Total serum testosterone levels at each of the peaks rise to just below or within the range of concentrations in adult men. Sex hormone binding globulin concentration is very low at birth, but rises rapidly, such that the unbound level of testosterone is high at the first postnatal peak but low at the second postnatal peak [7]. This is reflected in the high concentrations of testosterone in the saliva of one-week old infants, followed by low concentrations during the infant peak detected in blood [12,13]. Conversely, concentrations of dihydrotestosterone are low in the first weeks after birth but very high during the second postnatal peak [14].

Given the complexities of early effects of testosterone on sex differentiation discussed above, and further complexities associated with the development of the short bones of the fingers, this review evaluates evidence bearing on the broader hypothesis that digit ratios, including 2D:4D, are influenced by perinatal androgen action, and if so, whether the effects are large enough and specific enough to merit their use as biomarkers. The word "prenatal" will be used where effects can be isolated as such, or if not, "perinatal" will be used. Moreover, in some cases it will be possible to refer specifically to "testosterone" or more generally to "androgens." While most research has focused specifically on 2D:4D, as will this review, I will also broaden the discussion to "digit ratios" where appropriate. The review continues with (1) a description of digit ratio measures, including methods of measurement, measurement error, and the development of sex differences, followed by (2) consideration of the potential involvement of perinatal androgens in the development of observed sex differences, then (3) evaluation of each established technique for studying effects of perinatal androgens, its applications to the validation of digit ratio measures, and the potential roles of digit ratio measures to complement the established approach, given their respective advantages and disadvantages for different types of research.

The introduction of digit ratios as possible markers for prenatal androgens [17] follows from the very long-established observation that adult men have longer ring fingers (fourth digits) than adult women relative to the lengths of other fingers [18-22]. Methods of digit measurement have followed standard protocols, with most studies measuring on the skin surface, from photocopies, or from digital scans, and a few using left side hand-wrist radiographs originally obtained to assess bone age. Most researchers have focused digit ratio research specifically on 2D:4D, with a few exceptions, including authors also measuring other sex-dimorphic ratios like middle-to-ring finger, 3D:4D [23-25], and researchers using an older method of assessment, distal extent [22]. In each case, female values are higher. For 2D:4D measured on the skin surface, lengths of each finger are determined from the most proximal flexion crease (as two flexion creases often appear at the base of the fourth digit) to the fingertip. The intra- and inter-observer repeatability of these measures have been established by a number of researchers employing 2D:4D [26].

In adults, the point-biserial correlation between sex and right-hand, skin-surface 2D:4D is not high, even in racially homogenous samples (r = 0.22, ~60% overlap between male and female distributions [17]). Furthermore, the large population and racial differences observed in both adults [27-29] and children [25,30] could introduce serious confounding. In mixed race samples, race accounts for more variation in 2D:4D than does sex. African-descended populations have lower 2D:4D at all ages, but substantial differences have also been reported among national or ethnic groups usually considered racially similar [28]. However, the magnitude of sex differences reported in right-hand, skin-surface 2D:4D have been similar in different populations, so sex differences appear to be independent from population differences. The problem of racial confounding can be avoided by careful study design, including collection of homogeneous samples and comparisons within families. Alternative digit ratios, like 3D:4D, might also reduce racial confounding [25].

The size of the sex difference in right-hand 2D:4D seems to be smaller in children than adults (r = 0.15, ~90% overlap between boys and girls [17]). However, a longitudinal study using left-hand, radiograph-measured 2D:4D found larger sex differences at 9 years-old (r = 0.23) than at 17 (r = 0.18) [24]. Sex differences in 2D:4D were shown to be unaffected by puberty using a cross-sectional design [17]. Serial measurements have corroborated the finding [24]. However, it is now well established that 2D:4D increases during childhood in both sexes [24,25,30,31]. The magnitude of the change is small, and in children older than five years old is certainly smaller than the sex difference, but between ages one and five the increase is greater than the sex difference. Age should be considered in planning studies with children.

