Worldwide, tuberculosis (TB) is an important problem in prison populations, especially in low and middle income countries 1 . In the Rio de Janeiro (RJ) State prison system, the TB incidence rate during the year previous to our study 2 was 2676 per 100,000, 33 times higher compared to the incidence rate of the general population 3 , and a 4.6% prevalence was observed in a X-ray systematic screening performed in one of the 32 prison units of the system 4 .

In most prisons of low and middle income countries, the TB control program is limited to the diagnosis of cases among inmates attending spontaneously the prison clinic for symptoms suggestive of TB (passive case-finding) and supervised short duration treatment using, in Brazil, the 2HRZ/4HR regimen as recommended by the Ministry of Health 5 . This strategy should remain the priority but should be associated, in highly endemic prisons, with systematic screening as recommended 1 5 6 7 8 9 but seldom applied.

To support this recommendation, we previously developed a mathematical model of TB epidemiology in a highly endemic prison 10 based on data obtained from previous studies 4 and on data extracted from the medical literature. Then, we used this model to simulate the impact of various control strategies. The predictions we obtained showed that the most effective strategy to obtain a rapid and sustained reduction of active TB prevalence in this setting associates basic control methods (passive case-finding and treatment), X-ray screening at entry in prison and annual systematic screenings.

As the impact of this strategy on TB incidence and prevalence in highly endemic confined settings has not been previously studied, we applied this strategy and measured its impact in a prison unit of high TB incidence which we followed up for two years.

A total of 4326 X-rays were performed: 1374 and 1244 during the initial and second systematic screenings, 754 at entry in the prison during the first year of study and 954 during the second year. These 4326 X-rays represents 97.7% of the total number of X-rays expected to be performed based on the nominal lists of inmates provided by the prison administration.

Characteristics of inmates examined during the 1^st systematic screening and screening at entry during the 1^st and 2d year respectively are shown in Table 1. Inmates screened during the 2d systematic screening were excluded from the comparison as they had been already examined during the 1^st systematic screening (n = 1374) or at entry (n total = 1708). Inmates in the first systematic screening and those screened at entry had similar characteristics: they were young adults and, on average, had poor schooling, substantial illiteracy rates and lived mostly in poor sections of RJ city and metropolitan area. Many of them had been previously incarcerated and more than ten per cent reported a previous treatment for TB. The mean total time in jail was significantly longer among participants in the systematic screening.

*Settlement with high density of population and partial urbanization.

For technical reasons, drug susceptibility testing was not systematically performed in TB cases diagnosed during the second year of study, but results were available for 89/127 (70.1%) culture-positive cases diagnosed during the first year of study. One case was resistant to RFM and INH (MDR) and five others were resistant only to INH (n = 2) or SM (n = 3). Another case, initially resistant to INH, was identified as being MDR 6 months later.

The main findings of our follow up study, carried out in a highly endemic prison, are: 1) a sharp decrease in active TB prevalence between the first and the second X-ray systematic screenings conducted at one year interval; 2) a decrease in TB incidence rates of cases identified by passive case-finding between these two years; 3) in inmates screened at entry, the stability of high TB prevalence rates between the first and the second year.

Among the different strategies of TB detection in highly endemic prisons, the detection of symptomatic cases can be done awaiting inmates to come forward spontaneously to the prison dispensaries or be initiated at regular interval by the health service using questionnaires 5 , with a better yield of cases but, still, with a limited and slow impact on incidence rates 10 . Our results suggest that, associated with basic control measures and screening at entry, regular periodic X-ray systematic screening (once a year is a common interval) may be a particularly efficient measure as it allows a sharp decrease in incidences and permits the identification of an important number of cases which, as shown by the actual and previous studies 4 13 14 15 , would have been missed by passive case-finding or systematic screening based on symptoms. Our present results are consistent with the predictions obtained by simulating, in our model, the same strategies 10 . In both instances, the prevalence rate of active TB was reduced by around half between the first and the second X-ray screening.

To tentatively explain the important impact of X-ray screening strategies which we observed in this highly endemic prison, we should consider the extensive circulation of MTB strains which we previously demonstrated 16 , suggesting that most active TB cases are due to new infections and not to reactivations of previous infections. Control strategies including X-ray screenings may be particularly efficient in this context.

The predominance of recent exogenous infection by strains circulating in the prison over reactivations may explain, as previously proposed 16 , the relatively low frequency of drug-resistant cases observed, in apparent contrast with the high frequency of history of TB treatment. The relatively high cure rate (around 75%) observed in the RJ prison system during the study period and during the years previous to the study 2 may be another explanation.

The diagnosis of TB at an early stage, when contagiousness is lower, particularly important in highly endemic overcrowded and confined environments, can better be achieved by X-ray screening 4 17 . This is consistent with the fact that, in our study, TB cases identified by passive case-finding were more often smear-positive as compared with cases identified by X-ray screeening.

The fact that high TB prevalences among entering inmates, similar to prevalences previously observed at entry in RJ prison system 11 , remained stable over the study period reflects the poor conditions of incarceration in the remand centres and other prisons from where entering inmates were transferred. According to international 6 7 8 and Brazilian recommendations 5 18 , the TB screening at entry in prison should be integrated in the mandatory medical examination of entering inmates for the individual benefit of inmates and in order to avoid the import of infectious sources in the prison population.

