Nature of study area and participants: The present cross-sectional study was carried out among 500 unrelated individuals (250 males and 250 females) belonging to the Rajbanshi population of West Bengal. The Rajbanshi caste group has been selected for the present study due to its numerically larger strength, and the fact that they also appear to satisfy the condition of moderate numerical size as compared to the surrounding populations [10,14]. The individuals selected for the present study were identified as belonging to the Rajbanshi population by observing their physical features, cultural features and by recording their surnames. This data was subsequently verified from the official records of the Gram Panchayat (a local village level governing authority). The Rajbanshi individuals who participated in the present study were unrelated to one another and were the residents of a homogenous Rajbanshi-inhabited village named ‘Kachuaboalmari’ located in Nandanpur Gram Panchayat, Sadar Block, Police Station Kotwali, District Jalpaiguri, West Bengal, India and were aged between 18 and 60 years (mean age of males = 41.63±11.60 years, mean age of females = 37.67±11.49 years). By unrelated individuals, it is meant that consanguine relatives of them such as brother-brother, sister-sister, mother-son, mother-daughter, father-son and father-daughter were not included in the study. The participants were selected using a stratified random sampling method. The participants were free from any physical deformity of hands and fingers. The nature of the study was explained to the participants in detail and an informed consent was taken from each of them prior to recording their measurements.

Furthermore, a total of 200 (100 males and 100 females) adult individuals belonging to the tribal Karbi population of Assam aged between 18 and 60 years were included to validate the accuracy of the statistical models that were derived in the present study. They were also selected using a random sampling method from two villages namely Rongriso and Plimplamlamgso located in the district Diphu Karbi-Anglong. The present study has been conducted in accordance with the ethical standards of human experiments as laid down in the Helsinki Declaration of 2000 [26]. Necessary permissions and clearances of the study protocol were obtained from the Department of Anthropology, University of North Bengal, the Nandanpur Gram Panchayat, Sadar block of Jalpaiguri district of West Bengal, the village level governing authorities of Rongriso and Plimplamlamgso villages of Karbi-Anglong, Assam.

Anthropometric measurements recorded: The anthropometric measurements were recorded following the procedures of Singh and Bhasin [27]. The measurements recorded were length of index finger of left hand (LIFL), length of index finger of right hand (RIFL), length of ring finger of left hand (LRFL) and length of ring finger of right hand (RRFL). The IFL and RFL were measured as the linear distance between the midpoint of the proximal most flexion crease of the base, and the anterior most points of the index and ring finger (tip) in the midline on the palmer surface respectively [Table/Fig-1]. A sliding caliper was utilized to measure the lengths of the fingers. Care was taken to observe that there was no abduction or adduction at the wrist joint. The index and ring finger ratio was calculated for both hands by dividing IFL by RFL (IFL÷RFL). All the anthropometric measurements were recorded to the nearest 0.1 cm. The technical error of measurement (TEM) was determined to check the consistency and reliability for the intra-observer and inter-observer differences in connection with the measurements [28,29]. TEM was calculated as TEM= √ (∑D²/2N), where ‘D’ is the difference between the measurements, and ‘N’ is the number of individuals measured. For estimating TEM in the present study, LIFL, RIFL, LRFL and RRFL were measured in 30 randomly selected Rajbanshi adult individuals other than those included in the present study by two of the authors (AG and NM). The coefficient of reliability (R) (R= {1−(TEM)²/SD²}, SD= standard deviation of the measurements) were subsequently determined from the measurements [Table/Fig-2]. Very high values of ‘R’ (>0.96) were obtained for all the anthropometric measurements and these were appreciably higher than the cut-off value of 0.95 [29]. Thus, the anthropometric measurements obtained in the present study by AG and NM was reliable, reproducible and free from any observer bias, and TEM was not taken into consideration for further statistical analysis. All the measurements for the present study were subsequently recorded by AG. The systematic errors due to the shift in the style or landmark interpretations or between instruments (e.g., sliding caliper) were also tested using the standard procedure [30]. Similarly, the differences in the anthropometric measurements were evaluated with repeated measures and sessions using one-way ANOVA and the mean differences were observed to be statistically not significant (p>0.05).

