Balancing the Scales: Optimizing Reporting of Infection Rates in Clinical Trials of Bispecific Antibodies in Multiple Myeloma

Abstract

Infections are a major complication of multiple myeloma (MM) and are a significant contributor to early mortality in patients with MM [1]. While the introduction of new treatments over the past several decades has led to improved survival of patients with MM, it is also associated with increased risk of infections [1]. This is particularly true as it pertains to the novel and highly efficacious B-cell maturation antigen (BCMA)-directed bispecific antibodies (BsAbs) [2]. Putative causes for the heightened risk of infections with BCMA-directed BsAbs are: (1) most patients treated with these agents have at-least triple refractory disease [3-5] and prior MM treatments further compromised their immune system against the setting of an existing background of MM-associated immunosuppression [1], and (2) the inherent immunosuppression associated with targeting plasma cells and potential for T cell dysfunction with ongoing BsAb therapy [2, 6]. To better understand the etiology and prevalence of infections in patients with MM treated with BCMA-directed BsAbs, and in order to provide appropriate treatment and prophylaxis to these patients, a comprehensive and consistent reporting of infections is critical [7-9].

Conventionally, clinical trials report the incidence of an adverse event (AE). Incidence is defined as the “proportion of people experiencing an event” [10] while receiving treatment or within a short time after discontinuing treatment. One limitation of the incidence metric is its inability to reflect multiple events per patient. An AE is counted only once for each patient, at first occurrence. This aspect of the incidence metric greatly compromises its ability to serve as a measure of AE burden, which in our case is the overall burden of infections associated with each of the BCMA-directed BsAbs treatments for MM. To illustrate this point, consider the following: 20% of patients each having one infection episode will translate to an incidence of 20%, whereas 10% of patients each having two infection episodes will translate to an incidence of 10%, although the total burden of infection in the two scenarios is the same. A concern regarding this limitation of the incidence metric was expressed in the CONSORT Harms 2022 statement [10].

A second limitation of the incidence metric is that the incidence of an AE will generally increase with longer study follow-up. Indeed, when reviewing monotherapy trials that tested the clinical benefit of various BCMA-directed BsAbs as treatment for MM [2, 6, 11], the incidence of all grade, and of grade ≥ 3, infections which ranged between 15% and 55%, tended to increase with increased duration of follow up, which ranged from 1.7 to 22.8 months. This correlation between duration of follow-up and incidence of infection makes the interpretation and comparison of these data between trials challenging.

To correct for differences in duration of follow up across studies, a recent research letter proposed the alternative metric of “infection rate per month per 100 patients” [11]. However, rather than removing bias due to differential study duration, this metric may introduce additional bias. Specifically, the metric divides the number of patients with infection by the total number of patients in the study; multiplies this number by 100; and then divides this latter number by median duration of follow-up (in months; in other words, this metric divides incidence by median follow-up). As such, although only the first infection (of the grades of interest, e.g., of any grade or of grade 3+) counts towards the numerator, follow-up time for patients that already had their first infection is still counted towards the denominator, even though these patients are no longer “at risk” for an event (i.e., for first infection) (see Figure 1a, patients 1 and 3). Further, study follow-up time could include time after patients have discontinued treatment and AE assessment has ceased (usually sometime after discontinuation) (see Figure 1a, patient 2). Taken together, with increasing follow-up, the denominator is continuously inflated with patient time for patients that are not at risk for the event, leading to an apparent decrease in infection rate. To illustrate the “behaviors” of the two metrics above, we collected longitudinal results reported by three BsAbs trials: the MajesTEC-1 study (teclistamab) [3, 12, 13], the MagnetisMM-3 study (elranatamab) [4, 14, 15], and the LINKER-MM1 study (linvoseltamab) [5, 16, 17]. With increasing follow-up time, infection incidence generally increased whereas the number of patients with infection per month per 100 patients decreased (Table 1).

