Digit ratios by computer-assisted analysis confirm lack of anatomical evidence of prenatal androgen exposure in clinical phenotypes of polycystic ovary syndrome

Polycystic ovary syndrome (PCOS) is a complex endocrine disorder, having no single diagnostic trait [1, 2]. Much controversy has surrounded the diagnosis of this condition but in 2003 experts proposed that a diagnosis of PCOS be based on the presence of two of three symptoms: 1) oligo or chronic anovulation (amenorrhea), 2) biochemical and/or clinical hyperandrogenism and 3) polycystic ovaries on ultrasonography [1, 2]. These criteria recognized the broad clinical spectrum of PCOS including, the manifestation four unique phenotypes [3]. Frank PCOS represents the most severe form of this condition and is characterized by the presence all three symptoms. Non-PCO PCOS is characterized by oligoanovulation, hyperandrogenism, but normal ovarian morphology. Ovulatory PCOS describes the presence of hyperandrogenism, polycystic ovaries and normal menstrual cycles, whereas Mild PCOS describes the presence of oligoamenorrhea and polycystic ovaries, but no hyperandrogenism. While the validity of these phenotypes is still being debated [4, 5], there is consensus among experts that PCOS imparts serious consequences for the long-term health and quality of life of patients and therefore should invite early identification and intervention [1‐6].

Despite familial clustering, the diverse manifestations of PCOS make it challenging to determine a single etiologic factor for this condition [7]. Experimental evidence in nonhuman primates has suggested that programming by prenatal androgens may contribute to variable susceptibility to PCOS in adult life and therefore lead to a heterogeneous clinical presentation (reviewed in [8]). Pregnant rhesus monkeys who received androgen treatment early in pregnancy (i.e. Day 40-44 of a 165-day pregnancy) gave arise to offspring that developed enlarged polycystic ovaries, hyperandrogenism, oligo-anovulation, increased basal luteinizing hormone (LH) secretion, insulin resistance, abdominal obesity, and hyperlipidemia. By contrast, the offspring of mothers that received treatment later pregnancy (i.e. after Day 90) did not demonstrate the same neuroendocrine alternations in LH secretion or changes in insulin sensitivity (reviewed in [8]). The differential effects of androgens during fetal development indicated that distinct programming windows existed for androgens to permanently modify future aspects of reproductive and metabolic function.

That prenatal androgens might contribute to development of PCOS has been retrospectively investigated by Cattrall et al. using a putative anatomical marker of in utero androgen exposure [9]. Exposure to androgens during fetal development affects finger length growth and leads to distinct discrepancies in male and female hand patterns (reviewed in [10]). Typically, men display a lower ratio between the index (2D) and ring (4D) fingers compared to women [11]. Cattrall et al. found a small, yet significant, difference in the right hand 2D:4D of women with PCOS compared to healthy female controls (98.3% of that in the controls) providing support for prenatal androgens in the etiology of PCOS [9]. More recently, we attempted to determine if clinical phenotypes of PCOS were associated with variations in 2D:4D but found no such difference [12]. The population studied by Cattrall et al. included only women with the most severe forms of PCOS, while our study encompassed a more varied patient population [9, 12]. Yet, when the women participating in our study were stratified by clinical phenotype, we failed to detect a difference in 2D:4D, even in women with Frank PCOS [12].

To date, studies investigating 2D:4D in women with PCOS have used Vernier calipers to measure finger lengths. Recent studies have indicated that use of Vernier calipers is not the most reliable method of acquiring 2D:4D [13‐15]. Vernier calipers demonstrate an interrater error of 1.2% or more [14], which has implications for comparing findings among studies [13]. Since the difference in right 2D:4D detected by Cattrall et al. may have fallen within the error range for Vernier calipers, we feel it important to reevaluate 2D:4D in women with PCOS using computer-based calipers which express an interrater error of <1% [14]. To that end, the aim of the current study was to re-evaluate the 2D:4D of our previous PCOS study population using the more reliable technique of computer-assisted analysis.

The goal of this study was to address the controversy over evidence of prenatal androgen exposure reflected in the digit ratios of women with PCOS. Recently our group showed that when 2D:4D were measured with Vernier calipers, women with PCOS did not demonstrate finger length patterns consistent with increased levels of in utero androgen exposure [12]. This was in contrast to a previous report that had also used Vernier calipers to measure 2D:4D in women with PCOS [9]. Since observed differences in 2D:4D are generally small, there is growing support that studies investigating potential effects of prenatal androgens use the most consistent and reliable technique available to measure finger lengths [13‐15, 18‐20]. In this study, we imaged the hands of women with four clinical phenotypes of PCOS, healthy female controls and men, and used computer-based calipers to measure their finger lengths since this method was recently validated to be the most reliable [14, 15]. Consistent with this being the most reliable technique, we obtained levels of intra-observer reliability when calculating 2D:4D (Mean ICC = 0.912) that were much higher than what we obtained when we used Vernier calipers to make measurements in this same study population (Mean ICC = 0.845) [12]. These levels of reliability were in range of what others have demonstrated in studies comparing the appropriateness of computer-assisted measurements for obtaining 2D:4D in women [14].