[Table/Fig-1]:

Landmarks of anthropometric measurements of IFL (A-B) and RFL (C-D)

The statistical analysis was done using the Statistical Package for Social Sciences (SPSS version 17.0, Chicago, Illinois, USA). A p-value of less than 0.05 was considered to be statistically significant. One-way analysis of variance (ANOVA) based on the general linear model was performed to compare sex differences in IFL, RFL, and index-ring finger ratio (left and right) between sexes. Two-way ANOVA using general linear model was also performed to control the influence of age-sexes on IFL, RFL and index-ring finger ratio (left and right). The bilateral differences were also evaluated between measurements in right and left hands using ANOVA. The study sample (n=500) was randomly split into two groups based on random number table; the first group comprised of 400 participants (200 males and 200 females) on which the statistical analysis was performed to derive sex discriminating models and the second group of 100 participants (50 males and 50 females) was designated as ‘hold out sample’ on which the validity of derived models were tested. The analysed group and test group were balanced for ethnic origin/age and sex variations. The binary logistic regression (BLR) analysis was employed to derive the predicting models for estimation of sex from IFL and RFL. The accuracy of models derived using BLR analysis from the 400 participants was tested on the hold out study sample from the same population. The accuracy of derived models were also tested on the other sample drawn from a different unrelated population group (in this case Karbi) comprising of 200 participants (100 males and 100 females) which was designated as ‘test group’. These samples were taken into consideration using the random sampling method to test the accuracy of the derived models in BLR analysis. Sectioning point in logistic regression analysis was 0.5. In BLR analysis for estimation of sex; all scores greater than 0.5 for the derived value of function (y) were classified as female, and scores below 0.5 as males. The correct prediction percentage was thus, obtained on the hold out sample and the other unrelated test sample. Besides, drawing prediction percentage based on BLR analysis, Receiver Operating Curve (ROC) analysis was performed on the predicting probabilities obtained from BLR analysis. Area Under the Curve (AUC) was considered as the predictive accuracy of IFL and RFL in sex estimation.

The descriptive statistics of IFL and RFL among male and female individuals in the present study are depicted in [Table/Fig-3]. The IFL and RFL lengths were observed to be significantly longer (p<0.001) among males than females in both right and left hands. The sex differences in finger length were observed to be statistically significant (p<0.001). Differences in lengths of LIFL and LRFL were observed to be higher than RIFL and RRFL. The ring finger was observed to be longer than index finger among both sexes [Table/Fig-3]. Statistically significant side differences were not observed between right and left hands for IFL among both sexes and RFL in females (p<0.05) [Table/Fig-4]. The descriptive statistics of IFL and RFL ratio among males and females in the present study are shown in [Table/Fig-5]. The mean values of IFL and RFL ratio were identical in left and right hands among both sexes. The sex differences in IFL and RFL ratio were not observed to be statistically significant (p>0.05) on right and left sides [Table/Fig-5].

A significant correlation was observed between sex and the different variables analysed. Predicting models derived for estimation of sex from IFL and RFL using BLR analysis are shown in [Table/Fig-6]. It is evident that the univariate and bivariate models derived for IFL and RFL estimate sex with mean correct prediction percentage ranging between 70.0% and 75.0% in the hold-out group and between 61.0% and 66.0% in the test group. Mean Correct Prediction Percentage (CPP) was thus, higher for the hold out group belonging to the same population from which the regression models were derived than the test group that constituted of a different population group from the region. CPP in the hold-out group was higher for IFL than the RFL in both hands [Table/Fig-7]. Based on the Receiver Operating Curve (ROC) analysis [Table/Fig-8], LIFL (79.5%) and LRFL (79.6%) shows greater potential for discriminating sex over RIFL (77.1%) and RRFL (78.1%). The IFL and RFL together [Table/Fig-9] have an increased potential of discriminating sex in the left (80.5%) and right hands (78.5%).

[Table/Fig-8]:

ROC analysis indicating the predictive probability from IFL (A) and RFL (B) on both sides for estimation of sex

[Table/Fig-9]:

ROC analysis indicating the predictive probability from combined IFL and RFL of Left (A) and Right (B) sides for estimation of sex

The morphological sex difference in the absolute length of the finger has been reported along with the fact that male fingers were significantly longer as compared to female fingers [19,24,25,31,32]. In the present study, RFL was observed to be longer than IFL among both hands of males and females. Besides, IFL and RFL of both hands of males were significantly longer than those of female individuals (p<0.05). Several researchers have reported the existence of significant sex differences in RFL and IFL in different ethnic populations [18,19,24,25,33]. Differences in finger length dimensions (e.g., IFL and RFL) seem to be indicative of a size-based difference along with a shape-based difference between sexes [19,33]. The distribution of finger length measurements showed the extent and range of overlapping in male and female values for IFL and RFL on right and left side among the Rajbanshi individuals. The present study has observed statistically insignificant bilateral variations in RFL and IFL among them. Bilateral differences in the IFL and RFL could influence the standards derived in the present study and the findings obtained from one side thus, may not be applicable to the other owing to side differences in both sexes among the study population. A number of studies have reported bilateral variations in IFL and RFL among individuals belonging to different Indian ethnic populations [18–20,24,25].

