Gender Disparities in Citations and Altmetric Attention Score in Oncology

Abstract

PURPOSE

Altmetric Attention Score (AAS) is a measure of the quantity of attention that a scholarly work receives, and evidence about gender gaps in AAS in oncology is lacking. Our objective was to analyze potential disparities in the AAS within oncology by comparing research publications authored by women first and last authors with those authored by men. Secondarily, we aimed to quantify the extent of over-/undercitation by gender.

MATERIALS AND METHODS

The initial data set was compiled from the Altmetric database through Application Programming Interface (API) using oncology-related search terms. Author gender categories were assigned on the basis of the Gender Guesser API. For example, those with first and last authors labeled woman were categorized as woman first author/woman last author (WW). Over-/undercitation was calculated using observed citations and expected citations. Analyses were completed both for the oncology literature as a whole and for prominent subspecialty peer-reviewed journals.

RESULTS

Our search yielded 652,834 articles published between January 1, 2009, and January 31, 2024. For AAS, women in the first author position had a 15.2% lower score compared with men counterparts and women in the last author position had an 8.3% lower score than men (P < .01 for both). Although the proportion of WW authors in oncology publications increased over time, the man first author/man last author combination was overcited (mean citation percentage difference [MCD] = +16.2%), whereas WW was undercited (MCD = –7.7%). There was variation in both proportion of WW papers and over-/undercitation among oncologic subspecialties.

CONCLUSION

Significant gender disparities in citation rates and AAS exist across various fields within oncology. This highlights a systemic issue where woman-authored research is undercited and receives less attention compared with man-authored work, with the potential to affect career advancement, funding opportunities, and academic recognition.

INTRODUCTION

Gender disparities in academia are a global concern, reflecting significant imbalances in the professional landscape of academic and research institutions.^1-3 Among the various indicators of these disparities, the differences in citation counts—citation disparity—are particularly troubling.^4,5 Previous studies have indicated that publications authored by men tend to garner more citations than those authored by women.^6-9 This discrepancy manifests the Matthew Effect, where men researchers' contributions are seen as more central and pivotal to their fields. In contrast, the Matilda Effect describes the relative under-recognition and undervaluation of scientific achievements when women are the principal researchers, and this phenomenon has been demonstrated in oncology specifically.¹⁰ The persistent citation disparity between men and women researchers has prompted various explanations about systemic barriers, including differences in resource allocation for open access (OA) publications, collaboration dynamics, academic rank, areas of specialization, and the topics addressed in published works.^11-13

CONTEXT

• Key Objective

• What gender disparities exist in research citations and Altmetric Attention Score (AAS) by first and last author gender in oncology and across oncology subspecialties?

• Knowledge Generated

• For AAS, manuscripts with women in the first or last author positions experienced significantly lower scores compared with their counterparts who were men. Furthermore, the proportion of women first/last author combination manuscripts increased over time, but the men first/last author combination was overcited, whereas the women first/last author combination was undercited.

• Relevance

• These results demonstrate a systemic issue within oncology where works authored by women receive less attention and are cited less than those authored by men. This can have significant impacts on career development and highlights the need for the academic community to identify measures to help narrow this gap.

Although citations are traditionally seen as a key indicator of academic productivity and often serve as a basis for research and academic promotions, the introduction of the Altmetric Attention Score (AAS) provides a more nuanced measure of research impact. This score not only assesses the volume of citations but also evaluates the quality of attention that scholarly articles receive. It includes a range of complementary metrics, such as article downloads, mentions in policies and patents, media coverage, and engagement within professional networks (eg, reference managers) and social media platforms like Twitter/X and mainstream media.¹⁴ This expansive approach offers a more comprehensive assessment of an article's influence.⁸ Despite its potential, AAS remains underused in examining gender disparities.

Previous studies have examined publication patterns within the field of oncology and found substantial gender differences, such that women were less likely to be senior authors or authors of clinical trials, and that citation counts were higher for men first and last authors. However, these studies were limited to certain fields within oncology and did not examine citation gap or AAS. Therefore, in this study, we analyzed potential disparities in the AAS specifically within the field of oncology. We examined the impact of research publications authored by women first and last authors compared with those authored by men. Furthermore, our study quantified the extent of over-/undercitation in oncology literature, with an additional focus on differences across various subspecialties.