We suggest that incidence rate would be a more appropriate metric to characterize infection risk while accounting for differences in the duration of time for which each individual is at risk for infection. A foundational measure of occurrence in epidemiology, incidence rate is calculated as the incidence of an event of interest divided by the total person-time at risk [19]. When calculating the incidence rate of infection, observation time should end (i.e., be censored) after a patient had the first infection, or completed the AE assessment period, whichever comes first (Figure 1b). A more comprehensive measure, in line with the recommendations of the CONSORT Harms 2022 statement [10], would count all infection events experienced by individual patients (Figure 1c), whereby follow-up should end at completion of the AE assessment period. Incidence rates calculated in this way can also address a related but separate question, which is to understand the time course of infection. For this purpose, period-specific incidence rates can be reported, considering the population at risk at the beginning of each such period.

As incidence rates are rarely included in current clinical trial publications, comparisons of infections across trials may be made based on incidences reported at similar follow-up times. For example, cumulative risks of all infections and grade 3–4 infections around 14 months of follow-up across MajesTEC-1 (76% and 45%) [3], MagnetisMM-3 (70% and 40%) [4], and LINKER-MM1 (74% and 36%) [5], respectively, appear largely similar (Table 1). In addition, studies may also be able to compare incidence during specific time periods (e.g., 0 to < 3, 3 to < 6, and 6 to < 9 months), such as those reported in MajesTEC-1 and LINKER-MM1 [5, 13], which provides additional insight into the time course of infection. Of note, this approach still does not accurately account for multiple infections per patient or time at risk for infection. Moreover, general limitations of cross-trial comparisons remain, including different study populations and patient characteristics. These include differences in study eligibility criteria, geographic representation and associated healthcare practices, and the effects of the COVID-19 pandemic. Further, it is important to report clinically meaningful safety outcomes. For example, reporting of serious infections and infections leading to hospitalization, treatment delay, or treatment discontinuation helps to assess treatment-related impact on patients and healthcare systems; additional details on duration of symptoms, duration of study therapy, and time to pathogen clearance may provide insight into the chronicity of the infections.

The proper account of at-risk time using incidence rate (all infection episodes) may be a simple change in infection-reporting, contributing to the efforts to optimize the ability of researchers to interpret, synthesize, and compare infection risks across time. This simple but significant change in infection-reporting may enable better infection-risk mitigation and improved benefit–risk ratios for patients with MM treated with BCMA-directed BsAbs. Although the proposed data presentation will require some new programming for study sponsors, the main challenge will be standardizing reporting across study sponsors. The goal of standardized reporting requires harmonization of methods to calculate rates of infections and broad acceptance by all stakeholders. First, a standardized definition of “at-risk for infection” period after the last dose of treatment (study or comparator) can be established through regulatory guidance or a cross-industry proposal for harmonization; second, the same harmonization effort could be adopted to require reporting of incidence rates that account for all infection episodes; and finally, a consistent methodology for graphical and tabular representation could be developed to allow for naïve inter-trial comparison. Whether all stakeholders would agree to participate is unknown.

The authors have nothing to report.

The authors have nothing to report.

J.R. reports consulting or participation in an advisory role for Takeda, Adaptive Biotechnologies, Karyopharm Therapeutics, Antengene, Sanofi, Genentech, Pfizer, Janssen, AbbVie, Bristol Myers Squibb/Celgene, and Regeneron; speakers' bureau engagement for Celgene, Janssen, Bristol Myers Squibb, Sanofi, and Adaptive biotechnologies; and reimbursement for travel, accommodation, and expenses related to meetings for Regeneron. M.V.D. reports consulting or participation in an advisory role for Roche/Genentech, Lava Therapeutics, Janssen Oncology, and Sanofi. M.L., M.A., C.H., K.K., G.K., and T.R. are employees of and own stocks/other ownership interests in Regeneron. N.B. reports consulting or participation in an advisory role for Janssen and Sanofi; and speakers' bureau engagement for Sanofi.