INTRODUCTION
The global prison population is gradually increasing on the basis of socioeconomic and political factors [Reference Masoud1]. Prisoners often have poor health conditions, such as alcoholism, smoking, malnutrition, drug use and human immunodeficiency virus (HIV) infection, all of which can lead to a depressed immune system. As such, they are at increased risk of developing active tuberculosis (TB) if infected with Mycobacterium tuberculosis. Additionally, overcrowding, poor hygiene, and poor ventilation in prisons are adverse factors for developing active TB as well as increased transmission if active TB cases are present [Reference Bick2]. The incidence of TB and latent tuberculosis infection (LTBI) in inmates and prison staff worldwide is much higher (5–70 times) than that in the general population [Reference Masoud1, Reference Kariminia3–Reference Abebe5]. Moreover, the incidence of multidrug-resistant tuberculosis (MDR-TB), a serious global problem, is higher in prisons than in the general population [Reference Masoud1, Reference Stuckler6, Reference Lönnroth7], and multidrug resistance is defined as resistance to isoniazid and rifampicin.
By the time a prisoner in the general prison community is diagnosed with TB, their infection is likely to have spread among their fellow inmates and prison staff. Once an active TB case is identified in a prison, a large-scale TB contact investigation is sometimes needed among inmates, their visitors, and prison staff. Moreover, the investigation should be promptly initiated, as delayed detection of secondary TB cases and delayed initiation of active TB treatment increases the opportunity of further spread of TB infection in prisons that are connected to the community. The WHO guidelines on TB control in prisons, recommend chest X-ray (CXR) examination for screening of new prisoners [Reference Masoud1]. However, CXR cannot detect LTBI, and generally only detects active pulmonary TB present at the time of the CXR. Therefore a negative CXR result does not rule out active TB, due to its moderate sensitivity, and its inability to detect LTBI that may progress to active TB during the period of the prisoner's incarceration [Reference Layton8]. Active TB often occurs in prisoners who had a negative CXR at entry examination. CXR also has poor specificity for detecting active TB and may lead to over-diagnosis and the costs associated with unnecessary follow-up examinations. Effective TB screening in prisons is important not only to control TB in that setting, but also to prevent the spread of TB in the general community. Early and precise diagnosis of LTBI, and chemoprevention for those positive, would limit development of active TB in prison inmates and staff, thereby enhancing TB control.
M. tuberculosis-specific interferon-gamma release assays (IGRAs) – QuantiFERON®-TB Gold In-Tube (QFT; Cellestis Ltd, Australia) and T-SPOT®.TB (Oxford Immunotec, UK) are widely available and provide a more accurate diagnosis of M. tuberculosis infection with higher specificity than that of the tuberculosis skin test (TST) especially in bacillus Calmette-Guérin (BCG)-vaccinated individuals, mainly because they are not affected by BCG vaccination. QFT is currently the only available IGRA in Japan.
TB services in prisons, which are usually funded by governments, are under economic constraints and the methods chosen should not only provide good medical care, but also be cost-effective [Reference Stuckler6, Reference Lönnroth7]. The reagent costs for TB screening using IGRAs are higher than those for TST and CXR, but numerous studies have demonstrated IGRAs are both more effective and less expensive on a programme basis in different settings [Reference Kowada9–Reference Hardy14]. In this study, we assessed the cost-effectiveness of QFT compared to TST, TST followed by QFT and CXR in order to evaluate the optimal entry TB screening method in prisons.
METHODS
Target population
Prisoners aged 20 years old were chosen as a hypothetical cohort on a lifetime horizon in developed countries. HIV status and MDR-TB rate were considered in our models. Four cohorts were established; (i) base case; (ii) inclusion of HIV infection; (iii) inclusion of MDR-TB; (iv) inclusion of both HIV infection and MDR-TB.
Markov models
The following four clinical states were included in our model to represent the possible clinical states in the target population: (i) healthy (no LTBI and no TB); (ii) LTBI; (iii) TB; (iv) death [Reference Kowada9]. Decision-analytical calculations were performed using Tree Age Pro Healthcare Module 2009 (Tree Age Software Inc., USA). Each cycle length was 1 year.
