Historically, clinical investigations have often overlooked test subject sex as a variable and failed to report test subject sex. However, advancements in research have highlighted the importance of sex differences in multiple biological systems—endocrine, biochemical, pharmacological—which reflect disparity in incidences of diseases, immunological reactions, and responses to drugs [1-3].

Males and females can have different biological properties even on the most basic cellular levels [4,5]. For example, microglia-to-neuron signaling plays a central role in chronic pain hypersensitivity in males, but not in females. For females, adaptive immune cells, such as T lymphocytes, take the place of microglia. Such sexual dimorphism in pain mechanisms indicates that male and female mice are not interchangeable in pain studies [6]. Such differences in basic properties are also reflected in oncology [7]. The majority of anticancer drugs, including methotrexate, paclitaxel, fluorouracil, capecitabine, gemcitabine, topotecan, and etoposide show sex differences in clearance and survival rate [8-10]. However, failure to report test animal sex remains widespread in preclinical research. This potentially results in poor reproducibility [11].

Investigations of test animal sex selection and reporting have been conducted in various fields of medicine [12,13]. However, previous studies did not determine whether researchers justified the selection of one sex over the other or include multiple experiments in the same study. Therefore, we addressed these shortcomings by manually analyzing data that were informative about the state of test animal usage in oncological research.

We surveyed a total of 382 articles. Of these, 50.3% failed to report the sex of the test animals (Figure 2). However, most (98.2%) reported genetic background information. In total, 67.1% of articles that reported sex used only female mice, 24.6% used male mice, and only 8.3% used both male and female mice. This trend was reflected in all five of the oncology journals, with each publishing similar percentages of studies that used males, females, or both sexes (Figure 3).

Of the single-sex studies, only 30.6% provided justifications for using only one sex, and 96.2% of all justifications were simply based on anatomical differences, while the remaining 3.8% were unique justifications. However, even after we excluded studies with justifications, the same trend persisted. Female mice were used most often, with only a small percentage of studies incorporating both sexes into their experimental groups (Figure 4). Of the 15.7 studies that used mice of both sexes as test subjects, only 2.3 studies used both male and female mice and included sexbased analyses (Figure 5).

Overall, our analysis revealed that researchers in the field of preclinical oncology do not pay adequate attention to test subject sex, as less than 50% of publications during 2015 reported subject sex, while the numbers of studies including both sexes and/or sex-based analyses were much lower. Ideal studies should include both sexes alongside sex-based subgroup analyses [15]. Studies that use both sexes without conducting sex-based analyses have ambiguous value in comparison to single-sex studies, because the use of both sexes can hinder the interpretation of results if there are differences between the sexes related to the topic of the study.

Another prerequisite of an ideal study is inclusion of justification(s) for not specifying sex or using a single sex. Female mice may be used for in vivo experiments due to evidence that female mice show more drastic effects than male mice after treatment [16], while male mice may be selected as test subjects because hepatocellular carcinomas are more prevalent in males [17]. Overall, only 14.4% of the studies we surveyed were ideal, either including both sexes and sex-based analyses, or single-sex studies with justifications.

Most studies, however, were less ideal. One of the articles we examined described an experiment using female mice to make prostate cancer xenograft models without commenting on the sex disparity between host and graft. This was puzzling, as female mice have different physiologic properties from male mice and prostate cancer xenografts may behave aberrantly when grafted from male donors to female hosts [18].

Some journals have their own policies regarding animal test subjects, while others recommend following pre-established guidelines, such as the Animal Research: Reporting In Vivo Experiments (ARRIVE) guidelines (Table 1). The ARRIVE guidelines consist of a checklist of 20 basic items describing the minimum information that all scientific publications involving animal experiments should report. These include recommendations regarding the number and specific characteristics of animals used such as species, strain, sex, and genetic background [19].

All five of the journals included in this study had policies recommending that the sexes of animal test subjects be delineated in the article. However, 28% to 64.5% of articles (50.3% average) left the sex of animal subjects unspecified, suggesting that journal policies are often not followed. A 2010 study examining reports in general biology, immunology, neuroscience, physiology, pharmacology, reproduction, endocrinology, behavioral physiology, behavior, and zoology (Journal Citation Reports subject categories) found similar results [12]. In that study, the authors found that subject sex was not reported in 22%–42% of articles in neuroscience, physiology, and interdisciplinary biology journals, or in more than 60% of immunology reports, while fewer than 8% of articles in the behavior, endocrinology, and pharmacology categories failed to specify the sexes of experimental animals or tissues. Another study published in the journal Pain revealed that 97% of all articles published from 1995 to 2005 reported test subject sex [13].

In comparison with the values observed in previous studies, the 49.7% of reporting test subject sex among the five oncology journals surveyed in this study was very low. A recent investigation found that cancer research articles using mice had the lowest rates of sex and age identification (48%) among many disciplines [20]. Those authors also noted that preclinical cancer research was more likely to use female mice as test subjects. Our findings agreed, with 57.5% to 77.9% of studies including only female mice. This was a surprising finding as it is often assumed that studies using male subjects dominate preclinical research. It was hypothesized that the explanation for the predominant use of male subjects was simply convention, reinforced by the assumption that female hormonal cycles act as a source of confounding variability in experimental studies. However, there is no documentation of female estrous cycles affecting experimental results in mice. Studies that include both sexes offer several advantages, including enhanced power of research, practical advantages, and conformation with moral imperatives [12,13]. The use of female mice in oncology research might be due to the greater ease of accommodating female mice in research facilities, but the present study does not provide any concrete answers to this question.

Our study has certain limitations. First, the scope of our investigation was limited to studies published in the top five oncology journals in 2015. It therefore illustrates a cross-sectional view of oncological research trends and does not reflect historical trends. Nonetheless, our findings are compliant with those of a recent large-scale study [20] and are therefore likely to be an accurate reflection of current research trends. Another limitation is that the criteria we used to identify justifications for single-sex studies were ambiguous, and we had difficulty in defining which studies adequately justified using test subjects of only one sex because there is no gold standard for classifications regarding justification [21]. Even so, when our results were adjusted to include justified single sex studies, the majority still lacked any explicit explanation for female preferences in test subject sex. Our manual text mining approach made the assessment of the presence of justifications in single sex studies possible, even if these assessments were not perfect. This allowed us to consider the hypothesis that because certain cancers are specific to sex, the disparity in the number of articles investigating female-specific cancers could be explained. Indeed, most studies that studied sex-specific cancers such as cervical, prostate, or even breast cancer utilized only one sex.

In the field of preclinical oncology, as represented by the five journals with the highest impact factors in the field, more than half of all analyzed articles did not report the sexes of test subject animals. This represents a widespread problem in oncological preclinical research, for these low levels of sex-reporting rates undoubtedly hinder the reproducibility of research and cloud accurate interpretations of results, especially since increasing evidence indicates that there are clinically significant differences between male and female physiologies. Based on our results, we suggest that researchers take care to report test subject sex and to conduct sex-based subgroup analyses, even at the preclinical level.