Published ahead of print on September 6, 2007, doi:10.1164/rccm.200703-435OC
© 2007 American Thoracic Society doi: 10.1164/rccm.200703-435OC
Heredity versus Environment in Tuberculosis in TwinsThe 1950s United Kingdom Prophit Survey—Simonds and Comstock Revisited1 Department of Infectious Diseases and 2 Department of Clinical Epidemiology, Leiden University Medical Center, Leiden, The Netherlands Correspondence and requests for reprints should be addressed to Jaap T. van Dissel, M.D., Ph.D., Department of Infectious Diseases, C5-P, Leiden University Medical Center, P.O. Box 9600, 2300 RC Leiden, The Netherlands. E-mail: j.t.van_dissel{at}lumc.nl
Rationale: In his 1978 article on tuberculosis (TB) in twins, Comstock concluded that the 2.5-fold higher concordance rate for TB among monozygotic versus dizygotic twins in the Prophit survey of the 1950s implicated inherited susceptibility as a major risk factor for TB in humans. His analysis did not take into account strong imbalance of variables within subgroups, underestimating possible confounding effects of environmental factors. Objectives: To reconsider the role of environmental versus hereditary factors in determining the concordance rate of TB among twin pairs. Methods: Reanalysis of the Prophit Survey. Measurements and Main Results: A known Mycobacterium tuberculosis–positive or M. tuberculosis–negative sputum in the index TB case markedly influenced the odds ratio (OR) of concordance in the twin pairs. In 87 pairs with co-twins exposed to a sputum-negative index case, monozygotic and dizygotic twins did not differ in concordance for TB (OR, 1.1; 95% confidence interval [95% CI], 0.4–2.8). A higher concordance rate for TB among monozygotic versus dizygotic twins was confined to 106 pairs with the co-twins exposed to a sputum-positive index case (OR, 3.4; 95% CI, 1.6–7.2), and was highest in adolescent twins living together. ORs of TB concordance were proportional to intensity of exposure (sputum smear positivity, physical proximity between twin pairs, contagiousness of disease, and living together) rather than to zygosity. Conclusions: In the Prophit survey of susceptibility to TB among twins, environmental factors (i.e., intensity of exposure to tubercle bacilli) outweigh the importance of hereditary factors. Environmental factors and the context of transmission should be given more emphasis when studying interindividual and population differences in susceptibility to infectious diseases such as TB.
Key Words: tuberculosis • human genetics • heredity • twins
Publications on the immunopathogenesis of tuberculosis (TB) often take for granted that an individual's susceptibility to tuberculous disease is to a large extent preset by genetically determined host factors (1–6). In this respect, reference is made to studies that in the past TB was more prone to occur in children of tuberculous parents, or in siblings, than in the general population. However, children exposed to tuberculous parents will develop progressive TB just as often as children exposed to household sources of infection who are unrelated genetically to the children (e.g., nannies, lodgers, or friends), and in the past a large percentage of tuberculous individuals came from families in which neither parent had active TB (6–8). One well-designed study of the Danish Adoptation Register indicated that premature death due to an infection in adults has a strong genetic background, but in that study very few individuals suffered from TB (9). Arguments for a hereditary influence are also based on reports on differences between racial groups (e.g., Native Americans and Inuits) in progression and mortality of TB (10). Although there is some evidence for interethnic differences in susceptibility to TB at the cellular level (11), the findings in many of the population studies may be confounded by exogenous factors, such as nutrition, the living environment, intensity of exposure, and the like. In studies performed in Brazil and Uganda that attempted to adjust for shared environment, apparently opposite conclusions were drawn with respect to the role for genetic determinants, such as the tumor necrosis factor (TNF) gene cluster in susceptibility to TB (12, 13). Moreover, interpretation of the population studies often is hampered by differences in racial mix, in diagnostic criteria for TB, and in the criteria used to identify control subjects. In the present context, it suffices to state, however, that, although ethnic groups may differ in their susceptibility to TB, such a finding cannot be put forward as definite proof of interindividual differences in susceptibility within racial groups or populations. Thus, the genetic argument depends particularly on studies of TB in twins. Given their usually identical social and economic background, studies in twin pairs are believed to provide definite proof of a dominant role of hereditary factors in the development of manifest TB, by showing a higher concordance of disease rate among monozygotic than genetically nonidentical, dizygotic twins (14, 15). The pivotal study on TB in twin pairs, the U.K. Prophit survey, was performed by Simonds in