The incubation period distribution of tuberculosis estimated with a molecular epidemiological approach

Transcription

1 Published by Oxford University Press on behalf of the International Epidemiological Association ß The Author 2011; all rights reserved. Advance Access publication 26 March 2011 International Journal of Epidemiology 2011;40: doi: /ije/dyr058 The incubation period distribution of tuberculosis estimated with a molecular epidemiological approach Martien W Borgdorff, 1,2,3 * Maruschka Sebek, 4 Ronald B Geskus, 1,3 Kristin Kremer, 5 Nico Kalisvaart 4 and Dick van Soolingen 5 1 Department of Clinical Epidemiology, Biostatistics and Bioinformatics, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands, 2 Center for Infection and Immunity Amsterdam (CINIMA), Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands, 3 Department of Infectious Diseases, Public Health Service Amsterdam, Amsterdam, The Netherlands, 4 KNCV Tuberculosis Foundation, The Hague, The Netherlands and 5 National Tuberculosis Reference Laboratory, Center for Infectious Disease Control (CIb/LIS), National Institute for Public Health and the Environment (RIVM), Bilthoven, The Netherlands *Corresponding author. Department of Infectious Diseases, Public Health Service Amsterdam, Nieuwe Achtergracht 100, 1018 WT Amsterdam, The Netherlands. mborgdorff@ggd.amsterdam.nl Accepted 28 February 2011 Background There is limited information on the distribution of incubation periods of tuberculosis (TB). Methods In The Netherlands, patients whose Mycobacterium tuberculosis isolates have identical DNA fingerprints in the period were interviewed to identify epidemiological links between cases. We determined the incubation period distribution in secondary cases. Survival analysis techniques were used to include secondary cases not yet symptomatic at diagnosis with weighting to adjust for lower capture probabilities of couples with longer time intervals between their diagnoses. In order to deal with missing data, we used multiple imputations. Results We identified 1095 epidemiologically linked secondary cases, attributed to 688 source cases with pulmonary TB. Of those developing disease within 15 years, the Kaplan Meier probability to fall ill within 1 year was 45%, within 2 years 62% and within 5 years 83%. The incubation time was shorter in secondary cases who were men, young, those with extra-pulmonary TB and those not reporting previous TB or previous preventive therapy. Conclusions Molecular epidemiological analysis has allowed a more precise description of the incubation period of TB than was possible in previous studies, including the identification of risk factors for shorter incubation periods. Keywords Tuberculosis, infectious disease, incubation period, DNA fingerprinting 964

2 INCUBATION PERIOD OF TB 965 Introduction The incubation period and serial interval are important determinants in the epidemiology of infectious diseases. The incubation period is defined as the period from infection to disease onset and serial interval as the period between a given disease state, for instance onset of symptoms, in the source case and secondary case. 1 Since the incubation period and serial interval of tuberculosis (TB) may be months to decades, 2 4 TB epidemiology evolves over centuries. 5 However, the incubation period and serial interval of TB have not been well characterized. 1 An indirect estimate based on mathematical modelling suggested that the incubation period and serial interval of TB in a given setting were likely to depend strongly on age and calendar year. 1 Measurement of the incubation period of TB is difficult because of the long time periods generally involved. In trials of preventive therapy among household contacts in the USA, contacts in the control group were followed over 10 years. 6 Of those developing TB within 10 years, up to 50% did so within 2 years and up to 80% within 5 years. 6 A limitation of this design is that some of these contacts may have been infected before or after the index case occurred, perhaps even outside the household as shown in South Africa, 7 thus blurring the distribution. The incubation period of TB was also determined in a BCG trial in the UK. 8 Individuals in the control group were checked annually for infection, and incident TB cases were recorded. 2,9 Of the cases occurring within 7 years, 82% did so within 2 years and 95% within 5 years. 9 A limitation of this study was the limited age range and that no risk factors associated with long or short incubation periods were identified. Finally, the incubation period and serial interval were previously estimated in The Netherlands using epidemiologically confirmed contacts of TB patients whose Mycobacterium tuberculosis strains had identical DNA fingerprints. 