Cost Effectiveness of Universal Hepatitis B Virus Screening in Patients Beginning Chemotherapy for Sarcomas or GI Stromal Tumors.

PURPOSE: The value of screening for hepatitis B virus (HBV) infection before chemotherapy for nonhematopoietic solid tumors remains unsettled. We evaluated the cost effectiveness of universal screening before systemic therapy for sarcomas, including GI stromal tumors (GISTs). PATIENTS AND METHODS: Drawing from the National Cancer Centre Singapore database of 1,039 patients with sarcomas, we analyzed the clinical records of 485 patients who received systemic therapy. Using a Markov model, we compared the cost effectiveness of a screen-all versus screen-none strategy in this population.
RESULTS: A total of 237 patients were screened for HBV infection. No patients developed HBV reactivation during chemotherapy. The incremental cost-effectiveness ratio per quality-adjusted life-year (QALY) of offering HBV screening to all patients with sarcomas and patients with GISTs exceeded the cost-effectiveness threshold of SG$100,000 per QALY. This result was robust in one-way sensitivity analysis. Our results show that only changes in mortality rate secondary to HBV reactivation could make the incremental cost-effectiveness ratio cross the cost-effectiveness threshold.
CONCLUSION: Universal HBV screening in patients with sarcomas or GISTs undergoing chemotherapy is not cost effective at a willingness to pay of SG$100,000 per QALY and may not be required.

INTRODUCTION

Hepatitis B virus (HBV) reactivation (HBVr) is a well-recognized complication of immunosuppressive therapy in patients chronically infected with HBV, defined as those positive for hepatitis B surface antigen (HBsAg). HBVr is associated with a range of complications, and in patients with cancer, this can lead to delay or premature discontinuation of chemotherapy and compromise oncologic outcomes. The use of particular therapies in certain tumors renders specific groups of patients susceptible to HBVr.[1,2] B lymphocyte–depleting agents, such as rituximab, the monoclonal antibody against CD20, are especially immunosuppressive. Their use in the treatment of B-cell non-Hodgkin lymphoma, in concert with steroids, causes HBVr in up to half of HBsAg-positive patients with lymphoma treated with rituximab-based regimens.[3-5] Several prospective trials have shown reductions in rates of HBVr and HBV flare with use of prophylactic antiviral therapy in this population.[6,7] Universal screening for HBV in treatment of B-cell non-Hodgkin lymphoma is thus effective. It has also been shown to be cost effective[8] and is now universally recommended when initiating therapy in patients with lymphoma.[9]

The data for universal HBV screening in solid tumors are less clear. A cost-effectiveness analysis by Day et al[10] using a model based on the treatment of solid tumors revealed a pooled cost-effectiveness ratio of nearly $150,000 per life-year saved. The study included only patients undergoing chemotherapy for early breast cancer or advanced non–small-cell lung cancer. The absence of directly immunosuppressive treatment and the less favorable natural history and treatment outcomes in advanced solid tumors compared with lymphoma have been offered as reasons for this disparity. However, the heterogeneity in biology, prognosis, and therapeutic regimens used for various solid tumors probably necessitates tumor-specific evaluations of the cost effectiveness of universal HBV screening. This is especially important to systematic evaluation in an HBV-endemic region like Singapore, where the high prevalence of chronic HBV infection (3.6%)[11] would, according to the latest ASCO provisional clinical opinion update, necessitate that all patients be screened before starting systemic therapy.[9]

Sarcomas are heterogeneous yet uncommon tumors of mesenchymal origin that comprise 1% of adult malignancies.[12] In the setting of advanced disease, they are often treated with myelosuppressive cytotoxics, either singly or in combination; objective response rates are modest, with median survival of only 12 months.[13] GI stromal tumors (GISTs) are a striking exception to this rule, having become the prototype for successful targeting of oncogene-addicted cancers. With the development of imatinib, a potent inhibitor of the cKIT oncoprotein constitutionally activated in the majority of GISTs, median survival in patients with advanced GISTs is now 5 years.[14] Other than several case reports documenting HBVr with the use of imatinib,[15,16] to our knowledge, there has been no systematic evaluation of the value of HBV screening when treating sarcoma. Compared with other solid tumors, the doses and drugs used for systemic therapy in treatment of sarcoma are typically much higher. It would thus be of clinical interest to study this population of patients who receive therapy that is expectantly more myelotoxic. In this study, we sought to evaluate the incidence of HBVr in patients receiving chemotherapy for sarcomas or GISTs using data from this database, and to assess the cost effectiveness of universal screening for HBV infection before treatment initiation.

