In our analysis, 4R was the least expensive regimen for the treatment of LTBI. Over the patient's lifetime, the 4R regimen was less expensive and more effective than the current standard of care (i.e., 9H) over a wide range of estimates for adherence and efficacy. Although more expensive than 4R or 9H, 3HP was more effective, at a cost of $48,997 per QALY compared with 4R and $25,207 compared with 9H, and would therefore be considered cost-effective using our threshold of $50,000 per QALY. Furthermore, this regimen became cost-saving for patients at very high risk or at extreme likelihood of treatment nonadherence. For example, patients with HIV infection have a risk of activation approximately 10 times the baseline estimate (16), so 3HP may be cost-saving in these individuals. However, the cost-effectiveness of 3HP at baseline risk was highly sensitive to its efficacy relative to 9H and to our estimates for adherence. If the ongoing large clinical trial shows that 3HP is less efficacious than 9H, its use would likely be restricted to these highest-risk individuals unless the price is reduced and self-administered therapy is considered an acceptable option.

Isoniazid given by DOT for 9 months was very expensive in our model and did not compare favorably to the other regimens. Nonetheless, this regimen may be useful in instances where nonadherence to self-administered therapy is likely and rifamycins are contraindicated, such as in patients taking boosted HIV protease inhibitors. Additionally, two studies looking at the cost-effectiveness of observed isoniazid in drug users found it to be cost-saving compared with no therapy (17, 18). In both instances, however, the costs of DOT were very low because the procedures were performed by clinic staff already seeing the patients for other purposes, and no shorter regimens were compared.

Our model is the first to directly compare all three CDC-recommended regimens for the treatment of LTBI. Our data are consistent with recent cost data obtained by Aspler and colleagues in a randomized, controlled trial of isoniazid versus rifampin (19). Based on their cost projections, 4 months of rifampin could be 20% less efficacious than isoniazid and still be cost-saving. Conversely, Kahn and colleagues compared isoniazid, rifampin, and rifampin plus pyrazinamide in recent immigrants and found that rifampin was more costly than isoniazid for most patients (but more effective), although their estimated drug costs for rifampin were substantially higher than our current data ($97.70/mo vs. $27.70/mo). A notable difference between this study and ours is their focus on drug resistance patterns. We assumed that everyone in our model was infected with organisms susceptible to all drugs. However, because isoniazid resistance is the most common resistance pattern seen in the United States (20), incorporation of susceptibility patterns into our model would have biased the results further toward rifampin monotherapy.

One significant limitation to our study is the fact that limited data exist to derive point estimates for the efficacy and adherence to 3HP. We used a wide range of parameters in sensitivity analyses, but adherence rates for this regimen without DOT have never been published. In one retrospective study comparing adherence to daily and weekly self-administered alendronate, patients taking the weekly regimen were more adherent over a 1-year period (21), but whether this pattern would be shared with 3HP in a population of patients with LTBI is unknown. If TB programs begin to use isoniazid plus rifapentine in the future without direct observation, publication of data on adherence would be necessary to evaluate its overall utility.

In our model, the costs for each regimen were driven largely by the costs associated with active cases. We used a conservative approach, attempting to reproduce the current costs of treating active TB as much as possible. Other studies have used data from Brown and colleagues (22) in their estimates of the cost of treating an active case, but in that study an average of 77% of patients were hospitalized for an average of 20 days. More recent data from Taylor and colleagues (23) suggest that many fewer patients are hospitalized and for shorter periods, and we used these estimates in our study. There is also some controversy over the total costs associated with active TB. Most studies only include the costs of acute treatment of active disease. However, Pasipanodya and colleagues (24) demonstrated significant respiratory impairment after the treatment of active TB (2), and the societal costs associated with this morbidity may be over $150,000 per case (23). If these costs are included, our model predicts that 3HP would be more effective and less expensive even for very low-risk individuals (data not shown).

Isoniazid remains the standard of care for treating LTBI in adults in the developed world. However, if the efficacy of isoniazid plus rifapentine is confirmed in the ongoing clinical trial, our model suggests that this combination will be a more effective and less expensive alternative for high-risk individuals, particularly those with HIV, and will be cost-effective even for lower-risk patients. Meanwhile, our results indicate that increased use of 4 months of rifampin in place of 9 months of isoniazid to treat LTBI would be cost-saving for most programs in the United States and Canada.