Our study reports epidemiology and risk factors of EPTB from a high-burden country where the prevalence of HIV is low [
13]. To our knowledge no similar study has been reported from other south Asian countries or other high TB burden countries. Younger age and female gender were strongly associated with EPTB, but smoking, diabetes mellitus, use of immunosuppressive drugs/steroids, and past history of pulmonary TB were associated with PTB.
Sites of EPTB
In our study, the lymph nodes were the most common site of EPTB. Earlier studies have suggested that localization of EPTB may be variable. In Hong Kong [
4], the genitourinary system and the skin were the common sites, whereas in the USA [
3], bones and/or joints were the most common sites. Results of our study are comparable to two studies from Turkey [
5,
6] which reported that lymph nodes accounted for nearly half the cases of EPTB. In some studies clinical diagnosis of lymph node tuberculosis has been strengthened by a positive culture of
Mycobacterium tuberculosis complex [
3,
6,
11]. However, due to lack of facilities for culture in our setting, lymph node TB is usually diagnosed by fine needle aspiration cytology and/or node excision followed by histopathological examination for granulomatous inflammation with caseous necrosis. Therefore, we cannot rule out some mis-diagnosis of lymph node TB.
Risk factors for EPTB
Earlier studies of risk factors for EPTB had a lower proportion of EPTB cases [
3,
11]. But in our study, proportions of PTB and EPTB were almost same, possibly because it was carried out in the main referral centre for the region and the proportion of EPTB may be higher due to availability of diagnostic facilities. Pulmonary TB cases are also diagnosed at primary health care centres because of the decentralization of diagnostic and treatment facilities under NTP. This means that they are often never referred to MTH. Gender differences observed in our study confirm the findings of previous studies in both developing [
16,
17] and developed countries [
18,
19]. This may be a consequence of gender differences in both exposures to TB infection and prevalence of susceptibility risk factors (e.g., smoking) [
16]. Other possible factors accounting for the difference are stigma associated with having TB and lack of access to health care, especially for females, in some developing countries like Nepal [
17].
The other main risk factor for EPTB relative to PTB that we identified was being younger than 25 years. This is consistent with studies from the USA [
11] and Europe [
12] which have reported that younger age was an independent risk factor for EPTB. Other studies from the USA [
3] and Turkey [
5], have reported that age was not associated with EPTB. These inconsistencies could be due to differences in prevalence of host-related factors or important co-exposures.
In our study, only 4.1% of the patients less than 25 years of age were "ever smokers" as compared to 74.8% of the patients who were aged 25 years or more. Proportions of "ever smokers" differed between males (46.8%) and females (27.1%). This raises the possibility that the age and sex differences between PTB cases (in whom smoking was more common) and EPTB cases could be a result of confounding by smoking. However, after adjusting for potential confounding factors (including smoking) by logistic regression analysis, younger age and female gender remained strongly associated with EPTB. Therefore, after primary infection in the lungs the probability of reactivation at an extra-pulmonary site may be higher at younger age. It would be useful to confirm the association of age and gender with EPTB in other high-burden countries. Our results suggest that at older ages reactivation of TB was common in the lungs. This may be due to decreased local immunity in the lungs in the elderly as a result of associated life-style factors (smoking) or co-morbid conditions which may predispose to re-activation in the lungs. A recent study from the UK has reported that co-morbid conditions like emphysema and bronchitis were independent risk factors for PTB [
20].
In our study, smoking was associated with PTB. This is consistent with a meta-analysis which reported that smoking is a risk factor for TB infection and for pulmonary TB disease [
21]. Another report has suggested that smoking is associated with relapse of TB and smokers are less likely to have isolated extrapulmonary TB [
22]. We also found that past history of TB was associated with PTB, although we could not identify if this was as a result of reactivation (relapse) or reinfection [
23]. However, evidence suggests that in high-burden countries reinfection is more common than relapse [
24]. Therefore history of smoking and contact with a case of TB should arouse a high degree of suspicion for active TB. Such information would be useful for screening the patients for TB. Currently there are no health education campaigns or legislation regarding sales of cigarettes in Nepal. This is significant public health concern in view of tuberculosis control.
Our results of comparison of EPTB and PTB by logistic regression analysis suggested that factors like diabetes mellitus, use of immunosuppressive drugs/steroids and past history of TB are associated with PTB compared to EPTB. Our results are consistent with other studies that have reported an association between diabetes mellitus and PTB [
25,
26]. However, study from Turkey [
5] examined the association of diabetes, use of immunosuppressive drugs/steroids and past history of TB with EPTB but found no association with any of these factors. A study from the UK [
20] reported that use of immunosuppressive drugs/steroids and co-morbid conditions were associated with PTB. Therefore it is important to periodically screen the patients with chronic conditions like diabetes, those on immunosuppressive drugs/steroids for occurrence of tuberculosis.
Studies from developed countries have reported an increasing trend of EPTB among HIV infected persons [
9,
10] and HIV infection is associated with EPTB [
3,
10,
12]. A recent study from a large tertiary hospital in south India reported that EPTB showed an increasing trend among HIV infected patients [
12]. In Nepal, the adult HIV/AIDS prevalence is 0.5% [
13]. Testing for HIV infection is carried out based on epidemiological and/or clinical suspicion and only a fraction of our cases were tested. Therefore, we did not have enough data to carry out a useful analysis of the association between HIV infection and EPTB. We carried out a logistic regression analyses excluding those cases which were HIV-positive. The results of our main logistic regression analyses did not change. The reported rate of HIV seropositivity among TB cases in Nepal was only 2.4% in the year 2001/02 [
13]. The current rate may be higher, but probably not high enough to make a substantive difference to our conclusions. However it is important to screen the HIV/AIDS patients for TB despite the low incidence of HIV/AIDS in Nepal. Further studies are required to study the association between TB and HIV/AIDS.
Our study had several other limitations. Information about life style factors was sometimes incompletely recorded in the files. Therefore, we could not study the effect of amount and duration of smoking and alcohol use on EPTB. The PTB and EPTB patients were selected from a tertiary-care hospital. Hence, the cases we studied may not be representative of those occurring in the general population. Information about nutritional status and microbial factors was not available from the case files in our setting. A further limitation of the study was that it only identified factors that differed between PTB and EPTB. There may be some risk factors common to both forms of TB that could not be identified by our analysis.