Some guidelines for BIAs have been published in an effort to standardize research procedures. To obtain a better understanding of existing recommendations regarding the scope of costs used in BIAs, we searched for and summarized these guidelines. Based on a previous review (18), 10 BIA guidelines were reviewed. These had been published by various organizations and countries including the ISPOR, the National Institute for Health and Clinical Excellence (NICE) in the United Kingdom, Canada, France, Ireland, Australia, the Netherlands, Belgium, Thailand, and Poland (16, 19–27). Microsoft Excel 2016 was used to summarize the various recommendations.

This systematic review was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (28). PubMed, EMBASE, Web of Science Core Collection, and the Cochrane Library were searched for studies published on BIAs for anticancer drugs from 1 January 2016 to 26 July 2021. The search terms and the detailed search strategies are presented in Appendix 1.

Studies were included if they (1) were original BIAs pertaining to anticancer drugs used to treat cancer patients, (2) reported the scope of costs, and (3) were published in English. Studies were excluded if they (1) did not focus on cancer patients, (2) did not include the scope of costs, (3) calculated the budget impact of biosimilars compared with that of the original drug (biosimilars usually have similar effects at a lower price compared with the original drugs, and thus a comprehensive scope of costs is generally considered unnecessary), or (4) were published in the form of a systematic review, meta-analysis, abstract, or dissertation. The literature search and screening were undertaken independently by two investigators. Any disagreements were adjudicated by senior investigators.

On the basis of the ISPOR Task Force guidelines (15, 16), we developed an evidence table summarizing how each study was designed, the scope of costs considered, and the results. The scope of costs was divided into two parts. The first part included the costs of intervention (target anticancer drugs), and was calculated by multiplying the unit price of the anticancer drug by the amount used in the target population. The second part included the impact on other costs, which consisted of two components: condition-related costs and indirect costs. In BIAs, condition-related costs usually include monitoring costs (costs of medical resources about monitoring disease progression or other events, e.g., imaging examination, laboratory examination and blood pressure monitoring), administration costs (costs of acquiring and using drugs, e.g., drug preservation and injection), adverse event (AE) costs (costs of drugs for AE management, e.g., leucocyte increasing agent for leukopenia), and subsequent treatment costs (costs of treatment after disease progression, e.g., immunotherapy after chemotherapy failure), while indirect costs usually include the costs of lost productivity and social services, referring to the working hours and productivity loss due to disease, disability, or death which includes the loss of salary for patients and their families/caregivers caused by discontinuing school, sick leave, and early death, etc., and in most cases are only considered when adopting the societal perspective. In addition to the scope of costs, 14 items regarded as essential for BIAs were included in the evidence table (29, 30): country, intervention, research funding, perspective, supported decision-making, target population, time horizon, market share, comparator(s), treatment duration, results of the BIA, uncertainty and scenario analyses, validation, and data sources. Then, we systematically extracted data and summarized the scope of costs considered in all of the included studies in evidence tables using Microsoft Excel 2016.

To illustrate the influence of the scope of costs on BIA results, we developed a Microsoft Excel-based budget impact model for an anticancer drug and estimated two cost-scope scenarios: scenario 1, which did not consider subsequent treatment costs, and scenario 2, which considered subsequent treatment costs. We then compared the results of these two cost-scope scenarios. The model was developed based on ISPOR guidelines (15, 16).

This model was built to estimate the budget impact of introducing geptanolimab as a treatment option for patients with relapsed or refractory peripheral T cell lymphoma (R/R PTCL) from the perspective of China's National Healthcare Security Administration. The target population was the annual number of new patients with R/R PTCL. The model conceptualized two distinct market scenarios: (1) a status quo scenario in which geptanolimab was not included in the National Reimbursement Drug List (NRDL), which only included chidamide for the treatment of R/R PTCL; and (2) an alternative scenario in which geptanolimab was included in the NRDL and offered as an alternative treatment to chidamide. The budget impact was the cost difference between the two market scenarios. The baseline year was 2021 and the time horizon was 3 years. The structure of the model is shown in Figure 1.

Demographic and epidemiological data were obtained from published studies, statistical yearbooks, and expert interviews (31–34).

