Comparative Effectiveness of Fidaxomicin for Treatment of Clostridium Difficile Infection

, , , ,
The American Journal of Pharmacy Benefits, July/August 2014, Volume 6, Issue 4

Our analysis models the clinical and economic value of fidaxomicin for the treatment of Clostridium difficile infection compared with metronidazole and vancomycin.

Clostridium difficile is a ubiquitous spore-forming organism and the most common identifiable cause of healthcare-associated diarrhea. Data from the National Inpatient Sample from 2000 to 2005 indicate that hospitalizations associated with C difficile infection (CDI) more than doubled, as did the case-fatality rate, from 1.2% to 2.3%.1 Severe complications, including colectomy, are also becoming more common.2 These widely reported increases in incidence, severity, and mortality of CDI are thought to be due at least in part to the emergence of a new virulent strain referred to as North American Pulsed Field type 1 and PCR ribotype 027 (ie, NAP1/027).3 The economic burden of CDI in the United States is estimated to exceed $1 billion annually.4

Current treatment options have significant limitations. Although not approved by the FDA for this indication, oral metronidazole is recommended as first-line therapy for mild to moderate disease by the Infectious Diseases Society of America (IDSA) guidelines.5 However, it is associated with a high rate of treatment failure, meaning a lack of clinical response requiring a change in therapy, of around 20%.6-8 Moreover, its systemic absorption prevents its use for prolonged periods of time. Oral vancomycin is the only FDA-approved treatment for CDI, and is the recommended treatment for severe CDI. Compared with metronidazole, oral vancomycin is significantly more expensive.

CDI recurrence rates (defined as complete resolution of symptoms while on appropriate therapy, followed by reappearance of symptoms after treatment has been stopped) are about 20% to 30% under both treatments, most likely caused by disruption of the intestinal microbiotia because of the broad spectrum of these antibiotics.7-10 Recurrent CDI imposes a significant economic burden on the healthcare system,11,12 particularly when patients require readmission. In addition, patients with multiple recurrences may require prolonged and repeated courses of therapy with oral vancomycin and sometimes other adjunctive therapies.

One additional concern—particularly for vancomycin but also for metronidazole13—is the possible selection of vancomycin-resistant enterococcus (VRE), which can lead to invasive infections with substantial risk to patients and cost to the healthcare system.14 Consequently, CDC15 and other organizations5 recommend against the use of vancomycin as first-line treatment for CDI.

Fidaxomicin is a narrow-spectrum macrocyclic antibiotic that has recently been approved by the FDA and other regulators as treatment for CDI.13 Cure rates with fidaxomicin are similar to those seen with vancomycin.16 Its main advantage is that it lowered the risk of recurrent CDI by almost 50%16 and of VRE acquisition by as much as 24%.17 Fidaxomicin is considerably more costly than metronidazole and more costly than vancomycin, raising questions about its value for money. Against this background, we set out to assess the differential clinical and economic impact of fidaxomicin compared with metronidazole and vancomycin with a microsimulation model.


Model Description

We developed a cohort-level Markov model to simulate the clinical and economic outcomes from the healthsystem perspective for a hypothetical population of adult inpatients at risk of developing CDI. We defined 4 treatment strategies (referred to hereafter as paths) for initial CDI based on current guidelines and evidence (Table 118). Treatment strategies were defined by various drug combinations for the initial infection, as well as first, second, and third (or more) recurrences. A schematic diagram of the model structure is shown in Figure 1. Health states include the initial CDI treatment episode, recurrent CDI treatment episodes, persistent CDI requiring retreatment, VRE colonization, and VRE infection. Patients could reside in both the CDI health states and the VRE health states simultaneously. The simulation ended when the entire population transitioned through the model into 1 of 5 final health states: (1) never developed CDI, (2) recovered from CDI, (3) persistent CDI requiring treatments not considered in the model, (4) death due to CDI, or (5) death due to VRE infection. The time horizon of the model was 1 year.

