Debating a testosterone “sex gap”

See allHide authors and affiliations

Science  22 May 2015:
Vol. 348, Issue 6237, pp. 858-860
DOI: 10.1126/science.aab1057

Embedded Image

An athlete competes in the women's triple-jump final in the 2012 Olympic games in London, UK.


Sexual dimorphism of testosterone (T) in elite athletes was at the center of a recent case at the “Supreme Court of Sport,” the Court of Arbitration for Sport in Switzerland, after teenage Indian sprinter Dutee Chand challenged a sports policy regulating competition eligibility of women with naturally high T. The idea of a “sex gap” in T is a cornerstone of this policy (1). Policymakers infer that men's higher T is the “one factor [that] makes a decisive difference” between men's and women's athletic performances (2)—so that women with naturally high T may unfairly enjoy a “massive androgenic advantage” over other women athletes (2). We report on an emerging scientific debate about whether the sex gap in T applies to elite athletes.

In 2011 and 2012, respectively, the International Association of Athletics Federations (IAAF) and the International Olympic Committee (IOC) adopted controversial policies that regulate levels of natural T in women athletes (1, 3). The IAAF policy sets a ceiling for women of 10 nmol/liter in blood, which it identifies as “within the normal male range,” whereas the IOC policy targets levels “within the male range” (1, 3). Women with high natural T, according to the IAAF, have an unfair advantage over women with lower natural levels (1). Unless they are androgen-resistant, women must lower their T in order to continue competing (1), which would require surgery or antiandrogens (4).

Appealing her exclusion under the IAAF policy, Chand told the Indian Express “At every level of my life … I have competed the way I am. I've been told the hormonal issue with me is natural so that's why we have decided [to appeal]” (5). Her March appeal was the first formal challenge to the policy; a decision is forthcoming.

The T policy is the latest attempt to use a biological marker to draw a bright line between women and men for sex-segregated sports. Decades of sex testing of all women athletes relied on biomarkers such as chromosomes. Officials dropped blanket testing in the 1990s, acknowledging that sex is irreducibly complex and that there is no scientific criterion for separating all men from all women ( 6). Nevertheless, they retained an ad hoc policy for when a woman's sex was questioned, which was criticized for continuing the doomed project of sex testing and for being arbitrary (7).

Still determined to find a biological way to regulate who can compete as a woman, policy-makers turned to testosterone, arguing that T is both the “performance enhancing hormone” (8) and a sharply differentiated trait between men and women (2, 3). In most studies, men's T levels are about 10 times those of women, and the highest levels seen in women are well below the lowest levels seen in healthy men. One policy-maker characterized this as “a huge no man's land” (9).

Recently, though, the idea of unequivocal sexual dimorphism in T levels has been challenged, at least in elite athletes. Only two large-scale studies of T in elite athletes exist, and they draw contradictory conclusions regarding a sex gap in T (10, 11). In the first, data are from 446 men and 234 women across 15 highly varied Olympic events (10). These data were collected as part of the GH-2000 study, an IOC- and World Anti-Doping Agency–funded project aimed at developing a test to detect human growth hormone abuse. The report states, “hormone profiles from elite athletes differ from usual reference ranges” in both men and women (10). In fact, there was “overlap between men and women, although the mean values differ.” Among women, 13.7% had T above the typical female range, and 4.7% were within the typical male range. In contrast to reference ranges, 16.5% of these elite male athletes had T below the typical male range, with 1.8% falling in the female reference range.

Not long after the GH-2000 report appeared, IAAF researchers published their own study on serum T in 849 elite women athletes in track and field from the 2011 Daegu World Championships (11). That study showed just 1.5% of women athletes with T above the female reference range, a sharp contrast with the 13.7% in the GH-2000 study.

DEBATING THE EVIDENCE. Three critiques of the GH-2000 report—raised by IAAF policy-makers in the Daegu study and in an IAAF-IOC rebuttal—bear on whether there is a sex gap in T (11, 12). The first issue is how sera were analyzed: The GH-2000 study used immunoassay (IA); the Daegu study used mass spectrometry (MS). IA overestimates T at lower values. There is no question that MS yields more accurate T readings at lower values. The use of IA in the GH-2000 study might have resulted in some inaccurately high readings among women, but it cannot explain the fact that a considerable proportion of men had very low T levels (in fact, IA underestimation would have countered the latter pattern). So the use of IA cannot account for the finding of a male-female overlap in the GH-2000 data. The Daegu study did not report men's values, so it can only shed partial light on the question of a gap.

The second disagreement concerns when to draw serum, because T changes in response to competition. IAAF-IOC policymakers suggest that the female-male overlap in T observed in the GH-2000 data may be an artifact of sampling within 2 hours after competition (12).

