Article Text

Download PDFPDF

Tuberculosis recurrence and its associated risk factors among successfully treated patients
  1. J-P Millet1,2,3,
  2. À Orcau1,
  3. P García de Olalla1,
  4. M Casals1,2,
  5. C Rius1,
  6. J A Caylà1,2
  1. 1
    Epidemiology Service, Public Health Agency of Barcelona, Barcelona, Spain
  2. 2
    CIBER de Epidemiología y Salud Pública (CIBERESP), Spain
  3. 3
    Departament de Pediatria, d’Obstetricia i Ginecologia i de Medicina Preventiva, Universitat Autònoma de Barcelona (UAB), Barcelona, Spain
  1. Correspondence to Dr J-P Millet, Plaza Lesseps, 1, PC, 08023 Barcelona, Spain; jmillet{at}; juampablomillet{at}


Background: Little is known about recurrent tuberculosis (TB) in developed countries. The objective of this study was to determine the probability of TB recurrence and the associated risk factors among cured patients in a city with moderate TB incidence.

Methods: A population-based retrospective longitudinal study was carried out in Barcelona, Spain. All patients with culture-confirmed TB and drug susceptibility testing were included between 1995 and 1997 and followed until December 2005. The authors defined recurrence as a new TB event after a patient was considered cured and had remained free of the disease for a minimum of 1 year. Kaplan–Meier and Cox regression were used in the statistical analysis. HRs with 95% CIs were calculated.

Results: Among the 681 patients studied, the authors observed 29 recurrences (recurrence rate 0.53/100 person-years of follow-up). The mean incidence of TB in Barcelona from 1995 to 2005 was 36.25 cases per 100 000 inhabitants. The incidence of recurrence was 14.6 times higher in the cohort than the incidence of a first TB episode in the general population. The factors associated with recurrence at bivariate level were being male, being an immigrant, being an intravenous drug user (IDU), having human immunodeficiency virus, smoking, being an alcoholic, being in prison, and having both pulmonary and extrapulmonary TB. At multivariate level, only being an immigrant (HR 3.2, 95% CI 1.2 to 9), an IDU (HR 2.9, 95% CI 1.3 to 6.4) and male (HR 4.3, 95% CI 1.3 to 14.6) were associated.

Conclusion: Having TB in the past is a risk factor for developing TB. Social policies must be implemented in populations at risk of recurrence, especially in immigrants and IDUs.

Statistics from

In recent years, additional problems in tuberculosis (TB) control have been observed worldwide: HIV infection, rises in drug resistance and rises in recurrence, particularly in countries with high rates of TB among patients with HIV.1 2 3 4 5 6 7 8 9 10 11 12 Apart from the associated morbidity and mortality, which are often underdocumented, these problems imply introducing longer, more complex treatment regimes, which in turn make adherence more difficult and facilitate the presentation of new episodes and development of drug resistance.8 9 13 14 15

In the midst of the AIDS and TB pandemics, the rise in migratory movements and international travel suggest that TB must be considered a global problem. Industrialised countries must reinforce their surveillance and control measures.7 10 16 17 18 19 20 21 Awareness of the risk factors of recurrent TB would identify the most susceptible populations to focus present and future efforts and interventions of treatment follow-up, TB control strategies and the application of public health measures.10 22

Although failure to adhere to treatment is a well-known risk factor for relapse, recurrence rates, causes and associated factors have still not been well described, particularly among patients already cured of their first TB episode.9 11 20 Moreover, it is postulated that the risk of having a recurrence is higher in certain vulnerable groups and that treatment outcomes may not reflect the long-term status of patients in some high-burden TB settings. This has negative impacts for patients and TB control.7 8 9 16 22 23 24 25 26

Therefore, the aim of the present study was to calculate the probability of TB recurrence and to determine the associated risk factors in the city of Barcelona. This cosmopolitan European city, where the TB incidence has fluctuated between 27 and 30 cases per 100 000 inhabitants in recent years, has an effective TB prevention and control programme.21 27

Material and methods

Study subjects

A retrospective cohort study was performed on all patients detected by the Barcelona TB Prevention and Control Programme (TBPCP) who had initiated treatment between 1 October 1995 and 31 October 1997, with TB confirmation by culture, and with drug susceptibility testing (DST) for the four first-line drugs plus streptomycin done. All patients were residents of Barcelona city, Spain; the population according to the 1996 census was 1 508 805 inhabitants.28 The follow-up period lasted until 31 December 2005.


