Evaluation of risk factors associated with the development of MDR- and XDR-TB in a tertiary care hospital: a retrospective cohort study

Ethics statement

The study was approved by the Office of the Medical Superintendent, Allama Iqbal Memorial Hospital (registration number: AIMTH/IEB/2018/564). The need for patient consent was waived by the Office of the Medical Superintendent, Allama Iqbal Memorial Hospital Sialkot, due to the retrospective nature of the study. All the relevant information of study participants was routinely collected, recorded and managed by attending physicians, directly observed treatment short course (DOTS) facilitator and PMDT data manager.

Study setting

Sialkot is one of the districts of the Punjab province of Pakistan. It is located in the north east of the province with an area of 3,016 km² with an estimated population of 3.894 million. Current study was conducted at Pulmonology Department of Allama Iqbal Memorial Teaching Hospital (AIMTH) in Sialkot. AIMTH is a 400 beds teaching hospital. The patients treated in this hospital are either self-referral or referred by the physicians from other public or private hospitals and community clinics of the city or other nearby districts.

Study population

All the patients with confirm diagnosis of DS- and DR- PTB attending pulmonology department of AIMTH during October 2014–January 2019 were included in the current study. This duration of patient’s recruitment was selected because Programmatic Management of drug resistant tuberculosis (PMDT), a step towards support of new STOP TB strategy, was established in 2014 in the hospital.

Inclusion/exclusion criteria

In present study, a case was considered any PTB patient of either sex, aged 11 years or above and diagnosed with Xpert MTB/RIF proven RR, DST culture proven RR-, MDR- or XDR-TB. A control was a PTB patient of either sex, aged 11 years or above and DST culture confirmed DS-TB patient (Ullah et al., 2016; Ahmad et al., 2012; Balaji et al., 2010). Sputum smear negative DS-TB patients, HIV patients, pregnant females or patients with incomplete medical and microbiological records were excluded from the analysis (Desalegn et al., 2018). However, sputum smear negative but culture positive resistant MDR- and XDR-TB patients were included in the final analysis because acid fast bacilli (AFB) sputum smear often shows negative results due to low count of bacteria and positive culture confirms the diagnosis (Linguissi et al., 2015). HIV patients were not included into the study due to referral of substantial number of patients to the other speciality care hospitals. Moreover, pregnant females were deliberately excluded to avoid any risk of bias related to physiological changes during pregnancy. The methodological flow diagram of the current study is described in Fig. 1.

Data collection and management

All the required data of patients were extracted from Medical and Electronic Nominal Recording Reporting System (ENRS). A pre-designed data collection form (DCF) was used to extract all the relevant information. The primary components of DCF were demographics, including gender, age, initial weight, occupation, marital status and residential status. Social history included tobacco use, alcohol and drug abuse. Clinical characteristics were comprised of previous treatment history of TB, outcomes of previous treatment, family history of TB, presence of cavity on chest X-ray (CXR), sputum smear positivity at onset and any co-morbid conditions including diabetes, hypertension, chronic obstructive pulmonary disease (COPD), gastric ulcer and hepatitis C.

Bacteriology and DST

In compliance with the guidelines of national TB control program (NTP), all “presumptive TB cases” with suspected CXR at TB outpatient department were subjected to direct sputum smear microscopy and rapid DST including Xpert MTB/RIF. Sputum smear microscopy was performed with two sputum samples (at the spot sputum sample and next morning sample) to identify the presence of AFB by using the Ziehl-Neelsen stain. Xpert MTB/RIF was performed to detect M. tuberculosis complex and RR simultaneously. Patients who were declared rifampicin sensitive were enrolled for further DS anti-TB treatment. The patients who were RR detected, referred to PMDT site for the treatment. Their specimen was sent to TB reference laboratory, where the culture test was carried out on Lowenstein-Jensen medium. All positive cultures were subjected to DST using the agar proportion method performed on enriched middle-brook 7H10 medium against all first line (except pyrazinamide) and second line anti-TB drugs. BACTEC mycobacterial growth indicator tube (MGIT) was used for DST of pyrazinamide.