The model was validated during and after development through the checking of code by a second researcher and by the examination of results under a range of input values. Costs and outcomes were not discounted because CDI treatment is on average relatively short in duration (ie, days). All costs are expressed in 2010 US dollars. Costs from the literature were adjusted to 2010 US using the Medical Care Services component of the Consumer Price Index.19 The model was programmed in Visual Basic for Applications, with an accompanying series of Microsoft Excel work sheets for displaying model inputs and results. We followed the modeling guidelinesof the International Society for Pharmacoeconomics and Outcomes Research.20

Evidence Review

We conducted a review of the CDI-related literature to identify model parameter values. The search was limited to the years 2000 through 2010 and to studies published in English. The search terms are listed in the


. Two researchers screened 627 identified abstracts; 327 articles of interest were reviewed in full text. Reference sections for articles of interest were reviewed to identify articles not captured by our searches. All relevant parameter data identified were abstracted into tables. A total of 36 studies were used in deriving the actual point estimates for use in the model. We used studies whenever possible that were from the United States and/or from randomized controlled trials. We also prioritized estimates that best fit the parameters we needed for the model (ie, required no manipulation to be used in the model). Parameter ranges were noted for all studies to help establish reasonable ranges for sensitivity analyses. To validate our search and summary of the literature, as well as to establish plausible point estimates and ranges for model parameters for which no strong evidence in the literature was identified, we conducted interviews with 12 national experts in CDI treatment, costs, and outcomes. We identified experts through seminal papers and guidelines. All experts had academic affiliations, most were clinicians, and some were receiving or had received funding from Optimer Pharmaceuticals in the past.

Because of data limitations, a number of assumptions were required. We assumed that metronidazole,vancomycin, and fidaxomicin were the only relevant CDI treatments, that physicians generally followed IDSA treatment guidelines, that no adverse events or side effects occurred (except for VRE acquisition and infection), and that patients were fully adherent to prescribed treatments. Efficacy assumptions included that vancomycin had the same efficacy regardless of dose and that efficacy was the same in both the inpatient and outpatient settings. When patients exited the model because of retreatment failure (ie, if they switched treatments and then failed yet again while on treatment), we did not assign them further costs because of lack of data for this population. Because of a lack of data on costs of CDI in settings such as skilled nursing facilities and long-term care facilities, we assumed that patients were only discharged from the hospital to their home. Lastly, we did not account for the impact of CDI on quality of life, and assumed that all CDI patients are the same (ie, no modeling of CDI patient subgroups).

Data Sources

Table 2

21-64 lists all key model inputs, their respective point estimates and ranges for sensitivity analysis, and the reference(s) upon which each estimate is based. The treatments for strategies 1, 2, and 3 were based on Society for Healthcare Epidemiology of America/IDSA guidelines, supplemented with information on typical treatment and prescribing practices collected during our expert interviews. 5 The fidaxomicin treatment path (path 4) was designed to most closely reflect the pivotal fidaxomicin clinical trials.16 Efficacy estimates for vancomycin and fidaxomicin were based on results from the pivotal fidaxomicin clinical trial17 and a variety of other data sources.6,8,10,42,46,51 Efficacy for metronidazole was based upon the best available effectiveness data in the published literature.6-9,39-42,47-49,51

Model Analysis

Our primary clinical outcome of interest was recurrences of CDI avoided. Other outcomes included persistent CDI requiring treatment change, readmissions due to CDI, CDI-related deaths, VRE colonization, and VRE infections avoided. We compared the clinical effectiveness of the different treatment strategies by comparing the total number of expected outcomes for each treatment path. We then generated the cost of each treatment path to generate the incremental cost per outcome of interest avoided. The focus of our analysis was a comparison of effectiveness and cost between the fidaxomicin path (path 4 in Table 1) and the metronidazole- and vancomycin-based treatment paths (paths 1, 2, and 3 in Table 1).

To assess the robustness of model results, we conducted 1-way sensitivity analyses and probabilistic sensitivity analyses on key model inputs (assuming a uniform distribution for all variables). The 1-way sensitivity analysis allowed us to inspect the sensitivity of model results to changes in each key model parameter. The probabilistic sensitivity analyses consisted of 1000 model runs where input parameters were simultaneously varied randomly across the ranges reported in Table 2 for each model run. All sensitivity analyses used the primary outcome of incremental cost per recurrence avoided.

Role of the Sponsor

This study was funded through a contract from Optimer Pharmaceuticals, makers of fidaxomicin. Optimer Pharmaceuticals reviewed the model during its development, but the authors made final decisions regarding model design. The sponsor had no other role in the conduct of the study or in the collection, analysis, or interpretation of the data. A draft manuscript was reviewed by Optimer Pharmaceuticals, but the manuscript was written and developed independently of the project sponsors by the authors, who made final decisions regarding the content and study conclusions.


