Lack of precision in the reporting of sex and gender in cancer clinical trials with a PRO endpoint: an exploratory literature review.

Introduction
Since 1993, the U.S. National Institutes of Health (NIH) has mandated the inclusion of women and minorities in research [1]. The concept has been extended to consider sex as a biological variable and gender as a psychosocial variable [2]. The scientific community acknowledges sex and gender as fundamental concepts in clinical research [3], but they are still often neglected in research design and scientific reporting, eventually leading to biases and limited generalizability of results [4]. Accurate consideration of the trial participants’ sex and gender in health research requires a clear understanding of the two concepts [5], which might be reflected in the correct use of the respective linguistic terms [3].
By definition, sex is a multilevel concept [6] and is used when reporting biological factors linked to, for example, reproductive anatomy or physiology [7] and gender when reporting gender identity, psychosocial, or cultural factors [8]. Gender describes a contextualized social and structural experience and expression of identity [9] encompassing a spectrum of male and female characteristics expressed to varying degrees [10]. Although gender is often seen as the psychosocial counterpart to sex, this perspective can be misleading, as it implies a direct causality—suggesting that gender is simply a consequence of sex—when, in fact, the relationship is much more nuanced [5].
Numerous authors emphasize the importance of incorporating sex and gender into health research to enhance the sensitivity, precision, and relevance of retrieved study results and the conclusions drawn from them [9], [11], [12], [13]. This could result in more effective policies and programs, fostering social justice and cost savings for the healthcare system [13]. Therefore, Ginsburg et al. propose gender-transformative actions to eliminate sex and gender health disparities, like educational programs and training of health professionals in research and health systems [10]. Therefore, analyses of similarities and differences are crucial to disclose [14], as the heterogeneity of demographic factors can influence health outcomes. In oncology specifically, sex- and gender-related differences are particularly pronounced as incidence and prognosis [15], [16], symptom burden [17], [18], [19], [20], [21], and treatment tolerability [22] often vary substantially between women and men. For decades, female patients have faced exclusion or underrepresentation in clinical trials [23] despite sex discrimination being widely recognized in human trials since the 1990 s [24]. The disparity leads to less evidence-based medicine for women compared to men [24], underpowered subgroup analyses, and a misleading generalization of results to all cancer patients [25], [26], [27], [28], [29].
One critical treatment outcome in cancer clinical trials are patient-reported outcomes (PROs), which measure Health Related Quality of Life, including symptoms, functional status, psychological well-being, and overall Quality of Life [30], [31]. PROs enable patient-centered healthcare since the patient’s health status is reported directly from the patient, without interpretation by a clinician or anyone else [32]. Numerous studies have already demonstrated PRO variations in cancer patients based on sex or gender. For example, male cancer patients reported better physical and social functioning [17], [18], less nausea/vomiting [19], less diarrhea [20], and less fatigue [18], [21] compared to female cancer patients. The use of PROs in clinical trials [3] enables the identification of sex or gender-specific differences and similarities within the respective study population. This would be particularly of interest for already known sex and/or gender biases in symptom management, for example, in pain management, where women are frequently undertreated [33] due to misjudgment of the symptom burden.
To meet the challenges of reporting sex and gender in research across disciplines, the European Association of Science Editors (EASE) instituted a Gender Policy Committee that assembled a multidisciplinary team of scientists and journal editors to formulate guidelines, resulting in the Sex and Gender Equity in Research (SAGER) guidelines [4]. The SAGER guidelines aim to standardize the reporting of sex and gender information throughout the entire study process, commencing with the study design and extending to the analysis and interpretation of the results. The goal was to enhance the quality of reporting of sex and gender aspects in scientific publications [4]. In 2022, the SAGER implementation checklist was published to encourage use in research practice [34].
Top-tier journals are highly selective in choosing which manuscripts to publish - rigorous peer review and experienced editorial boards ensure thorough checks on manuscripts considered for publication to maintain the high quality and integrity of published work. However, Stark et al. (2023) [35] report that only a few journals recommend the use of the SAGER checklist.
This exploratory literature review aims to clarify the following questions for cancer RCTs with a PRO endpoint:

To what extent does the current reporting practice in cancer RCTs with a PRO endpoint comply with the SAGER guidelines before the availability of the SAGER Implementation Checklist (2022) [35].