the 1950s (16), but reference is often made to the reanalysis of the data by Comstock (17). Various studies have addressed the incidence of TB in identical and nonidentical twins (17–21). Most of the earlier reports on TB in twins were insufficient in numbers, provided an incomplete and unrepresentative sample, or were retrospective (22, 23). Hence, these studies were subject to criticism (16, 22). In an effort to resolve the matter, a carefully conducted twin study was performed by Simonds and reported to the Prophit Committee of the Royal College of Physicians of London (16). Simonds identified twin pairs in whom at least one of the twin couple had been diagnosed as having clinically active TB. In a prospective follow-up of over 200 twin pairs, investigating the rate of TB in the co-twin, the Prophit survey demonstrated a higher concordance rate for clinical TB in monozygotic twin pairs (i.e., about 32%) than in dizygotic twin pairs (i.e., about 14%). Simonds held environmental rather than hereditary factors responsible for this difference in concordance rate and this conclusion was presented in the report written by her physician husband after Simonds' untimely death (16). In 1978, Comstock reanalyzed the data from the Prophit survey, using advanced statistical methods to control for confounder variables other than zygosity. He concluded that the higher concordance rate among monozygotic versus dizygotic twins indicated that inherited susceptibility is a major determinant for development of clinical TB among humans (17). It is ironic that a study that was taken to suggest only a small role for hereditary factors in the development of clinical TB by those who conducted the investigation is presently cited as a pivotal study demonstrating the influence of heritable factors in susceptibility to this disease (1–6). The 2.5-fold difference in rate of concordance for clinical TB among monozygotic versus dizygotic twin pairs is remarkable because, if accepted as proof of genetic influence, the degree of concordance should be conditioned by the underlying genetic mechanisms—that is, frequency of resistance-to-disease genes and their mode of expression. In recent years, various candidate genes that, in humans, may help control development of tuberculous disease have been proposed, including the following: HLA, natural resistance-associated macrophage protein-1 (NRAMP-1). and vitamin D receptor. None of these has been consistently implicated in TB susceptibility, nor has the influence of individual genes been particularly strong (9, 10, 12, 13, 24–34) or anywhere near the influence as suggested in the twin study analysis. This apparent discrepancy prompted us to reanalyze the dataset of the Prophit survey. To test in more detail the Simonds' hypothesis that environmental factors might be sufficient to explain differences in TB rate among twins, we paid special attention to the intensity of exposure to tubercle bacilli.
The dataset of the Prophit survey was published in full by her husband (16) after Simonds' premature death. The aim of the Prophit survey was to attempt to find all the twins in a given tuberculous population and determine if the co-twins of monozygotic index cases had a higher incidence of tuberculosis than the co-twins of dizygotic index cases. In short, in the mid-1950s, Dr. Simonds sent questionnaires to 21,840 patients documented in the TB registration of 13 chest clinics in London and the provinces, to identify all twins within this population diagnosed by a chest physician as suffering from TB. She identified 415 twin pairs among the responding patients (more than 97% send back the questionnaire); in this sample, the proportion of twins (1.95%) was only slightly less than the expected, corrected national twin rate (about 2%). By this method, the investigators averted a common pitfall in twin surveys—that is, restricting the investigation to those patients stated to be twins in case history only and not investigating each individual with the condition, or relying on the reported family history of TB (14, 22). No absolute distinction between monozygotic and dizygotic twin pairs could be made at that time, but Simonds carefully classified zygosity of twins, through hair color and texture, eye color, facial characteristics, blood grouping, and so forth. A total of 210 twin pairs had to be excluded because one of their members had died in infancy due to a cause other than TB (n = 172), the determination of zygosity was inconclusive (n = 27), or a diagnosis of TB was made in both twin members at the same time (n = 11). All index cases were observed, on average, for 8.5 years (up to study closure in 1956) either before or after their entry in the survey, whereas co-twins with TB were observed for at least 2 years. This observation period was used to confirm the diagnosis of TB and follow up on the treatment results. The fate of co-twins in 205 evaluable twin pairs was determined: All forms of clinically manifest TB were sought for and confirmed by X-ray and/or bacteriology. Details on the Prophit survey report that included the dataset of the twins can be found elsewhere (16, 17).