Clinical Outcomes

The fidaxomicin path (path 4) was superior for all of theclinical outcomes considered (

Table 3

). Initial treatment with fidaxomicin had the strongest effect on recurrences, with a reduction of anywhere from 43.65% (16.2 recurrences per 100 CDAD patients) compared with vancomycin, to 45.8% (17.8 recurrences per 100 CDAD patients) compared with metronidazole. As expected, fidaxomicin also reduced the number of hospital readmissions for CDI. In addition, fidaxomicin reduced the number of VRE colonizations, particularly compared with vancomycin (8.3 per 100 CDI patients vs 18.6 per 100 CDI patients).

Economic Outcomes

Initial treatment of CDI with fidaxomicin increased the cost per CDI case (including the initial episode and any related recurrences) by $990 to $2318 compared with other treatment pathways based on vancomycin and/or metronidazole (

Table 4

). Although fidaxomicin reduced nondrug costs (because of reduced recurrence rates) and VRE-related costs, the higher cost of the drug meant that overall treatment costs were higher.

The incremental cost per recurrence avoided, which was our primary end point, was $6109 to $13,027 higher for the fidaxomicin-based pathway compared with all others (Table 4). Similarly, incremental costs per other outcomes avoided (including persistent CDI, CDI-related death, VRE outcomes, and readmissions) were all higher for fidaxomicin.

Sensitivity Analyses

Our 1-way sensitivity analysis found results to be most sensitive to changes in the probabilities of recurrent CDI for the initial CDI episode for both metronidazole and vancomycin (and to a much lesser extent, for fidaxomicin), all VRE clinical-related probabilities, probability of recurrence (beyond the first 2 recurrences) if treated with vancomycin, cost of an invasive VRE infection, and the fidaxomicin drug cost. In only 1 of the 1-way sensitivity analyses did the incremental cost per recurrence avoided change in terms of the direction of the relationship (ie, in all other instances fidaxomicin was more costly, but also more effective, than the 3 comparator paths). The 1 exception was for the probability of recurrent CDI when treated with vancomycin for the initial CDI episode; when set to the minimum of its range, the fidaxomicin-based path (path 4) became cost saving (ie, less costly, more effective) relative to the vancomycin-based path (path 3). Our probabilistic sensitivity analysis results are shown in

Figure 2

. The red dot reflects the incremental cost per recurrence avoided as estimated by the model when parameters were set to the base case values shown in Table 2. The blue dots reflect estimates of incremental cost per recurrence avoided as estimated from 1000 model runs where input parameters were varied randomly across the ranges reported in Table 2 for each model run. From the scatter plots, we concluded that our results were reasonably robust to simultaneous changes in the input variables. For example, in comparing path 4 with path 1, path 4 was more costly but also more effective in 95.5% of simulations, less costly but more effective in 2.8% of simulations, and more costly but less effective in 1.5% of simulations.

We conducted a threshold analysis between path 4 and paths 1, 2, and 3 by varying the price of fidaxomicin downward to determine the price at which the fidaxomicin path becomes cost saving (ie, less costly, more effective) relative to at least 1 of the 3 comparators. The daily drug cost at which the fidaxomicin-based path became cost saving versus the other treatment paths (assuming the length of therapy as outlined in Table 1) occurred at a price of $190 (for two 200-mg fidaxomicin tablets) for path 3, $90 for path 2, and $67 for path 1. In addition, some patient subgroups are likely to have higher risk of recurrence and higher cost of recurrences, so we conducted a threshold analysis to ascertain at what cost of recurrence (based on attributable hospitalization cost for recurrence, previously set at $5979) fidaxomicin would become cost saving. We found that the cost of recurrenc would need to be greater than $48,000 (compared with path 1) and $44,000 (compared with path 2) and $25,000 (compared with path 3).


We constructed a health economic model to evaluate the comparative effectiveness of fidaxomicin for the treatment of CDI. We found that initial treatment with fidaxomicin produced the best clinical outcomes, but also the highest costs because savings resulting from lower utilization of medical care were outweighed by the higher drug costs. The estimated incremental cost to avoid a CDI recurrence using fidaxomicin ranged from $6109 to $13,027. Fidaxomicin would be cost saving inpatients whose cost of recurrence is between $25,000 and $48,000. Our findings were robust to changes in key model inputs under 1-way and probabilistic sensitivity analyses, with the exception of 1 variable (probability of recurrent CDI if treated with vancomycin for the initial CDI episode).