PATIENTS AND METHODS

We identified 1,039 patients who were diagnosed with biopsy-proven bony or soft tissue sarcomas or GISTs between January 1, 1992, and December 31, 2013, who were receiving medical treatment at the National Cancer Centre Singapore. Patients who did not receive any systemic therapy during this period were excluded from the study, leaving 274 evaluable patients with sarcomas and 211 with GISTs (Table 1). The medical records of these patients were reviewed.

Table 1

Demographic and Clinical Characteristics of Patients With Sarcomas or GISTs Who Received Chemotherapy (N = 485)

Definition of HBV Screening and HBVr

Patients were considered to have been screened when HBsAg testing was performed at any time before or up to within 6 months of initiation of systemic therapy. Chronic HBV infection is defined as being positive for HBsAg. Within the screened population, patients who were found to be HBV core antibody positive and HBsAg negative were considered to have had past HBV infection.

On the basis of a definition previously described by Lok et al,[17] hepatitis was defined as an abrupt rise in serum ALT of more than three-fold the upper limit of the laboratory reference range or an absolute increase of ALT to more than 100 U/L compared with prechemotherapy values.

Hepatitis attributable to an HBVr was defined as the presence of hepatitis as described earlier, as well as a rise in HBV DNA of 10-fold or more compared with prechemotherapy values or an absolute increase of more than 105 copies/mL.[18] Base-case values for HBVr and transition probabilities of hepatitis were estimated from our clinical cohort of 485 patients. The ranges of these parameters were derived from review of the literature.

Modeling Approach

We created a Markov model (Fig 1) to examine the cost effectiveness of a screen-all strategy versus a screen-none strategy in patients with sarcomas or GISTs who were beginning neoadjuvant, adjuvant, or palliative chemotherapy. The sarcoma and GIST populations were analyzed using separate models.

Fig 1

Model structure. (A) Diagram illustrates the structure of the Markov model of hepatitis B virus (HBV) screening in patients with sarcomas receiving chemotherapy. Square node denotes the two strategies in this study: universal screening and no screening. Patients under both strategies were classified according to intent of chemotherapy and type of drug combination used according to expected immunosuppressive effect. Regimens included doxorubicin only (D only), ifosfamide only (I only), doxorubicin and ifosfamide combination (DI), other doxorubicin or other ifosfamide combinations (other D/other I), gemcitabine and taxane combination (GemTax), other single-agent regimens, and combination regimens of two or more agents. We further subdivided the other D/other I group into various subtypes as follows: rhabdomyosarcoma, Ewing sarcoma, osteosarcoma, and others. Each subgroup of patients was further categorized depending on whether patients were chronically infected with HBV, had HBV infections that resolved, or had never been infected with HBV. Only one such breakdown is shown in the diagram because the rest shared the same structure. All patients were observed until death. The difference between the universal screening arm and no-screening arm are was that patients with chronic HBV were treated with lamivudine or entacavir prophylaxis. Circle M indicates the time point when follow-up started. Only the universal screening arm is shown because the other arm has an identical structure. (B) Diagram illustrates the structure of the Markov model of HBV screening in patients with GI stromal tumors (GISTs) receiving chemotherapy. The only difference from (A) is that patients with GISTs were classified according only to intent of chemotherapy. (C) All patients, except for patients without HBV infection, may develop hepatitis, which may be followed by discontinuation of chemotherapy, continuation of chemotherapy despite reactivation, resolution of hepatitis, or death. Patients without hepatitis may die as a result of cancer or other causes. Patients with hepatitis flare may die as a result of hepatitis, cancer, or other causes. The Markov cycle length was assigned to be 3 weeks, which is the duration of one cycle of chemotherapy. TKI, tyrosine kinase inhibitor.

Patients under both strategies were categorized according to clinical indications for chemotherapy (ie, neoadjuvant, adjuvant, or palliative chemotherapy) and further categorized based on chemotherapy regimen according to the myelosuppressive effect expected. The population of patients who received palliative chemotherapy was subdivided into various chemotherapy groups similarly.

Within the screen-all strategy, patients were screened for HBsAg before initiation of chemotherapy. HBV prophylaxis was then administered to patients who were chronic HBV carriers using either oral lamivudine (100 mg once per day) or entecavir (0.5 mg once per day) at the start of chemotherapy and continued for 6 months beyond completion. None of the patients within the screen-none strategy received antiviral prophylaxis.