As this was a modeling study with all inputs and parameters derived from published literature, ethics approval was not required.
Markov models were developed for four strategies: QFT, TST, TST followed by QFT, and CXR. The prisoners were stratified by BCG vaccination for TST strategies (Fig. 1).
Simplified illustration of the Markov models for entry tuberculosis screening in prisons. A square node represents the decision node. An open circular node (○) represents a chance node. Branches from a chance node represent possible outcomes. A solid circular node (•) represents a Markov node. CXR, Chest X-ray examination; QFT, QuantiFERON-TB Gold In-Tube; INH, standard 9-month INH chemoprophylaxis protocol for latent tuberculosis infection; TST, tuberculin skin test; TST/QFT, TST followed by QFT strategy.
QFT strategy
By this strategy the prisoner undergoes QFT testing. If the QFT is positive, active TB is detected by CXR, and the sputum smears and/or cultures are positive, the prisoner is treated according to the standard 6-month protocol for active TB. If the QFT is positive and active TB is not detected by CXR, the prisoner is treated for 9 months with isonicotinyl hydrazide (INH) chemoprophylaxis. We also considered the adherence and complication rates of chemoprophylaxis. If the QFT is negative, a CXR is not required and there is no need for follow-up. We used published estimates of sensitivities and specificities of QFT with and without HIV infection from a meta-analysis of studies of developed countries [Reference Hoffmann and Ravn15, Reference Pai, Zwerling and Menzies16]. Published estimates of sensitivity and specificity of CXR were also used [Reference Tattevin17, Reference Cohen18].
TST strategy
The prisoner undergoes TST testing. If the TST induration diameter is ⩾5 mm in a non-BCG-vaccinated prisoner and ⩾10 mm in a BCG-vaccinated prisoner, the prisoner undergoes CXR. If active TB is detected by CXR, and the sputum smears and/or cultures are positive, the prisoner is treated according to the standard 6-month protocol for active TB. If active TB is not detected by CXR, the prisoner is treated according to the standard 9-month INH chemoprophylaxis protocol for LTBI. If the TST induration diameter is <5 mm in a non-BCG-vaccinated prisoner and <10 mm in a BCG-vaccinated prisoner, a CXR is not required and there is no need for follow-up. We used published estimates of sensitivity and specificity of the TST with and without HIV infection from a meta-analysis of studies of developed countries [Reference Hoffmann and Ravn15, Reference Pai, Zwerling and Menzies16].
TST followed by QFT strategy
The prisoner undergoes TST testing. If the TST induration diameter is ⩾5 mm in a non-BCG-vaccinated prisoner and ⩾10 mm in a BCG-vaccinated prisoner, the prisoner undergoes QFT. If the QFT is positive, active TB is detected by CXR, and the sputum smears and/or cultures are positive, the prisoner is treated according to the standard 6-month protocol for active TB. If the QFT is positive and active TB is not detected by CXR, the prisoner receives 9 months of INH chemoprophylaxis treatment. If the TST induration diameter is <5 mm in a non-BCG-vaccinated prisoner and <10 mm in a BCG-vaccinated prisoner, neither QFT nor CXR are required, and there is no need for follow-up.
CXR strategy
The prisoner undergoes CXR. If active TB is detected by CXR, and the sputum smears and/or cultures are positive, the prisoner is treated according to the standard 6-month protocol for active TB. If active TB is not detected by CXR, there is no need for follow-up.