The index and ring finger ratio was observed to be similar among both sexes for right and left hands (p>0.05). Considerable overlapping in the frequency distributions of index and ring finger ratios were observed among both sexes in the study population. Finger ratio is generally considered to be a sexually dimorphic trait that is independent of body size and the ratio is not significantly related to height and age of the either sexes [19,25,33]. Studies have observed IFL and RFL ratio as a sexually dimorphic trait that has been established early in life and remains fairly stable postnatal and remained consistent with age and growth in a population [22,23,34]. The sexual dimorphism in IFL and RFL ratio develops between the 13th and 14th week of gestation [35,36] under the influence of parental androgens and estrogens [37,38]. In general, a lower IFL and RFL ratio has been reported among females [19,33,39,40]. Manning reported that IFL was generally about 96.0% of RFL [40]. So the average IFL and RFL ratio for males would be 0.96. Hence, index and ring finger ratio may serve as an important index for estimation of sex of an individual. Studies among different ethnic populations have observed that index and ring finger ratio can be used for sex estimation [18,19,24,25]. Its utility however, is shown to be limited in forensic case work and is shown to vary between different populations. In the present study, no statistically significant sex differences were observed in index and ring finger ratio which can be attributed to population variations.

A number of previous studies have utilized sectioning point analysis to determine sex of individuals. Sectioning points utilized in these studies however, was derived from mean male and female values. The present study uses a more robust BLR analysis in sexing and the sectioning point in logistic regression analysis are taken as 0.50. Accordingly, the results of the present study indicate LIFL, RIFL, LRFL and RRIL as sex discriminating variables among adult Rajbanshi and Karbi individuals. Here estimation of sex was performed using independent linear anthropometric measurements of hand dimension. The index and ring finger ratios were not taken into consideration in BLR models due to their derived and dependency in nature showed insignificant sex differences. Recent studies on sexing in anthropological casework have utilized ROC analysis to predict the accuracy of variables in sexing based on AUC [19,41–43]. Our observations based on AUC curve indicate that predictive probability increases when IFL and RFL are taken together when compared to IFL and RFL taken individually. Predicted probabilities are observed to be higher on left than on right side, and LIFL and LRFL appear to be better predictors of sex than RIFL and RRFL in estimation of sex of the adult Rajbanshi individuals. Our observations with regard to predictive probabilities of IFL and RFL together are similar to that reported from a North Indian population [19]. It is evident that the BLR models derived in the present study on the Rajbanshi population show a higher prediction percentage when applied on the similar adult Rajbanshi population in the hold out group than on the test group of different population comprising of tribal Karbi individuals. This emphasizes the fact that regression models derived in a population group should ideally be applied to the same population.

Limitations of the study are related to the fact that the study was conducted on live adult population. The findings of the present study thus, should not be applied on children, adolescents or elderly individuals. The dimensions of finger lengths are likely to alter after death with rigor setting in or with putrefactive changes occurring later. Thus, the observations of the present study can be applied only in cases of dismemberment in which the human remains are relatively fresh and do not exhibit postmortem changes that can alter finger lengths. The findings may not be applicable for desiccated, decomposed, and bloated bodies that affect hand dimensions.

Estimation of sex from the IFL and RFL measurements is a supplementary approach when extremities or other body parts are not available for examination. The present study has highlighted the application of IFL and RFL to determine sex among individuals belonging to the Rajbanshi population of North Bengal, India. It is suggested that finger lengths are relatively useful predictors of sex and both IFL and RFL could be utilized to estimate sex of the individuals when more reliable means of sex estimation are not available during medico-legal investigations. This study further observes that the models derived for a population group have better applicability when applied on the same population. Studies on estimation of sex from finger lengths in different age groups and among different populations need to be encouraged.