Which terms are currently used for the categorization of sex and/or gender in the study population? How many trials report sex and/or gender divided in more than two categories?

Does the SAGER compliance of articles published in journals with a high impact factor (IF > 10) differ from those published in journals with a lower impact factor (IF ≤ 10)?

Does the sex ratio represented in the study reflect the sex ratio of disease incidence?

Methods

Selection of references
The literature search was conducted in PubMed, applying a time filter from 2019 until February 2022 to represent recent literature before the availability of the SAGER Implementation Checklist (2022) [35]. An example of the search strategy is included in the Supplementary Appendix. The literature review software DistillerSR [36] was used throughout the review process to provide full transparency of the 2-level study selection process (Level 1: Abstract screening, Level 2: Full-text screening) and subsequent data extraction (Level 3). At least two reviewers decided independently on the eligibility of references. Conflicts were resolved by discussion or consultation with a third reviewer if needed.
The reviewers considered the following set of inclusion and exclusion criteria shown in Table 1 to decide on the eligibility of references:

The selected cancer diagnoses are among the most common solid tumor types by incidence [23], affecting people of all sexes.
The eligible references were linked according to their trial registration number (e.g., NCT, EudraCT) as the reviewers extracted the data of interest on a trial- and not on an individual publication basis, resulting in a holistic evaluation of a trial. For matters of data extraction (Level 3), a pre-defined case report form (CRF) was used. The CRF included items to assess general trial characteristics and the SAGER guidelines checklist for evaluating studies with human participants [4]. To describe the trials evaluated, the following characteristics were collected: cancer type, involvement of trial organizations and/or industry, trial phase, treatment(s) evaluated in the trial, type of control group (i.e., active comparator vs. placebo-controlled), PRO endpoint (i.e., primary, secondary, exploratory), PRO instrument, PRO assessment mode, sample size, statistical analysis method, blinding, disease stage, and the impact factor of the journal the main publication was published in.
Furthermore, the terms and categories used to describe sex and/or gender were extracted to provide answers for review question b. The cut-off value of 10 used in exploratory question c. is based on the general view that an impact factor of 10 is considered excellent. For reference: In 2020, only 3.65% of journals had an impact factor of 10 or higher [37]. To answer review question d. the percentage of females/women in the study population was extracted and compared to the respective disease incidence [23].

Appraisal of reporting quality
To answer review questions (a) and (b) about compliance with the SAGER guidelines [4], the items of the SAGER guidelines checklist were evaluated by two authors under consideration of pre-specified reviewer instructions (see Supplementary Table 1). The SAGER score was calculated as a metric representing the total count of items receiving a “yes” rating. The maximum possible score was 17 (as no epidemiological studies were included) but varied if only one sex and/or gender was included in the respective trial. The following categorization was established to aid interpretation: 0–5 points (low compliance), 6–11 (medium compliance), 12–17 (high compliance).

Synthesis
For variables of interest, frequencies and corresponding percentages are reported. To answer review question b., a Mann-Whitney-U test was calculated to compare groups. SPSS version 29.0.0.0 was used to analyze the dataset.

Results
The PubMed search resulted in 2,450 identified references for all three cancer types of interest. Of those, 205 publications met the inclusion criteria, referring to 184 trials. See Fig. 1 for a full overview of the reference selection procedure.

Trial characteristics
Eighty-four (45.7%) of the trials examined patients with lung cancer, 82 (44.6%) with colorectal cancer, and 18 (9.8%) with bladder cancer. Industry involvement was declared in 94 (51.1%) of the included trials. Targeted therapy (26.6%) and chemotherapy (26.1%) were the most frequently evaluated treatments. The majority of trials had an active control group (76.6%) and investigated PROs as a secondary endpoint (50.5%), with the majority using the EORTC measurement system (41.8%). A comprehensive overview of the trial characteristics is available in Table 2.