As chidamide was the only drug listed in the NRDL for treatment of R/R PTCL, we assumed that the market share for chidamide was 100% in the scenario without geptanolimab, and in the scenario in which geptanolimab was included in the NRDL, we assumed that geptanolimab would gradually replace chidamide over time. The various market shares were based on available sales data for chidamide (35).

Based on the study perspective, in scenario 1 we only considered drug costs and condition-related costs, which included drug acquisition costs, inspection and testing costs, disease management costs, and AE management costs. As mentioned above, in scenario 2, we also considered subsequent treatment costs. All interventions and condition-related costs and reimbursement-related parameters were based on the Menet Medical Information database, published literature, China's National Healthcare Security Administration website, expert interviews, and assumptions (36–39). All costs were expressed in 2021 US dollars (USD) using the prevailing exchange rate of 1 Chinese yuan (CNY) = 0.1547 USD (40).

In addition, we assumed that the median progression-free survival rate was equal to the average treatment duration using geptanolimab and chidamide. We also used the rate of disease progression in 1 year under treatment with geptanolimab and chidamide to estimate the number of patients who were likely to experience further progression and require further treatment each year. Clinical efficacy data for geptanolimab and chidamide were obtained from published clinical trials (38, 39). Subsequent treatment regimens and the proportions of cancer patients who received each subsequent treatment regimen were obtained from clinical guidelines and expert interviews (41).

A summary of all model inputs and their sources is presented in Table 1, and detailed information is presented in Appendix 2.

The study perspective is related to what resources should be assessed in BIA. The majority of the guidelines included recommend that the perspective should be that of the healthcare payer or the budget holder (16, 19, 21–27). However, the choice of perspective might differ under some specific circumstances. For example, The Netherlands recommends a wider societal perspective when the healthcare payer is the government (24). For Canada, the guideline recommends the assessment of the budget impact related to drugs only with public purchaser perspective (20). Regarding time horizon, all the guidelines examined recommend that it depends on the duration of the budgeting period, target intervention diffusion rate and type of intervention and condition (16, 20–27). The guidelines of ISPOR and Thailand recommends 1–5 years of time horizon (16, 26), of Canada and Poland recommend 2–3 years (20, 27), of French suggests 3–5 years (21), of Australia suggests over 6 years (23), of Belgium recommends a minimum of 3 years (25) and NICE recommends 5 years (19).

The existing recommendations regarding the scope of costs are summarized in Table 2. Regarding direct costs, all 10 guidelines examined recommend that intervention costs, AE management costs, and subsequent treatment costs should be considered in BIAs (16, 19–27), and eight of the 10 guidelines recommend that administration costs and monitoring costs should also be considered (16, 19, 21, 22, 24–27). Regarding indirect costs, none of guidelines recommend to consider them routinely in BIAs, and only four guidelines recommend that they should be considered in specific circumstances whereby indirect costs significantly influence the results or are able to be reasonably estimated (16, 21, 25, 27). For example, the guideline of Belgium suggests that indirect costs should not be included in a BIA as these are not generally relevant to the budget holder, however they can be included in a BIA as a complementary analysis if they are significant (25).

A total of 1,367 articles were initially identified, of which 29 articles were included in the final analysis (42–70). Figure 2 shows a flowchart of the literature screening process.

The characteristics of the included studies are summarized in Table 3 and Appendix 3. Most of the included studies originated in the United States (n = 18, 62%) (42–46, 48, 50–59, 64, 66), followed by Italy (n = 2, 7%) (69, 70) and one from each of Brazil (60), the Netherlands (62), France (47), Japan (49), Norway (65), Saudi Arabia (61), Spain (63), and Thailand (67), as well as one multi-country study (58). The studies covered 11 types of cancer including non-small-cell lung cancer (n =8, 28%) (43, 45, 47, 54, 55, 57, 62, 65), prostate cancer (n =5, 17%) (42, 44, 48, 59, 64), colorectal cancer (n = 4, 14%) (60, 67, 68, 70), ovarian cancer (n = 3, 10%) (46, 53, 63), breast cancer (n = 2, 7%) (56, 61), myeloma (n = 2, 7%) (49, 58), melanoma (n = 1, 3%) (52), head and neck cancer (n = 1, 3%) (69), cell carcinoma of the urothelium (n = 1, 3%) (50), gastroenteropancreatic neuroendocrine tumor (n = 1, 3%) (66), and epithelial ovarian, fallopian tube or primary peritoneal cancer (n = 1, 3%) (51). Most of the interventions in these studies involved innovative anticancer drugs, including selective poly ADP-ribose polymerase (PARP)-1 and PARP-2 inhibitor (e.g., niraparib) (51, 53), epidermal growth factor receptor (EGFR) and human epidermal growth factor receptor 2 tyrosine kinases inhibitor (e.g., afatinib) (55, 62), immune checkpoint inhibitors (e.g., nivolumab, pembrolizumab) (47), and vascular endothelial growth factor inhibitor (e.g., bevacizumab) (60).

