Treatment patterns and costs among US patients with diffuse large B-cell lymphoma not treated with 2L stem cell transplantation

Diffuse large B-cell lymphoma (DLBCL) is an aggressive B-cell non-Hodgkin lymphoma and is the most common histologic subtype in the USA, accounting for roughly a third of all lymphoma diagnoses [1]. The estimated age-adjusted annual incidence was 5.6 per 100,000 men and women based on US cases from 2015 to 2019; age-adjusted death rates have remained relatively stable over the last 10 years; and the 5-year relative survival rate was approximately 64.6%, with a slightly higher rate for those with early-stage disease and who are ≤64 years of age [2]. Though the therapeutic landscape for DLBCL continues to grow rapidly, first-line (1L) therapy for DLBCL, i.e., rituximab plus cyclophosphamide, doxorubicin, vincristine and prednisone (R-CHOP) therapy or similar, has remained the most effective standard of care for nearly 2 decades. Despite the severity of the disease, approximately 60–70% of patients are cured with 1L immunochemotherapy [3,4]. However, some of the remaining patients treated with 1L R-CHOP will have primary refractory disease, and others will experience relapse after an initial response; survival outcomes for patients with relapsed or refractory (R/R) disease are poor, ranging from 6 to 17.5 months [5–7].

For those patients whose disease fails to respond to 1L R-CHOP, often the approach is to treat with platinum-based immunochemotherapy as salvage treatment, followed by high-dose chemotherapy and hematopoietic stem cell transplantation (HSCT) for chemosensitive patients [8]. However, approximately 50% of patients with R/R DLBCL are not intended for transplant [3,9]. Though the treatment landscape has rapidly evolved over the past few years, to date, there is a lack of effective second-line (2L) treatment options for patients with DLBCL who are not intended for HSCT. Clinical practice guidelines [8,10] suggest novel agents, such as chimeric antigen receptor (CAR) T-cell therapies, tafasitamab-cxix in combination with lenalidomide, or treatments undergoing clinical trial evaluation, or rituximab-based chemotherapy regimens, such as rituximab plus gemcitabine and oxaliplatin (i.e., R-GemOx) be used with palliative intent [4,11,12].

While treatment patterns and costs across lines of therapy among patients with DLBCL were previously explored [13,14], recent studies have focused on populations with advanced age [15,16] and comorbidities [17]. However, currently there is a paucity of real-world research for patients with R/R DLBCL who are not intended for HSCT, an important subpopulation of patients for whom there are limited standard treatment strategies, and management of disease remains a challenge. Therefore, to understand the context of the clinical and economic burden of this subpopulation, the objective of this study was to examine demographic factors, treatment patterns and an estimation of the healthcare resource utilization (HCRU) and total healthcare costs among patients in the USA with DLBCL whose 1L treatment has failed and who did not receive HSCT as 2L treatment.

Patients & methods

Study design & data sources

We conducted an observational study that used US administrative claims data from 1 January 2009, to 30 September 2020. The IBM^® MarketScan^® database is a large US commercial health claims database containing enrollment and demographic information, pharmacy and medical (inpatient and outpatient) claims of employees and their dependents, and Medicare supplemental plans. The database includes claims from all census regions of the USA. MarketScan data are de-identified and comply with the Health Insurance Portability and Accountability Act. Due to the use of de-identified patient administrative data, this study was exempt from institutional review board oversight.

Patient selection

Patients with a diagnosis code for DLBCL (International Classification of Diseases [ICD], Ninth Revision, codes: 200.0×, 200.5× and 200.7×, 200.8×, 202.8×, 202.9×; ICD, Tenth Revision, codes after 1 October 2015: C83.30–C83.39) on ≥1 inpatient claim or ≥2 outpatient claims were identified. The index date was defined as the start of 2L systemic treatment between 1 January 2010, and 31 March 2020. Patients were included if they were ≥18 years of age at the index date, received ≥1L treatment for DLBCL, and had ≥12 months of continuous enrollment before the index date and ≥6 months of continuous enrollment after the index date (Figure 1). Furthermore, patient pharmacy claims must have included evidence of 1L systemic treatment with an anthracycline and rituximab or another CD20-targeted agent before index and evidence of initiation of 2L systemic therapy. Patients were excluded if they had a claim of any other primary malignant neoplasm within 1 year before the index date, received a HSCT at any time during the study period, or participated in a clinical trial during the study period.