It should be kept in mind, however, that our model takes the health-system perspective, implying that individual stakeholders within the healthcare system will face different trade-offs. Under the current payment system, there is a distortion in incentives to reduce recurrences. Hospitals receive a fixed payment per admission under the diagnosis-related group payment system. Thus, if a patient is being treated for CDI, the hospital must absorb the full incremental cost of any drug that costs more than the cheapest alternative (eg, fidaxomicin compared with metronidazole). The benefits of a more expensive drug such as fidaxomicin (eg, reduction in recurrences) are not seen by the hospital except in the case of those few patients who had already been admitted when the recurrence occurs. In addition, hospitals would forgo additional revenue they would have received for patients who would otherwise have been readmitted for recurrent CDI. Thus, most of the benefit would accrue to payers. Novel payment approaches might better align those distorted financial incentives. For example,the Medicare program could extend its policy of not providing additional payments for avoidable complications such as central-line associated bloodstream infections65 to CDI recurrences. Accountable care-type payment arrangements, under which payers and providers share savings derived from superior care, would further align incentives.66 Finally, a new technology add-on payment for fidaxomicin from the CMS will provide hospitals with a payment, in addition to the standard-of-care diagnosisrelated group reimbursement, of up to 50% of the cost of the drug for a period of 2 to 3 years, effective in the fiscal year starting on October 1, 2012. This will also positively impact the value proposition of fidaxomicin for hospitals, at least in the near future.

Even under current conditions, fidaxomicin might be the primary treatment choice for high-risk subgroups such as elderly patients, immunocompromised patients, and those on concomitant antibiotics—if the risk of recurrence and the associated cost of treating the recurrence are high enough. As stated, we estimated that fidaxomicin would be cost saving if the cost of a recurrence were to exceed $25,000 to $48,000 (depending on the treatment being compared). For example, immunosuppressed oncology patients are more likely to require a prolonged hospital stay for a CDI recurrence, suggesting that the higher upfront cost of fidaxomicin could be recouped. In addition, recent data indicate fidaxomicin resolves diarrhea in oncologic patients 2 days faster than vancomycin,46 which might allow patients to resume chemotherapy more quickly. Patients with inflammatory bowel disease or frail elderly patients might represent other high-risk subgroups in which primary use of fidaxomicin could be cost neutral. Data for differential recurrence rates and costs in such high-risk subgroups are currently not available, but ongoing research is likely to provide a better basis for decision makers.

Given the lack of data, there were other potential benefits for which we could not account. At this time, there is no literature to inform the use of quality-adjusted life-years (QALYs) for health states related to CDI. In particular, patients who struggle with repeated recurrences over many months are likely to have a markedly diminished quality of life and increased suffering,67 but this has not been quantified in a way that could be used by our model. In order to truly understand the value of fidaxomicin, a cost-effectiveness analysis that accounted for human suffering by incorporating QALYs is needed. We did consider the indirect benefits of fidaxomicin, such as avoidance of VRE colonization and subsequent infection. Although we were not able to capture the savings of avoiding VRE outbreaks (and the attendant infection control costs), such costs would have to be very high to counterbalance the cost of fidaxomicin.

We also made some simplifying assumptions for our model, which resulted in some study limitations. For example, caution should be used in applying our results, which were based on the average CDI patient, to specific CDI patient subgroups (eg, the very elderly) because the costs and benefits of different treatment strategies may differ significantly from those of specific subgroups of CDI patients. We used a mix of efficacy and effectiveness data for metronidazole (given the lack of a single seminal trial to use as a base case estimate) and efficacy (clinical trial) data for vancomycin; thus, our results may be biased against metronidazole. We were unable to incorporate the outcomes of CDI patients who repeatedly fail the treatment because of data limitations. After the second recurrence, we assumed that the probability of treatment success and recurrence remained the same for every subsequent recurrence, because there is no literature or even expert consensus to support data for these differential probabilities. Finally, given the quite recent introduction of fidaxomicin, there is little empiric information on how clinicians are using the drug, and no guidelines from clinical societies exist yet for its use. We did not, for example, include a pathway in which fidaxomicin is used as second-line treatment (primarily because of lack of data).

In conclusion, we developed a comparative effectiveness model of 4 CDI treatment strategies, 1 of which included the use of a newly approved drug, fidaxomicin. We found that a treatment path beginning with fidaxomicin produced superior clinical outcomes but higher costs relative to 3 other treatment paths. Fidaxomicin would be cost neutral or cost saving for patient subgroups with higher costs of recurrence. However, even in scenarios in which fidaxomicin costs more than the current alternatives, stakeholders may need to make a value judgment about whether the clinical benefits of fidaxomicin are worth the additional cost. Our simulations provide guidance for decision makers to optimize use of this novel and effective treatment while also controlling costs.