Each patient moved from various states within a Markov model, as illustrated in Figure 1. These health states included mortality risks associated with other medical conditions, HBV infection, and cancer. The input data for mortality related to other medical conditions was obtained from a life table, whereas cancer-specific mortality was obtained from trial data. Mortality rate from HBVr was modeled as a cause of death independent of cancer or other medical conditions.

Model Inputs

The proportion of patients falling into each category was estimated with retrospective analysis of our cohort of patients with sarcomas or GISTs. The clinical probabilities used were based on our clinical cohort of 485 patients. Probability estimates of HBVr for sensitivity analysis were derived from systematic review of the literature (Table 2).

Table 2

Clinical Event Probabilities and Utilities

Cost-Effectiveness Analysis

All costs were adjusted to 2015 Singapore dollar values (cost details listed in Appendix Table A2). Effectiveness was quantified in terms of quality-adjusted life-years (QALYs), which are the sum of products of health state utility and the duration in each health state. Costs and QALYs were discounted at 3%.[30] We calculated incremental cost-effectiveness ratios (ICERs), defined as the additional cost in Singapore dollars per gain in QALYs in patients using the more expensive strategy over the less expensive strategy. We used an ICER of SG$100,000 per QALY instead of $50,000 per QALY as the cost-effectiveness threshold, because it was thought to reflect inflation and economic growth in the past two decades,[31] and it has been used in other published cost-effectiveness analyses.[32] Cost effectiveness was calculated from the societal perspective limited to direct medical costs.

Table A2

Cost Estimates

To account for uncertainty in parameter estimates, we conducted a one-way sensitivity analysis by examining ICERs with different inputs of each parameter within plausible range. We also performed a probabilistic sensitivity analysis by simultaneously sampling values for all parameters from their plausible ranges and calculating the distribution of ICERs for 10,000 iterations. Beta and gamma distributions were used to represent probability parameters and cost parameters, respectively.[33] All analyses were performed using TreeAge Pro 2015 (TreeAge Software, Williamstown, MA).

RESULTS

There were 274 patients with sarcomas and 211 patients with GISTs who received systemic therapy. Among all 485 patients who received chemotherapy, 237 (48.9%) were screened for HBV infection before initiation of chemotherapy (Table 1). Of the screened population, 13 patients (5.5%) were found to be chronic HBV carriers, and 28 (11.8%) were found to have had past HBV infection (subgroup analysis summarized in Appendix Tables A1 and A3). This is consistent with the national incidence of HBV in Singapore.[11]

Table A1

Subgroup Data of Patients With Sarcomas Who Received Chemotherapy (n = 274)

Table A3

Patients Screened for HBV

Base-Case Analysis

Table 3 lists the results of the base-case analysis. If HBV screening and prophylaxis were not offered, patients with sarcomas or GISTs who received chemotherapy would be expected to survive 4.3 and 11.5 years, respectively, which translated into 0.831 and 2.98 QALYs, respectively, when adjusted for utility and discounted for gains in the future. The difference in QALYs is attributable to the longer survival time and higher health state utility of patients with GISTs compared with patients with sarcomas. However, costs related to chemotherapy and treatment of HBVr are also higher for patients with GISTs than for those with sarcomas (SG$120,881 v SG$14,926), which is attributable to longer duration of chemotherapy and higher accumulated risks for HBVr among patients with GISTs.

Table 3

Cost, Effectiveness, and ICER at Base Case

In the base case, screening for both patients with sarcomas and patients with GISTs exceeded the SG$100,000 threshold by at least two-fold. Offering HBV screening led to greater improvements in QALYs for patients with GISTs than for those with sarcomas (0.007 v 0.002 QALYs), but at a higher incremental cost (SG$2,567 v SG$314); the net effect was that the ICER of offering HBV screening to all patients with sarcomas was smaller than that of offering screening to those with GISTs (SG$226,771 v SG$393,900 per QALY).

When the cost effectiveness of screening subgroups of patients based on therapeutic intentions was examined, screening those who received chemotherapy for adjuvant or neoadjuvant intentions was much more cost effective than palliative intentions (SG$138,071 v SG$1,855,517 per QALY for patients with sarcomas and SG$216,138 v SG$805,121 per QALY for patients with GISTs). However, even the smallest ICERs of these subgroup analyses was still higher than the cost-effectiveness threshold of SG$100,000 per QALY.