A TB case was defined as a case of TB confirmed by culture and with DST done. Recurrence rather than relapse was used because DNA strain techniques were not available to determine whether recurrences were due to relapse of the same TB episode or reinfection with a new strain. Recurrence was defined as a new clinical and/or microbiological diagnosis of TB in any patient who had correctly completed treatment for their first episode, and who had been free of the disease for at least 1 year. Otherwise, patients would have more probability of having a relapse.29 Patients who had correctly completed their TB treatment were considered as cured, even if there was no microbiological confirmation.

Variables and sources of information

Demographic and clinical data were obtained from the epidemiological questionnaire used by TBPCP. Smear test and culture results, as well as DST, came from microbiological reports provided by the Barcelona public health laboratories and laboratories of the Penitentiary Institute. Sociodemographic variables for each patient included age, sex, country of birth, place of residence (categorised as ‘inner city’ or ‘other’) and homelessness. Clinical history variables included HIV infection and route of infection (injected drug use, sexual or transfusion), diabetes, renal pathology, immunosuppressor treatment, previous TB treatment, alcoholism, smoking status and prison history. Regarding the TB episode, variables were type of TB, sputum smear test result, chest radiograph, resistance or multidrug resistance to first-line drugs, type of treatment, whether it was completed correctly and whether it was directly observed (DOTS).

In order to minimise loss to follow-up and to control for inconsistent information, hospital clinical records, the local census, the Barcelona city mortality registry, primary health care records, and the Barcelona city Drug Information System (SIDB) were consulted. In the data collection and checking process, information from the TB and AIDS registers of the city were compared. Since most of the people who move residence usually go to other areas of Catalonia, the TB register and the mortality register for our region (Catalonia) served to check for possible recurrence events among patients who had moved away from Barcelona during the follow-up period.

Statistical analysis

General and specific rates of recurrence were calculated for the groups at greatest risk. The TB incidence rate in the city over the cohort recruitment period was compared with the incidence of recurrence. A descriptive analysis of the cohort was conducted which consisted of calculating medians and interquartile ranges (IQRs) for the quantitative variables, since none of the quantitative variables had normal distributions. χ2 tests were carried out for categorical variables, or two-sided Fisher’s exact tests when pertinent. The duration of follow-up was calculated as the interval from the date when TB treatment was initiated until recurrence, death, moved away (transferred) or the end of the study. In calculating density of recurrence incidence, the denominator was person-years of follow-up (PY). Patients who had a recurrence were compared with the rest of the cohort. The Kaplan–Meier method was used to construct survival curves at univariate level, with differences between curves assessed with the log-rank test. A Cox regression was performed at univariate and multivariate levels, introducing the variables which either showed an association (with p<0.05) at univariate level or had epidemiological interest such as age and country of birth. A forward inclusion approach was used. Hazard ratios (HRs) were used as the measure of association, and 95% confidence intervals (95% CIs) were calculated. The likelihood ratio test was used to confirm that variables complied with the proportional risk hypothesis. Statistical analyses were conducted using the statistical packages SPSS 13.0 and Egret 2.0.3.


Description of the cohort

Of the 1865 cases of TB diagnosed during the inclusion period, 999 cases formed part of the cohort to be followed up (fig 1). The analysis included the 681 cases of the cohort who were at risk of TB recurrence according to the adopted definition. At the end of the follow-up period, all cases were classified as TB recurrence, moved away from Catalonia, dead or still alive.