Regarding the perspective, three studies (10%) considered the societal perspective (57, 67, 69), four studies (14%) considered the healthcare system perspective (60, 62, 63, 70), and all other studies (n = 22, 76%) considered the budget-holder perspective (42–56, 58, 59, 61, 64–66, 68). Regarding the budget-holder perspective, one study calculated the budget impact from the hospital perspective (66), five adopted the health-plan perspective (45, 46, 48, 55, 64), and 16 adopted the third-party payer perspective including public health insurance, Medicare, and commercial insurance (42–44, 47–54, 56, 58, 59, 61, 65, 68). Five studies reported results from more than one perspective (43, 50, 54, 58, 68).

The focus of our study was the scope of costs included in the BIAs of the sample studies. All of the studies considered the costs of anticancer drugs, most of which were calculated based on the unit prices and treatment duration assumptions. In 12 studies (41%), the authors assumed a treatment duration on the basis of treatment effect data, including the median progression free survival, the average metastasis-free survival, the median time to treatment failure, and the time until relapse (43, 45, 48–50, 53, 55–57, 62, 63, 69). In 10 studies (34%), they assumed a treatment duration based on actual patient treatment durations including days spent receiving therapy, treatment in clinical trials, and prescribed treatment durations (44, 46, 51, 54, 58–60, 64, 68, 70). In six studies (21%), the treatment duration was based on either the drug administration instructions or clinical guidelines (42, 47, 52, 61, 65, 67). Only one study (3%) based the treatment duration on hospital data (66).

Regarding condition-related costs, 25 studies (86%) considered these, most of which included administration costs (e.g., physician visits, injections, consumption of materials, and hospital pharmacy costs), AE management costs (e.g., grade ≥3 AE), gene-testing costs (e.g., BRCA testing and EGFR testing), and hospitalization costs. Of these studies, 12 (41%) provided a clear reason why they considered condition-related costs, while the other 13 (48%) reported condition-related costs directly. For example, Stargardter et al. (45) considered condition-related costs because they adopted ISPOR task force recommendations, while Appukkuttan et al. (42) considered the costs associated with prostate cancer in the United States without providing any reference framework. In addition, although there were 25 studies considering condition-related costs, only 11 (38%) considered subsequent treatment costs. Of those, in five studies (14%), the authors took subsequent treatment costs based on guideline recommendations into account, while in the other six studies (24%), the authors took the subsequent treatment costs based on their own model structure and relative considerations into account. For example, Kongnakorn et al. (50) calculated the costs of post-progression (on subsequent 3 line active treatment) and post-progression/off-treatment (best supportive care) as subsequent treatment costs, while Stellato et al. (52) developed a Markov model and considered the costs of subsequent recurrence events in patients. Bly et al. (54) and Mistry et al. (56) estimated the costs of patients receiving subsequent lines of therapy after the first-line therapy using their target drugs. Graham et al. (55) assumed that progressive disease costs comprised monthly continuing-care costs that were applied each month between progression and the final year of life, and end-of-life costs that were applied each month during the final year of life. Delgado-Ortega et al. (63) considered all treatment lines in their model and assumed that patients received maintenance treatment until disease progression, and then received chemotherapy.

Four studies (14%) did not consider condition-related costs. Monirul et al. (47) and Goldstein et al. (57) assumed that all costs other than for drugs were equivalent under the two strategies, while Silva et al. (60) and Ortendahl et al. (66) did not mention costs other than those for drugs in their studies.

Two studies (7%) considered indirect costs, which included loss of productivity or absence from work of either the patient or a caregiver (69, 70). The main reason why they considered indirect costs was the societal perspective they adopted in their studies.