METHODOLOGY

Sample Creation

The initial data set was compiled from the Altmetric database through Application Programming Interface (API) obtained through Drexel University focusing on scholarly published articles that include specific keywords that are relevant to oncology (Appendix, online only). Designation for funding was first measured as yes/no and then classified as pharmaceutical versus not. The search was focused on articles published between January 1, 2009, and January 31, 2024, and yielded 652,834 records. Information on data enrichment and preprocessing can be accessed in the Appendix.^15,16

Name-Based Assignment of Author Gender Categories

We established author gender categories on the basis of the first names of the first and last authors of the papers using the Gender Guesser API (Appendix). For example, those with first and last authors labeled woman were categorized as woman first author/woman last author (WW), even if the other authors' genders were undetermined. Gender Guesser API, much like NamSor, and Gender API and Genderize have shown to achieve 92.8% accuracy in prediction on first names alone.^17,18 Limitations related to assigning gender categories are discussed in the Appendix.

AAS Score

AAS was analyzed by gender of the first and last authors. We used an Ordinary Least Squares regression approach to establish a model for investigating the dynamics between the logarithmically transformed AAS (independent variable) and the volume of Dimensions citations (dependent variable; Appendix).

Subspecialty Delineation

We compiled a roster of peer-reviewed publications across various oncological subspecialties for analysis with a minimum of three journals per specialty. The selection of journals was predicated on the availability of comprehensive data and their respective impact factors. To gauge citation impact, a baseline citation score of 1 was established, reflective of the significance of receiving at least one citation within the first 2 years post-publication. In total, 64 journals were classified into 11 categories, with 451,881 total articles.

Subspecialty Metrics

Over-/undercitation was calculated as a percent difference of observed citation from the gender-blind expectations model (Appendix).¹⁹ The degree to which papers authored by man first author/man last author (MM) pairs are over- or undercited can be quantified using the following formula: mean citation percentage difference (MCPD) = (oMM – eMM)/eMM × 100, where oMM represents the actual observed number of citations received by MM papers from the citing articles and eMM is the expected number of citations for MM papers, as forecasted by the gender-neutral model. Details regarding all these parameters and accompanying tests of significance can be found in documentation of the mgcv package in R.²⁰

RESULTS

A total of 652,843 papers were retrieved and the distribution of gender categories is shown in Table 1 and Appendix Figure A1.

TABLE 1.

Number of Publications by Gender Combination per Year (by 1,000)

Year	Gender Combination
	MM	MU	MW	UM	UU	UW	WM	WU	WW
2009	8.1	1.4	1.5	2.0	7.6	0.6	3.7	0.8	1.7
2010	6.3	1.1	1.3	1.8	7.1	0.5	3.3	0.7	1.7
2011	7.1	1.3	1.5	2.0	8.6	0.6	3.8	0.8	2.0
2012	8.3	1.6	1.8	2.4	8.9	0.8	4.6	1.1	2.7
2013	8.7	1.8	2.0	2.6	9.7	0.9	4.9	1.2	3.0
2014	11.0	2.1	2.6	3.3	12.0	1.1	6.0	1.4	3.8
2015	11.1	2.3	2.8	3.5	11.9	1.3	6.3	1.6	4.0
2016	12.2	2.6	3.2	3.8	12.2	1.4	6.9	1.8	4.6
2017	11.7	2.6	3.2	3.8	11.1	1.4	6.9	1.9	4.6
2018	11.7	2.7	3.3	4.0	11.0	1.6	6.9	1.9	4.8
2019	11.1	2.8	3.4	4.0	10.6	1.7	6.9	2.0	4.8
2020	12.3	3.1	3.8	4.4	10.5	1.9	7.7	2.3	5.5
2021	13.8	3.7	4.2	5.5	14.3	2.3	8.6	2.9	6.2
2022	13.2	3.8	4.3	5.6	15.5	2.4	8.5	2.9	6.4
2023	10.3	3.0	3.6	4.4	11.2	2.1	6.9	2.3	5.3

Abbreviations: MM, man first author/man last author; MU, man firrst author/unknown last author; MW, man first author/woman last author; UM, unknown first author/man last author; UU, unknown first author/unknown last author; UW, unknown first author/woman last author; WM, woman first author/man last author; WU, woman first author/unknown last author; WW, woman/woman.

Gender Differences in Overall Attention and Source of Attention

The analysis shows a notable difference in the overall AAS and its individual components between women and men serving in the first and last author positions (Figs 1A and 1B). Women in the first author position have a 15.2% lower AAS compared with their men counterparts (P < .001). The discrepancy is slightly less for last authors, with women exhibiting an 8.3% lower AAS than men (P < .001).