10 This study, however, had a limited number of 69 source-secondary case pairs and follow-up was limited to 4 years only. In The Netherlands, nationwide DNA fingerprinting started in Active follow-up takes place among patients whose isolates have identical DNA fingerprints to determine whether contact was likely during the infectious period. Using this DNA fingerprint and contact information to identify epidemiological links, we determined the time distribution of the incubation period and serial interval of TB over a 15-year study period and identified risk factors for relatively short incubation periods and serial intervals. Methods Patient data were obtained from The Netherlands Tuberculosis Register (NTR) held by the KNCV Tuberculosis Foundation. It comprises data on TB patients since 1993 and is estimated to be 499% complete for notified TB patients. It has been estimated that 485% of TB patients are notified in The Netherlands. 12 Cultures were done in 485% of the notified patients with pulmonary TB and in 475% of those with extra-pulmonary TB. We used data on age, sex, year of diagnosis, country of birth, urban residence, TB localization, previous TB and preventive treatment, and co-morbidities presumed to affect the immune response, including HIV infection, diabetes, malignancies, kidney failure, organ transplants and immunosuppressive therapy. All M. tuberculosis complex isolates of patients diagnosed with TB in The Netherlands are sent to the National Institute for Public Health and the Environment (RIVM) for species identification, DNA fingerprinting and drug susceptibility testing. DNA fingerprinting was until 2009 performed with standard restriction fragment length polymorphism (RFLP) typing using IS6110 as a probe. 13 Strains with fewer than five bands in their RFLP pattern were subjected to sub-typing using the polymorphic GC rich sequence (PGRS) as a probe. 14 Information of the RIVM and NTR was matched using sex, date of birth, postal area code and year of diagnosis as identifiers. If a patient had an isolate with a DNA fingerprint identical to that of another patient, they were both interviewed by staff of the respective municipal health services to determine possible contact. If such contact was found, these patients were considered an epidemiologically confirmed couple, provided that at the time of contact the reported source case had pulmonary TB and the reported secondary case was not (yet) symptomatic. In some instances, group infections, for instance among clients of a pub, precluded the identification of couples and such cases were excluded from the analysis. The serial interval was defined as the time period between onset of symptoms in the reported source case and that in the secondary case. The incubation period was defined as the time period between infection and the onset of symptoms in the secondary case, where the time of infection was estimated as the halfway point between onset of symptoms of the source case and either the date of diagnosis of the source case or onset of symptoms of the secondary case, whichever came first. If the serial interval or incubation period was negative, these couples were excluded from the analysis. Negative serial intervals or incubation periods might be the result of misclassification of sources and secondary cases by the public health services, or of errors in the estimated onset of disease in the source or secondary case either due to imprecise recall or after imputation of missing values. If a secondary case had been detected before the onset of symptoms through active case finding (usually presenting with radiological abnormalities consistent with TB and a positive culture), data were considered right censored and the symptom-free period

3 966 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY contributed to the incubation period estimate in survival analysis. As done previously, 10 we weighted observations by the inverse of the probability to observe a couple, which was estimated as the duration of the total study period (5479 days) minus the number of days between the isolates. 10 Risk factors for short incubation periods were identified with Cox proportional hazards analysis. In this analysis, a hazard ratio 41 represents a shorter incubation period. Only main effects were considered. Since the effect of age was nonlinear, this was modelled using splines. We corrected for possible correlation between incubation periods, if individuals had been infected by the same source case by using a robust sandwich estimate of the standard error. 