Data sources, data, outcomes, and assumptions
Using Medline, we undertook a search of the literature published from 1980 to 9 September 2012. Age-specific all-cause mortality rates were obtained from Japanese life tables [19]. The risk of TB-related mortality was assumed to increase with age, based on data from the Japanese Ministry of Health, Labour and Welfare and from other Japanese studies [20]. Cohort studies performed in Japanese individuals were used to obtain the adherence rate (the proportion of patients who accept LTBI treatment) of the standard 9-month INH chemoprophylaxis protocol, the probability of INH-induced hepatitis, and the efficacy (preventing progression from LTBI to TB) of the standard 9-month chemoprophylaxis protocol [Reference Yoshiyama21]. We assumed the probability of successful active TB treatment [Reference Chee22] (Table 1). Data from meta-analyses, which included studies from numerous countries, were used for determining the sensitivities and specificities of QFT with and without HIV [Reference Hoffmann and Ravn15, Reference Pai, Zwerling and Menzies16], and CXR [Reference Tattevin17, Reference Cohen18]. The sensitivity and specificity of TST with and without HIV were also assumed [Reference Hoffmann and Ravn15, Reference Pai, Zwerling and Menzies16].
Baseline estimates for selected model variables
BCG, Bacillus Calmette-Guérin; CXR, chest X-ray examination; HIV, human immunodeficiency virus; INH, isonicotinyl hydrazide; LTBI, latent tuberculosis infection; MDR-TB, multidrug-resistant tuberculosis, QFT, QuantiFERON®-TB Gold In-Tube; TB, tuberculosis; TST, tuberculin skin test.
On Monte Carlo stimulation distributions, costs are in lognormal distribution and utilities are in β distribution.
All costs were adjusted to 2012 Japanese yen, using the medical care component of the Japanese consumer price index [23], and were converted to US dollars, using the OECD purchasing power parity rate in 2009. Cost data were collected from various published sources [Reference Yoshiyama21, 23–Reference Resch25]. Direct costs, such as inpatient and outpatient costs, and indirect costs arising from loss of productivity, as reported by the Japanese Ministry of Health, Labour and Welfare in 2009, were included. The cost of QFT screening included the screening kits, one physician visit, and the labour cost for laboratory technicians. The cost of TST screening included the labour cost for two physician visits and the TST reagents. The cost of CXR included the material cost of CXR, one physician visit, and the labour cost for radiological technicians. The cost of treating active TB, MDR-TB and INH chemoprophylaxis was determined based on the published literature. [Reference Yoshiyama21, Reference Resch25] The costs of the smear and culture examinations of sputum were also considered when active TB was detected by CXR. The cost of standard 9-month INH chemoprophylaxis, as well as the treatment of adverse effects was considered [Reference Yoshiyama21] (Table 1).
The main outcome measure of effectiveness was quality-adjusted life-years (QALYs) gained. The incremental cost-effectiveness of each screening arm was applied and compared. All costs and clinical benefits were discounted at a fixed annual rate of 3%.
Health state utilities were calculated using a utility weight of 0·80 for active non-MDR-TB, 0·58 for active MDR-TB, 0·85 for non-MDR-TB taking chemoprophylaxis (9 months) with complication, and 1·00 for non-MDR-TB taking chemoprophylaxis with no complication [Reference Resch25, Reference Marra26].
Sensitivity analyses
Using one-way and two-way sensitivity analyses, the range of cost-effectiveness was explored by comparing all strategies simultaneously to determine which strategy yielded the greatest benefits. Each model variable was assigned a distribution based on the values in the literature or assumptions. We also conducted probabilistic sensitivity analysis with Monte Carlo simulation, using the cost-effectiveness acceptability curve. The type of distribution used for each variable, their median and 95% confidence interval (if available) are given in Table 1. On Monte Carlo stimulation distributions, costs are in log-normal distribution and utilities are in β distribution.