Figure 1

Flow chart for selection and evolution of patients.

The median age was 36 years (IQR 28–52) and there was a predominance of men (67.7%). The majority of patients were autochthonous (93.1%), 18.3% of them lived in the most deprived district of the city (Ciutat Vella), 54.5% were smokers, 27.3% alcoholics, 21.9% HIV positive and 16% were intravenous drug users (IDUs). Among the patients who were IDUs, 92.6% of them were HIV positive and their mean age was 32.6 years. Among the HIV-positive patients, 70.7% were IDUs, 17.9% heterosexuals and 11.4% were male homosexuals who were non-IDUs. The predominant type of TB was pulmonary (76.5%); of those patients with pulmonary TB, 62% were smear-positive, 96.5% had an abnormal chest radiograph and 38% presented cavitation. Regarding drug resistance, the percentages of primary and secondary resistance were 4.4% and 1.8%, respectively, and that of multidrug-resistant (MDR) TB was 1%. Of the 681 patients, 81 (12.3%) had undergone some prior treatment for TB. The clinical and demographic characteristics of the cohort, and whether or not they presented a recurrence, are shown in table 1.

Table 1

Distribution of demographic and clinical variables for the cohort of 681 patients by whether they presented tuberculosis recurrence or not and the probability (raw risk) of having tuberculosis recurrence, Barcelona 1995–2005

Recurrence and its risk factors

The median length of follow-up from the time of starting treatment to becoming disease free for the 681 patients studied was 8.9 years (IQR 7.8–9.6 years). The outcomes were as follows: 29 (4.3%) had a recurrence, 136 (20%) had died, 32 (4.7%) had moved away and 484 (71%) of the patients were still alive. The probabilities of recurrence after 3, 6 and 8 years were 1%, 3.1% and 4.4%, respectively. The incidence rate of recurrence was 0.53/100 PY (an incidence of 530 per 100 000 inhabitants per year) and the median time to recurrence was 4.6 years (IQR 3.2–7.2). The incidence rate among immigrants was 1.17/100 PY, whereas that among autochthonous patients was 0.47/100 PY. Among HIV-positive patients, recurrence incidence was 1.06/100 PY and among HIV-negative patients, 0.39/100 PY. Given that the mean of TB incidence in the city from 1995 to 2005 was 36.25 cases per 100 000 inhabitants, the incidence of recurrence in the cohort was 14.6 times higher than the incidence of a first TB episode in the general population.

The factors which influenced recurrence at bivariate level were being male, being an IDU, being an immigrant, being HIV infected, having a prison history, being an alcoholic, being a smoker and having mixed TB (pulmonary/extrapulmonary) (table 1 and figs 2–4). At multivariate level, the only variables which remained significant were being an immigrant, being an IDU (which conferred a recurrence risk of almost three times) and being male (men had up to four times the risk of experiencing a recurrent episode of TB compared with women) (table 2).

Figure 2

Risk of recurrence by country of origin, Barcelona 1995–2005. Log-rank test, p = 0.058.

Figure 3

Risk of recurrence by injecting drug user status, Barcelona 1995–2005. Log-rank test, p = 0.001.

Figure 4

Risk of recurrence by sex, Barcelona 1995–2005. Log-rank test, p = 0.009.

Table 2

Probability (adjusted risk) of having a recurrence, by different variables, Barcelona 1995–2005


This study shows that a patient with TB, despite having completed treatment and being considered cured, has a much higher probability of a recurrent episode than a person in the general population has of presenting their first TB episode. Moreover, it was observed that being male, an immigrant or an IDU are risk factors for having a recurrent episode of TB. In contrast, age, alcoholism, HIV-positivity, pattern of radiological affectation, being incarcerated or having a prison history, type of TB, type of treatment and duration of treatment have not been shown to be risk factors for TB recurrence.