Treatment patterns

Treatment lines were characterized by identification of eligible therapies for DLBCL based on a combination of National Comprehensive Cancer Network guidelines, input from clinicians regarding clinical practice and previous research studies [10,18]. Specific anticancer systemic agents included bendamustine, brentuximab, carboplatin, cisplatin, cyclophosphamide, cytarabine, etoposide, fludarabine, gemcitabine, ibrutinib, ifosfamide, lenalidomide, mitoxantrone, oxaliplatin, polatuzumab (pola), procarbazine, rituximab, vincristine, the anthracyclines daunorubicin, doxorubicin, epirubicin, idarubicin, mitoxantrone and CD20-targeting agents obinutuzumab, ibritumomab, rituximab, tositumomab, ofatumumab, veltuzumab, ublituximab and ocrelizumab. These agents were extracted by the corresponding codes derived from the Healthcare Common Procedure Coding System and/or National Drug Code (NDC). Front-line therapy was defined as the first episode of an eligible therapy that was given after or up to 14 days before the first claim for DLBCL. The definition of a line of therapy is generally the first eligible drug episode plus other eligible drugs given within 28 days. The line was advanced if a patient switched to a different treatment regimen (the addition/substitution of new agents [except dexamethasone and prednisone] within 28 days after starting initial eligible therapy), or if the same regimen had a treatment gap of >120 days (except for rituximab, where a gap of 270 days was required to advance the line). Rituximab, obinutuzumab, ofatumumab, and lenalidomide were considered maintenance therapy when one of the drugs were used in combination with other drug(s) in a previous regimen, and then subsequently used as monotherapy started within 180 days after the previous regimen end date. For oral prescriptions billed by NDCs, we determined the duration of use based on the days' supply value recorded on the outpatient prescription claims. The last pharmacy claim end date for each line was considered as the line end date. Treatment patterns were described for the overall population. The descriptive summary of patients' treatment patterns included number and percent of patients who received lines of systemic therapy (1L, 2L, third line [3L] and fourth line [4L] or beyond), overall treatment sequence by agent(s), duration of therapy (defined as time between the start date and end date) for 1L, 2L, 3L, time from 1L to 2L (defined as index [2L] start date -1L end date +1) and treatment sequence by line.

Healthcare resource utilization & costs

HCRU was reported for all patients who received inpatient (hospitalizations), outpatient, emergency and pharmacy services; costs were reported for all patients with complete cost information. HCRU and costs were assessed in total, as well as by DLBCL-related and non–DLBCL-related visits and cost; pharmacy claims were assessed in total. HCRU and costs were assessed over the entire post-index follow-up period on a per-patient-per-year (PPPY) and per-patient-per-month (PPPM) basis. Metrics assessed were the number of visits, length of stay and duration of days per visit for inpatient hospitalizations, as well as number of pharmacy claims.

Statistical analysis

Descriptive statistics were generated to assess differences in characteristics among the study population. Continuous variables were summarized using means and standard deviations (SD), and medians and ranges; categorical variables were summarized using frequencies and percentages. Missing data were not imputed and were reported as unknown or not documented. Overall costs (medical and pharmacy costs) were reported and defined as costs divided by the entire number of days in the study follow-up period independent of HCRU. All costs were inflation adjusted to 2020 cost levels (US dollars) using the Consumer Price Index medical care component.

Data sharing

Bristol Myers Squibb policy on data sharing may be found online at www.bms.com/researchers-and-partners/independent-research/data-sharing-request-process.html.

