Lymphoma is the most common blood cancer1 and is experiencing a therapeutic renaissance with cellular therapies, bispecific antibodies, antibody drug conjugates, and other targeted agents, resulting in a rapidly changing treatment landscape. Early-phase trials are key to this evolution and are carefully designed to assess, report, and mitigate harmful toxicities, with meticulous recording of grades, rates, timing, and causality to determine safety and dosage for new treatments. Subsequent drug development decisions are guided by such data. However, a major disconnect exists between the tight regulatory environment mandating strict toxicity definitions, grading, causality, and regulatory reporting, and the publication of toxicity findings from early-phase trials. No standardized methods exist for the publication of phase 1 trial results, with significant variation in details included in publications. Publications have seen a recent rise in the use of “minimizing-terms” in concluding statements.2 However, in lymphoma, as in other cancers, no guidance exists to define an “acceptable safety profile” based on rates or severity of toxicity seen.3 “Acceptability” thresholds for toxicity must also consider disease severity, life expectancy, and likelihood of therapy success or cure. Compounding these issues, as few as 35% of phase 1 trials reach publication.4, 5 Data are mainly disseminated by conference abstracts. Even conference presentations and posters are publicly inaccessible, thus published abstracts, despite significant limits on information permitted, are the main formal citation in justification and discussion of further development. Toxicity data are presented to stakeholders in heterogenous, and often incomplete formats, which may have significant consequences: misinterpretation of safety, duplication of research, subsequent failure of later phase trials, delayed regulatory approvals, and post-approval dose changes, surveillance and/or reporting due to excessive toxicity.6 In order to quantify variations in toxicity reporting and use of minimizing terms, we analyzed lymphoma early-phase trials published as international conference abstracts. First in human (FIH), phase 1, phase 1b/2, and phase 1/2 trials in lymphoma were identified manually from published abstract books from seven international conferences between 2022 and 2024 American Society of Hematology (ASH) annual scientific meeting, European Hematology Association (EHA) annual meeting and American Society of Clinical Oncology (ASCO) and International Conference on Malignant Lymphoma (ICML). Data were collected from published conference abstracts, with no data collected from posters, oral presentation slides, or recordings. Encore abstracts reporting identical data were excluded; those reporting a different number of patients or more advanced data cut-off were included separately. Abstracts were analyzed for reporting of toxicity outcomes: all grade toxicity, grade 3 or higher (G3+) toxicity, adverse events of special interest (AESIs), serious adverse events (SAEs), and deaths. Use of minimizing terms was recorded, defined as those that may diminish toxicity, and included “safe,” “well-tolerated/tolerable,” “manageable,” or “acceptable.” The latter is in contrast to the use of an “accepted” rule-based safety finding linked to a dose finding primary endpoint of phase I trials. Descriptive statistics were used to report proportions. The Fisher's exact test was used to calculate the difference between proportions.7, 8 Ethics approval was not required for review of published conference abstracts. Two hundred and eighty-six abstracts reporting 214 different trials were identified for analysis. (Supporting Information S1: Appendix Table S1) The majority, 232/286 (81%), enrolled patients with B-non Hodgkin lymphomas, and 259/286 (90%) enrolled patients with relapsed/refractory disease. The study design features were as follows: First in Human (FIH) 9%, phase 1 33%, phase 1B or phase 1B/2 11% respectively, and phase 1/2 36%. Industry trials comprised 69%, investigator-led 20%, and 11% not stated. We recorded the timing of the analyses relative to the trial, with 70% reporting preliminary or interim results, 18% final results, and 12% long-term follow-up. Proportions of studies reporting the major toxicity outcomes (all-grade toxicity, G3+, AESI, SAEs, and deaths) are shown in Figure 1. Only 2% of studies reported all these outcomes. Less than 50% of studies reported any of G3+ toxicity, SAEs, or AESIs. Of studies that included dose escalation (n = 154) only 103/154 (67%) reported the occurrence of any dose-limiting toxicities (DLTs); which were separated by dose level in only 21/154 (14%). Dose discontinuations were reported in 107/286 (37%), with cause specified in 49/107 (46%), and dose reductions were reported by a quarter of analyzed abstracts, 65/286 (23%), with cause specified in just 16/65 (25%). Minimizing terms were used in the conclusions of 248/286 (87%) abstracts. Of the 248 abstracts including minimizing language, only 164/248 (66%) reported all-grade toxicity and 142/248 (57%) reported G3+ (Table 1). SAEs were described in 52/248 (21%) and deaths due to AEs in only 35/248 (14%) of the abstracts. AEs leading to dose reduction were included in 47/248 (19%), with cause specified in 15/47 (32%). AEs leading to dose discontinuation were included in 95/248 (38%), with cause specified in 43/95 (45%). In studies using no minimizing terms reported rates of G3+ toxicity (Table 1), and deaths due to AEs (19%) were similar to those using minimizing terms. None reported SAEs. Studies using minimizing terms were more likely to report all-grade toxicity in the published abstract than those not using minimizing terms P = 0.02, Table 1). However, these studies that used minimizing terms actually reported higher rates of all-grade toxicity than those that did not use any minimizing terms, regardless of the threshold for all-grade toxicity we evaluated (all-grade toxicity rate of >50% P = 0.01, >70% P = 0.01 and >90% P < 0.001, respectively, all in Table 1). There was no significant difference in use of minimizing terms based on sponsor type (industry vs. investigator-sponsored, both 88%, respectively), timing of analysis (preliminary 88% vs. final/long-term 83%, P = 0.422), or specific study phase (FiH + Ph1; 88% vs. Ph1/2; 86%, P = 0.834). Our study provides a detailed analysis of toxicity reporting in published abstracts from early-phase lymphoma trials presented at major hemato-oncology conferences over the last 5 years. Only 2% of studies reported all relevant toxicity types, and most individual toxicity descriptors were reported in less than 50% of studies. Concerningly, there was a high use of minimizing terms (87% of abstracts) despite high rates of toxicity, SAEs, and even deaths reported. No concordance between the use of minimizing terminology and rates of Grade 3+ toxicity or deaths was seen, and higher rates of all grade toxicity in studies using minimizing descriptors compared to studies without minimizing terms. The deficiencies identified are consistent with analyses of later-phase trials and in other diseases, where the use of minimizing terms ranged from 15% to 86%, and were used despite very high rates and severity of toxicities.2, 9-13 Published data from lymphoma randomized controlled trials (RCTs) presented at ASH identified minimizing language in 37% of abstracts, despite rates of G3+ or serious AEs in experimental arms ranging between 59% to 90%.10 A myeloma study found no association between use of minimizing terms and G3 + AE rates in a study of RCTs.12 Our study supports these later-phase trial data; however, it uniquely analyzes early-phase lymphoma studies where safety is the most common primary endpoint. The CONSORT guidelines exist for reporting RCTs.14-16 CONSORT-DEFINE extended guidance to early phase dose-finding trials in 2023, including abstracts, specifying all important harms should be included, and that author interpretation should be consistent with results, balancing reported benefits and harms.3, 6 However, they lack details on grade, type, or rates of toxicities to be included. None of the four conferences we analyzed incorporates these guidelines into their abstract publication guidance.3, 17 Our analysis demonstrates poor adherence to these guidelines and recent recommendations for use of neutral language,18 and highlights the necessity of efforts to improve adverse event collection, analysis, and reporting. The Lancet Haematology Commission on Adverse Events has identified several priority areas that require action.19, 20 One significant limiting factor in adequate reporting is strict yet varied guidelines on limits for words, figures, or tables. Consistency across conferences and the provision of links to supplementary information may assist. However, stronger obligations from regulators for investigators to publish full manuscripts, and incentives for journals to publish phase I data even in the event of a negative study would overcome these limits. A key limitation of our study is the restriction to abstract data review. However, this was a conscious and key element of the study design; as published abstracts, not conference presentations, are the publicly available