High (feminized) digit ratio (2D : 4D) in Danish men: a question of measurement method?

Sir,

We read with interest the advance-access publication of Bang et al. (2005) in Human Reproduction. After review of the article, we have the following comments. The second-to-fourth digit ratio (2D : 4D) is a sexually dimorphic trait in humans: men tend to have lower 2D : 4D values than women. In late 1998, it was suggested in an article published in Human Reproduction (Manning et al., 1998) that 2D : 4D may be a useful biomarker for prenatal androgen action. This hypothesis has generated much interest since then and, as is now discernible, has started off an expanding research program. 2D : 4D has been shown to be a correlate of a suite of sex-dependent, hormonally influenced variables, including behavioural, cognitive, personality and somatic traits, adult-onset diseases and measures related to fertility and sexuality (Manning, 2002; Putz et al., 2004; Cohen-Bendahan et al., 2005).

Bang et al. investigated associations between right 2D : 4D, various testicular, semen and sex-hormone parameters in a sample of 360 young men from the Danish general population. Out of 14 correlations, they found only two to be statistically significant, but mutually contradictory. Hence, assuming that these were spurious (type I errors) rather than real effects, the authors concluded that ‘measurements of finger lengths do not have the power to predict the testicular function of adult men’ (Abstract).

Two features in the Bang et al. study attracted our attention. First, average 2D : 4D was 1.02 in the total sample; and second, finger lengths were measured from hand outline drawings (an unvalidated 2D : 4D measurement method). Here, we present arguments and evidence on the following: (i) observing mean 2D : 4D levels beyond unity in males, particularly from Northern European descent, is an unexpected and interesting finding; (ii) the 2D : 4D measurement method of Bang et al. is unlikely to be responsible for this finding; (iii) single measurements of sex hormones might be too unreliable to capture presumably subtle associations with 2D : 4D.

Presently, there are no other 2D : 4D data from Danish men that could serve as a reference point. Geographical (population or ethnic) variation in typical 2D : 4D levels is marked (Manning et al., 2000, 2003, 2004a; Peters et al., 2002) and currently poorly understood. Importantly, the entire 2D : 4D literature (more than 80 reports, with samples from more than 25 different countries or ethnic groups from five continents) does not contain a single male sample with a mean 2D : 4D larger than unity. Not even female samples with a mean 2D : 4D as high as 1.02 are known. Sample standard deviations for 2D : 4D generally are about 0.03. Hence, contrary to the assertion of Bang et al. that the observed average 2D : 4D value of 1.02 in Danish men was only ‘slightly higher’ than averages found for men in the UK (0.99, according to Bang et al., but most UK studies yielded 0.97 or 0.98), this constitutes a very large group difference of about one and a half standard deviation units. As for evidence from countries in proximity to Denmark, male mean 2D : 4D levels are 0.93 in Finland (Manning et al., 2000), 0.95 in Sweden (Sanders et al., 2005), 0.95 in Lithuania (Manning, 2002), 0.955 in Belgium (Millet and Dewitte, in press), 0.96 in Germany (Manning et al., 2000; Kempel et al., 2005) and 0.99 in Poland (Manning et al., 2000).

From the existing data, it appears that Northern European males of Scandinavian descent have comparatively low 2D : 4D. A typical male 2D : 4D level of 1.02, as reported by Bang et al. for Denmark, must be considered as exceptionally high. Such evidence for feminization in Danish men is consistent with their previously reported low sperm quality and quantity and high incidence of testicular cancer (Jørgensen et al., 2002), but perhaps less so with the inconspicuous sex ratio at birth in Denmark (Grech et al., 2002; Voracek and Fisher, 2002). Bang et al. stated that sample or national differences in 2D : 4D levels are ‘not likely to influence the interpretation of our results. The effects of increasing or decreasing ratios will be the same’ (Discussion). However, this only holds true to the extent that reliable 2D : 4D measurements can be obtained from hand outline drawings, to which central question we now turn.

Hand outline drawings have never been used before in 2D : 4D research and thus are not validated against the commonly used measurement methods (namely, direct finger measurements, or from photocopies, or from printouts of digital scans of the palm, made with vernier callipers; or from scan files or digicam pictures, made with image processing or graphics software tools). Bang et al. wrote in their Methods section that ‘the men had an outline of the ventral surface of their right hand made by the examining physician. These drawings were used for the measurement of the 2nd and 4th finger length from the basal crease of the finger to the tip’.

Unfortunately, from this description, it remains unclear what exactly was done. Conventional, palm-down hand outline drawings do not capture ventral surface landmarks, such as the basal finger creases. Conversely, palm-up hand outline drawings can only be made via transparent materials, such as glass or plastic plates, and are therefore possibly subject to distortion by the reflection and refraction properties of the material, and must be drawn by eye, without direct tissue contact, which might introduce a further margin of error. More general problems associated with these procedures might be individual differences in fingernail length, size of distal finger fat pads and degree of curvature of the finger tips (Manning et al., 2005), pressure exerted on the supporting surface, and webbing of the fingers (i.e. the location of the web intersection between fingers, which need not concur exactly with the location of the basal finger crease). In addition, one recent report indicated that different ascertainment methods may yield divergent 2D : 4D measurements (Manning et al., 2005).