Results

Patient characteristics

Of the (N = 182,441) patients with DLBCL identified with ≥1 inpatient claim or ≥2 outpatient claims, a total of 750 patients were included in the final analytic sample after eligibility criteria were applied (Figure 2). The average length of follow-up was 26.7 months (SD: 21.7) and the study population was (52.3%) male, with a median age of 62.0 years (range: 19.0–91.0); 68.5% were <70 years of age. The average Charlson Comorbidity Index (CCI) score was 4.1 (SD: 2.8) and 38.9% had a CCI score ≥5 (Table 1). Timing by year of index date (start of 2L treatment) can be found in Supplementary Figure 1.

Treatment patterns

In this patient population, where 2L therapy was an inclusion criterion, most patients (71.7%) did not have evidence of continued treatment beyond 2L therapy. Figure 3 illustrates the treatment flow across increasing lines of therapy for the most common regimens. In 1L treatment, 75.2% of patients (564/750) received R-CHOP and 11.6% (87/750) received R-CHOP plus another agent. The median duration of 1L therapy was 4.1 months (range: 0.03–47.2), and patients initiated 2L therapy approximately 9 months after 1L treatment (median: 9.4 months, range: 0.3–109.2; Supplementary Tables 1 & 2). The most common 2L treatments were rituximab monotherapy (39.5% of patients [296/750]) and bendamustine ± rituximab ± other (16.3% of patients [122/750]; Supplementary Table 2). The median duration of 2L treatment was 4.1 months (range: 0.03–74.9). The most common 3L and 4L treatments were also rituximab (54.2 [115/212] and 62.5% [40/64] of patients, respectively; Supplementary Tables 3 & 4).

Healthcare resource utilization

The average number of total medical visits was 68.1 (SD: 55.2) PPPY (Figure 4). Approximately half (50.4%) of patients had at least one inpatient visit, and the average length of stay was 16.5 days (SD: 25.8) PPPY (1.4 days [SD: 2.2] PPPM). All patients had at least one outpatient visit, though the average number of visits was 66.0 (SD: 53.5) PPPY (5.5 days [SD: 4.5] PPPM) and the average number of pharmacy claims was 26.2 (SD: 18.6) PPPY (2.2 days [SD: 1.5] PPPM).

About 89% of the study population (PPPY) had a DLBCL-related outpatient medical visit, averaging 28.7 visits (SD: 39.2) PPPY (2.4 visits [SD: 3.3] PPPM; Supplementary Figure 2). The prevalence of inpatient hospitalizations related to DLBCL was 36.3% PPPY (versus 26.8% for non–DLBCL-related inpatient hospitalizations), with an average of two admissions (SD: 2.2) PPPY (0.2 admissions [SD: 0.2] PPPM) and an average length of stay of 17.5 days (SD: 27.2; 1.5 days [SD: 2.3] PPPM; Supplementary Figure 3). The prevalence of emergency department visits and the average number of visits are summarized in Supplementary Figure 4.

Overall costs

Over the entire follow-up period, patients had an average total inpatient hospitalization cost of US$80,197 (SD: $157,268) PPPY ($6678 [SD: $13,096] PPPM), outpatient costs were $85,299 (SD: $101,611) PPPY ($7103 [SD: $8462] PPPM) and emergency department costs were $2898 (SD: $7974) PPPY ($241 [SD: $664] PPPM; Table 2). Whereas inpatient hospitalization costs for DLBCL-related admissions were $91,456 (SD: $174,923) PPPY ($7616 [SD: $14,567] PPPM), outpatient costs were nearly half that at $48,257 (SD: $77,088) PPPY ($4019 [SD: $6419] PPPM), and emergency department costs were $4817 (SD: $12,598) PPPY ($401 [SD: $1049] PPPM). Overall, the DLBCL study population had an average medical and pharmacy cost of $141,532 (SD: $189,579) PPPY ($11,786 [SD: $15,787] PPPM) during the follow-up period (Supplementary Table 5).