RESULTS
Base case
In the base-case analysis, QFT [US$1477·92, 27·91988 quality-adjusted life-years (QALYs)] was more cost-effective than TST (US$1890·20, 27·89481 QALYs), TST followed by QFT (US$1515·38, 27·91999 QALYs), and CXR (US$8911·10, 26·55811 QALYs). When HIV infection was added to the base-case analysis, QFT (US$1791·31, 27·90787 QALYs) remained more cost-effective than TST (US$2156·64, 27·88342 QALYs), TST followed by QFT (US$1837·21, 27·90785 QALYs), and CXR (US$9240·92, 26·53962 QALYs). Similarly, addition of MDR-TB to the base-case analysis found QFT (US$1552·84, 27·91540 QALYs) more cost-effective than TST ($US1971·02, 27·89033 QALYs), TST followed by QFT (US$1590·16, 27·91552 QALYs), and CXR (US$9006·08, 26·55363 QALYs). When both HIV infection and MDR-TB were added to the base-case analysis, QFT (US$1866·05, 27·90339 QALYs) remained more cost-effective than TST (US$2236·32, 27·87895 QALYs), TST followed by QFT (US$1911·03, 27·90338 QALYs), and CXR (US$9353·27, 26·53518 QALYs) (year 2012 values) (Table 2). On the analysis of all prisoners exposed to MDR-TB, QFT (US$2548·27, 27·85583 QALYs) remained more cost-effective than TST (US$3044·75, 27·83089 QALYs), TST followed QFT (US$2583·60, 27·85608 QALYs), and CXR (US$10267·99, 26·49407 QALYs). Incremental cost-effectiveness ratio of QFT-alone strategy resulted in a cost saving of US$141580·09/QALYs gained compared to the TST followed by QFT strategy.
Cost-effectiveness of four strategies for TB screening of prisoners (BCG vaccination rate = 0·977)
CXR, Chest X-ray examination; HIV, human immunodeficiency virus; ICER, incremental cost-effectiveness ratio; MDR-TB, multidrug-resistant tuberculosis; QALYs, Quality-adjusted life-years; QFT, QuantiFERON®-TB Gold In-Tube; TB, tuberculosis; TST, tuberculin skin test; TST/QF, TST followed by QFT strategy.
Sensitivity analyses
One-way and two-way sensitivity analyses were performed over the ranges for each variable. The cost-effectiveness was not sensitive to BCG vaccination rate (Table 3), LTBI rate, TB rate, HIV infection rate and MDR-TB rate.
Sensitivity analysis of BCG vaccination rate
BCG, Bacillus Calmette-Guérin; CXR, chest X-ray examination; HIV, human immunodeficiency virus; ICER, incremental cost-effectiveness ratio; MDR-TB, multidrug-resistant tuberculosis; QALYs, quality-adjusted life-years; QFT, QuantiFERON®-TB Gold In-Tube; TST, tuberculin skin test; TST/QFT, TST followed by QFT strategy.
Probabilistic sensitivity analyses
The cost-effectiveness acceptability curve of 20-year-old prisoners by Monte Carlo simulations for 10 000 trials demonstrated that QFT strategy was the most cost-effective, with a value of 100% at all willingness-to-pay levels compared to TST, TST followed by QFT, and CXR strategies (Fig. 2).
Cost-effectiveness acceptability curve. The curve uses willingness to pay to chart the changing percentage of interventions for which QFT strategy (◆) is cost-effective relative to TST, TST followed by QFT, and CXR strategies. The cost-effectiveness acceptability curve demonstrates that the QFT strategy was the most cost-effective, with a value of 100% at all willingness-to-pay levels compared to TST, TST followed by QFT, and CXR strategies. QFT, QuantiFERON-TB Gold In-Tube; QALY, quality-adjusted life-year.
DISCUSSION
In this study, we demonstrated that the QFT strategy was the most cost-effective for entry TB screening in prisons in developed countries. We previously reported the cost-effectiveness of IGRA compared to TST and CXR for screening high-risk group such as healthcare workers (HCWs) and rheumatoid arthritis patients prior to initiation of tumour necrosis factor-α antagonist therapy, and also showed that QFT-alone yielded the greatest benefits at the lowest cost [Reference Kowada9, 11–Reference Kowada13]. de Perio and colleagues demonstrated that use of the QFT leads to superior clinical outcomes and lower costs than the TST. They also concluded that QFT should be considered for screening non-BCG-vaccinated and BCG-vaccinated new HCWs for LTBI [Reference de Perio10]. Hardy and colleagues showed that QFT blood testing followed by CXR is feasible for TB screening, cheaper than screening using the NICE guideline and identifies more cases of LTBI in immigrants from high-risk countries [Reference Hardy14].