The simplest understanding of estrogen-receptor-mediated effects on metaphyseal growth is that they promote fusion of growth plates by the accelerated degeneration of hypertrophic chondrocytes [35]. In addition, sex steroids, again apparently through an estrogen pathway [36], produce the growth spurt associated with the early stages of puberty. The growth spurt seems to result from the systemic role of sex steroids in the increasing secretion of growth factors, especially IGF-1, during puberty [39]. Growth-inhibiting effects of estrogens seem to be dependent on their growth-promoting effects, in that hypertrophic chondrocytes are stimulated by high levels of estrogens, become exhausted, and degenerate as a result [35]. This effect of estrogens on bone growth plays a key role in the normal pattern of linear growth during puberty, including growth in stature. Men attain a higher stature (and generally longer fingers) than women through hypermorphosis, an extended period of growth owing to later onset of the pubertal growth spurt, not through a higher rate of growth. The few men who have been identified with either estrogen receptor deficiency or aromatase deficiency present with continued growth in stature far beyond sexual maturation, and low bone mineral density [40].

Sex steroid effects on anlagen or metaphyseal growth could influence observed differences in relative finger bone lengths in any of a large number of ways, including (1) differences among bones in sex steroid sensitivity, or (2) differences among bones in the sequence of growth processes, overlaid on temporal fluctuations in testicular activity. Sex differences in digit ratios could arise if bones from different fingers are differentially receptive to sex steroids (because they differ in receptor activity, aromatase activity, or present different conditions for interaction between steroid complexes and growth factors). Sex differences could also arise even if the bones of different fingers have similar responses to sex steroids but differ in their temporal patterns of growth. Differences among bones in the timing of periods of peak growth or onset of chondrification or ossification, together with changing sex differences in steroid production, would result in different effects of sex steroids, despite similar patterns of sensitivity and exposure among bones. Differences in the timing of growth events have been extensively documented, both among short bones of the hand [41], and between the sexes [42,43].

These potential explanations are not exhaustive. The effects of perinatal steroids might predispose tissues to divergent patterns of development in later periods. Whether the role of perinatal androgens in promoting sex differences in digit ratios are more direct or indirect could have substantial bearing on the validity of 2D:4D as a specific marker for perinatal androgens.

One study found that 13 females and 7 males with postnatally-treated CAH had lower 2D:4D (but only on the right hands of females and left hands of males) than 44 female and 28 male relatives unaffected by CAH [82]. Another study found that 27 females and 9 males with postnatally-treated CAH had lower 2D:4D (on both hands of females and only the right hands of males) than 52 female and 52 male age-matched controls unaffected by CAH [83]. However, a study employing 2D:4D measured on radiographic films of the left hand failed to find a significant difference between 66 CAH females and 69 age-matched control females and 77 males, though 2D:4D was intermediate between means for unaffected males and females [31]. All three studies used subjects of widely varying ages, but two studies matched cases and controls by age. Given that 2D:4D levels increase with age in both male and female children, it is possible that additional error introduced by using subjects of such varying ages, non-consanguineous controls, or left-hand measures could account for the different results.

Although [31] used the largest number of cases, their design was otherwise less ideal than the two studies finding effects, as they included subjects from ages 1 to 20, encompassing the full period of growth in 2D:4D, apparently matched some cases and controls with age differences greater than one year, and measured only left-hand 2D:4D. [83] included newborns to 13 year olds, also a problematic range of ages, but matched age to within two weeks, and performed analyses separately in subjects younger and older than two years of age. [82] did not use age matching, but all subjects were older than five, by which time 2D:4D is largely established and sex dimorphic, and cases were matched with related controls, further increasing detection power.

Higher 2D:4D has been associated with more CAG elements [86], further supporting a direct association of sex differences in digit ratios with androgen action. As 2D:4D was assessed among adults in this study, effects cannot be definitively isolated to the perinatal period, but given the developmental pattern of 2D:4D, pubertal period effects are unlikely. A study relating CAG repeats and 2D:4D among children would serve as more conclusive evidence for perinatal effects of androgens.

(1) Sex differences in digit ratios develop early in childhood and are unaffected by pubertal growth. This seems to be firmly established.

(2) High amniotic testosterone-to-estradiol ratio predicts low 2D:4D by two years old. While current evidence is supportive, further studies using amniotic testosterone or other direct methods should be conducted.

(3) CAH affected girls have lower 2D:4D than unaffected controls. This observation is not firmly established and further studies could help to resolve inconsistencies in the literature.

(4) The CAG repeat polymorphism of the androgen receptor gene influences 2D:4D as predicted. Further studies could help to confirm this finding.

Questions remain that might also be answered through animal or in vitro experimentation, provided that the utility of particular models can be established. For example, which periods of androgen exposure during the perinatal period produce observed sex differences in digit ratios? How do the development of the right and left sides differ to yield such varying results (even among laboratory mice)?