15 All survival analyses were performed using the R statistical package version Data on delay (from onset of symptoms until diagnosis) were often missing. Data were not missing completely at random. For instance, older people were less likely to report the duration of delay, and if they did, tended to report longer delays. We created five data sets in which the missing values were imputed using the Multivariate Imputation by Chained Equations technique as implemented in the mice package. 17 In addition, we analysed the subset of couples with complete data on delay. Results In the period , TB patients were registered in NTR and (70%) were culture confirmed and genotyped at the RIVM. Matching the NTR and RIVM data sets yielded patients, 6456 (53%) of whom were clustered. Among these, 1095 epidemiologically confirmed secondary cases were identified, uniquely attributed to 688 source cases with pulmonary TB, thus yielding 1095 couples of secondary cases and their source case. Information on delay from onset of symptoms to diagnosis was available for 439/688 (64%) of the sources and 691/1095 (63%) of the secondary cases, yielding 422 couples of symptomatic sources and secondary cases with full information on delay and a positive serial interval. The characteristics of the patients in the full data set, 6456 clustered patients, 1095 epidemiologically confirmed couples and 422 couples with complete delay data are presented in Table 1. Compared with all patients, clustered patients were more likely to be male, young, born in The Netherlands and to have pulmonary TB. Clustering was not associated with HIV infection or other co-morbidities. Compared with all clustered patients, those with epidemiological confirmation and considered source cases were more likely to be young adults, whereas the proportion considered as secondary cases declined with age throughout (Table 1). Among those developing disease within 15 years, the median incubation period was 1.26 years [95% confidence interval (95% CI) years] with a range of years. The Kaplan Meier probabilities of developing disease within 1, 2 and 5 years were 45% (95% CI 42 48%), 62% (95% CI 58 65%) and 83% (95% CI 80 85%), respectively (Figure 1). The median serial interval was 1.44 years (95% CI years) with a range of years. The incubation time increased with age of the secondary case (Figure 2, P < ), and was shorter in secondary cases who suffered from extra-pulmonary TB than in those with pulmonary TB only (Table 2). More detailed analysis by ICD-9 code did not reveal associations with particular extra-pulmonary sites (data not shown). The incubation period was shorter in secondary cases who were male and who did not report previous TB or previous preventive therapy (Table 2). Risk factors among source cases for shorter incubation periods in their secondary cases were urban residence, having diabetes and not reporting previous TB. The Beijing genotype was not associated with the duration of the incubation period. The hazard ratios of risk factors among secondary cases were not influenced much by taking risk factors among source cases into account (Table 2). Risk factors for short serial intervals were very similar to those for short incubation periods, with <5% difference for the hazard ratios of each of the risk factors (data not shown). In the subset of couples with complete information on delay, the median incubation period was 1.34 years (95% CI ) and the median serial interval 1.53 years (95% CI ). Risk factors for short incubation periods and serial intervals were similar to those reported above, with hazard ratios tending to be slightly further removed from 1 and confidence intervals being somewhat wider (data not shown). Discussion This article expands a previous analysis that was limited to a study duration of only 4 years and a limited number of cases. 10 It confirms the expected skewed distribution of the incubation period and serial interval and shows that the median incubation period in epidemiologically confirmed secondary cases is 1.26 years. The incubation period was shorter in secondary cases with extra-pulmonary TB, for males, young patients and those who did not report previous TB or previous preventive therapy. This study has included many more secondary cases than previous studies and has allowed the identification of risk factors for short incubation periods. The study confirms that the benefit of preventive therapy is largest if targeting those having been infected recently, i.e. contacts of infectious patients, rather than targeting just anyone with latent tuberculous infection (LTBI).