Discussion

This observational administrative claims study assessed treatment patterns, HCRU and healthcare costs among patients with DLBCL whose 1L treatment failed and who did not receive HSCT for the duration of the study period. The median age of our study population was 62.0 years, and nearly 40% of the study population had a CCI score ≥5. The prevalence of hospitalizations was higher for DLBCL-related claims, and on average, patients were admitted two-times a year for 17.5 days per admittance. The mean cost of hospitalizations and emergency department visits for DLBCL-related claims was approximately three-times that of non–DLBCL-related claims. In terms of treatment patterns, R-CHOP remains a frontline mainstay for patients with newly diagnosed DLBCL, with very few exceptions. However, among this patient population with no evidence of HSCT starting in 2L, therapeutic journeys become more heterogenous, and the quantity of regimens increase.

We found that after 2L therapy, 28% of patients moved forward with 3L therapy, whereas most patients no longer had evidence of a treatment claim. Across all lines, moving from 2L to 3L displayed the largest percentage of patients with no evidence of further treatment during the study period. Generally speaking (with change of insurance aside), this indicates different paths: either death due to disease, refusal of further treatment, or, alternatively, no progression during the prior line of therapy and can be considered in remission. About 30% of those who started 3L therapy moved forward to receive 4L+ therapy. This observed trend in sample size attritions with each line of therapy in our study was similar to a previous real-world study using US claims data. The study found that among those who received 2L therapy, 23% received 3L and 21% received 4L therapy, although their study population differed in that it included patients who received HSCT [19].

The Sankey plots displayed in Figure 3 demonstrate several points regarding the treatment journey for this patient population. For one, in accordance with standard of care, R-CHOP remains the most common treatment regimen for patients with newly diagnosed DLBCL, with very few exceptions. However, starting in 2L, the quantity of different treatment regimens attempted, as well as the prevalence of recycling regimens in 3L and beyond, increased in this patient population. This suggests an absence of a clear standard of care, as well as a high unmet medical need for this patient population [20].

The median duration of therapy for 1L, 2L and 3L was 2 to 5 months for each line, and R-CHOP was the most common 1L therapy regimen. However, beyond 1L, rituximab followed by rituximab in combination with bendamustine were the most common components of 2L and 3L therapy, with ≥50% of patients receiving those treatments. These data were consistent with previous real-world studies across different databases [21–23]. Furthermore, this is in alignment with suggested treatment options for this patient population, as this regimen is typically a therapy given with palliative intent for patients not intended for HSCT (or CAR T-cell therapy) [10,23,24]. Other real-world studies including patients with R/R DLBCL had slightly different common components [25–27], and a smaller prevalence of patients being treated with bendamustine ± rituximab due to either a demographically or a clinically different patient population; patients who were intended for HSCT were retained in these studies. Recent evidence suggests an improved progression-free survival benefit for 1L treatment using pola with rituximab, cyclophosphamide, doxorubicin and prednisone (pola-R-CHP) over R-CHOP (hazard ratio 0.73 [95% CI: 0.57–0.95]) [28]. While this may influence the preferred 1L treatment selection, ORR (pola-R-CHP 85.5 vs R-CHOP 83.8%) and OS (hazard ratio 0.94 [95% CI: 0.65–1.37]; p = 0.75) results were similar. Future studies will be needed to ascertain the impact of a new 1L treatment option on 2L or later treatment patterns.

HCRU was high among patients in this study due to the length of hospitalizations and a relatively high prevalence of outpatient visits in the 2L therapy setting; the average number of total medical visits was 68.1 PPPY, and the average length of stay for inpatients was 16.5 days (SD: 25.8), driven largely by DLBCL-related hospitalization. All patients had an outpatient visit, which was also mostly DLBCL related. High HCRU was noted in previous studies of relapsed DLBCL [29] and compared with non-relapsed disease with regards to hospital admission, emergency department visits and other services [30].