After comparing recurrence incidence with that in other cosmopolitan European countries such as the UK (where rates are around 2.3/100 PY)11 or Australia30 and with countries with high TB incidence such as South Africa, Uzbekistan or Vietnam,6 22 24 31 32 the incidence of recurrence observed in the present study (0.53 cases/100 PY) is low, especially considering the moderate TB incidence in Barcelona and the high percentage of IDUs and foreign-born patients in the cohort. The terms recurrence, relapse and reinfection are sometimes used with different meanings in published papers,4 22 26 making it more difficult to compare different recurrence rates. The low recurrence rate in this study could be the result of the strict definition used, which excludes patients whose treatment was not correct and who relapsed in under 1 year from the study, thus reducing the chances of relapses related to the same episode because of poor treatment compliance.22 26 The low rate of recurrence could also be due to the influence of the high mortality, the low percentage of MDR TB and the effectiveness of TBPCP, which has over 20 years of experience, especially in ensuring compliance with TB treatment by applying the DOTS strategy in some cases.8 26 33 During the follow-up period, the TB incidence rates in Barcelona were declining, and fell from 58 cases/100 000 inhabitants in 1995, when the study began, to 29.7 cases/100 000 in 2005, the final year of the study. Comparing incidence rates of TB recurrence among cured patients with incidence rates of having an initial episode shows that the former has up to 14 times higher risk of developing TB than patients in the general population. Moreover, the observed probability of recurrence increases consistently to 4.4% in 8 years, a significant number that shows the risk of relapse and reinfection according to the time patients have had TB. Higher rates of exogenous reinfection found in other studies among patients with a history of TB than among those who have never had it could also have an influence.31 34 It is clear that the higher TB incidence corresponds to the higher proportion of recurrences due to reinfection.12 35 Based on the formula established by Wang et al,12 the expected proportion of reinfection in our 1997 population (incidence of 48.5/100 000) among those with recurrent lapsing would be about 32%.

Failure to adhere to the treatment is associated with a higher risk of relapse,22 especially among immigrants, IDUs, HIV-positive patients and those over 65 years of age.4 8 9 14 However, despite the important implications for control strategies and TB treatment follow-up, very little is known about the factors associated with recurrent TB among cured patients.26 33 Given that TB has always been associated with poverty, the associations observed between TB recurrence and immigration status or being an IDU could be explained by the unhealthy living conditions seen in these groups. The majority of recently arrived immigrants come from low-income countries with a high TB burden and consequently many of them are already infected. On arrival in Spain, they often live in conditions which are detrimental to the immune system and favour the development of TB. Overcrowding, poor nutrition and the stress of migration are probably the additional reasons for contracting TB and TB recurrence, not the fact of being an immigrant in itself.36 Owing to the important rise in the immigrant population in Barcelona city (a percentage which has increased from 1.9% in 1996 to 15.6% in 2007),28 a rise in TB recurrence rates is expected in the next few years, particularly in this group. Hence, all those working, directly or indirectly, with the TBPCP should be on the alert.

Despite the longer follow-up time, no multivariate-level association between recurrence and smoking, nor between recurrence and the existence of pulmonary cavitation, was found in the present cohort study.11 14 22 25 There was no association found with age, alcoholism or type, or duration of treatment, in contrast to findings of an earlier study in South Carolina, USA. Perhaps this is because of the high mortality rate and low statistical power of only 29 recurrences.9 22 Although an association between HIV and recurrence was found in the bivariate analysis, it was not confirmed at a multivariate level, perhaps because of the high mortality rate of the cohort at the beginning of follow-up, as with other mentioned variables.22 24 The free access to highly active antiretroviral treatment (HAART) therapy in Spain since 1996–7 could also be responsible for the low frequency of recurrence in this group. The present study did not find any association between antituberculosis drugs and recurrence, even though such an association has been reported by some studies in countries of low7 and high TB incidence.14 22 24 25 26 Apart from the influence of second-line treatments, and of DOTS, which is currently very common in these cases, the low number of cases with MDR TB in the cohort is probably the result of the high mortality rate among MDR patients during the pre-HAART era and that some of the MDR TB cases could relapse in less than 1 year.