4 INCUBATION PERIOD OF TB 967 Table 1 Characteristics of all TB patients, those being clustered, those reporting epidemiological contact and those included in the analysis All patients (n ¼ ) Clustered patients (n ¼ 6456) Epidemiological confirmation Sources (n ¼ 688) Secondary (n ¼ 1095) Couples of symptomatics with known delays Sources (n ¼ 290) Secondary (n ¼ 422) Sex Male (56) 467 (11) 688 (17) 191 (41) 255 (37) Female (48) 221 (9) 407 (17) 99 (45) 167 (41) Age (years) < (66) 15 (6) 104 (44) 8 (53) 35 (34) (61) 202 (14) 290 (20) 79 (39) 131 (45) (56) 189 (10) 260 (14) 80 (42) 95 (37) (58) 135 (11) 193 (16) 58 (43) 74 (38) (57) 80 (11) 122 (17) 29 (36) 43 (35) (49) 36 (8) 73 (16) 23 (64) 26 (36) (36) 13 (4) 38 (12) 7 (54) 13 (34) (26) 18 (6) 15 (5) 6 (33) 5 (33) Country of birth Netherlands (57) 278 (12) 605 (26) 119 (43) 228 (38) Foreign born (51) 410 (10) 490 (12) 171 (42) 194 (40) Localization of TB Pulmonary TB (PTB) (57) 638 (16) 773 (20) 272 (43) 310 (40) PTB þ EPTB (52) 50 (7) 108 (15) 18 (36) 46 (43) Extra pulmonary TB (EPTB) (46) 214 (12) 66 (31) HIV infection Yes (55) 17 (5) 26 (8) 8 (47) 12 (46) No/not reported (53) 671 (11) 1069 (17) 282 (42) 410 (38) Other diseases Yes (48) 48 (7) 66 (9) 20 (42) 23 (35) No/not reported (54) 640 (11) 1029 (18) 270 (42) 399 (39) Reported delay in days, median (25th, 75th percentile) Number reporting Patient s delay 32 (18 88) 18 (3.5 60) 32 (18 88) 18 (3.5 46) Doctor s delay 18 (3.5 46) 11 (3.5 32) 18 (3.5 53) 11 (3.5 32) Total delay 74 (32 130) 46 ( ) 81 (39 131) 39 (3.5 95) The incubation periods in this study were similar to those among household contacts in the USA, 6 whereas those reported in the UK BCG trial were on average somewhat shorter. 8 The shorter incubation periods in the UK study may be attributed in part to the relatively young study population, and perhaps also in part to some bias resulting from ignoring losses to follow-up. The longer incubation period in patients aged 430 years is intriguing. Whereas in earlier modelling studies a longer incubation period in older people was predicted, this was explained by their high risk of infection decades ago. 1 Since the current study was restricted to a 15-year period, this explanation is unlikely to hold here. In birth or generation cohort studies, young adults appeared particularly at risk for TB mortality It might be that this increased risk of death among young adults can be explained, perhaps in part, by immunological factors, which are also associated with shorter incubation periods. Another factor which might explain the role of age is that of previous infection. We found that those reporting a history of TB treatment or preventive therapy had longer incubation periods. Perhaps these

5 968 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY individuals had a better prepared immune system due to previous infection helping to control a new (re-) infection. Previous infection is often undetected, and the probability it ever occurred increases with age, thus providing a possible explanation for the effect P (without symptoms) Time since infection (years) Figure 1 Kaplan Meier tuberculosis-free survival probabilities among 1095 epidemiologically confirmed secondary cases with tuberculosis in the period (shaded area represents 95% confidence interval) of age on the incubation period. An alternative or additional explanation might be postulated as well: the elderly are often infected by other elderly, 21 whose TB may be attributable to endogenous reactivation. Bacteria leading to endogenous reactivation or having undergone a long latent period might be less likely to lead to rapid progression to disease. However, this latter explanation is not supported by our data, since age of the source case was not a risk factor for short incubation periods in the secondary cases. We found that patients with extra-pulmonary TB had shorter incubation periods. We found no association with any specific extra-pulmonary site. However, since the number of cases for each particular site was limited, this analysis was not powerful and we cannot exclude such associations. Male patients had shorter incubation periods than female patients. Perhaps men are more intensively exposed or more susceptible to the development of TB than women, as also seems to be suggested by the predominance of male TB patients in the notifications in most countries. 