FIG 1.

Gender differences in overall attention and sources of attention, including (A) relative differences in AAS for woman first author compared with man first author by specific source of attention; (B) relative differences in AAS for woman last author compared with man last author by specific source of attention. AAS, Almetric Attention Score.

In terms of different sources of attention, first author women received less coverage by 20.0% in news (P < .001), 26.3% in policy discussions (P < .001), 29.8% in patent references (P < .001), 29.3% in Weibo mentions (P = .02), and 16.0% in Twitter/X mentions (P < .001). Similar trends were observed for women last authors, with overall decreases in mentions across news (P = .006), policy discussions (P < .001), patents (P < .001), Wikipedia references (P < .001), and citation counts (P < .001). No single metric showed a statistically significant preferential mention of women first or last authors over men.

For the longitudinal analysis (2009-2023), there was an upward trajectory in engagement metrics across both genders. Nonetheless, our data reveal a persistent gender gap. On Twitter/X, men first authors outpaced women with an average of 12.5 mentions compared with 8.6 (P < .001; Fig 2A). A similar trend emerged in news media, with men first authors receiving slightly more mentions than women by a margin of 0.2 (Fig 2B). Further examination of altmetric score highlights a gender discrepancy favoring men first authors, who scored an average of 1.8 points above women first authors (Fig 2C).

FIG 2.

Analysis of AAS over time and by source of attention, including (A) mentions of first and last author manuscripts on Twitter/X by gender over time; (B) first and last author news mentions by gender over time; (C) first and last author AAS by gender over time. AAS, Almetric Attention Score.

Within last authorship, manuscripts with last authors who were men were more frequently mentioned in Twitter/X conversations, averaging 13.2 mentions in contrast to 9.5 for their women counterparts (P < .001; Fig 2A). This gender imbalance was echoed in news coverage, where men last authors were mentioned nearly twice as often as women last authors (0.8 v 0.4 mentions, respectively; P < .001; Fig 2B).

Citation Imbalance by Specialty

The proportion of papers in each author category can be seen in Figure 3B. Woman first author/man last author (WM) and WW were the categories with the lowest proportions of manuscripts, and both also experienced minimal growth over the time period of interest (WM: 10.2% in 2009 to 13.0% in 2024; WW: 9.5% in 2009 to 14.4% in 2024; Fig 2B).

FIG 3.

Analysis of gender combinations (gender of first author and gender of last author) and citations, including (A) proportion of WW papers over time by oncologic specialty; (B) gender combination proportions over time (including WW, WM, MM, MW); (C) over-/undercitation by gender combination (including MM and WW); (D) over-/undercitation by all gender combinations (including WW, WM, MM, MW); (E) over-/undercitation by oncologic specialty and gender combination (including MM and WW); (F) graph illustrating imbalance between mean difference in citations for WW versus MM with each point illustrating a specific journal which is color-coded according to the oncologic field in which it belongs. MM, man first author/man last author; MW, man first author/woman last author; WM, woman first author/man last author; WW, woman first author/woman last author.

The representation of WW authorship in oncology subspecialties has relatively modest growth, with the exception of breast oncology and gynecologic oncology, which showed a marked increase in WW authorship. Gynecologic oncology, for instance, nearly doubled its representation of women authors (WW 17.7% in 2009 to 35.4% in 2023). Overall, none of the subspecialties experienced a decline in women's authorship, with the majority showing at least a nominal increase over the past 14 years (Fig 3A).

Figure 4 elucidates distinct interjournal variations in the representation of WW authorship across various oncology subspecialties. Notably, within each field, there are instances of wide variability between specific journals.

FIG 4.

Proportion of WW papers over time by oncologic specialty, including specific high-impact journals within each category. WW, woman first author/woman last author combination.

Citation Imbalance

The MM combination experienced overcitation, with an MCPD of +16.2% (Fig 3C). In comparison, WW papers were found to be undercited, with an MCPD of –7.7%. Furthermore, this trend generally persisted among all four gender combinations (Fig 3D). Interestingly, man first author/woman last author papers experience the highest overcitation with an MCPD of +21.3%. However, the overall pattern of WW papers being routinely undercited and authorship by men facilitating increased citation were observed. This trend persisted even when unknown authors were included in the over-/undercitation analysis, where the combinations of unknown/man and man/unknown experienced higher overcitation compared with unknown/woman and woman/unknown combinations (Appendix Fig A2).