Sensitivity Analyses

One-way sensitivity analyses were conducted to determine the input parameters to which the results were most sensitive. Our analysis of screening all patients with sarcomas revealed that only changes in mortality rate secondary to HBVr could make the ICER cross the cost-effectiveness threshold. Screening became increasingly cost effective with a high rate of death resulting from HBVr, and the rate at which it crossed SG$100,000 was 18% (Fig 2A). Similarly, the cost effectiveness of screening all patients with GISTs depended only on mortality rate secondary to HBVr. When mortality rate secondary to HBVr was greater than 38%, screening patients with GISTs for HBV became cost effective (Fig 2B).

Fig 2

One-way sensitivity analysis. Diagram illustrates the range of incremental cost-effectiveness ratios (ICERs) of hepatitis B virus (HBV) screening in patients with (A) sarcomas or (B) GI stromal tumors (GISTs) when the value of each parameter is varied within plausible range when keeping the other variables constant. The axes cross at the base-case ICER (SG$226,771 per quality-adjusted life-year [QALY] for patients with sarcomas and SG$393,900 per QALY for patients with GISTs). Although the ICERs remained greater than the cost effectiveness of SG$100,000 per QALY when the values for most parameters were changed, HBV screening became cost effective when mortality risk resulting from HBV reactivation (HBVr) was greater than 18% for patients with sarcomas and 38% for those with GISTs. D, doxorubicin; I, ifosfamide.

Because there are no estimates of HBVr in patients with sarcomas or GISTs receiving chemotherapy in the literature, we examined conservative scenarios in which patients who received prophylaxis manifested no reactivation, whereas those who did not receive prophylaxis were subject to reactivation risk from 0% to 100%. Holding the other parameters at base-case value, the ICERs of conducting HBV screening in all patients with sarcomas and all patients with GISTs were greater than $190,000 and $280,000 per QALY, respectively, even when the risk of reactivation without prophylaxis was 100%.

Even at the higher bound of chronic HBV prevalence (4.2%), ICERs of screening patients with sarcomas or GISTs did not cross the cost-effectiveness threshold. The result was relatively robust to rates of reactivation with antiviral prophylaxis and utility states in both the palliative and neaoadjuvant or adjuvant arms. Probabilistic sensitivity analysis revealed that at a cost-effectiveness threshold of $100,000 per QALY, conducting HBV screening in patients with sarcomas and patients with GISTs before chemotherapy was not cost effective in 91.6% (Fig 3A) and 99.8% (Fig 3B) of the iterations, respectively.

Fig 3

Probability sensitivity analysis. Diagram illustrates distributions of incremental cost-effectiveness ratios in 10,000 iterations of probabilistic sensitivity analysis in patients with (A) sarcomas or (B) GI stromal tumors (GISTs). With an increasing cost-effectiveness threshold, the universal screening approach is more likely to be cost effective. At a cost-effectiveness threshold of SG$100,000 per quality-adjusted life-year (QALY), 91.6% and 99.8% of the simulations suggested hepatitis B virus screening to be not cost effective for patients with sarcomas and GISTs, respectively.

DISCUSSION

The findings of this study lend support to the ASCO provisional clinical opinion that for patients who neither have HBV risk factors nor anticipate cancer therapy associated with a high risk of reactivation, current evidence does not support HBV screening before initiation of cancer therapy.[9] Even in a population in which HBV infection is endemic, we have shown that it is not cost effective to practice universal screening for patients with sarcomas or GISTs.

Our analysis showed that the cost effectiveness of conducting HBV screening among patients with sarcomas or GISTs before chemotherapy depends on the probability of dying as a result of HBVr. The proportion of HBVr cases resulting in death may be affected by quality of health care infrastructure and availability of tertiary care that would vary geographically.

Our study used real-world data from a prospective clinical database with 485 patients, providing a more realistic application and analysis of the screen-all versus screen-none strategy. This is in contrast to previous cost-effectiveness analyses, which were purely based on input values derived from published literature. Our population was identified for study because there are no current available data regarding HBVr among patients with sarcomas or GISTs, making this the first comprehensive analysis to our knowledge of performing HBV screening in the sarcoma and GIST population. The population studied also encompassed a wide range of histologic subtypes, including both bony and soft tissue sarcomas, to accurately reflect the diversity of biologies and therapies. The systemic treatment of sarcomas involves the use of cytotoxic chemotherapy, either singly or in combination, associated with varying degrees of myelotoxicity. Conversely, GISTs are treated with small-molecule tyrosine kinase inhibitors. There is no evidence from our analysis that such therapies predispose patients with sarcomas or GISTs to HBVr, consistent with what has been shown for other solid tumors,[10] as distinct from the demonstrably immunosuppressive therapies used in the treatment of lymphoma and other hematologic malignancies.