22 Whereas the Beijing genotype has been shown to be associated with more rapid progression to disease in The Gambia, 23 this was not confirmed in the present study. There appears to be no immediately obvious explanation for the risk factors among source cases for short incubation periods in their secondary cases. Whether these were incidental findings or biological risk factors may be confirmed by future studies Relative hazard Age (years) Figure 2 The age of secondary cases as a risk factor for short incubation periods of tuberculosis (shaded area represents 95% confidence interval)

6 INCUBATION PERIOD OF TB 969 Table 2 Risk factors for short incubation periods: multivariable analysis Model 1 a Model 2 a Hazard ratio b (95% CI) P-value Hazard ratio b (95% CI) P-value Source case Male 1.11 ( ) 0.37 Urban 1.39 ( ) 0.01 Born in The Netherlands 1.04 ( ) 0.71 PTB þ EPTB (reference PTB) 1.25 ( ) 0.32 HIV positive 1.14 ( ) 0.78 Diabetes 1.71 ( ) 0.01 Other co-morbidities 1.40 ( ) 0.39 Previous TB 0.59 ( ) 0.01 Previous preventive therapy 0.91 ( ) 0.78 Smear negative, culture positive 0.75 ( ) 0.21 Secondary case Male 1.28 ( ) ( ) 0.02 Urban 1.05 ( ) ( ) 0.12 Born in The Netherlands 1.01 ( ) ( ) 0.97 PTB þ EPTB (reference PTB) 1.43 ( ) ( ) 0.00 EPTB (reference PTB) 1.71 ( ) ( ) HIVþ 1.35 ( ) ( ) 0.87 Diabetes 1.04 ( ) ( ) 0.86 Other co-morbidities 1.05 ( ) ( ) 0.86 Previous TB 0.65 ( ) ( ) 0.04 Previous preventive therapy 0.62 ( ) ( ) 0.02 Smear negative, culture positive 0.83 ( ) ( ) 0.20 Beijing genotype 0.74 ( ) ( ) 0.19 a Model 1: taking into account variables of secondary cases only; and Model 2: taking into account variables of source and secondary cases. Age was taken into account using splines. b Reference groups are those without the characteristic of interest, unless specified otherwise. A hazard ratio 41 represents a relatively short incubation period compared with the reference group. The most important limitation of the study is the questionable validity of delay data. However, the serial interval of isolates was similar to that of the onset of symptoms (data not shown), suggesting that inaccuracy of delay may have had limited influence. Missing data were rather frequent for reported delay. Therefore, we used imputation of missing data. Whereas the results with respect to the median incubation period in the full data set and those with complete data were similar (1.34 and 1.26 years, respectively), the analysis of risk factors was more powerful using the full data set. Another limitation is the duration of the study period. Epidemiological confirmation may be difficult to obtain after long incubation periods. Therefore, the method used in this study appears suitable to characterize rapid progression within years, and probably not for the long-term risk of reactivation disease over decades. Finally, we were unable to assess some additional risk factors, which would have been of interest, such as smoking, family relationship between source case and secondary case and intensity of exposure. Since data on the latter two variables are now being collected, such an analysis may be possible in the future. Funding This study was financially supported by the participating institutes without external funding. Acknowledgements The study was initiated by D.v.S., designed by M.W.B., data were collected and quality assured by M.S., K.K., and N.K., data were analysed by M.W.B. and R.B.G. with contributions of all authors in data interpretation, the paper was drafted by M.W.B., and critically revised for important intellectual content and finally approved by all co-authors. M.W.B. will