SAGER checklist results
Review question a. To what extent does the current reporting practice in cancer RCTs with a PRO endpoint comply with the SAGER guidelines?
The terms sex and gender were used correctly by 39.1% of the reviewed trial publications. In the trials (N = 6) that examined only one sex or gender, this specificity was reported by 50% in the title and 83.3% in the abstract. The majority of the analyzed studies with more than one sex or gender category included did not provide a breakdown of the study population by sex and/or gender (96.6%) in the description of the study sample. In 34.2% of the studies not only included one sex or gender presented data that had been disaggregated by sex and/or gender and described differences and similarities between the sexes and/or genders. Six trials (3.3%) provided PRO data disaggregated by sex or gender. For an aggregated detailed breakdown of the SAGER checklist results, refer to Table 3. For the SAGER checklist results on an individual study level, refer to Supplementary Table 2.

Across all trials, a mean overall SAGER score of 2.5 (sd = 1.9) (median = 2; range = 0–12) was reached. For a comprehensive overview of the SAGER scores classified as low (0–5), medium (6–11), and high (12–17), please refer to Table 4.

Terminology used (concept and categories)
Review question b. Which terms are currently used for the categorization of sex and/or gender in the study population? How many trials report sex and/or gender non-binary?
The term sex was the most prevalent, used by 101 (54.9%) trials, followed by gender (39, 21.2%), and both (32, 17.4%). Twelve trials (6.5%) did not mention any of the terms. Refer to Table 5 for a comprehensive overview of the terms used to categorize the respective concepts of sex and/or gender. The numbers highlighted in bold signify compliance with the recommendations for bias-free language by the American Psychological Association [38]. The rating of unclear is assigned in cases where no sex and/or gender categories were reported in the manuscript. None of the herein-examined trials reported sex or gender with more than two categories.

Impact factor dependency of SAGER checklist compliance
Review question c. Does the SAGER compliance of articles published in journals with a high impact factor (IF >10 points) differ from those published in journals with a lower impact factor?
About a third (35.33%) of reviewed RCTs were published in journals with an impact factor above 10.
The mean SAGER score for RCTs published in a journal with an impact factor above 10 is 3.1 (sd=2.0) (median = 3; range = 0 - 12). In contrast, RCTs published in a journal with an impact factor below 10 demonstrate lower compliance with the SAGER checklist, with a mean score of 2.2 (sd = 1.8) (median = 2.0; range = 0 - 8) (see Figure 2). The observed difference was found to be statistically significant (Mann-Whitney-U-Test U = 2674.00, Z = −3.513, p <.001).

Study participants ratio related to the sex distribution in the incidence of the disease
Review question d. Does the sex ratio represented in the study relate to the sex ratio of disease incidence?
The representation of females in the examined trial populations is below the global incidence in bladder and colorectal cancer trials and above the global incidence in lung cancer trials. When considering the lower and upper ends of the range, it becomes evident that there is a considerable discrepancy between the disease incidence and the representation of females in some of the studies. See Table 6 for more details.

