests that the majority of the preenrollment rotavirus infections were acquired postneonatally, others have reported that neonatal infections also provide substantial protection against subsequent rotavirus disease in both an industrialized [31] and a less industrialized [32] country. For a single asymptomatic rotavirus infection before the end of the immunization series ( 6 months), the effectiveness was 45%, which increased to 100% for two such infections. This latter number is based on only 10 instances of early double infections; still, these findings are consonant with the results of Velazquez et al. [15], who noted steadily increasing effectiveness against future infections associated with number of infections. The time-to-episode Cox models showed relatively high effectiveness (72%) of an initial symptomatic rotavirus infection against further episodes. This is very close to the 77% effectiveness of a symptomatic episode reported by Velazquez et al. [15] in a Mexican population with comparable levels of rotavirus infections (about 8% infected by 2 months of age, identical to our population, and 67% infected by 12 months of age compared with our 86%) and using similar statistical methods. We observed even higher homotypic protection (for serotype G3, the predominant strain in this population), which was estimated to be 91%. We had insufficient data to make definitive statements about heterotypic protection. Our homotypic protection results accord with those of many researchers in diverse populations. Ward and Bernstein [30], Bernstein and colleagues [11, 13], Ukae et al. [12], and Linhares et al. [14] all reported 190% protection conferred by an initial rotavirus infection. Reves et al. [16] reported no overall protective effect of an initial infection, but in that Egyptian population there appeared to be greater serotypic diversity and exposure than in many studied populations. It has been suggested that multiple infections may result in heterotypic immunity [12, 33]. In spite of an apparent high level of circulating rotavirus, there were only 5 infants with more than two detected rotavirus infections (symptomatic and asymptomatic combined) among our placebo recipients. This may be a consequence, however, of high homotypic protection, lack of exposure to other serotypes, aging effects, or incomplete surveillance. Velazquez et al. [15] reported some evidence of increased homotypic protection, but definitive results with respect to heterotypic immunity are difficult to obtain; they require very large studies in populations with much greater serotypic variation than is seen in many settings. We had evidence of significant waning of the protective effect of a symptomatic rotavirus episode. Although Chiba et al. [34] reported a similar result, Bernstein et al. [11] demonstrated high homotypic protection through the year following the first infection. These discrepancies could be due to differences in degree of exposure or levels of circulating rotavirus, to withinserotype genetic diversity, or to immune system or nutritional status differences. There are two principal sources of bias in our analyses. The first is that they ignore variability of exposures and susceptibility to rotavirus among the infants. As noted by Reves et al. [16], there are pockets, or subsets, of infants at high risk who tend to experience more reinfections. Infants who have experienced one infection by a given point in time may be more highly exposed in the future than are those who have had no infections. This will bias effectiveness estimates toward zero. The other potential bias is that there may be more intervening infections than we have detected, with priming effects boosting the apparent effectiveness of an observed infection (symptomatic or asymptomatic). A third potential source of bias derives from a lack of specificity in our definition of rotavirus episode; rotavirus was the only pathogen for which we had data and, although detected, may not have been the cause of a given gastrointestinal illness episode. Accurate assessment of the degree of protection afforded by natural rotavirus infections is rendered difficult by the above sources of bias and by factors such as age- and season-related differences in susceptibility and exposure. We have accounted for these factors to the extent possible through careful attention to age- and time-related risk sets and relevant comparison

5 1566 Moulton et al. JID 1998;178 (December) groups, leading to conclusions of high homotypic protection and relatively high overall protection in this American Indian study population. The data suggest that if we take the protective effectiveness of naturally conferred immunity as an upper limit, there is still room for improvement of multivalent rotavirus vaccines with multidose protocols. Acknowledgments We thank Janné Croll and Robert Weatherholtz for their organization of the data collection and management and Donna Sander and Douglas Knowlton for conducting the laboratory analyses. References 1. Jin S, Kilgore PE, Holman RC, Clarke MJ, Gangarosa EJ, Glass RI. Trends in hospitalizations for diarrhea in U.S. children from 1979 through 1992: estimates of the morbidity associated with rotavirus. Pediatr Infect Dis J 1997;15: Cook SM, Glass