Citation gaps were not uniform across all subfields (Figs 3D and 3E). In urologic oncology, MM experienced a percentage overcitation rate of 0.32%, whereas WW works were significantly undercited by 24.2%, bringing the total gender-based citation gap to 24.5%. Medical oncology stands out with the largest disparity, where MM pairs receive 19.0% more citations than anticipated and WW pairs 22.9% fewer (total citation gap = 41.9%).

Gynecologic oncology and radiation oncology demonstrated exceptions to this pattern. In these subspecialties, both WW and MM authorship configurations receive citations at rates exceeding expectations, with MM pairs at 59.5% and WW at 39.7% in gynecologic oncology, and 47.6% for MM and 39.5% for WW in radiation oncology.

Sources of Research Funding

Across all of oncology, WW combinations were 5% more likely to report any funding source, which includes institutional, NIH, pharmaceutical, or other sources (MM 35.6% v WW 40.5%; Appendix Table A1). In contrast, MM were more likely to report funding from pharmaceutical companies (MM 8.2% v WW 5.5%; Appendix Table A2).

DISCUSSION

In our study encompassing 652,842 published manuscripts between 2009 and 2024, we sought to evaluate gender differences in proportions of papers published, citation metrics, and AAS across the field of oncology. The data from our study reveal that publications authored by men, particularly those with men first and last authors (MM), are consistently cited more frequently than those with women first and last authors (WW). This trend was observed across various oncology subspecialties but was most pronounced in medical oncology. This could reflect a combination of cognitive biases where scholars are more likely to cite works from their peers who are men, as well as structural biases within the research community that might favor principal investigators who are men for clinical trials, research topics, or methodologies more commonly pursued by men. Furthermore, the proportion of women physicians within each field likely plays a role, particularly in fields where the number of women physicians are increasing over time but where women are still less likely to be senior authors and thus not yet contribute significantly to the WW category.

The undercitation of women-authored papers is troubling because it can directly affect the visibility and influence of women's research contributions. Citation metrics can influence a variety of professional outcomes, including graduate opportunities, funding success, career positions, awards, distinctions, and the critical processes of tenure and promotion.²¹ These trends have been substantiated by research, including a critical evaluation of the Altmetric Top 100 lists from 2015 to 2019, which demonstrated a significant citation gap disadvantaging women authors.²² Furthermore, an in-depth examination of more than 5,000 articles from high-impact medical journals showed a recurring pattern of fewer citations for papers where women serve as first or last authors.⁷ Particularly pronounced was the finding that papers with women occupying both the first and last authorship positions experienced a citation rate halved compared with their counterparts who were men. These studies echo our own findings, pointing to an average gender citation gap (as defined by MCPD) of 23% in articles authored by WW when contrasted with those authored by MM, highlighting a critical issue that the academic community must address to promote equitable recognition of all researchers.

Our study examined possible consequences of undercitation using the AAS, which has provided perspective on how research is perceived and valued across different platforms.⁸ The AAS findings are indicative of a persistent gender gap when it comes to the visibility and recognition of academic work. Our results indicate that papers with women in the first author position were found to have an AAS that was 15.2% lower than those of their men counterparts—a significant difference that carries important implications for the reach and impact of their research. Last authors who are women also faced a deficit, with an 8.3% lower AAS compared with men. One possible explanation for the lower AAS experienced by women first/last authors is the finding that women were less likely to be published as primary (36%) and senior (26%) authors in high-impact factor journals.⁷ Because articles in high-impact journals are more likely to be discussed in mainstream media, reach larger Twitter/X audiences through retweets, and have other downstream effects, these articles presumably experience higher AAS scores which may contribute to the gender disparity observed.

Our findings indicated that manuscripts by men first and last authors consistently received more mentions across the majority of platforms over time. The gender gap in social media mentions cannot be solely explained by differential use of social media. Indeed, one study of health science researchers demonstrated equal Twitter/X use among men and women physicians.²³ Additionally, a survey study published in JAMA reported that men and women physicians reported similar use of social media in a professional capacity.²⁴

Although there is a scarcity of research on AAS gender disparities specifically within oncology, other medical disciplines have begun to shed light on this phenomenon. Studies from cardiology, for instance, reveal that articles authored by women first authors are significantly less likely to achieve high AAS compared with those by men.⁸ Yet, when women occupy senior authorship roles, this gap in AAS does not seem to persist, and the difference in AAS between WW and MM authored articles is not significant.