For the entire study population herein, the average total medical and pharmacy costs exceeded US$140,000 PPPY (with medical costs making up ~90% of the total). One study that evaluated costs for a similar patient population found that mean costs for inpatient care increased with each line of therapy, from US$44,451 during 1L to $55,068 during 2L and $75,094 during 3L, and the mean cumulative costs for healthcare was >$300,000 during 3L treatment [19]. Another study showed that all-cause healthcare costs PPPM were also significantly higher in the relapsed versus non-relapsed group (US$6566 vs $1951, respectively; p < 0.001) [30]. R/R DLBCL has been associated with high costs driven mostly by lengthy duration of stay for the intensive care unit and other inpatient hospitalizations, as well as many outpatient office visits [22,30]. Of note, with respect to resource use and costs, we found a large amount of variation in the data, likely due to how individualized the medical journeys were for this patient population in terms of duration and treatment intensity; similar results were seen in other studies as well [21,29].

Indeed, the research base has reported that about 50% of patients with R/R DLBCL are not eligible to receive high-dose chemotherapy and subsequent stem cell rescue [5,31]. Moreover, even when some patients with chemosensitive DLBCL are eligible for transplant, about half of patients will also relapse after transplantation. More recently, CAR T-cell therapy has shown to be an effective treatment option for patients, as it is currently approved for R/R large B-cell lymphoma in the 2L+ setting; randomized clinical trials are evaluating whether CAR T-cell therapy could possibly replace HSCT [9,32]. Patients with R/R DLBCL have shown improved outcomes [33,34]; however, research suggests that there are barriers to treatment for many patients, such as severe life-threatening toxicities, individual- and institutional-level access complexities and economic considerations [13,35]. We recently published a claims study that estimated the average total healthcare expenditures for patients with DLBCL treated with CAR T-cell therapy, and costs ranged from US$380,000 to $526,000 [13]. While that analysis looked at CAR T-cell therapy across databases, patients with CAR T-cell claims in MarketScan alone (n = 60) were too sparse in the current study for meaningful analyses given the recent approval of CAR T-cell therapy; however, the costs reported in the current study remain relevant as they highlight the economic burden of this disease among those treated with chemoimmunotherapy, who may not be eligible for, or currently have access to HSCT or CAR T-cell therapy. Future real-world research with a larger sample size of patients receiving CAR T-cell therapy will be necessary for comparative effectiveness studies.

There are limitations to consider in this study; for one, pharmacy claims data do not include ICD coding, so the distinction between DLBCL-related and non–DLBCL-related costs for drugs could not be captured. Other relevant clinical characteristics, such as International Prognostic Index risk factors and genetic information, could not be captured. In addition, this analysis used commercial claims data supplemented with Medicare/Medicaid claims data, which may not provide a complete picture of patients' treatment history, and there may be inaccuracies in categorization of treatment patterns and lines of therapy. The data sources could not provide intentions for HSCT and thus, the results should not be generalized into a single patient population. Further, information on the patients' race and socioeconomic statuses, which may be associated with a patient's diagnosis, treatment and costs, were not available in the IBM MarketScan claims database. It is likely that the disease-related costs associated with this disease are an underestimation, as we have not captured indirect costs, such as productivity loss and other out-of-pocket expenses. Lastly, previous DLBCL studies have shown that many patients do not receive guideline-concordant treatment during their treatment journey [4,27,36]. Our criterion was more restrictive, requiring treatment with an anthracycline and rituximab or other CD20-targeted agent, and the attrition table suggests that most patients with DLBCL in the database did not receive R-CHOP or a similar type of treatment in 1L. Though there are limitations to this approach, we did this to best capture patients who had 1L R-CHOP failure or similar using claims data. There may be limited selection bias and misclassification, which is inherent to developing lines of therapy in claims data. The direction of the bias is hard to determine; however, the restrictive criterion suggests that our estimates may be an underestimation.

Conclusion

The results in this study highlight that treatment for patients with R/R DLBCL is resource-intensive and costly. The data highlight a wide variation in treatment patterns, a lengthy duration of hospitalizations, and a high amount of outpatient visits. Despite recent treatment advancements for such patients, there is still a need for more effective therapy options to reduce clinical and economic burden in this population.