Cost-Effectiveness of Subtype-Based Treatment Strategies for Diffuse Large B-Cell Lymphoma Patients (DLBCL)

BACKGROUND Patients (pts) with DLBCL may have widely divergent outcomes despite harboring histologically similar tumors. Gene expression profiling (GEP) and immunohistochemistry (IHC) algorithms can assign pts to the germinal center B-cell-like (GCB) or activated B-cell-like (ABC) subtypes, with the latter carrying a less favorable prognosis. While IHC is widely available, GEP is more accurate in characterizing the exact subtype, however it is largely limited to academic institutions due to cost and feasibility. The addition of novel agents such as lenalidomide to the standard regimen of rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone (RCHOP) demonstrates the potential to improve outcomes for ABC DLBCL compared to historical controls (Nowakowski et al. J Clin Oncol. 2015).  We examined the cost-effectiveness of using subtype-specific treatment strategies compared to RCHOP with or without a novel agent.

METHODS We developed a Markov model to compare cost and effectiveness of 3 treatment strategies for pts 18-65 years of age with newly diagnosed DLBCL: (1) administering RCHOP to all pts, (2) administering lenalidomide+RCHOP (R2CHOP) to all pts or (3) performing subtype testing and administering RCHOP to pts with GCB and R2CHOP to pts with ABC. We calculated the costs and effectiveness of each strategy based on a clinical scenario with data derived from Nowakowski et al. 2015. The model utilized GCB/non-GCB-specific overall survival (OS) and progression-free survival (PFS) data for historical controls treated with RCHOP (strategy 1), and data from pts treated with R2CHOP from the phase 2 study (strategy 2). Strategy 3 utilized RCHOP outcomes for GCB pts and R2CHOP outcomes for non-GCB pts.  Next, we conducted an exploratory analysis comparing these strategies in a hypothetical scenario. We used composite PFS and OS survival curves from our previous systematic review (Read et al. CLML 2014) for GEP-defined GCB and ABC pts, respectively, with RCHOP (strategy 1), and assumed various hazard ratios (HRs) representing hypothetical improvements in PFS and OS by adding a novel agent to RCHOP for ABC pts and no benefit (HR=1) for GCB pts (strategy 2 and 3). In addition to the gold standard GEP test, we considered practical IHC testing with potential subtype misclassification. Health outcomes were measured in life years (LYs) and quality-adjusted life years (QALYs). Drug and administration costs were based on average wholesale price and 2015 Medicare physician fee schedule. Model robustness was addressed in probabilistic sensitivity analyses (PSAs).

RESULTS. In clinical scenario analysis, Strategy 1 provided 5.2 QALYs (6.7 LYs) at a cost of $69,920; strategy 2 improved health outcomes by providing 6.5 QALYs (8.5 LYs) at a cost of $143,753; and the subtype-based strategy 3 provided the greatest benefit of 6.9 QALYs (8.8 LYs) at a cost of $97,200. The incremental cost-effectiveness ratio (ICER) for strategy 3 was $16,881/QALY ($13,340/LY) compared with standard RCHOP, and strategy 2 was dominated since subtype-based treatment was more effective and less costly. PSA demonstrated 99.7% probability that subtype-based treatment was the most cost-effective strategy, at a willingness-to-pay value of $50,000/QALY. In the hypothetical scenario, with HR of PFS and OS for ABC pts with novel treatment varying from 0.5 to 0.9, strategy 3 produced ICERs ranging from $16,971/QALY ($13,992/LY) to $123,509/QALY ($108,000/LY) when compared with RCHOP. Similar findings were observed for models involving IHC subtype testing (see Table).

CONCLUSIONS. There are multiple subtype specific diagnostic and treatment strategies emerging for pts with DLBCL. The gains in QALYs and LYs, in addition to favorable ICERs suggest that subtype-specific treatment strategies for DLCBL can provide meaningful clinical benefit and value and should be explored in future trials.

Table. Incremental cost-effectiveness of subtype-based treatment

	GEP test (sensitivity =specificity=1.00)				IHC Hans algorithm (sensitivity=0.82, specificity=0.90)
PFS/OS HR for ABC	Incremental LY	Incremental QALY	ICER($/QALY)		Incremental LY	Incremental QALY	ICER($/QALY)
Nowakowski et al. 2015  HR(PFS)=0.35, HR(OS)=0.24	2.42	2.0	16,971		2.20	1.81	20,526
HR=0.5 (OS/PFS)	1.74	1.49	20,255		1.59	1.35	25,047
HR=0.7 (OS/PFS)	0.97	0.83	36,475		0.88	0.76	45,015
HR=0.9 (OS/PFS)	0.28	0.25	123,509		0.26	0.23	148,391