Cost-Effectiveness of isoniazid preventive therapy among HIV-infected patients clinicaly screened for latent tuberculosis infection in Dar es Salaam, Tanzania: A prospective Cohort study

This study was a prospective cohort study conducted in six urban-based clinics providing HIV/AIDS treatment services in Dar es Salaam between February 2012 and March 2014. The six HIV care and treatment clinics were Muhimbili National Hospital (MNH), Amana, Temeke, Pastoral Activities and Services for people with AIDS in Dar es Salaam Archdiocese (PASADA), Mwananyamala and Buguruni. All the clinics provided primary HIV care. The clinics at MNH, Amana, Temeke and PASADA were among the initial 14 centers participating in a separate pilot program to evaluate the feasibility of countrywide provision of IPT to HIV-infected patients in Tanzania. We used this opportunity to conduct the present study whereas the four clinics constituted an IPT group. Patients from Mwananyamala and Buguruni facilities were not part of the National IPT pilot program and were thus enrolled to form a control group. All the clinics run for five days per week and each attended about 70-100 patients daily, making a total of 8400-12,000 patients monthly. Patients’ visits in these clinics were scheduled once monthly, bimonthly or trimonthly but participants of the present study were requested to attend monthly.

Detailed methods on data collection procedures are described elsewhere [15, 16]. In brief, this was an observational study nested in the then ongoing pilot study on IPT provision in the country. Patients were consecutively enrolled to receive IPT (IPT group) or did not receive IPT (control group, whose patients came from sites which were not under the pilot study). Inclusion criteria were age ≥ 10 years, HIV-infected, both ARV naïve or experienced and who were willing to stay in Dar es Salaam for at least 2 years. Patients were excluded if they i) were known alcohol abusers (ii) had a history of current or past hepatitis (iii) medical contraindications to IPT, (iv) current or recent (within the past two years) receipt of TB treatment, (v) pregnant (vi) had WHO HIV clinical stage 4 AIDS (vii) had a history of treatment non-compliance [17]. The National Tuberculosis and Leprosy Program (NTLP) symptom-based screening tool was used for screening of active TB. This tool is comprised of 5 questions; presence of cough for ≥2 weeks, fever for ≥2 weeks, hemoptysis of any duration, excessive night sweats for ≥2 weeks and noticeable weight loss, or weight loss of ≥3 kg within 4 weeks [10]. Our previous research among HIV-infected population found that this tool had a sensitivity of 71.4% and a specificity of 75.9% for identifying active TB. Although the false negative rate was as high as 28.6%, the negative predictive value was 98.4% ensuring that about 98% of patients who screen negative to the tool are truly free of active TB. The positive predictive value was 11.4% [18].

Patients who presented with none of the symptoms in the screening tool were declared as not having active TB and were evaluated for IPT eligibility. Presentation of any of the symptoms in the screening tool raised suspicion for active TB and thus patients with symptoms underwent standard TB screening. These were eligible for enrollment upon exclusion of active TB. We defined TB as per the NTLP definition which requires the presence of two of the following: (i) Symptoms of TB (cough, fever, night sweats, loss of weight for more than 2 weeks), (ii) AFB visible by direct Ziehl-Nielsen staining of sputum specimen or M. tuberculosis cultured from sputum in Lowenstein–Jensen media, (iii) Chest radiograph independently interpreted as highly suggestive of TB (iv) TB in organs other than the lungs proven by one culture positive specimen from an extra-pulmonary site or histopathological evidence from a biopsy or a strong clinical evidence, including macroscopic evidence of specimen inspection, consistent with active extra-pulmonary TB and the decision by a medical doctor to treat with a full course of anti-tuberculous therapy (v) A clinical response to anti-TB medication in patients with culture- negative TB [17]. For both treatment groups, we collected baseline socio-demographic and clinical data.

Patients in the two groups were under standard care for HIV as per National guidelines [10]. Additionally, patients in the IPT group were provided with a monthly supply of INH 300 mg to be taken orally daily for a total of 6 months. Both groups were followed up monthly for a total of 24 months. Isoniazid (INH) side effects were monitored during treatment.

Primary end points were incident active TB, all-cause mortality, lost to follow-up and transfer out. The secondary end point was death from TB defined as death from any cause except accident related deaths occurring in a patient diagnosed with TB. All-cause mortality was defined as death resulting from any cause after recruitment into the study. Lost to follow-up was defined as absence from the clinic for two consecutive scheduled visits with failure to track the patient. Patients who transferred out were patients who shifted care from an HIV clinic of enrollment to the study to another clinic which was not a study site.

We used the number of TB cases prevented and number of deaths averted for the assessment of effectiveness. To diagnose TB, the TB symptom-based screening tool was repeated monthly for the two groups. TB suspects were provided with two 50 mL falcon tubes for sputum collection. Sputum was collected in an open air, on the spot for the 1st specimen and at home on the following morning for the 2nd specimen. Sputum samples were transported immediately to a TB laboratory in the Infectious Diseases Control (IDC) center in the city for Mycobacterium tuberculosis microscopy after Ziehl-Nielsen (ZN) staining and for culture in Lowenstein–Jensen (LJ) media. Patients with smear negative results underwent a chest X-ray examination as per NTLP guidelines. All deaths that occurred during follow up were recorded. Assumptions were made that the treatment effect of the IPT was additive to the treatment effect of the ART. Confounding variables to the effects were age, levels of immunosuppression, HIV disease stage, and socioeconomic status of the patients. The treatment effects were presumed to be variable over time as the benefit of IPT wanes off within 2 years of treatment. TB incidence and death rates among patients in the two groups were calculated as the number of TB cases or deaths divided by the total person years of follow-up. Averted TB cases and averted deaths attributable to IPT provision were calculated as the difference between these rates in the control group and the IPT group [19], presented as averted cases or deaths per 100,000 population.