Based on demographic and epidemiological data, it was estimated that there would be 7,885, 9,050, and 10,017 newly diagnosed patients with R/R PTCL in China in 2022, 2023, and 2024, respectively. These patients would receive treatment with either geptanolimab or chidamide. The resulting market shares are shown in Table 4. Based on the target population, market shares, and disease progression data, it was estimated that in the scenario without geptanolimab in the NRDL, the number of chidamide patients receiving subsequent treatment was 7,048, 8,089, and 8,953 in the 3 years, and in the scenario with geptanolimab in the NRDL, the number of geptanolimab patients receiving subsequent treatment was 1,886, 2,384, and 2,831 in the 3 years and the number of chidamide patients receiving subsequent treatment was 5,015, 5,519, and 5,902 in the 3 years.

When not considering subsequent treatment costs (scenario 1), comparing the two market scenarios (without/with geptanolimab in the NRDL), the model estimated that the total annual reimbursement budget would increase by $1,458,842, $1,844,493, and $2,190,023 in 2022, 2023, and 2024, respectively. These increases were mainly driven by the longer period for which patients receiving geptanolimab were progression-free compared with that for patients receiving chidamide.

When considering subsequent treatment costs (scenario 2), there was a shift from cost increases to savings, with the model estimating that the total annual reimbursement budget would decrease by $38,087,822, $64,021,668, and $84,387,363 in 2022, 2023, and 2024, respectively. These decreases were mainly driven by (1) the higher annual disease progression rate in chidamide patients compared with that in geptanolimab patients, and (2) the higher average annual costs of subsequent treatment for chidamide patients compared with that for geptanolimab patients.

The differences in the budget impact of the two cost-scope scenarios were significant, at $39,546,664, $65,866,161, and $86,577,386 in 2022, 2023 and 2024, respectively. The results of the comparative study are shown in Figure 3 and Appendix 4.

The results of our systematic review revealed that although BIA guidelines from the ISPOR, the NICE, and various countries indicate that costs related to decision-making should be included in BIAs (15, 16, 19–27), most BIAs for anticancer drugs did not include a rational scope of costs. In particular, they tended to ignore the costs of subsequent treatment. This finding is similar to that of a previous systematic review by Han et al. (17), although they only reviewed BIAs for drugs used in the treatment of lung cancer and did not focus on the scope of costs. Nonetheless, more than half of the studies they examined did not consider subsequent treatment costs. This incomplete consideration of the scope of costs may generate inaccurate and unrealistic BIAs. For example, Appukkuttan et al. (42) aimed to compare the budget impact of the use of darolutamide for the treatment of non-metastatic castration-resistant prostate cancer. Although they emphasized that patients were assumed to continue to be treated with the target drugs until disease progression, they did not include subsequent treatment costs after disease progression, which they noted as a limitation of their study.

The results of our comparative study showed that for some anticancer drugs, the inclusion of subsequent treatment costs can result in significant differences in the budget impact, even replacing costs with savings in some cases. There are two main reasons for this: first, innovative anticancer drugs usually have a better treatment effect, albeit over a longer treatment duration at a higher price (72). Thus, if BIAs do not include subsequent treatment costs, the higher costs involved in longer treatment duration at a higher price using a comparator will not be captured. Second, innovative anticancer drugs generally reduce the rate of progression in patients compared with traditional anticancer drugs (75), meaning that patients receiving innovative anticancer drugs require less subsequent treatment. These savings can counteract the additional costs of innovative anticancer drugs used prior to disease progression, sometimes even reversing the impact on the budget, resulting in cost savings for the payer. The example presented in our comparative study illustrates this well. The treatment duration under geptanolimab is longer than that under chidamide, and geptanolimab is more expensive than chidamide. Thus, if we do not consider subsequent treatment, the costs for patients treated with geptanolimab will certainly be higher than those for patients treated with chidamide. However, patients treated with geptanolimab had a lower progression rate and lower average treatment costs for subsequent treatment regimens than those treated with chidamide, resulting in lower overall costs for patients treated with geptanolimab when the costs of subsequent treatment were included.