The definition of recurrence has conditioned the low number of recurrences. Perhaps some association has remained hidden owing to low statistical power or because some variables, such as prison, are related only to exposure and could change over time. However, one positive aspect, despite the fact that a limitation of cohort studies is losses to follow-up, has been the low proportion of such losses—2.8%. This is the result of the individualised search for patients in various registries. Another positive aspect was that none of the cases with positive culture were excluded, but a conclusion of the study results would be limited to those cases for whom culture testing and DST had been carried out. Among the patients excluded from the study, there was a higher proportion of prison inmates. This is perhaps due to not having DST available in some cases. Under these circumstances, our data could be underestimating the associations found between risk factors and recurrence, because prison inmates are usually men with various TB risk factors. Another limitation was that DNA strain genotyping techniques were not available; therefore, it was not possible to determine whether recurrences were due to reinfection by a different strain, rather than relapse with the same strain.4 Several studies observed differences between reinfection and reactivation according to whether the country’s TB rate was high or low.5 6 30 31 34 35 It has been seen in cities with low incidence in Europe and the USA that recurrences are probably due to endogenous reactivation of the bacillus.7 17 27 The incorporation of such techniques in our cohort study in the future will permit determining whether recurrences are due to exogenous reinfection or to endogenous reactivation of the same strain.4 22 34 37 38

We conclude that the highest probability of TB recurrence occurs among men and among the most vulnerable populations, namely IDUs and immigrants. This reinforces the notion that the risk of having a recurrence is higher in certain vulnerable groups and that treatment outcomes may not reflect the long-term status of patients in some TB settings, which has a negative impact on patients and TB control.7 8 9 16 22 23 24 25 26 39 This population would benefit from stricter follow-up and stricter post-treatment controls. It should be noted that the probability of TB recurrence is higher than that of having a first episode; hence, those patients with a history of TB must be considered as a group at risk of having TB despite having correctly completed treatment.

What is already known on this subject

Although failure to adhere to treatment is a well-known risk factor for tuberculosis relapse, the recurrence rates, causes and associated factors have still not been well described among patients already cured of their first tuberculosis episode. Recognising the risk factors of recurrent tuberculosis should help us to determine the most susceptible population group.

What this study adds

As a result of this study, we know that the highest probability of tuberculosis recurrence is found among men and the most vulnerable populations, namely injecting drug users and immigrants. The probability of recurrence is also greater than that of having an initial episode. Hence, patients with a history of tuberculosis, despite having correctly completed treatment, ought to be considered a group at risk of having tuberculosis. These findings have implications in directing efforts and interventions in treatment follow-up, tuberculosis control strategies and the application of public health measures.


Special thanks to the health services in Barcelona which collaborate with the TBPCP, to the Barcelona Microbacteria Study Group (Drs Núria Martín, Pere Coll, Margarita Salvadó, Julià González), to Dr Anna Rodés and Elsa Plasència (Generalitat de Catalunya), to Dr Pilar Estrada (Casc Antic Primary Health Team), to Dr Teresa Brugal (Public Health Agency of Barcelona, ASPB) and Dr Josep M Jansà (Generalitat de Catalunya) as well as to all the health and administration personnel of the ASPB Epidemiological Service, and to Dr Mónica Guxens without whom this study would not have been possible. Thanks also to Gloria Ribas of the Mortality Register of Catalonia, Servei d’Informació i Estudis de la Direcció General de Recursos Sanitaris del Departament de Salut de la Generalitat de Catalunya. Thanks to Dave MacFarlane and Jeanne Nelson for their support and help in the translation. Finally, thanks to Dr Ronald Geskus of the Amsterdam Health Service/Academic Medical Center for his support and help with the survival analysis.



  • Funding This study was supported by the Centro de Investigación Biomédica en Red de Epidemiología y Salud Publica (CIBERESP), Spain.

  • Competing interests None.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.