We chose the provider perspective for the analysis because the provision of IPT would ultimately be done in both the government and private owned hospitals which provides HIV care and treatment. Healthcare costs (cost related to the consumption of resources in the health care system) were obtained from records, government standards and market price and through interviewing respective healthcare workers. We adopted a micro-costing approach which is a cost estimation method that involves a direct enumeration and costing out of every input consumed in the treatment of a particular patient [20, 21]. For the analysis of cost effectiveness we used financial costs (actual monetary expenditures) [21]. Personnel cost was derived from the government salary scales obtained from the office of the Human resources management in the health facilities. Patients in the IPT group could be obtained from 2 clinics only but to hasten recruitment of patients in this arm, 4 clinics were involved. Therefore, in order to avoid spurious inflation of the cost of clinic space and personnel costs in the IPT group, we made an assumption that all the 1283 patients who received IPT were seen in two clinics instead of four to match the two clinics in the control group. For this case, the total number of staff used in cost estimation is that of two clinics in IPT group and two clinics in the control group. Recurrent cost was derived from the sum of the cost for training and that for the clinic space which was estimated as a rented space using the market price for renting 1 square meter per month. All the nurses and doctors in the HIV clinics underwent training on intensified case finding, infection control and Isoniazid Preventive therapy (3Is) organized by the National AIDS Control Program (NACP). The cost data for the training was thus obtained from the NACP. As the training was for 3 days, we assumed each topic used one third of the training days. From this assumption we therefore apportioned 1/3 of the cost of training to IPT training. Prices for INH and other anti-tuberculosis drugs were obtained from the Stop TB Partnership Global Drug Facility (GDF) prices of 2013 [22]. Costs for laboratory tests were taken as an average of the market prices in the year 2012-2013. Other costs considered were salaries for the staff in the HIV clinics (we apportioned the cost to be 10% for IPT services, guided by the average time the staff used to offer IPT-related services to a patient). The costs of investigations to undergo standard screening for tuberculosis were also included (sputum for acid fast bacilli microscopy, full blood count, erythrocyte sedimentation rate and culture in Lowenstein- Jensen media). In this case, number of active TB suspects during follow-up was determined for each group and the total costs of investigations to rule out or diagnose TB was estimated by multiplying the cost of each investigation by the number of suspects. Although for the purpose of this study we had transported sputum for analysis in the Infectious Disease Control (IDC) laboratory, we omitted transportation cost because the facilities had the capacity to process the sputum within their facilities thus the transportation cost was unrealistic to the facilities. For cost of utilities (water and electricity), we apportioned 10% of the expenses in the clinics to IPT services. Except for drug costs which were already in US$, all other costs were estimated in Tanzanian shillings, and converted to US$ using the average annual exchange rate for the year 2012 obtained from the Bank of Tanzania which was Tsh. 1562.4 for 1 US $ [23].

The cost data was entered into and analyzed with Microsoft excel 2010. The analysis was annualized in the 1st and 2nd year and a discount of 3% was applied to the cost of the 2nd year. The time horizon for the analysis was 2 years. ICER was calculated as the difference of the total cost for each group divided by the difference in the effectiveness (TB cases prevented and deaths averted). A univariate sensitivity analysis of cost was performed and it involved analysis of one cost parameter after another when others were kept constant. A sensitivity range of +/−10% was used to vary the base case values of the incremental cost parameters for the entire cohort while observing the change in the ICER per TB case prevented and per deaths averted. The choice of the sensitivity range was guided by an average annual inflation rate which is around 7.31% from 1999 until 2017 [24]. There were five cost categories analyzed; personnel, laboratory, medication, utilities and recurrent costs. We also checked the robustness of the effects, number of TB cases prevented by IPT and that of deaths averted by IPT on the ICER. Patients’ characteristics and outcomes were analyzed using SPSS statistical package version 20. Baseline characteristics were analyzed using proportions, medians ± interquartile range (IQR) and means ± standard deviation (± SD). Chi square was used to compare proportions, student t-test to compare means while medians were compared using K independent samples.

Effectiveness data used in this analysis came from a cohort study with heterogeneous patients in the two treatment groups, which might have contributed to the difference in effectiveness data observed. However, the sample size was large enough to reach significant levels even in very slight differences in socio-demographic and clinical data. Furthermore, in the same study population Cox proportional hazard analysis found that none of the baseline characteristics could predict an occurrence of tuberculosis while ARV use for 2 or more years and IPT use predicted all-cause mortality [15]. As the screening of patients was through the use of a symptom-based screening tool, care should be taken in comparing the findings of this study to similar studies which used TST for screening.