With the rise of social media as a pivotal platform for the dissemination of scientific and medical knowledge, the relevance of AAS and their associated sources of attention have surged, given their potential correlation with citation rates. In the field of oncology, this correlation has been particularly noted in randomized phase III cancer trials, where a higher AAS is often an indicator of a greater number of citations.^25,26 This trend is mirrored in gynecologic oncology research, where a robust positive relationship has been identified between median citation counts and median AAS.²⁷ The potential association between citation metrics and AAS underscores the importance of amplifying the presence of women-authored research in prominent journals to bolster visibility and citation impact.²⁸

Social media has emerged as a critical driver of academic visibility, with increasing evidence suggesting that online presence significantly influences citation metrics and academic recognition.²⁹ Men are more likely to be recognized as Digital Opinion Leaders on platforms like Twitter/X, achieve social media verification, and experience fewer barriers to online engagement. In contrast, women face systemic disadvantages, including lower rates of verification and a higher likelihood of experiencing harassment.³⁰ These challenges likely exacerbate gender disparities in the dissemination and impact of research, as online discourse increasingly drives news coverage, blog discussions, and broader engagement with academic work. This study faces several limitations. Using an online database to determine the gender of authors introduces the possibility of misclassification and incomplete data due to some authors being labeled as unknown or excluded. Additionally, this approach does not account for the nuances of gender identity and self-identification, which transcend binary categories. Such complexities are beyond the capacity of bioinformatic tools to discern, as they typically rely on name-based gender inference. Nonetheless, the use of author gender categories serves as a practical approximation for our study since names significantly influence the perception of gender identity. Such perceptions can affect judgments of scientific merit and, by extension, citation practices—regardless of the true gender identity of the authors in question. The study also does not delve into the specifics of topics or publication types—such as videos or manuscripts—which can influence AAS. For instance, studies have found that content related to treatment and quality of life, or the use of video abstracts, tends to attract more online engagement.³¹ Large clinical trials also generally receive more attention, and given that men are more likely to receive pharmaceutical funding and be chosen to lead such clinical trials, studies with last authors who are men may receive more attention for this reason. Future research into the impact of gender on AAS might consider these factors more closely, including examining the gender makeup of citation sources and the proportion of women authorship in papers, which could provide deeper insights into the nuances of academic influence and recognition. Finally, critics of AAS point out the lack of transparency in methodology and the weights of each component, as well as poor reproducibility.³² Despite these limitations, this scoring system still likely represents the best way to determine the attention an article receives.

Despite existing challenges, there are several effective measures that can be implemented to narrow the gaps in citation rates and AAS related to gender disparities. Establishing transparent and equitable citation practices is crucial. Journals can play a significant role by encouraging or requiring the inclusion of diverse work, particularly by integrating more women and under-represented groups into their reviewer and editor pools. Additionally, there should be a deliberate effort to promote the work of women researchers through social media and press releases, which could involve partnerships with academic institutions to showcase a broader range of research through various communication channels and support of OA publishing. Furthermore, funding bodies are encouraged to consider AAS and citation data within the context of existing biases when assessing grant applications. By accounting for the influence of gender bias on these metrics, it is possible to reduce their impact on funding decisions and support fairer research evaluation practices.

As the proportion of women in medicine steadily grows over time, we should see a corresponding increase in representation on panels, editorial boards, and so on, which may subsequently improve gaps in citation rates and visibility. Indeed, the Association of American Medical Colleges reported an increase in the proportion of women practicing medicine from 28.3% in 2007 to 36.3% in 2019, with 2019 also representing the first year that the majority of US medical school students were women (50.5%).³³ Despite under-representation of women in the current workforce, gender differences in expected citation rates should not exist, and these differences cannot be explained by the proportion of women physicians alone. This is exemplified in the field of gynecologic oncology—although 57% of academic physicians in obstetrics and gynecology are women,³⁴ our study demonstrated that MM papers are still overcited.

In conclusion, significant gender disparities in citation rates and AAS exist within oncology. Our findings highlight a systemic issue where women-authored research is undercited and receives less attention compared with works authored by men, potentially affecting career advancement, funding opportunities, and academic recognition. As we advance, it will be essential to continue monitoring these metrics and implementing robust policies that not only recognize but also combat the biases that perpetuate gender disparities in academic research.