Discussion
This exploratory literature review demonstrates that the majority of cancer RCTs conducted before the publication of the SAGER Implementation Checklist [34] in bladder, colorectal, and lung cancer lack precision in the reporting of sex and/or gender, highlighting the need for these different concepts to be considered more rigorously in the research process of clinical trials with PRO endpoints. However, still being low, descriptive group comparisons indicate that compliance with SAGER guidelines is higher in journals with a high impact factor. The term“sex” was the most used, followed by “gender”, and both. All herein-examined trials reported sex and/or gender in two categories. Finally, the sex ratio of the study sample did not accurately reflect the sex ratio of the incidence population of the respective diseases, as the overall percentages showed either an under- or over-representation of females. The results of this review align with previous work by Hall and colleagues (2022) [39], which reported inconsistent use of terminology regarding the concepts of sex and/or gender in 76% of evaluated RCTs, leading to FDA approval of anticancer therapies. The results of this review indicate that parity in research concerning sex and gender has not yet been achieved, even though this was one of the major targets of Nature editorials in 2010 [24] and 2022 [40]. Moreover, research has shown that PROs can differ significantly across sex and/or gender groups, with women often reporting greater distress related to body image and sexual dysfunction following treatment, while men may underreport such concerns due to gendered norms around emotional expression [41], [42]. However, many clinical trials conflate sex and gender or fail to define them altogether, making it difficult to disaggregate whether observed differences stem from biological factors, sociocultural influences, or both. Our findings highlight the need for more precise and inclusive operationalization of sex and gender in trial design and reporting, particularly when evaluating PROs that are inherently shaped by gendered experiences.
The emerging evidence on the lack of consistent reporting and analysis of sex and gender in cancer clinical trials establishes a robust baseline against which future impact of the SAGER Implementation Checklist can be compared and builds the ground for future efforts, including various aspects of the identified issue (e.g., the policy review by the European Commission [43]). First, at the initial stages of study design and protocol writing (against current practice, e.g., in head and neck cancer clinical studies [44]), it is critical to decide which concept (i.e., sex and/or gender) is relevant for the respective research question to be answered in the study. For example, if the research question is targeting biological, genetic, pharmacokinetic, or hormonal aspects, sex will likely be the concept of interest. However, if psychosocial or behavioral aspects are the focus of the study, researchers may wish to consider assessing gender instead or in addition to sex. Another important aspect to consider at this initial stage of study design is to incorporate sex and/or gender analyses. This is exemplified by the European Commission’s requirement for grant recipients within the Horizon Europe program [45]. To ensure clarity in reporting sex and gender in clinical trials, it is critical to carefully decide how these concepts will be categorized to avoid assuming a simple male-female dichotomy. In several trials, one category of sex, usually male, was reported. This incomplete reporting practice reflects an assumption of dichotomy. About the reporting of findings, the SAGER guidelines place considerable emphasis on the necessity of considering sex and/or gender aspects in every section of a manuscript [4]. The American Psychological Association (APA) provides a style guide on the reporting of gender and sex in the 7th edition of their publication manual, intending to encourage and guide researchers to use language that is free of bias [38]. Moreover, Nature Portfolio journals have taken the initiative of raising the standard for the reporting of sex and gender in their published manuscripts by introducing policies and enforcing the application of the SAGER guidelines [4]. Our findings highlight the importance of sex and gender reporting in cancer clinical trials. A first practical step toward more transparent reporting is to evaluate the current status, raise awareness among researchers and clinicians, and use the results for educational purposes. Improving sex and gender reporting can facilitate the identification of differences in symptom burden, treatment response and health outcomes. These findings can guide researchers and clinicians in designing future cancer trials with improved sex and gender considerations, contributing to more equitable and precise patient care.
The herein-reported results of an exploratory literature review offer insights into the status of sex and/or gender reporting in cancer clinical trials with a PRO endpoint; however, it should be noted that there are some limitations to this study. One such limitation is that the SAGER guideline was primarily designed to standardize the reporting of sex and gender information rather than to evaluate existing research. To provide a metric for comparing the compliance of SAGER guideline use, a SAGER score was calculated and categorized for ease of interpretation. The score and the score categorization were conducted by the authors and are not based on previous literature. Multiple items in the guideline may not be applicable for all evaluated studies, as some items only refer to trials that included one sex and/or gender, this may lead to potential biases in the SAGER score. Some SAGER guideline items are unclear for evaluation of proper use, as they need to be interpreted. In particular, Item 1: "The terms sex/gender used appropriately" requires a clear operationalization and leaves in its current state a wide range of interpretations. Additionally, the journal’s impact factor cut-off at 10 points is based on a common generalization in the scientific community. Furthermore, we only included publications available on PubMed in our review and focused on trials addressing three specific cancer entities. However, PubMed coverage for colorectal cancer, and lung cancer studies was shown to be around 80% [46]. Lung, colorectal, and bladder cancers were selected based on their clinical relevance, and the availability of a substantial number of trials in recent years. These cancers also span different organ systems and affect both sexes, allowing for a more nuanced assessment of sex and gender operationalisation in clinical trial reporting. Future reviews could expand on database use and include additional cancer types to provide a more comprehensive overview to ensure broader applicability of findings. Finally, the exploratory comparison of the sex ratio observed in the included trials with that of disease incidence was not adjusted for potential confounders such as disease stage or other clinical variables, which should be considered when interpreting these results.
In conclusion, the appropriate consideration of the sex and gender in cancer clinical trials with PRO endpoints is vital for understanding and addressing the unique biological and psychosocial aspects of cancer in different populations. It enables the development of more effective, personalized treatments and promotes equity in cancer care and research.

Supplementary Information