The dataset was entered, validated, and analyzed by SPSS 14.0 (SPSS, Inc., Chicago, IL) using descriptive statistics
Dataset Subtracted from Prophit Survey We observed minor differences between data on the twin pairs in the Prophit survey report (e.g., tables in text) and its accompanying comprehensive appendix ("The Twin Samples"), and the dataset subtracted from this report and/or its appendix in the reanalysis by Comstock (17). Of note, Comstock recoded cases with missing values for some characteristic by classifying them as "others" (Table 1). In the multivariate analysis, such cases (e.g., in which sputum smears were not available) were included as negative for that characteristic. He excluded three twin pairs from the multivariate analysis because not all variables could be determined (missing values for some characteristic). In all, differences in data in the Prophit survey report, its accompanying appendix, and the dataset subtracted by Comstock are minor and are indicated in Table 1.
Primary Multivariate Analysis of the Dataset In his analysis of the data, Comstock applied multivariate analysis to control for confounder variables other than zygosity. We applied the standard multivariate logistic regression technique of the statistical package SPSS; the results of this analysis, together with those presented by Comstock, are given in Table 2. Like Comstock, we had to exclude three twin pairs because of missing values for some characteristics. Of note, to obtain the adjusted percentages in Table 2, the constant term of the logistic regression coefficient had to be back-calculated so that the average rate of concordance was found when average values for the variables in the regression analysis were used. This aspect was not mentioned in the article by Comstock (17). However, even with small corrections on numbers in the dataset (Table 1), the adjusted incidence rates for TB found by us did not differ much from those reported by Comstock (17) (Table 2).
From Table 2, it could be concluded that sputum smear positivity in the index twin had no apparent effect on the subsequent rate of TB in co-twins: Both before and after adjustment for the effects of the other variables, rates in sputum smear–positive and smear-negative index cases did not differ, nor did sputum smear status affect the 2.5-times higher concordance rate among monozygotes than among dizygotes. This latter finding in particular was taken by Comstock to refute the original interpretation by Simonds. Of note, the proportion of sputum smear–positive index patients was similar in the groups of monozygotic and dizygotic twins (just over 51%). The adjusted incidence rates (Table 2) (i.e., calculated from adjusted ORs) were obtained by comparing individuals who differed only in the characteristic of interest and had constant values in all other variables. Such an adjustment estimates what might be observed had the subjects indeed differed only on the particular characteristic being examined, with all other variables having identical distributions within the two levels of outcome. The adjusted analysis "averages out" the effect of one determinant over the other, but in the face of strongly different subgroups, this may create the wrong impression that the effect of one variable is the same for all subgroups. This is a well-known phenomenon: In the presence of "interaction"—that is, a widely different result in subgroups—an analysis that does not take such differences into account and merely enters all variables in a model as confounders will give the impression of a uniform effect over all strata. Departures from the assumptions underlying logistic regression analysis may render the adjustment as presented in Table 2 problematic. This can be viewed most clearly by assessing consistency of concordance rates separately among the relevant (and biologically plausible) categories (e.g., in twin pairs with sputum smear–positive and smear-negative index cases).
Known Positive or Negative Sputum Smear Index TB Case Markedly Influenced the OR of Concordance in Twin Pairs
By contrast, in 106 twin pairs, the index patient was sputum smear positive; in these pairs, the concordance amounted to 40.0% (12/30) in monozygotic twins and 11.8% (9/76) in dizygotic twins (OR, 3.4; 95% CI, 1.6–7.2; P = 0.002 for difference). The percentage of sputum smear-positive findings were about equal in monozygotic and dizygotic twin pairs (i.e., 54.5 and 50.7%, respectively), fulfilling the requirement for equal exposure of the co-twin. However, in dizygotic twins, exposure of the co-twin to positive sputum from the index case was similar whether he or she had TB or not, whereas in monozygotic pairs, 72% of the co-twins with TB had been exposed to positive sputum from the index case. The concordance rate in same-sex dizygotic twins was higher than in dizygotic twins of opposite sex (Figure 1). In sputum smear–positive cases, the concordance in female twins (or index case female) amounted to 47.1% in monozygotic twins and decreased to 18.2% in same-sex dizygotic twins, and to 7.4% in dizygotic twins of the opposite sex. In male twins (or index case male), the concordance amounted to 30.8% in monozygotic twins and decreased to 14.3% in same-sex dizygotic twins and to 7.7% in dizygotic twins of the opposite sex. No such dependency on sex was observed for concordance rate in twin pairs with a sputum smear–negative index case (21.7, 18.5, and 19.4% in monozygotic and dizygotic twins, same sex and opposite sex, respectively). Of note, on basis of the dates of diagnosis of TB given in the appendix of Simonds' study (16), it can be concluded that, in six of the concordant pairs, a diagnosis of TB in the co-twin had actually been made before the index presented with TB. In most of these cases, the co-twin had sputum smear–positive TB and thus could have infected the index twin member. This aspect that concerns a small subgroup was not discussed in the Prophet survey report, nor in the paper by Comstock (16, 17). Exclusion of these six twin pairs did not significantly alter the outcome of the analysis, showing that the higher TB concordance in monozygotic as opposed to dizygotic twins was confined to those pairs in which the index case could have infected the other twin member. Because we chose to perform a conservative "intention to treat" analysis of the database, not to create an unnecessary break with the analyses by Simonds and Comstock, we did not leave out these six cases.