To our knowledge, this study is the first cost-effectiveness analysis of IGRA compared to TST, TST followed by QFT and CXR for TB screening in prisons using Markov models. Our models also considered influences of HIV infection and MDR-TB in prisoners and found that QFT remained the most cost-effective strategy. Our finding that QFT provided superior cost-effectiveness probably results from the higher specificity of QFT compared to TST and CXR. Cost-effectiveness was not sensitive to the BCG vaccination rate. These findings may be applicable to other developed countries regardless of the BCG vaccination rate. Effective TB screening at diagnosis also enabled the modelling to avoid the thorny issue of ‘churn’ that occurs in many prisons, where prisoners on remand are moved rapidly between a number of prisons within the prison estate. This inevitably can also lead to increased transmission as well as difficulty in follow-up of any infected prisoners.
Baussano and colleagues examined the high incidence of LTBI and TB in prisons compared to the incidence in the corresponding local general population by systematic review [Reference Baussano27]. They concluded that improved TB control in prisons could potentially protect prisoners and staff from within-prison spread of TB and also reduce spread to the community, and would significantly reduce the national burden of TB. Reichard and colleagues [Reference Reichard28] assessed TB screening and management activities in large US jail systems. They concluded that completion rates and timeliness of TB screening, diagnostic, and treatment measures should be evaluated to identify areas needing improvement for documenting and monitoring inmate healthcare related to TB. They also suggested promoting therapy completion for inmates as a means of preventing TB transmission in the community. Lee and colleagues [Reference Lee29] demonstrated that TB services delivered in prisons have increased in the last decade, but systematic information on funding levels and gaps, services provided, and cost-effective delivery models for delivering TB services in prisons were lacking. Reid and colleagues [Reference Reid30] argued that improved implementation of conventional TB control activities in a broader range of public health interventions and prevention for HIV was needed for TB and HIV control in sub-Saharan African prisons. Several studies have demonstrated that the TB case-finding rate is greatly increased by utilizing mini-CXR screening [Reference Puisis31–Reference Jones and Schaffner33]. Cost-effective TB screening in prisons should be reconsidered for control of TB in prisons by public health intervention using tax, not only to significantly reduce TB in prisons, but also to avoid the spread of MDR-TB in the local community throughout the world. Both prison resource and containment strategies vary across countries. Optimal TB strategies across countries will be needed globally in the near future.
Our study had several limitations. First, only HIV infection and MDR-TB were considered in this model, but prisoners have more complex health problems in their society, such as alcoholism, smoking, and drug abuse. Second, there are few published data on the rates of LTBI and active TB, as well as transmission, in prisons. Third, the high costs of purchasing and maintaining equipment for CXR examination was not calculated in our model. Fourth, the period of imprisonment is not calculated in our model and we could find no data on the median time period of imprisonment. Fifth, each model variable was assigned a distribution based on the values in the literature or assumptions, performing one-way and two-way sensitivity analyses. However, there were only sparse TB and cost data for prisons globally. Different countries have different policies and resources for TB screening of prisoners. Further studies are needed to identify more specific TB policies in each country. Sixth, this analysis was for entry TB screening in prisons. Further cost-effectiveness studies for annual LTBI screening of prisoners using IGRAs are needed. Seventh, the time lag of the wait for sputum culture results of MDR-TB is not included in this Markov model. Finally, only a 9-month daily isoniazid regimen was considered in our model. For example, in a randomized controlled trial Chan and colleagues [Reference Chan34] demonstrated that a 4-month daily rifampicin regimen was safer and had a higher completion rate than the standard 6-month daily isoniazid regimen as a treatment for LTBI in male prisoners.
We conclude that QFT-alone is the most cost-effective strategy for entry TB screening in prisons in developed countries. Entry TB screening of prisoners using an IGRA should be considered on the basis of its cost-effectiveness by public health intervention, not only to significantly reduce TB in prisons, but also to better control the spread of TB infection in the community.
DECLARATION OF INTEREST
None.