and cited versions of the data, and therefore should adequately represent the data for stakeholder decision-making. Published manuscripts were not reviewed to avoid bias in data analysis due to the low rates of full publication of phase I studies and “positive” results being favored for publication by industry. Another limitation is that we did not specifically analyze whether patient-related outcomes (PROs) were reported in abstracts. Additionally, we cannot determine whether the absence of toxicity within an abstract is due to a lack of occurrence or a lack of reporting. However, systematized phase I study reporting in abstracts would overcome this. Our data demonstrate problematic levels of heterogeneity in toxicity reporting of phase 1 lymphoma trials by major international conference abstracts. This occurred across all toxicity descriptors required by regulatory bodies, with the majority of abstracts presenting incomplete reporting with near-universal use of minimizing terms. This has direct impact on future patients quality of life, treatment adherence, and financial costs to patients, health systems, and wasteful pursuit of poorly tolerated treatments.3 There is an urgent need for minimum toxicity reporting requirements and consistency in guidelines for toxicity reporting, with greater enforcement. We would recommend including rates and clinical types for all-grade, grade 3, AEs, SAEs, and deaths due to AEs, as a minimum. Ultimately, “tolerability” is a patient-assessed measure and should not be used to define clinician-assessed “safety” as this creates potential bias. Maximum grade toxicity has varied impact on patient safety, quality of life, and drug efficacy, depending on the toxicity itself, with some high grade toxicities being easily reversible, minimally impactful, and potentially in some cases a target effect. However, minimizing language should be avoided, and terms relating to patient tolerability must be substantiated with patient-reported outcome data. Standardized abstract toxicity reporting for phase 1 trials and improved adherence to new guidelines are solutions that can be easily implemented, which allow transparency in drug and dose development. A prior analysis of this work was published only as a poster abstract at ASH 2024. This manuscript represents an updated analysis of this original work and has not been published elsewhere. Marsali Maclean: Methodology; data curation; formal analysis; writing—original draft; writing—review and editing. Niamh Waters: Data curation. Dawn Swan: Formal analysis; writing—review and editing. Arina Martynchyk: Formal analysis; writing—review and editing. Charmaine Smith: Software; data curation; project administration. Darcy Vickers: Data curation. Denise Lee: Data curation; formal analysis; writing—review and editing. Geoffrey Chong: Data curation; formal analysis; writing—review and editing. Eliza A. Hawkes: Conceptualization; methodology; data curation; formal analysis; supervision; writing—review and editing; visualization. Lee: Roche: Honoraria; Gilead: Honoraria; Beigene: Honoraria. Chong: Amgen: Research Funding; AstraZeneca: Research Funding; Bayer: Research Funding; Bristol Myers Squibb: Consultancy, Research Funding; Dizal Pharma: Research Funding; HUTCHMED: Research Funding; Incyte: Research Funding; Innate Pharma: Research Funding; Merck: Research Funding; Pfizer: Research Funding; Pharmacyclics: Research Funding; Regeneron Pharmaceuticals, Inc.: Consultancy, Research Funding; Roche: Research Funding; Takeda: Consultancy. Hawkes: Research funding (paid to institution): Roche, Bristol Myers Squibb, Merck KGaA, AstraZeneca, TG Therapeutics, and Merck. Consultant or advisory roles (*Paid to institution): Roche*, Merck Sharpe & Dohme*, Astra Zeneca*, Gilead, Antengene*, Novartis*, Regeneron, Janssen*, Specialised Therapeutics*, Sobi*, GSK*. Travel expenses: AstraZeneca, Genmab, AbbVie. This research received no funding. Open access publishing facilitated by La Trobe University, as part of the Wiley - La Trobe University agreement via the Council of Australasian University Librarians. No individual participant data were collected during this study, which used publicly available data from published abstracts of international conferences. Qualified researchers may request access to the dataset for academic purposes under a data transfer agreement by contacting the corresponding author: email protected. The data that support the findings of this study are available from the corresponding author upon reasonable request. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.
Maclean et al. (Mon,) studied this question.