Living under the Same Roof Markedly Influenced the OR of Concordance in Twin Pairs with a Positive Sputum Smear Index TB Case
OR of Concordance for TB in Twin Pairs and Age Cohort
Figure 2 also gives the number of sputum smear–positive twin pairs among all those tested for this characteristic; as expected, this ratio is low in the cohort of children and rises sharply in adolescence. In the older age cohorts, the concordance rate in both types of twin pairs becomes about equal again, despite a significant proportion of the cases at later age being sputum smear positive. Figure 2 includes the number of twin pairs living together and indicates that, in the older age cohorts, the number has decreased. Thus, the lack of difference in concordance rate between elderly monozygotic and dizygotic twins may now be explained by, for instance, inability to infect one another because of greater physical distance.
Multivariate Analysis Revisited: Separate Analysis in Positive and Negative Sputum Smear Index TB Cases
In 87 evaluable twin pairs with a sputum smear–negative index case, we observed no influence of zygosity, living together, on TB concordance among twin pairs. Only known exposure to an exogenous TB contact had a strong influence on the occurrence of concordance of TB among the twin pairs (Table 4). In this subgroup, exogenous tuberculous contacts included, for example, a case of adult TB in a parent or siblings (besides the index co-twin who was sputum smear-negative in this group) or another source such as potential exposure at work (e.g., sanatorium staff). By contrast, in 106 evaluable twin pairs with a sputum smear–positive index case, a clear influence on concordance rate was observed in the twin members living together, whereas a strong association was found with type of zygosity (i.e., reflecting differences in physical proximity between twin pairs; P = 0.004; Table 4). Also, known exposure to an exogenous source remained a significant predictor though less strongly so (P = 0.012). Because Dr. Simonds included sputum smear–positive index twins within the group of positive exogenous contacts, a definite differentiation between exogenous and within-twin transmissions cannot be made, but likely, both played a role here.
In this study, we reanalyzed the Prophit survey data on rate of concordance for TB in monozygotic and dizygotic twin pairs and addressed the question of how important, relatively, heredity is as compared with environmental influences that may affect exposure in development of TB. Our findings indicate that environmental determinants of intensity and duration of exposure to tubercle bacilli, such as sputum smear positivity, type of zygosity (reflecting differences in physical proximity between twin pairs), age cohort, and living under the same roof at the time of contagious TB in the index case, outweigh any influence that hereditary factors may exert in twins. Of all twin studies in TB, the Prophit survey seems to be particularly well designed and conducted, and some potential pitfalls that might have affected complete recruitment of twins and individual classification of type of twins and of diagnosis of TB were averted (16, 17). In all cases, the observation of the twin pairs in the study exceeded at least 3 years and therefore was sufficiently long to define the rate of clinically manifest TB, which typically lags behind exposure up to about 2 years. Although in the 1950s, surveillance for TB in contacts of patients, by skin reactivity and roentgenogram, was proven effective, the point we need to consider here is whether underreporting would be expected to occur equally among all groups. Because most surveys of twins generally recruit relatively more endpoints in identical twins than nonidentical twins, because of reporting bias (14, 22), such an error would be expected to enlarge rather than diminish a difference in concordance rate between monozygotic and dizygotic twins; however, in the present prospective follow-up, this should be small. The difference in concordance rate between monozygotic and dizygotic twins was in agreement with previous data on TB in twin pairs (e.g., see References18 and 19). Comstock concluded previously on the same dataset that the higher concordance rate among monozygotic over dizygotic twins indicated that inherited susceptibility is an important risk factor for development of clinical TB among humans (17). In his analysis, sputum smear positivity in the index twin had no apparent effect on the subsequent rate of TB in co-twins; both before and after adjustment for the effects of the other variables, rates in sputum smear–positive and smear-negative index cases did not differ, nor did sputum smear status affect the approximately 2.5-times higher concordance rate among monozygotes than among dizygotes. In his adjusted analysis, he averages out the effect of one determinant over the other, which may be correct when the effect of one variable is the same for all subgroups. We show that, in the face of strongly different subgroups (i.e., sputum smear–positive and smear-negative cases), it is not correct to assume that this variable has an identical distribution among both monozygotes and dizygotes. In the present comparison of bacteriologically positive and negative twin pairs, only in sputum smear–positive cases was the concordance rate