APPENDIX. Supplementary Methods and Data

Oncology-Related Search Terms for Almetric Database Through Application Programming Interface

“sarcoma,” “radiation therapy,” “stem cell transplant,” “targeted therapy,” “immunotherapy,” “melanoma,” “malignancy,” “lymphoma,” “chemotherapy,” “carcinoma,” “neoplasm,” “tumors,” “oncology,” “leukemia,” “carcinogenesis,” “metastasis,” “adenocarcinoma,” “squamous cell carcinoma,” “glioma,” “myeloma,” “cancer genetics,” “precision medicine,” “cancer screening,” “cancer prevention,” “cancer diagnosis,” “palliative care,” and “cancer epidemiology,” and “sources of funding”

Alternative Metrics Collected for Data Enrichment

In addition to the gender of the authors, our data set was enriched with a variety of fields to support a multifaceted analysis. These include the “Altmetric Attention Score,” “Title,” “Journal/Collection Title,” “Journal ISSNs,” “Authors at my Institution,” “Departments,” “Output Type,” “OA Status,” “OA Type,” “Subjects (FoR),” “Sustainable Development Goals,” “Affiliations (GRID),” “Funder,” “Publication Date,” “DOI,” “ISBN,” “National Clinical Trial ID,” “URI,” “PubMed ID,” “PubMedCentral ID,” “Handle.net IDs,” “ADS Bibcode,” “arXiv ID,” “RePEc ID,” “SSRN,” “URN,” “News mentions,” “Blog mentions,” “Policy mentions,” “Patent mentions,” “Twitter mentions,” “Peer review mentions,” “Weibo mentions,” “Facebook mentions,” “Wikipedia mentions,” “Google+ mentions,” “LinkedIn mentions,” “Reddit mentions,” “Pinterest mentions,” “F1000 mentions,” “Q&A mentions,” “Video mentions,” “Syllabi mentions,” “Number of Mendeley readers,” “Number of Dimensions citations,” “Details Page URL,” “Badge URL,” “Publisher Names,” as well as authorship details such as “first_author_full,” “last_author_full,” “first_author_first,” “last_author_first,” and the corresponding gender identification fields. These additions aim to provide a comprehensive understanding of the research articles’ reach and impact, transcending beyond traditional citation metrics.

Funding Information

Funding status was classified for the analysis based on the presence or absence of funding (yes/no). Additionally, data on the funding mechanism were cross-referenced with the Find Drugs & Conditions³⁵ Application Programming Interface (API) and further analyzed to delineate funding sources based on whether they originated from the pharmaceutical industry or nonpharmaceutical sectors for those previously categorized as yes.

Dataset Preprocessing

The data set was rigorously preprocessed to maintain data accuracy and pertinence to our research objectives. This involved normalizing author names and correcting discrepancies within the metadata. To enhance the scope of papers suitable for gender classification based on first names, we used the DOI crossref API for comprehensive name citation matching.¹⁶ This process enabled us to associate abbreviated names, often represented by initials in certain journal entries, with their full-name counterparts. For instance, an entry such as L.B. Smith, accompanied by a DOI, was matched to its full form, Laura Bernice Smith, facilitating the application of the GenderGuesser API.¹⁷

Gender Guesser API to Assign Author Gender Categories

Using the Gender Guesser API—a commercial tool providing statistical name-gender correlations across 191 countries—we assigned gender to authors whose first names had at least a 99.9% probability of being associated with a particular gender, labeling them as man or woman accordingly.

Our comprehensive data set consisted of 653,400 papers, and we were able to confidently assign gender (first, last or both) in 75.07% of the papers. Furthermore, for 20.89% of the papers, gender could only be ascribed to either the first or last author (Appendix Fig A1).

Limitations of Assigning Gender Categories Using Gender Guesser API

It is important to clarify that author gender category is a statistical tool and does not necessarily reflect the actual gender identities of the individuals. It is a proxy indicating a correlation between a given name and gender identity as commonly recognized on social media or official documents. The actual gender identity of authors could only be discerned through self-identification or personal knowledge, which lies beyond the scope of an automated analytical process such as ours.

Furthermore, our study concentrated on the first and last authors due to a well-established norm in life sciences authorship. Typically, the first author of life science articles is the junior author responsible for conducting the research, whereas the last author is usually the senior author who conceptualizes and finances the research. Thus, a practical assumption was that first and last authorship were the most important contributors to garnering citations—although all authors play important roles in publication of an article, analysis of every authors’ gender was outside the scope of the current study.