Given the design of our study, we cannot evaluate the respective impact of screening at entry versus systematic screening on TB prevalence and incidence. However, in highly endemic settings such as the prison we investigated, efficacy of screening at entry on TB prevalence may be limited due to the extensive intra-institutional circulation of strains 16 ; in this context, the contribution of new sources of infection from outside to the maintenance of a high level of TB endemicity may be limited. The impact of screening at entry may gain its full efficacy after the level of TB endemicity is decreased by intensive control measures 10 .

Given the HIV-seroprevalence we observed among TB cases (2.9%), the overall HIV-seroprevalence in our study population is expected to be low. The impact of the X-ray based screening strategy we have implemented should be assessed as well in prisons where both HIV and TB are highly prevalent. In such prisons, this impact may be different as we expect, among HIV-infected TB cases, a higher percentage of cases resulting from the reactivation of previous infections 19 , a lower contagiousness of TB cases 20 and differences in X-ray signs, including TB cases with normal X-ray 21 .

The implementation of this strategy of X-ray screening, well accepted by inmates 4 17 , may be facilitated, as emphasized by Leung et al. 22 , by using a mobile X-ray unit and filmless techniques such as digitalisation or fluoroscopy with a limited exposure to X-ray hazards. Film reading can be performed by adequately trained non-medical personnel which was found by Hoog et al. 23 to be highly sensitive in detecting TB cases. To select subjects who will have a bacteriological examination, the criteria “presence of any radiological abnormality” used in our study should be preferred to the criteria “presence of abnormalities suggestive of TB” in order to limit difficulties of radiological interpretation 13 23 . Thus, X-ray is used as a screening tool to identify TB suspects and not as a diagnostic tool.

The screening procedure we used and the difficulty, for organisational reasons within the prison, to collect early morning sputum samples as recommended 5 , may explain the relatively high proportion of bacteriologically-negative cases which did not declare any symptom and were put under TB treatment solely on the basis of chest X-ray findings. In most instances, the X-ray lesions observed in these bacteriologically-negative cases were minimal. The yield of sputum examination is likely to be higher by using the Cepheid Xpert MTB/RIF 14 .

If X-ray systematic screening benefits to communities and individuals in general populations, including populations with a high prevalence of TB, remains unclear. However, systematic screening in confined populations at very high TB risk may be cost-efficient as compared with other control strategies as suggested by recent modelling studies 14 . In any case, the implementation of this strategy necessitates an initial investment which, in many low and middle income countries, may be considered as disproportionate given the limited resources dedicated to the prisons health system; in this case, periodic symptom based screening should be implemented. The place of PCR technics in screening algorithms remains to be established 8 14 . Whatever the screening strategy chosen, its implementation would be inappropriate unless good quality passive case-finding and treatment supervision services are provided 8 24 .

The current study has two limitations. 1) to measure the impact of systematic X-ray screening on TB prevalence and incidence, a randomized trial would have been much more appropriate, comparing the impact of this strategy to that of alternative interventions 24 , but the realization of such a study would be difficult in prison settings for operational and ethical reasons; 2) the implementation of the selected strategy could increase the awareness of TB in the prison and, therefore, the passive case detection rate. But this beneficial collateral effect can be considered as an integral part of any intervention process aimed at increasing the detection of TB.

In any case, in highly endemic prisons, any TB control strategy, even the most intensive, should be associated, to be fully efficient, with drastic reductions of overcrowding and architectural measures aimed at improving ventilation and solar illumination 25 26 in order to limit the extensive strain circulation 16 , measures which are actually jointly promoted in Brazil by the Ministries of Justice and Health, and the Global Fund 25 27 28 .

Among the strategies to control TB in prisons such as the prison we investigated, detection at entry as part of the mandatory medical examination on admission and passive case-finding are basic strategies which should be applied, whatever their cost is. As well unquestionable, as an efficient contribution to TB control, are the prevention, diagnostic and treatment of HIV infection - including TB chemoprophylaxis in PPD-reactive HIV infected inmates. Under the condition that its costs in terms of material, staff and functioning does not limit the implementation of the above mentioned measures, the X-ray systematic screening should be considered in highly endemic prisons in order to impact more rapidly on TB prevalence. The impact of environmental interventions which can be implemented without considerable cost 27 is difficult to evaluate, but such interventions should be mandatory in order to comply with international conventions and national laws on prisoner rights 29 30 . Unfortunately, this last strategy and most other strategies mentioned above are not or only partially applied in developing countries in the absence political willingness and will not be fully efficient without a drastic reduction of overcrowding, a key factor 31 which depends for a large part upon the criminal policy and the slow functioning of the justice system as expressed by the high percentages, in most countries, of not sentenced inmates 32 .

The authors declare that they have no competing interests.

AS and BL drafted the protocol, coordinated the field work, the data entry and management, realized the data analysis and, with MS, drafted the manuscript. VM contributed to the data management, realized the statistical analysis and contributed to manuscript drafting. AB contributed to data collection. GG contributed in the revision of X-rays and revised the manuscript. LBC reviewed the study protocol and contributed to the manuscript. All authors read and approved the final manuscript.