Evaluation and forecasting methods to estimate number of patients with non-Hodgkin lymphoma: a systematic literature review.

backgroundNon-Hodgkin lymphoma (NHL) is the most common hematologic cancer in the US. Validated projections of NHL cases are important for various stakeholders. The study aimed to identify and characterize methods forecasting NHL incidence, prevalence, and number of treatment eligible patients with NHL by line of therapy (LoT). In addition, methods evaluating the performance of cancer forecasting methods were also identified and utilized in selecting the most robust projection method applicable to NHL disease setting.

methodsA comprehensive search was conducted in MEDLINE and EMBASE databases, covering January 2002 to April 2024 for English-language studies reporting methods evaluating cancer count estimation and NHL projection methods. Study characteristics were extracted and described. Criteria was developed to identify the most appropriate methods for evaluating projection methods. The identified methods of evaluation were then adopted to measure the accuracy of NHL projection methods.

resultsTwenty-nine articles met the inclusion criteria for methods of evaluation, with 58.6% evaluating projection methods through calculating relative difference between observed and predicted case numbers. The most appropriate methods found for evaluating cancer incidence and prevalence projection were the average absolute relative deviation (AARD) and percent variation (VAR%), respectively. These methods were applied to projection methods identified through literature review to determine the robust method to project incidence and prevalence. Among twenty-six articles met the inclusion criteria for NHL projection methods, the joinpoint regression model was determined as the most robust method for projecting NHL incidence in the US, with the lowest AARD (1.6%). The projection method with assumptions of a 52.8% cure rate, a cure beginning ten years post-diagnosis, and all surviving patients cured after 20 years was identified as the most robust method for projecting NHL prevalence, with the lowest VAR% (8.3%). Unfortunately, due to the limited number and quality of studies, no robust method was identified for projecting the number of treatment-eligible NHL patients by LoT.

conclusionThis review identified the most appropriate method of evaluating projection methods, and identified methods for projecting NHL incidence and prevalence in the US. Nevertheless, further research is needed to validate and project the number of treatment-eligible NHL patients by LoT.