Case definitions

DS-TB case was defined as TB patient who had DST confirmed susceptibility to anti-TB drugs (Ahmad et al., 2012). RR-TB: A Disease detected as resistance to rifampicin with or without resistance to other anti-TB drugs. MDR-TB; A disease caused by resistant strains of M. tuberculosis to two important first line anti-TB drugs rifampicin and isoniazid with or without concomitant resistance to other anti-TB drugs (Wang et al., 2014). XDR-TB; A disease associated with resistance to at least isoniazid, rifampicin, a fluoroquinolone and any one of three injectable second-line drugs (amikacin, kanamycin, or capreomycin) (He et al., 2017). Registration group of patients at PMDT site were based upon previous treatment outcomes. These treatment outcomes were defined with minor modification to the definitions given by WHO and International Union against TB and Lung Disease (IUATLD) guidelines (World Health Organization, 2011; Laserson et al., 2005). These definitions were; New TB case: An incident tubercular infection case that met the suspected or confirmed case criteria and who had denied any previous TB diagnosis or anti-TB drugs treatment for 30 days or less (Tang et al., 2013); previously treated TB case: a patient who had received anti-TB drug treatment for at least one month in past and who had documented evidence of previous treatment in medical and electronic records (Laserson et al., 2005; Liu et al., 2011); relapse: a patient who had previous anti-TB treatment declared cured or completed but reported again with a positive sputum smear; treatment after failure: a patient who had a positive sputum smear at the end of 5th month of treatment for category I patient and who had 2 or more positive cultures of total five cultures in the final 12 months of treatment or any one positive cultures of the final three cultures for DR-TB patient; defaulter: a patient in which previous treatment was interrupted by 2 or more consecutive months. Others previously treated registration group: A patient who had previously been treated against TB but outcome of his previous treatment was unknown or not documented.

Statistical analysis

All data were analysed using SPSS 22.0 for windows (SPSS Inc., Australia). Quantitative variables were intimated as mean ± SD, while qualitative data were denoted as number of observations with percentages. Comparisons of categorical variables were performed by using χ² test or Fisher’s exact test, where appropriate. To evaluate the strength of association between 2/2 and greater than 2/2 variables, Phi coefficient and Cramer’s V were used. The association was considered strong, moderate or weak, if Phi coefficient and Cramer’s V values were (0.5–1), (0.3–0.49) and (0.1–0.29), respectively (Mallhi et al., 2015). For the comparison of continuous variable, student-t test was performed. A logistic regression model was used to determine risk factors independently associated with MDR- and XDR-TB. Variables with p value < 0.05 in univariate analysis were subjected to multivariate logistic regression. Adjusted OR and 95% CI were calculated. Receiver operating curve (ROC) analyses was performed to observe how the predicted model can distinguish between true positives and negatives. To assess the calibration of final multivariable logistic regression model, hosmer-lameshow test was probed. A double sided p-value < 0.05 was used to consider variables as statistically significant.

Socio demographic characteristics of DS-, MDR-and XDR-and RR-TB patients

According to DST criteria, DS-TB was diagnosed in 198 (34.1%) patients while MDR- and XDR-, RR-TB were diagnosed in 195 (33.6%), 11 (1.9%) and 176 (30.3%) patients, respectively. Out of total 198 DS-TB controls, 42.4% were females with a mean age of 40.94 ± 19.05 years, whereas majority (n = 106 (54.4%)) of MDR-TB cases were male with a mean age of 35.87 ± 15.93 years. With the exception of gender, statistically significant differences were observed among the demographic profile of both groups (Table 1). About 63% cases of XDR-TB were males, with a mean age of 27.18 ± 10.08 years. The characteristics of XDR-TB cases and DS-TB controls were compared and significant differences were observed in age and occupation with XDR-TB cases being younger than DS-TB controls 241 (Table 1).

In contrast to MDR- and XDR-TB, majority of RR-TB cases were females (53.4%) with a mean age of 36.52 ± 17.4 years. A statistically significant relationship was observed in gender, mean age, unemployment, other occupations, residential area and tobacco use among RR-TB cases.

Comparison of clinical features between DS-, MDR- XDR- and RR-TB

Among the resistant cases, 13.3% of MDR- and 27.3% of XDR-TB cases were smear negative. 1 +ve sputum smears were more profound in DS-TB, 2+ smears were most detected in RR-TB TB and 3+ sputum smears were significantly associated with XDR-TB. DR-TB patients were most observed with cavitation on CXR accounting for 35.2% of RR-, 87.2% of MDR- and 90.9% of XDR-TB cases. Almost an equal proportion among DS- (21.2%) and MDR-TB (23.1%) patients was observed with family history of TB while RR- and XDR-TB had shown slightly greater percentage (33.5% and 36.4% respectively). Greater number of the DR-TB patients belonged to previously treated group. Out of DR-TB patients with previous treatment history, majority were registered as others previously treated group followed by relapse and lost to follow up. However, failure group was not prevalent among XDR-TB patients. Diabetes, hypertension and gastric ulcer were found to be equally distributed among DS-TB controls and RR-TB cases while other comorbidities were more profound among DS-TB patients. We also observed no comorbidities among patients with XDR-TB (Table 2).