RI, LeBaron CW, Ho MS. Global seasonality of rotavirus infections. Bull World Health Organ 1990;68: Institute of Medicine. The prospects for immunizing against rotavirus. In: New vaccine development: establishing priorities. Vol 2. Diseases of importance in developing countries. Washington, DC: National Academy Press, 1986: Bernstein DI, Glass RI, Rodgers G, Davidson BL, Sack DA. Evaluation of rhesus rotavirus monovalent and tetravalent reassortant vaccines in US children. JAMA 1995;273: Perez-Schael I, Guntinas MJ, Perez M, et al. Efficacy of the rhesus rotavirus based quadrivalent vaccine in infants and young children in Venezuela. N Engl J Med 1997;337: Santosham M, Moulton LH, Reid R, et al. Efficacy and safety of high-dose rhesus-human reassortant rotavirus vaccine in Native American populations. J Pediatr 1997;131: Rennels MB, Glass RI, Dennehy PH, et al. Safety and efficacy of high-dose rhesus-human reassortant rotavirus vaccines report of the National Multicenter Trial. Pediatrics 1996;97: Lanata CF, Midthun K, Black RE, et al. Safety, immunogenicity, and protective efficacy of one and three doses of the tetravalent rhesus rotavirus vaccine in infants in Lima, Peru. J Infect Dis 1996;174: Linhares AC, Gabbay YB, Mascarenhas JD, et al. Immunogenicity, safety and efficacy of tetravalent rhesus-human, reassortant rotavirus vaccine in Belem, Brazil. Bull World Health Organ 1996;74: Joensuu J, Koskenniemi E, Pang XL, Vesikari T. Randomised placebo-controlled trial of rhesus-human reassortant rotavirus vaccine for prevention of severe rotavirus gastroenteritis. Lancet 1997;350: Bernstein DI, Sander DS, Smith VE, Schiff GM, Ward RL. Protection from rotavirus reinfection: 2-year prospective study. J Infect Dis 1991;164: Ukae S, Nakata S, Adachi N, Kogawa K, Chiba S. Efficacy of rhesus rotavirus vaccine MMU against gastroenteritis due to serotype 1 rotavirus. Vaccine 1994;12: Bernstein DI, Smith VE, Sander DS, Pax KA, Schiff GM, Ward RL. Evaluation of WC3 rotavirus vaccine and correlates of protection in healthy infants. J Infect Dis 1990;162: Linhares AC, Gabbay YB, Mascarenhas JD, et al. Epidemiology of rotavirus subgroups and serotypes in Belem, Brazil: a three-year study. Ann Inst Pasteur Virol 1988;139: Velazquez FR, Matson DO, Calva JJ, et al. Rotavirus infections in infants as protection against subsequent infections. N Engl J Med 1996;335: Reves RR, Hossain MM, Midthun K, et al. An observational study of naturally acquired immunity to rotaviral diarrhea in a cohort of 363 Egyptian children. Calculation of risk for second episodes using age-specific personyears of observation. Am J Epidemiol 1989;130: Matson DO, O Ryan ML, Herrera I, Pickering LK, Estes MK. Fecal antibody responses to symptomatic and asymptomatic rotavirus infections. J Infect Dis 1993;167: O Ryan ML, Matson DO, Estes MK, Pickering LK. Anti rotavirus G typespecific and isotype-specific antibodies in children with natural rotavirus infections. J Infect Dis 1994;169: Clemens JD, Ward RL, Rao MR, et al. Seroepidemiologic evaluation of antibodies to rotavirus as correlates of the risk of clinically significant rotavirus diarrhea in rural Bangladesh. J Infect Dis 1992;165: Ward RL, Clemens JD, Knowlton DR, et al. Evidence that protection against rotavirus diarrhea after natural infection is not dependent on serotypespecific neutralizing antibody. J Infect Dis 1992;166: Ward RL, Bernstein DI, Young EC, Sherwood JR, Knowlton DR, Schiff GM. Human rotavirus studies in volunteers: determination of infectious dose and serological response to infection. J Infect Dis 1986;154: Ward RL, McNeal MM, Clemens JD, et al. Reactivities of serotyping monoclonal antibodies with culture-adapted human rotaviruses. J Clin Microbiol 1991;29: Bernstein DI, Smith VE, Sherwood JR, et al. Safety and immunogenicity of live, attenuated human rotavirus vaccine Vaccine 1998;16: Stata. Version 5.0. College Station, TX: Stata, Thomas DG, Gart JJ. Improved and extended exact and asymptotic methods for the combination of 2 2 tables. Version Comput Biomed Res 1992;25: Cox DR. Regression models and life tables (with discussion). J R Statist Soc [B] 1972;34: Moulton LH, Dibley MJ. Multivariate time-to-event models for studies of recurrent childhood diseases. Int J Epidemiol 1997;26: SAS system. Version Cary, NC: SAS Institute, Harrell FE Jr, Lee KL, Pollock BG. Regression models in clinical studies: determining relationships between predictors and response. J Natl Cancer Inst 1988;80: Ward RL, Bernstein DI. Protection against rotavirus disease after natural rotavirus infection. J Infect Dis 1994;169: Bishop RF, Barnes GL, Cipriani E, Lund JS. Clinical immunity after neonatal rotavirus infection. A prospective longitudinal study in young children. N Engl J Med 1983;309: Bhan MK, Lew JF, Sazawal S, Das BK, Gentsch JR, Glass RI. Protection conferred by neonatal rotavirus infection against subsequent rotavirus diarrhea. J Infect Dis 1993;168: Pichichero ME, Losonsky GA. Asymptomatic infections due to wild-type rotavirus may prime for a heterotypic response to vaccination with rhesus rotavirus. Clin Infect Dis 1993;16: Chiba S, Nakata S, Ukae S, Adachi N. Virological and serological aspects of immune resistance to rotavirus gastroenteritis. Clin Infect Dis 1993; 16(suppl 2):S