In the current study, a comparison of 2D:4D among healthy female controls and women diagnosed with PCOS by the Rotterdam criteria showed no significant difference among groups. Unlike a previous report of lower right-hand 2D:4D in women with PCOS [9], we again noted a tendency toward lower 2D:4D in healthy female controls which would contradict a theory supporting excess prenatal androgen exposure in women with clinically defined PCOS. A tendency for lower 2D:4D in female controls contributed to our paradoxical finding that 2D:4D were positively, albeit weakly, associated with clinical markers of androgen excess in women. Our female control subjects were carefully selected following a complete clinical assessment for PCOS and they showed no indications of androgen excess, menstrual cycle irregularity or ovarian dysmorphology. The inclusion of men to our study population afforded us the possibility to compare 2D:4D between men and women with or without PCOS. This comparison substantiated that our female control group had 2D:4D that were consistent with what one would expect when comparing 2D:4D among sexes [11]. Moreover, our finding of lower 2D:4D in men compared to women with PCOS further supported the notion that women with PCOS do not demonstrate androgenized finger lengths.

Since the PCOS criteria we used spanned a broad clinical spectrum - with a potentially variable etiology - we stratified subjects by phenotype and compared 2D:4D among PCOS groups. Our hypothesis that variable manifestations of PCOS might reflect variations in 2D:4D was not supported even when using this more reliable technique to obtain finger length measurements. That is, even women with severe forms of the syndrome had finger length patterns that were no different than milder variants of PCOS or female controls. It is important to acknowledge that the interpretation of these finding is limited by the small number of subjects available for analysis in each of the phenotypic groups. Comparisons of 2D:4D among all women with PCOS and controls were sufficiently powered in that they had greater than 80% power to detect a difference of 0.019 in 2D:4D at an alpha level of 0.05 assuming a standard deviation of 0.035 [9, 12]. By contrast, comparisons of 2D:4D in women with Frank, Non-PCO, Mild or Ovulatory PCOS all had less than 22% power to detect differences among groups. Assuming the afore mentioned effect size and standard deviation, a minimum of 55 subjects per group would be required to perform an adequately powered assessment of digit ratios among PCOS phenotypes. There were several obstacles that limited our ability to recruit the desired number of subjects. First, we were limited in the number of women with PCOS that could be identified in a small geographical area. The prevalence of PCOS is estimated at 5 - 10% [3] and while that may appear to represent a relatively common condition, our ability to locate women that were sufficiently concerned over PCOS symptoms and willing to participate in a clinical study was limited. Second, the identification of PCOS phenotypes is an incredibly involved process requiring participation from various clinical specialists. Difficulty in coordinating access to resources limited the number of subjects we could assess at any given time. Last, it may not have been even possible to identify the necessary number of subjects meeting the criteria for each PCOS phenotype. This was likely the case for the Non-PCO PCOS group. Diagnostic criteria for PCOS are controversial as are the manifestation of these individual phenotypes [4, 5]. Over a two-year period, we screened over 125 women with concerns of PCOS symptoms and only identified 4 women meeting the criteria for Non-PCO PCOS. At this rate, it would have taken us over 25 years to identify the appropriate number of subjects. The low prevalence of this phenotype calls into question its validity and the need to verify ultrasonographic criteria for polycystic ovarian morphology [21‐23].

There is growing concern that 2D:4D may represent only a weak, or perhaps even useless, measure of prenatal androgen exposure [24‐27]. It is difficult to know for certain whether our negative findings represent a lack of involvement of prenatal androgens in PCOS or whether this represents a lack of an effect of androgen exposure on 2D:4D per se. That prenatal androgens directly contribute to the development of PCOS in humans seems increasingly unlikely. Experimental evidence showing that pregnant monkeys receiving exogenous androgens gave rise to female offspring with PCOS symptoms is the best data supporting a role for prenatal androgens in PCOS [8]. However, very little is known about the mechanism(s) whereby prenatal androgen excess leads to PCOS-like alterations in reproductive and metabolic function. In humans, it is unlikely that maternal androgens sufficiently cross the placental barrier to affect the fetus unless placental function is compromised in some capacity in PCOS mothers [8]. Another possibility is a fetal origin of androgen excess in which increased androgen production from the fetal adrenals and/or ovaries act during development to program future organ systems [8]. There is no evidence to support either of these possibilities and the most recent study of opposite-sex twin pairs did not support an increased risk of PCOS in the female twin despite exposure to intrauterine androgens of the male fetus [28]. Given that our study was limited by small sample sizes, it will be important to revisit this issue with an adequately powered study using non-controversial criteria for PCOS - if and when they come to fruition.

Computer-based measurements of 2D:4D verified that women with clinical phenotypes of PCOS do not exhibit anatomical indications of increased prenatal androgen exposure. The current study also extends our previous observations to show that 2D:4D in women with PCOS were higher than those in men, consistent with PCOS patients having relatively lower levels of in utero androgen exposure. We caution that these results do not entirely rule out the possibility that in utero androgens play a role in the development of PCOS; however, androgenized 2D:4D are not a characteristic of PCOS.