BIAs for anticancer drugs need to rationally consider the scope of costs in accordance with BIA guidelines, for instance, the ISPOR good practice guidelines (15, 16), before they are presented to the appropriate budget holder and/or published. Researchers conducting BIAs should divide the costs into three categories: the costs of the drug itself, condition-related costs, and indirect costs. The costs of the drug can be calculated based on the unit price and the estimated treatment duration. Condition-related costs include monitoring and testing costs, administration costs, management costs, and subsequent treatment costs. Our results showed that subsequent treatment costs are significant and need to be considered in BIAs for anticancer drugs. If researchers do not consider subsequent treatment costs in their base-case analysis, they should state the reason for their decision and include these costs in accompanying scenario analyses. However, because of the complexity of the cancer treatment regimen, there is no optimal approach to calculating the subsequent treatment costs for all anticancer-drug BIAs. Researchers can base their calculations on reasonable assumptions (e.g., the proportion of patients in whom the disease progresses based on clinical trials) or real-world data (e.g., the subsequent treatments received by patients). These assumptions and data need to be applied consistently to both the target drug and its comparator(s). Indirect costs should not be considered routinely because these costs are not relevant to the payers and budget holders at most of time. However, they should be considered when they can be predicted and are estimated to be significant. Because in some countries, payers are also responsible for the increased health care costs (e.g., community health care costs) due to rehabilitation and decreased hospital stay of patients.

In addition, it should be noted that although a comprehensive scope of costs is important in BIAs for anticancer drugs, not all BIAs need to consider all related costs. When the target drug and its comparator(s) display no differences in terms of treatment effects and only differ in price, such as biosimilars, researchers do not need to consider a comprehensive scope of costs. Because the consumption of medical resources during subsequent treatment regimens are similar, this will not have a significant effect on the BIA. This is why we excluded BIAs for biosimilars from our systematic review (76–78).

There is another important point that researchers need to ensure the transparency of BIAs for anticancer drugs, because some published BIAs are too simple and do not provide enough cost information, such as detailed parameters and assumptions related to costs (65). These will make decision makers confused about accuracy and reliability of BIA results, and they cannot make decisions based on these evidences. Researchers should (1) clearly describe model structure and its logical relations with costs, (2) state all assumptions of cost calculation and reasons for them, (3) demonstrate all costs parameters with their values, units and sources, and (4) illustrate the calculation methods or formulas used.

BIAs and affordability estimates for anticancer drugs, especially innovative drugs, have become increasingly important for reimbursement decision-makers in many countries in recent years. Our study is the first systematic review of BIAs for anticancer drugs and also the first to discuss the scope of costs in relation to BIAs. Among the previous studies, Jahn et al. (79) published a methodological review of BIAs for cancer screening, Abdallah et al. (80) published a methodological quality assessment of BIAs for orphan drugs, and Han et al. (17) published a review of BIAs for antitumor drugs used in the treatment of lung cancer. Although these reviews mentioned the scope of costs, they only provided brief summaries, and did not focus on this aspect. In contrast, in our study, we clearly identify the cost-scope limitations of previous BIAs, reminding researchers and decision-makers of the need to pay more attention to cost-scope issues to improve the accuracy and reliability of their BIAs. In addition, we compared the results of BIAs either considering or ignoring subsequent treatment costs. The results of this comparative study revealed the importance of the scope of costs.

Our systematic review has several limitations. First, BIAs for anticancer drugs are usually undertaken as a supplementary exercise in addition to the CEA that is submitted to budget holders, instead of being prepared independently. Therefore, the BIAs included in our systematic review might only represent a subset of all anticancer-drug BIAs, and some relevant studies may have been overlooked. Second, because the treatment regimens for various types of cancer differ and are generally complex, there is no optimal approach to calculating costs, and thus different approaches may have resulted in different outcomes in terms of budget impact. However, we did not consider this issue in our comparative study. Third, in addition to the scope of costs, there are several other factors influencing the accuracy and reliability of BIAs, such as uncertainty and scenario analyses, validation, and data sources. However, because these were not the focus of our study, we did not consider these factors. Fourth, although nowadays almost each jurisdiction has their own BIA guideline, we only included 10 guidelines as reference. Because some BIA guidelines are not available in public databases or published in other than English. This limitation restricted us from comprehensively reviewing all cost-scope recommendations.