We admit that, in view of these elaborations, we surmised that the high 2D : 4D levels obtained by Bang et al. could be suggestive of measurement bias. In order to achieve more knowledge on these issues and to evaluate their method, we performed a validation study of various 2D : 4D measurement methods.

The sample was comprised of 30 unselected Austrian men, aged 19–39 years (mean = 26.6 and SD = 5.4 years). Two trained investigators independently made single measurements of the men’s right 2D and 4D with digital vernier calipers accurate to 0.01 mm (we thank Clara Ertl and Barbara Reimer from our laboratory for their help with this research). Investigators were blind to each other’s measurements, blind to the research hypothesis outlined above, unaware of the Bang et al. article and experienced in the measurement of 2D : 4D. Measurements (one each from C.E. and B.R.) were secured in random order from four methods: directly from the fingers (DIM hereafter, for direct measurement method); from scan file printouts (SPM); from dorsal (palm-down) hand outline drawings (DHM) made with a sharpened pencil on white paper; and from ventral (palm-up) hand outline drawings (VHM) made with a thin permanent marker on an overhead transparency fixed onto a transparent plastic plate. Measurement landmarks were the ventrally located proximal-most (boundary) metacarpophalangeal flexion crease that divides the finger from the palm region (for DIM, SPM and VHM) or a line drawn through the appropriate between-finger web intersections (for DHM) and the fingertip, excluding possibly protruding fingernails (for all methods).

Measurement repeatabilities between investigators were evaluated with the appropriate type of intraclass correlation coefficient (ICC; two-way mixed-effects model with absolute-agreement definition; see Case 3 ICC in McGraw and Wong, 1996). For 2D, 4D and 2D : 4D, ICCs were, in order, 0.948, 0.983 and 0.860 for DIM; 0.991, 0.991 and 0.959 for SPM; 0.940, 0.868 and 0.830 for DHM; and 0.958, 0.932 and 0.780 for VHM (all Ps < 0.001). We concluded that interobserver repeatabilities were highest for SPM, although somewhat lower for the other established method (DIM) and both methods of hand outline drawings. To compare the agreement of 2D : 4D measurements obtained by different methods, corresponding (same-method) measurements were averaged, with the four methods representing the within-subject factor levels in a repeated-measures analysis of variance.

We had the following results: mean (SD) 2D : 4D was 0.953 (0.031) for DIM, 0.963 (0.033) for SPM, 0.961 (0.045) for DHM and 0.960 (0.038) for VHM. Mean 2D : 4D values obtained via DIM were significantly lower (Bonferroni-corrected P < 0.001) than those for SPM, which was exactly the opposite of what Manning et al. (2005) found. All other methods yielded very comparable average levels, which furthermore dovetailed with previous evidence showing that the average right 2D : 4D of male Austrians is in the 0.960s (Fink et al., 2004; Voracek et al., 2005), thus attesting both to the typicality of our sample and the validity of measurements from hand outline drawings. Further, measurement concordance (ICC) between established methods (DIM and SPM) was 0.890, whereas DIM was less concordant with DHM (0.694) and VHM (0.754) than SPM with DHM (0.871) and VHM (0.903).

In all, the results suggest that the 2D : 4D levels obtained by Bang et al. are unlikely due to biased 2D : 4D measurements from hand outline drawings. Their finding of a very high, feminized 2D : 4D in Danish men is interesting in itself and worth further investigations.

On a final note, however, we have come to believe that the studies, including Bang et al. on the association of 2D : 4D, taken as a proxy for prenatal testosterone exposure, with adult circulating testosterone, so far have been less than optimal methodologically. Apart from Bang et al., four further studies (McIntyre et al., 2003; Neave et al., 2003; Manning et al., 2004b; Kempel et al., 2005) failed to replicate, and only one (Benderlioglu and Nelson, 2004) replicated the original finding of Manning et al. (1998) of a significant negative association of 2D : 4D with adult circulating testosterone in men. All of them relied on single measurements of testosterone, mostly taken in the morning. It is known that in men, testosterone is secreted in spurts into the bloodstream, thereby causing measured testosterone levels to change noticeably within just a few minutes. The circadian rhythm of testosterone secretion does not only bring the highest, but also the most variable testosterone levels in the morning hours. This in tandem produces intraindividual reliabilities of measured testosterone, hovering as low as about 0.50 (Mazur and Booth, 1998). In general, associations of 2D : 4D with its theorized correlates are of small size (Manning, 2002), and unreliability of measurement inevitably attenuates effects. Ensuring high reliabilities not only in 2D : 4D measurements but also in the criterion variables therefore matters greatly in this research area. We welcome the opinions of Dr Bang and colleagues to the arguments and empirical evidence presented here.