Because of the scarcity of literature on HBVr in patients with sarcomas or GISTs and lack of reactivation within the database of data collected over 11 years, it was necessary to rely on data derived from published studies conducted among patients with lymphoma for our model inputs. We are aware of the clear association of anti-CD20–based therapy in lymphomas with HBVr as opposed to solid tumors, where such therapy is not used. As such, we could be confident that the ICERs we derived for sarcomas and GISTs would be underestimations, thus reinforcing the lack of cost effectiveness of HBV screening in sarcoma and GIST management.

Our study has a number of limitations. Sarcomas are a heterogeneous group of tumors with a wide spectrum of disease progression, response to treatment, and overall survival. In spite of this heterogeneity, sarcomas do share some common clinical features (eg, hematogenous rather than lymphatic spread and proclivity for lung metastasis) that lend clinical value to studying and managing them as one entity. To further account for this heterogeneity, we opted to divide patients into chemotherapy groups, within which patients may have had differing histologic subtypes. To address this limitation, we further classified one subgroup of patients who received other doxorubicin or ifosfamide combinations into histologic subtypes to better represent their disease characteristics (details provided in Modeling Approach in Appendix). The resultant small subset of chemotherapy groups may have masked the effect of more immunosuppressive therapies on reactivation rates. However, this was accounted for by sensitivity analysis in which we analyzed the results using the upper limits of reactivation rates described in the available literature. As such, the model output may be viewed as an estimate of the average cost effectiveness in any patients with sarcomas or GISTs. Our findings were consistent across both the neoadjuvant and palliative groups. The structure of our model is shown in Figure 1.

Although we found that the cost effectiveness of HBV screening was sensitive to the HBVr rate, it is notable that even at the highest reactivation rate, the ICERs of HBV screening were still much higher than the cost-effectiveness threshold (Fig 2). We also performed a scenario analysis in which we assigned all the parameters taken from lymphoma literature to the extremes of their possible ranges to make screening as cost effective as possible. Even at the highest possible risk of HBVr with prophylaxis, we derived an ICER of SG$172,197 per QALY for sarcoma and SG$174,602 per QALY for GIST. In addition, it is worth noting that relative risks of HBVr (ie, effectiveness of prophylaxis in preventing HBVr) had similar or greater influence on ICERs compared with corresponding HBVr rates. However, range of ICERs yielded by different relative risk estimates stayed above the cost-effectiveness threshold. These results indicate that our conclusion would likely remain unchanged with a more accurate estimate of HBVr risk or effectiveness of prophylaxis.

We recognize that the prevalence of HBV carriage may differ across various populations, and certain at-risk groups such as intravenous drug abusers may have higher prevalence of HBV carriage and thus be at higher risk of HBVr. The prevalence of chronic HBV carriage in a needle-sharing community has been reported to be as high as 40.0% in Chinese populations.[34,35] Using these input values in our model generated an ICER of SG$168,167 per QALY with a prevalence of 40% for patients with sarcomas and SG$381,333 per QALY for patients with GISTs. Additionally, we conducted a sensitivity analysis by allowing the prevalence of chronic HBV infection to be as high as 88% (not 100%, because infection was resolved in 11.8% of patients who had a history of acute chronic infection). We found that the marginal reduction in ICER decreased as the prevalence of chronic infection rose and leveled off at more than SG$160,000 per QALY and SG$380,000 per QALY for sarcomas and GISTs, respectively, which were higher than the threshold of SG$100,000 per QALY. Therefore, we are confident that our conclusion will not be changed by the prevalence of chronic HBV infection.

The only parameter whose change may make HBV screening cross the cost-effectiveness threshold is the mortality risk resulting from HBVr. We found that if this risk exceeded 18% for patients with sarcomas and 38% for those with GISTs, conducting HBV screening would be cost effective. On the basis of clinical experience, the mortality rate resulting from HBVr is unlikely to be higher than 5% in our local contexts because of close monitoring and prompt treatment once HBVr is detected. We recognize, however, that HBVr-related mortality cannot be considered constant around the world.

Our study suggests that universal screening before chemotherapy for patients with sarcomas or GISTs is not cost effective at a willingness to pay of SG$100,000 per QALY and should not be advised as part of routine prechemotherapy assessment. An exception to this recommendation may be made in settings where mortality resulting from HBVr is substantial or if other risk factors exist.