Upon statistical comparison between DS-TB controls and RR-TB cases, significant differences were observed in radiological findings at onset (p < 0.001), family history of TB (p = 0.007), registration group (p < 0.001), previous treatment history (p < 0.001), COPD (p = 0.012) and hepatitis C (p = 0.002). Moreover, we found significant differences in 1+ sputum smear (p = 0.013), radiological findings at onset (p < 0.001), registration group (p < 0.001), previous treatment history (p < 0.001), diabetes (p = 0.043), hypertension (p < 0.001) and COPD (P = 0.007) when compared between DS- controls and MDR-TB cases. Current analysis found that the development of XDR TB among PTB patients was significantly associated with 1+ (p = 0.004) and 3+ (p = 0.031) sputum smears, radiological findings at onset (p < 0.001), registration group (p < 0.001), and previous treatment history (p < 0.001) as compared to those patients who did not develop DR-TB (Table 2).

Pattern of resistance to anti tuberculosis drugs among MDR- and XDR-TB patients

We assessed pattern of resistance to both first line and second line anti-TB drugs among MDR-and XDR-TB patients. Almost an equal proportion of resistance to ethambutol was observed among both groups while no MDR-TB patient had shown resistance to amikacin. A significant difference was observed in resistance to second line anti-TB drugs (amino glycosides and flouroquinolones) (Table 3).

Risk factors associated with MDR-TB

Logistic regression analysis was performed to assess the association between various factors with MDR-TB. Unadjusted univariate analysis revealed ten independent predictors of MDR-TB which were further subjected to multivariate analysis (Table 4). The adjusted analysis was created using the variables with p < 0.05. Multivariate regression model demonstrated cavity on CXR as strongest factor (OR: 30.1) of MDR-TB, followed by previous treatment (OR: 19.2) and single marital status (OR: 11.1). Other factors significantly associated with the MDR-TB were age ≤38 years (OR: 2.5), tobacco use (OR: 2.9) and failure to the previous treatment (OR: 9.2) (Table 4). Hosmer-Lameshow test statistics have shown χ2: 11.250, degree of freedom: 8 and p value = 0.188. The p value > 0.05 demonstrated that model is good fit to data (there is no difference between observed and predicted (model) values of the dependent variables). ROC curve analysis with AUC as 0.968 (95% CI [0.953–0.982], P < 0.001) demonstrated excellent predictive ability for regression model of MDR-TB (Fig. 2).

Risk factors associated with XDR-TB

Aiming to evaluate the risk factors independently associated with development of XDR TB, logistic regression analysis was performed. Statistically significant variables were subjected to unadjusted univariate analysis. The variables with p < 0.05 in univariate analysis were considered candidates for adjusted multivariate analysis. We observed that the strongest risk factors independently associated with XDR-TB was cavity on CXR (OR: 59.6), followed by age group of ≤38 years (OR: 13.6), students (OR: 13.0), previous treatment (OR: 12.5) (Table 5). Hosmer-Lameshow test statistics have shown χ2: 4.195, degree of freedom: 5 and p value = 0.522, demonstrating model is good fit to data (there is no difference between observed and predicted (model) values of the dependent variable). ROC curve analysis with AUC as 0.972 (95% CI [0.941–1.00], P < 0.001) demonstrated excellent predictive ability for regression model of XDR-TB (Fig. 3).

Risk factors associated with RR-TB

Table 6 Shows results for risk factors associated with RR-TB, amongst diagnosed PTB cases. As would be expected, unemployment (aOR: 41.5) and history of previous TB treatment (aOR: 38.2) are greater risk factors of RR-TB among our study population. Other risk factors of RR-TB included females (OR: 5.7), Failure (OR: 4.9), cavity on chest X-ray (OR: 4.3) and age ≤38 years (OR: 2.8). The predictive accuracy of model was assessed by using Hosmer-Lameshow test statistics. The model is good fit to data (there is no difference between observed and predicted (model) values of the dependent variable) with χ2: 8.4, degree of freedom: 8 and p value = 0.39. ROC curve analysis with AUC as 0.962 (95% CI [0.946–0.978], P < 0.001) demonstrated excellent predictive ability for regression model of RR-TB (Fig. 4).

Limitations

Despite first study underscoring the association of various factors with DR-TB in Pakistan, current study is accompanied by few limitations. The retrospective and monocentric study design possesses selection bias. In addition, adults were included in the study population and the results cannot therefore be generalized to paediatric patients. The small number of XDR-TB in the present study may make statistical power quite small for identification of risk factors associated with XDR-TB. However, three previous studies with limited number of DR-TB cases (N = 75 (2.1%), 30 (4.3%) and 333 (5%)) evaluating factors of resistance to anti-TB drugs support the findings of the present study (Shen et al., 2009; Suarez-Garcia et al., 2009; Lv et al., 2017). No association was checked between HIV patients and DR-TB due to less number of HIV cases treated in the study setting owing to referral of these critically ill patients to specialized TB care hospitals. Last but not least, current study is strengthened by the first ever investigation on risk factors of XDR-TB in the country and the inclusion of heterogeneous group of patients from NTP. Additionally, the present study improves awareness about risk factors associated with resistance and highlights the need for more future 400 investigations to reduce disease burden.