Embedded Image

Close-up of a baton before women's 4 x 400-meter relay final, 2012 Olympic games, London, UK.


This criticism requires a selective reading of the evidence on the effect of competition on T levels. The IAAF-IOC critique cites a single report showing male T levels dropped and female levels were steady or modestly rose after an Ironman competition (13). The broader literature shows that T may rise, fall, or remain unchanged after competition, and the main factors determining the response seem to be the type and duration of competition—not the individual's sex (14, 15). Intense resistance exercise and short-duration exercise are associated with increase in T, whereas endurance exercise (especially lasting greater than 3 hours) is associated with decrease in T (14, 15). There are few data on endogenous T in women athletes, but the most recent review again indicates a great variety of responses to exercise are possible—including a large and lasting increase in resting T from long-term resistance training (14).

The timing of serum collection in the GH-2000 study makes sense in the antidoping context, because of the need to understand hormone profiles “after competition when anti-doping tests are usually made” (10). Doping tests are often how women with high natural T are flagged, so understanding how natural T responds to competition is important.

The tussle over timing may obscure the important point that T is dynamic. Recent research shows that, in both sexes, T dramatically responds to physical situations as well as social cues and contexts, diurnal rhythms, training, and other factors (1417). For example, positive feedback from a coach can cause a rapid doubling of T level (17). The “correct” time to sample T depends on the purpose of the study, but the timing of blood draws seems unlikely to determine whether a study finds overlap in T between the sexes.

The third issue raised by the IOC–IAAF critiques of the GH-2000 study is the most fundamental: the rules for subject inclusion and exclusion. Both scientific groups agree that subjects who have doped should be excluded. Where they part is whether women with naturally high T should be excluded.

The two camps take opposite views on whether to include these women—a decision that bears directly on whether their findings support or undermine the policies. The GH-2000 study includes all women with high natural T in the sample. The Daegu study included women with high T of unknown etiology, but excluded as “confounding factors” all women whose high natural T can be traced to diverse sex development, also known as intersex (DSDI). In simple terms, some of the biological characteristics of women with DSDI would be classified as female and others as male. This challenges common ideas about sex, but it is widely recognized in medicine, law, and the social sciences that when people are born with mixed markers of sex (e.g., chromosomes, genitals, gonads), the medical standard is that gender identity is the definitive marker of sex—there is no better criterion (18).

What, then, is the logic that classifies women with DSDI as confounders? The Daegu report consistently pairs clinical language, such as “diagnosis” and “disorder,” with hyperandrogenism for the women with DSDI, and in their rebuttal to the GH-2000 paper, IAAF-IOC policy-makers use the phrase “hyperandrogenic disorders of sex development” (12). This signals their judgment that women with DSDI are not healthy and, therefore, should be excluded from reference ranges. But DSDI women are not necessarily unhealthy. High T can be associated with health issues but is not, in and of itself, a health problem for women (4).

An a priori understanding of women with DSDI as unhealthy and, thus, outside normal variation creates a rationale for their exclusion both in reference ranges and the policies. But it is also circular: Because women with DSDI are a priori excluded when the reference ranges are created, the findings from the Daegu study—that women athletes have T levels no different from nonathlete women—reinforce their values as outsiders and justify the policy.

There is a strong scientific argument for including DSDI women in the sample. These studies aim to establish T reference ranges for elite athletes: i.e., the focus is on physiological ranges not clinical ranges. This calls for descriptive statistics, and in this case, there is no valid basis for discarding some values as outliers. In both studies, if the full range of values for women's endogenous T is included, there is an overlap in T.

CALCULATING FAIRNESS. What looks like a controversy rooted firmly in science is ultimately a social and ethical one concerning how we understand and frame human diversity. These assessments are not trivial: They shape not only the research methods and findings but also how we understand what is at stake in this policy. And this has very real consequences for people's lives.

Policy-makers, among others, claim that the problem is that women with naturally high T have unfair advantage, despite having acknowledged in their Daegu study that “there is no clear scientific evidence proving that a high level of T is a significant determinant of performance in female sports” (11). Others see a very different problem: Women who have lived and competed as women their whole lives suddenly find themselves having to undergo medical interventions in order to remain eligible to compete in a category to which everyone agrees they belong.

Calculating what counts as a fair and level playing field for women must take all women athletes into account, including those with naturally high T and/or DSDI. We could return to a consensus reached decades ago, where policy-makers faced these same concerns and concluded that women “who were raised as girls and classify themselves as female should not be excluded from competition as women” (19). In other words, ensuring that women with high endogenous T and/or DSDI “have the same rights to participation in athletics as all women” (20) would be a good place to start.

References and Notes

  1. Acknowledgments: Supported by NSF grants 1331115 and 1331123.


Navigate This Article