Validation and Forecasting Methods to Estimate Number of Patients with Non-Hodgkin Lymphoma: A Systematic Literature Review

Introduction: Non-Hodgkin lymphoma (NHL) is the most common hematologic cancer type in the US, with an estimated 80,620 new cases in 2024. Treatment options for NHL are changing rapidly with major variations by subtype and line of therapy (LoT), leading to improved survival rates in recent years. Scientifically robust data on projections of NHL cases are important for various stakeholders such as policy makers and drug developers to identify unmet need, plan and facilitate advancements in NHL treatment. However, there are variable projection methods that are being used, and the reliability and validity of these methods lack clarity. In this review, we first identified methods that are used in validating the performance of cancer count estimation strategies. Additionally, we identified and characterized methods estimating the incidence, prevalence, and number of treatment-eligible patients with NHL by LoT.

Methods: By following the guide of Cochrane Methodology Protocol and Review, the Study design, Data, Methods, Outcomes (SDMO) framework was used to develop the search strategy and structure the reporting of the eligibility criteria for this methodology review. A comprehensive search was conducted in the MEDLINE and EMBASE databases, covering January 2002 to April 2024. Eligible studies are those published in English that reported 1) any standard validating methods that were used to estimate or forecast counts of any cancer, and 2) any projection method estimating or forecasting the incidence, prevalence, or number of treatment-eligible patients with NHL using population-based data. References from relevant studies were additionally reviewed to identify potential studies fulfilling the eligibility criteria. Study characteristics were extracted and described. Organizational frameworks were developed to classify standard and projection methods respectively according to statistical analysis used.

Results: Out of 1,078 references, 21 met the inclusion criteria for standard method of validation: 19 articles (90.5%) validated projected incidence or prevalence while the rest validated projected number of treatment-eligible cancer patients. Of the 19 articles, twelve articles focused on total or multi-cancer sites including NHL, five articles focused on multi-cancer sites without NHL, one focused on prostate cancer, and another one focused on thyroid cancer. The most common standard method for validating incidence or prevalence was calculating the relative difference between observed data and projected values (n=12). For articles focused on treatment-eligible patients, comparisons with earlier studies and real-world sources (e.g., Kantar database) were used as the standard validation method.

Out of 167 references that were screened for projection method, 26 articles met the inclusion criteria for methods of NHL counts projection. Among these, 23 articles (88.5%) focused exclusively on incidence estimation, one article (3.8%) estimated prevalence alone, and two articles (7.7%) provided estimations for both incidence, prevalence, and number of treatment eligible patients. Among 25 studies estimating NHL incidence, nine distinct methods were identified. Eight studies used more than one method to project NHL incidence. The most common method was age-period-cohort (APC) model (n=12), followed by joinpoint regression model (n=8), constant incidence rate assumption method (n=5), generalized linear regression model (n=4), and other five methods (n=4). All three studies estimating NHL prevalence calculated prevalence within limited time intervals by incorporating specific annual incidence rates and annual survival rates. Among two studies estimating the number of treatment-eligible NHL patients, only one evaluated counts by LoT. No studies have reported estimated projected counts by treatment regimen.

Conclusions: In this systematic literature review, we documented that there is currently no standard method for evaluating projection performance specifically in NHL settings. Furthermore, forecasting methods to estimate NHL prevalence and treatment-eligible NHL patients by LoT are limited due to the lack of available treatment-related data. Given varying evaluation methods and estimation methods for NHL projection can impact projection performance capability, an important next step is to compare these methods to better inform more robust NHL projections.

Lin:Merck & Co, Inc.: Current Employment. Shao:Merck & Co, Inc.: Current Employment, Current holder of stock options in a privately-held company. Rogers:Merck & Co, Inc.: Current Employment, Current holder of stock options in a privately-held company. Burcu:Merck & Co, Inc.: Current Employment, Current holder of stock options in a privately-held company. Sineshaw:Merck & Co, Inc.: Current Employment, Current holder of stock options in a privately-held company.