Methods for Analyzing AAS Score

We used an Ordinary Least Squares (OLS) regression approach to establish a model for investigating the dynamics between the logarithmically transformed AAS (independent variable) and the volume of Dimensions citations (dependent variable). In this construction, we accounted for variables such as the first author's gender, open access (OA) status, and the publication year, enabling a measured examination of their collective influence on the citation count. Our regression framework incorporated an intercept to capture the baseline effect beyond the scope of the explanatory variables. A constant term was introduced, facilitating the interpretation of the model's coefficients with respect to a null baseline of the independent variables. Upon fitting the OLS regression model, we evaluated the statistical significance of each coefficient to determine the impact of gender, OA status, and AAS on the predicted number of citations.

Gender-Blind Expectations Model

Gender-blind expectations using generalized additive model on the binomial outcomes (man first author/man last author [MM], woman first author/woman last author; generalized additive model) previously described was used.¹⁸ Briefly, gender-blind null model predicts the probability of citations based on publishing journal, categorization of research, year of publication, and combined number of authors in the data set. For each article, one can define the over-/undercitation of each author gender category as the percent difference of observed citations from gender-blind expectations. The degree to which papers authored by MM pairs are over- or undercited can be quantified using the following formula: mean citation percentage difference = a (oMM – eMM)/eMM × 100, where oMM represents the actual observed number of citations received by MM papers from the citing articles and eMM is the expected number of citations for MM papers, as forecasted by the gender-neutral model.

FIG A1.

Number of articles by gender category (total articles = 652,843). MM, man first author/man last author; MU, man first author/unknown last author; MW, man first author/woman last author; UM, unknown first author/man last author; UU, unknown first author/unknown last author; UW, unknown first author/woman last author; WM, woman first author/man last author; WU, woman first author/unknown last author; WW, woman first author/woman last author.

FIG A2.

Over-/undercitation by gender combination (including combinations of men, women, and unknown). MM, man first author/man last author; MU, man first author/unknown last author; MW, man first author/woman last author; UM, unknown first author/man last author; UU, unknown first author/unknown last author; UW, unknown first author/woman last author; WM, woman first author/man last author; WU, woman first author/unknown last author; WW, woman first author/woman last author.

TABLE A1.

Analysis of Research Funding (all types) by Gender Combination

Gender Combination	Unfunded Percentage	Funded Percentage
MM	64.388485	35.611515
MU	59.718076	40.281924
MW	59.276442	40.723558
UM	56.001571	43.998429
UU	63.416507	36.583493
UW	54.865125	45.134875
WM	56.288146	43.711854
WU	56.257468	43.742532
WW	59.517286	40.482714

TABLE A2.

Analysis of Funding (pharmaceutical v not) by Gender Combination

Gender Combination	Non–Pharma-Funded Percentage	Pharma-Funded Percentage
MM	91.725208	8.274792
MU	93.440373	6.559627
MW	92.220682	7.779318
UM	94.554035	5.445965
UU	96.580450	3.419550
UW	94.802561	5.197439
WM	93.665998	6.334011
WU	95.109701	4.890299
WW	94.465034	5.534966

Abbreviations: MM, man first author/man last author; MU, man first author/unknown last author; MW, man first author/woman last author; UM, unknown first author/man last author; UU, unknown first author/unknown last author; UW, unknown first author/woman last author; WM, woman first author/man last author; WU, woman first author/unknown last author; WW, woman first author/woman last author.

PREPRINT VERSION

https://www.medrxiv.org/content/10.1101/2024.04.26.24306437v1

AUTHOR CONTRIBUTIONS

Conception and design: Andres Correa, Philip Abbosh, Adam Calaway, Amanda Nizam, Sarah P. Psutka, Nazli Dizman

Financial support: Philip Abbosh

Administrative support: Philip Abbosh

Data analysis and interpretation: Rebecca A. Campbell, Emma Helstrom, Lauren Chew, Renu Eapen, Elizabeth Plimack, Andres Correa, Alexander Kutikov, Amanda Nizam, Shilpa Gupta, Sarah P. Psutka, Pedro Barata, Mohit Sindhani, Christopher J. Weight, Laura Bukavina

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Gender Disparities in Citations and Altmetric Attention Score in Oncology

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/op/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).