in monozygotic twins three- to fourfold higher than in dizygotic twins, and in all comparisons was higher in female twins (or index case female) than in male twins, and in same- versus opposite-sex dizygotic twins. By contrast, among twin pairs with a sputum smear–negative index twin, the rate of concordance did not differ significantly between monozygotic and dizygotic twins and was about 20% in both groups. The logistic regression analysis of a main-effects–only model and an interaction model confirmed that the effect of zygosity differed depending on the sputum smear status of the index case. Therefore, our analysis indicates that the difference in concordance rate may be explained by likeliness of transmission of TB between twin pairs. This interpretation is backed up by analysis of the effect on concordance of the age cohorts, reflecting contagiousness of tuberculous disease and living together. Together, our findings suggest that intensity and duration of exposure to the tubercle bacilli (i.e., due to greater physical contact of monozygotic than dizygotic twin pairs with a sputum smear–positive index case, and more so in females than in males) outweigh in importance the influence that possible genetic determinants may exert. The present findings suggest an identical differential rate of physical proximity between twin pairs as previously reported in studies on potential confounders of genetic determinants of intelligence measures (IQ tests) and behavioral characteristics. In those studies, the highest co-twin physical closeness, a well-recognized confounder in hereditary aspects of behavioral research, was measured in female monozygotic twins, and the lowest amount of time spent in each other's company was measured in dizygotic twins of the opposite sex (35–38). In this respect, it is of interest that, for other diseases of infectious etiology and transmissible by air droplets and aerosol (39) (e.g., measles and chicken pox), the concordance rate in twins is also higher in monozygotic as compared with dizygotic pairs (18). In these highly contagious children's diseases, heredity is not believed to play a role in the disease becoming manifest (i.e., all individuals exposed to infectious droplets get measles or chickenpox provided they have not previously had the disease). Hence, in these diseases, differences in concordance rate between monozygotic and dizygotic twins directly reflect intensity of exposure and underscore the differences in the success of droplet/airborne transmission between identical and nonidentical twins. Our findings are not meant to disprove that genetic factors may play a role in the immunopathogenesis of TB. Clinical expression of disease and severity of immunopathology depend on the cross-talk between M. tuberculosis, with its specific virulence characteristics and invasiveness, and the individual's host immune response comprising both innate, preformed, as well as adaptive elements, the activity of many of which is genetically preset. Interindividual variability of clinical outcome results in part from variability in the genes that control the host defense. In fact, the diversity of pathogens, including M. tuberculosis, likely presents a driving evolutionary pressure for generation of interindividual variation among humans (1–3, 40, 41). However, the influence of host genetic factors as weighed against environmental factors on an individual's susceptibility to develop TB disease is a matter of debate, although genetic factors can be of decisive importance in the extreme susceptibility of rare, selected cases (42–46). In TB, the outcome of infection in over 90% of humans is a subclinical disease and lifelong immunity, and thus the vast majority of humans are resistant to this disease. Our findings suggest that, as far as the appearance of detectable clinical symptoms in those who become ill is concerned, environmental rather than hereditary factors determine the concordance rate in twins and do not lend support to the notion that the main factor governing the degree of TB morbidity in the population at large lies within the genetic make-up of the individual. This has implications for population-based surveys into genetic determinants of susceptibility to TB, because such studies are necessarily performed in TB-endemic countries in which—paradoxically—the importance of inherited differences in individual resistance may well be subordinate to environmental factors. By consequence, the finding that many of the associations of gene variants with TB from one population cannot be replicated in other countries and have kept the basis of genetic predisposition to TB elusive may well reflect differences in the environmental context of transmission rather than a suggested polygenic nature of genetic susceptibility (24, 30, 44–46).
Originally Published in Press as DOI: 10.1164/rccm.200703-435OC on September 6, 2007 Conflict of Interest Statement: None of the authors has a financial relationship with a commercial entity that has an interest in the subject of this manuscript. Received in original form March 16, 2007; accepted in final form September 5, 2007
Related articles in AJRCCM:
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
AT A GLANCE COMMENTARY
2 analysis and logistic regression. First, we repeated the multivariate analysis as presented by Comstock (