7 970 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY act as guarantor for the paper. Preliminary results of this study were presented at a meeting of the Tuberculosis Surveillance and Research Unit in March 2010 in Berlin, Germany. We thank the Municipal Health Services in The Netherlands for their essential contribution to data collection. Conflict of interest: None declared. KEY MESSAGES There is limited information on the distribution of the incubation period of tuberculosis and on risk factors for short incubation periods. In this large molecular epidemiological study in The Netherlands, the median incubation period of tuberculosis was 1.3 years, with Kaplan Meier probabilities of disease after 1, 2 and 5 years, of 45, 62 and 83%, respectively. The incubation time was shorter in secondary cases who were men, young, those with extra-pulmonary TB and those not reporting previous TB or previous preventive therapy. The study confirms that preventive therapy should be targeted on those having been infected recently (i.e. contacts of infectious patients), rather than on all individuals with latent tuberculous infection. References 1 Vynnycky E, Fine PE. Lifetime risks, incubation period, and serial interval of tuberculosis. Am J Epidemiol 2000; 152: Styblo K. Epidemiology of tuberculosis. Selected papers vol 24. The Hague: KNCV Tuberculosis Foundation, Rieder H. Epidemiologic Basis of Tuberculosis Control. Paris: International Union Against Tuberculosis and Lung Disease, Lillebaek T, Dirksen A, Baess I, Strunge B, Thomsen VØ, Andersen AB. Molecular evidence of endogenous reactivation of Mycobacterium tuberculosis after 33 years of latent infection. J Infect Dis 2002;185: Blower SM, McLean AR, Porco TC et al. The intrinsic transmission dynamics of tuberculosis epidemics. Nat Med 1995;1: Ferebee SH. Controlled chemoprophylaxis trials in tuberculosis. A general review. Bibl Tuberc 1970;26: Verver S, Warren RM, Munch Z et al. Proportion of tuberculosis transmission that takes place in households in a high-incidence area. Lancet 2004;363: BCG and vole bacillus vaccines in the prevention of tuberculosis in adolescence and early adult life. Bull World Health Organ 1972;46: Sutherland I. The ten year incidence of clinical tuberculosis following conversion in 2550 individuals aged 14 to 19 at the time of conversion. TSRU Progress Report The Hague: KNCV Tuberculosis Foundation, ten Asbroek AH, Borgdorff MW, Nagelkerke NJ et al. Estimation of serial interval and incubation period of tuberculosis using DNA fingerprinting. Int J Tuberc Lung Dis 1999;3: van Soolingen D, Borgdorff MW, de Haas PE et al. Molecular epidemiology of tuberculosis in the Netherlands: a nationwide study from 1993 through J Infect Dis 1999;180: van Hest NA, Smit F, Baars HW et al. Completeness of notification of tuberculosis in The Netherlands: how reliable is record-linkage and capture-recapture analysis? Epidemiol Infect 2007;135: Van Embden JD, Cave MD, Crawford JT et al. Strain identification of Mycobacterium tuberculosis by DNA fingerprinting: recommendations for a standardized methodology. J Clin Microbiol 1993;31: Kremer K, van Soolingen D, Frothingham R et al. Comparison of methods based on different molecular epidemiological markers for typing of Mycobacterium tuberculosis complex strains: interlaboratory study of discriminatory power and reproducibility. J Clin Microbiol 1999; 37: Therneau T, Grambsch P. Modeling Survival Data: Extending the Cox Model. New York: Springer, R Development Core Team. R: A Language and Environment for Statistical Computing. Vienna: R Foundation for Statistical Computing, Van Buuren S, Groothuis-Oudshoorn K. MICE: Multivariate Imputation by Chained Equations. R. J Statistical Software 2010 (in press). 18 Andvord KF, Wijsmuller G, Blomberg B. What can we learn by following the development of tuberculosis from one generation to another? Int J Tuberc Lung Dis 2002;6: Andvord KF. Continued studies of tuberculosis considered as a generation illness Int J Epidemiol 2008;37: Frost WH. The age selection of mortality from tuberculosis in successive decades Am J Epidemiol 1995; 141: Borgdorff MW, van den Hof S, Kremer K et al. Progress towards tuberculosis elimination: secular trend, immigration and transmission. Eur Respir J 2010;36: World Health Organization. Global Tuberculosis Control: a Short Update to the 2009 Report. Geneva: World Health Organization, de Jong BC, Hill PC, Aiken A et al. Progression to active tuberculosis, but not transmission, varies by Mycobacterium tuberculosis lineage in The Gambia. J Infect Dis 2008;198: