A New Method to Assess Asymmetry in Fingerprints Could Be Used as an Early Indicator of Type 2 Diabetes Mellitus

Abstract

Background:

Inexpensive screening tools are needed to identify individuals predisposed to developing diabetes mellitus (DM). Such early identification coupled with an effective intervention could help many people avoid the substantial health costs of this disease. We investigated the hypothesis that fluctuating asymmetry (FA) in fingerprints is an indicator of type 2 diabetes mellitus (T2DM).

Methods:

Participants with T2DM, with T1DM, and without any indication or known family history of diabetes were fingerprinted with a Crossmatch Verifier 320 LC scanner. Asymmetry scores for each finger pair were assessed using both pattern analysis (ridge counts), and a wavelet-based analysis.

Results:

Both methods for scoring asymmetry predicted risk of T2DM for finger pair IV, controlling for gender and age. AUC scores were significantly greater than the null for pattern asymmetry scores (finger IV AUC = 0.74), and wavelet asymmetry scores for finger pair IV (AUC = 0.73) and finger pair V (AUC = 0.73), for predicting T2DM. In addition, wavelet asymmetry scores for finger pair IV (AUC = 0.80) and finger pair V (AUC = 0.85) significantly predicted risk of T1DM.

Conclusions:

A diagnostic tool based on FA in the fingerprints of finger pair IV, measured using a wavelet analysis could be developed for predicting risk prior to associated health problems for both T2DM and T1DM. In addition, given that that the prints for fingers IV and V develop during the 14-17 weeks of gestation, we predict that interventions during this time period of pregnancy will be most successful.

Keywords

asymmetry diagnostics fingerprints risk T2DM T1DM

Type 2 diabetes mellitus (T2DM) is a long-standing metabolic disease that results in early complications from decades of maladaptive physiological processes. One-third of people diagnosed with T2DM have complications at the time they are diagnosed, with the duration of hyperglycemia being directly related to the extent of the complications.^1,2 Even before insulin deficiency manifests as hyperglycemia and a diagnosis of T2DM is made, pancreatic beta cells decline has already occurred.³ Aggressive lifestyle intervention can delay or prevent T2DM in those at high risk.^4-6 Therefore, the sooner it is possible to detect who is at risk for developing diabetes, the sooner the necessary preventive measures can be taken to prevent or delay the serious complications related to the disease and improve health outcomes.^1,2

Diabetes is a complicated disease with multiple genetic and environmental influences. Models for determining “Risk for T2DM” can be quite successful,⁷ but include phenotypic characteristics that are often manifested after the development of hyperglycemia (ie, high BMI, waist circumference). Models of risk for T2DM that incorporate genetic variants can provide insights into risk prior to the development of associated health problems,^8,9 however, the genetic polymorphisms that have been identified still account for only a small proportion of the variation that is present (ie, less than 10% of cases). In addition these methods remain costly and hard to apply to the general population.

We tested the hypothesis that fluctuating asymmetry (FA) in fingerprints in humans will be an “early” predictor of an individual’s propensity to develop Type 2 Diabetes (T2DM). Similar to diabetes,^10-12 fingerprints are influenced by both genetics and the gestational environment.^13,14 Developmental studies have determined that fingerprints form during early gestation (6-17 weeks),¹⁵ and thereafter remain unchanged.¹³ Fingerprints have been previously used in the diagnosis of several different medical disorders,^16,17 including diabetes.^18-20 For this study, we specifically focused on FA in fingerprints, or random asymmetry about a zero mean value,²¹ calculated as differences in fingerprints between the corresponding fingers on the right and left hands. FA has been hypothesized to be a particularly sensitive indicator of developmental instability due to increased growth rates²² and of degenerative diseases.²³ FA in some aspects of fingerprints has also been found to be associated with T2DM.²⁴ An additional benefit to considering fingerprints as an indicator of T2DM risk is our knowledge of the timing of the development of each finger pair during gestation. The ridge pattern of fingerprints starts developing with the thumb (7-8 weeks) and progresses toward the little finger (16-17 weeks).^13,15 Given the current uncertainty about precisely when during gestation the environment impacts the metabolism of the adult,²⁵ comparison of FA across fingers could help identify when preventive measures during pregnancy would be most beneficial.

The current study had 4 goals. First, we tested the hypothesis that fingerprint patterns of individuals with T2DM would be more asymmetrical than individuals without T2DM, controlling for gender and age. Second, we wanted to determine if wavelet analysis, a technique already used in forensics for fingerprint archival and matching, but not in previous studies of fingerprints as disease markers, would give results similar to the traditional ridge count or pattern analysis. If wavelet analysis could be used to assess fingerprint asymmetry, the potential for using fingerprints to identify individuals at risk for T2DM increases. Third, we tested the hypothesis that some stages of gestational development are more important in the onset of T2DM. To this end, we compared the degree of FA of the different fingers to determine which finger was most predictive of DM. And fourth, we examined the possibility that fingerprint asymmetry might be predictive of T1DM as well. Even though T1DM is a considerably different disease than T2DM, like T2DM, T1DM is influenced by multiple environmental and genetic factors,²⁶ including a relationship between environmental exposures in utero,²⁷ and is associated with dermatoglyphic features.²⁸

Methods

Participants

Participants were patients either under the care of physicians at the UMA Diabetes/Endocrine Center or the Family Medicine Center in Athens, Ohio. The participants were a convenience sample, approached at the time of clinical care.

Individuals were included in the control group if they did not currently have indications of diabetes, had no personal or family history of diabetes or any insulin resistant syndrome (such as polycystic ovarian syndrome, metabolic syndrome). This data was obtained both by participant interview and confirmed by physician chart review. We also used the patient’s medical records to exclude those individuals with chromosomal syndromes (Down, Turner, Klinefelter), monogenic disease (cystic fibrosis), polygenic morbidity (cleft palate and cleft lip with or without cleft palate), and females suffering from endometrial carcinoma or carcinoma of cervix, all known to be correlated with fingerprint asymmetries.¹⁷ T2DM and T1DM individuals had been diagnosed with fasting glucose and AC1 tests, respectively.

Data Acquisition and Processing

Rolled fingerprints were collected using a Crossmatch Verifier 320 LC scanner at a resolution of 500 ppi and 256 level grayscale. The prints of all fingers on both hands were stored as uncompressed digital images (bmp file format) on a laptop dedicated to the project with an encrypted storage device. Fingerprints were scored for similarity between corresponding fingers on each hand (symmetry) using both ridge counts (pattern analysis) and a wavelet based method. Pattern analysis is the method that has been used in previous studies of fingerprints and FA. This method generates ridge counts that are then compared between prints (see Figure 1). We used the software package RIDGECOUNTER.²⁹ For fingertips with a loop pattern, the ridge count equal to the number of ridges crossing the single straight line between the core and the triradius, and for fingertip patterns with 2 triradial points (whorl and double loop pattern), ridge counts equaled the counts crossing both the lines. We used the absolute difference between the scores for the homologous fingers as the asymmetry score. As fingertips with an arch are standardly assigned a ridge count of zero, finger pairs with 2 arches would have produced an “artificial” score of symmetry, and with 1 arch, and inflated score of asymmetry. Therefore, we did not include arch fingerprints in the pattern-based data.

Figure 1.

(A) Two loop fingerprints (top and middle) and 1 whorl patterned fingerprint (bottom) with lines between core and triradius along which the ridge counts were taken. The Δrc values are the respective difference in ridge count numbers determined for these prints. (B) For wavelet analysis, images of the prints are cropped around the core point to a size of 64 × 64 pixels. Haar wavelet decompositions are computed for each of 4 image quadrants to obtain a 36-element feature vector (afv, bfv, and cfv), which represents the global pattern of the fingerprint. Differences between the feature vectors are calculated as Euclidian distance and shown here as Δe. Note how this method better captures the subjectively larger difference between the fine lines of the top fingerprint and the coarse pattern of the one on the bottom.

There are some clear limitations to the pattern based method of assessing asymmetry. It is a very coarse measurement (only about 30 different values), and depending on the fingerprint, the count can be fairly subjective, easily varying by ±2.²⁹ However, we wanted to compare the asymmetry scores from the more sophisticated wavelet based analysis to those from the pattern analysis that has been used in previous studies.

For wavelet analysis, we processed a 256 pixel square subimage derived from the center of each print to calculate a feature vector (based on Haar wavelets) using MATLAB and the Wavelet Toolbox extension. Several related techniques have been proposed for fingerprint matching in forensics (Gabor filters, discrete wavelet transform, and Hartley transform).^20-32 All these methods have in common that they analyze the fingerprint on a global level (ie, treat the print as a texture with features such as ridge density or uniformity), rather than focusing on local details. While the different methods stress different features, they all compute a less complex description of the print, called a feature vector. This feature vector can then be compared with a second print, for example, by calculating the Euclidean distance³³ (see, eg, distances between 36 element feature vectors in Figure 1). Wavelet-based methods have several advantages over pattern analyses. First, they detect global (overall) differences and make very few a priori assumptions about the nature of the “trait” they need to detect. Second, unlike pixel-by-pixel image comparison, wavelet based methods are mostly immune to variation in acquisition, that is, finger placement on the scanner. Third, they provide a similarity score that we can use as a score of symmetry. While these techniques have been used to archive and match fingerprints for forensics, they have to our knowledge never been applied to measure symmetry.

Statistical Analyses

An age relationship is expected for T2DM as the risk for developing T2DM increases with age. In addition, females are less likely to develop diabetes at the same BMI that males develop diabetes.³⁴ Therefore, we controlled for both age and gender in our analyses. We conducted multinomial logistic regressions for each finger pair to determine if asymmetry scores could be considered a risk factor for diabetes, and to determine which fingers were the most predictive (highest coefficient). Parameters entered into all models in addition to the asymmetry scores included age, gender and an interaction between age and asymmetry scores. The discriminative accuracy of the asymmetry scores for each finger pair was also assessed using the area under the receiver operating characteristic curves (AUCs).

Results

Participants

A total of 340 individuals were finger printed, 83 were classified as controls (average age = 31.3 years ± 16.8), 200 had been diagnosed with T2DM (average age = 59.2 years ± 13.59) and 57 with T1DM (average age = 42.4 ± 16.93). In this sample there were 133 males (26 = controls; 78 = T2DM; 29 = T1DM) and 207 females (57 = controls; 122 = T2DM; 28 = T1DM).

Ridge Counts (Patterns) to Determine Asymmetry

Individuals that did not have a complete set of prints (all 10 fingers) that could be scored for ridge counts, due to either poor quality prints or an arch type print) were removed from the pattern analysis. We collected ridge count data for 85 females and 51 males. Of these individuals, 44 were classified as “controls,” 62 with T2DM, and 21 with T1DM. In the multinomial regression analyses of each pair of fingers, asymmetry scores were only significant in the model for finger pair IV (Table 1A). Gender was not significant for any of the finger pairs, however a significant influence of age was detected for all finger pairs except II, and an interaction between age and asymmetry score for finger pair IV (Table 1A). These results suggest that some of the younger controls may not be “true” controls (ie, individuals with a high risk of developing diabetes, but that have not yet developed the disease). Therefore, to remove the potential bias of “false” controls, we split the data set into 2 groups (>40 years of age and <40 years of age) and conducted the receiver operating characteristic (ROC) analyses on the subset of individuals over 40 years of age. The AUC score for the subset of individuals over 40 for finger IV was the most predictive, and was significantly better than the null (Table 2A).

Table 1.

Multinomial Regression Analyses for Diabetes State by Finger Pair: (A) T2DM as Compared to Controls, Using Ridge Counts to Assess Asymmetry; (B) T2DM as Compared to Controls, Using Wavelet Analysis to Assess Asymmetry; (C) T1DM as Compared to Controls, Using Wavelet Analysis to Assess Asymmetry.

(A)	Asymmetry score (ridge count)		Gender		Age		Age × asymmetry score
Finger pair	B	P	B	P	B	P	B	P
I (97)	−0.026	.88	0.323	.60	−0.145	.0001	0.001	.90
II (90)	0.638	.12	−0.290	.97	−0.069	.06	−0.021	.05
III (87)	0.653	.30	0.306	.65	−0.107	.006	−0.018	.24
IV (95)	0.921	.05	0.221	.76	−0.074	.02	−0.027	.03
V (96)	0.043	.90	0.254	.67	−0.139	.0001	−0.001	.99
(B)	Asymmetry score (wavelet)		Gender		Age		Age × asymmetry score
Finger pair	B	P	B	P	B	P	B	P
I (136, 72)	0.012	.10	0.070	.88	−0.050	.21	0.0001	.03
II (142, 78)	0.005	.53	−0.109	.81	−0.089	.018	0.0001	.35
III (151, 81)	0.013	.11	0.085	.85	−0.049	.13	−0.001	.02
IV (141, 74)	0.027	.003	−0.70	.88	−0.002	.95	−0.001	.0001
V (102, 63)	0.030	.006	0.091	.87	0.006	.87	−0.001	.0001
(C)	Asymmetry score (wavelet)		Gender		Age		Age × asymmetry score
Finger pair	B	P	B	P	B	P	B	P
I (47, 72)	0.003	.64	0.909	.03	−0.022	.54	0.0001	.27
II (47, 78)	0.007	.32	0.963	.02	0.011	.78	0.0001	.09
III (46, 81)	0.014	.05	1.025	.02	0.041	.23	−0.001	.007
IV (38, 74)	0.020	.02	1.140	.03	0.074	.09	−0.001	.002
V (36, 63)	0.038	.005	1.324	.02	0.17	.01	−0.002	.002

Sample sizes (DM, control) are in parentheses. Coefficient (B) and Wald significance (P) are reported for each parameter included in model. Statistically signficant P values are in bold.

Table 2.

Comparisons of AUC Scores for Each Finger in the Subset of Individuals Over 40: (A) Asymmetry Measured With Ridge Counts, Predicting T2DM; (B) Asymmetry Measured With Wavelet Analysis, Predicting T2DM; (C) Asymmetry Measured With Wavelet Analysis, Predicting T1DM.

(A) Pattern asymmetry scores	ROC analysis (over 40 years of age)
Finger pair (control, T2DM)	AUC	95% CI	Asymptotic sig.
Thumb I (7, 53)	0.46	0.275-0.638	.71
Index II (6, 47)	0.73	0.599-0.858	.07
Middle III (6, 51)	0.58	0.314-0.850	.55
Ring IV (7, 52)	0.74	0.433-0.910	.04
Pinky V (7, 53)	0.46	0.272-0.641	.71
(B) Wavelet asymmetry scores	ROC analysis (over 40 years of age)
Finger pair (control, T2DM)	AUC	95% CI	Asymptotic sig.
Index I (9, 121)	0.56	0.340-0.784	.54
Second II (12, 128)	0.54	0.355-0.71	.68

Middle III (13, 13)	0.65	0.492-0.800	.08
Ring IV (11, 127)	0.73	0.593-0.85	.01
Pinky V (10, 90)	0.73	0.587-0.875	.02
(C) WAVELET asymmetry scores	ROC analysis (over 40 years of age)
Finger pair (control, T1DM)	AUC	95% CI	Asymptotic sig.
Index I (9, 20)	0.53	0.281-0.786	.78
Second II (12, 22)	0.64	0.421-0.859	.18
Middle III (13, 20)	0.72	0.538-0.908	.03
Ring IV (11, 18)	0.80	0.638-0.958	.008
Pinky V (10, 15)	0.85	0.703-1.0	.003

Sample sizes for each finger pair are provided in parentheses in the first column. Statistically significant P values are in bold.

Wavelet Analysis of Asymmetry

More of the prints could be analyzed using wavelet analysis than the ridge count analysis, as the wavelet method is less reliant on having a clear print or a particular fingerprint pattern. In the multinomial regression analyses, in which we assessed the ability of asymmetry scores for each pair of fingers to predict T2DM diabetes state, asymmetry scores significantly predicted T2DM for finger IV and V (Table 1B). Gender was not significant for any of the fingers, however an interaction between age and asymmetry score was significant for the models of all but 1 of the finger pairs (Table 1B). Therefore, to remove potentially “false” controls (younger individuals that have a high risk of diabetes but have not yet developed diabetes) we analyzed only individuals over the age of 40 years in the ROC analyses. Area under the curves ranged from 0.54 to 0.73, and the AUC for finger IV and V were significantly greater than the null (0.5) (Figure 2, Table 2B).

Figure 2.

ROC curves showing the performance of the wavelet asymmetry scores in predicting T2DM for fingers (A) IV and (B) V, and for predicting T1DM for fingers (C) IV and (D) V. See Table 2 for AUC scores and 95% CIs.

Asymmetry scores were significant in predicting T1DM diabetes for finger pairs III, IV and V in the multinomial regression analyses (Table 1C). Gender was significant for every finger pair; a higher proportion of males with T1DM than females were detected in our sample. In addition, there were significant interactions between age and asymmetry scores for the models for all of the finger pairs except finger II (Table 1C). Therefore, we examined the ROC curves for the subset of individuals over 40 years of age. Area under the curves ranged from 0.53 to 0.85, and the AUC scores for finger pairs III, IV, and V were significantly greater than the null (0.5) (Figure 2, Table 2C).

Discussion

Assessing asymmetry of fingerprint scores has the potential to become a valuable new tool for health providers in their assessment of an individual’s risk of developing diabetes. Asymmetry scores calculated using both a wavelet-based method and the standard ridge count method, suggested that finger IV is the most predictive of T2DM in this population. Given that scoring asymmetry using the wavelet-based method produced more significant results than the scores obtained form standard pattern analyses (ridge count), we suggest that diagnostics incorporating asymmetry scores from the more easily obtained fingerprint scans can be developed that avoid the error and other limitations of pattern analyses. The AUC scores for finger IV based on wavelet scores (0.73) and pattern analyses (0.74) are lower than those from other diabetes risk models (eg, AUC = 0.85).³⁵ However, these previous models included traits that are themselves indicators of reduced health (eg, high waist circumference, hypertension). As compared to models based on amino acid profiles, which can predict risk of diabetes up to 12 years before the clinical onset of diabetes,³⁶ fingerprints can predict from birth and are fixed for life. And finally, compared to genetic testing, that can also predict as early as birth, AUC scores for the fingerprint asymmetry of finger pair IV that we detected are slightly higher than previously reported for genetic polymorphisms alone (AUC = 0.60, 95% CI = 0.57-0.63).³⁷

Understanding why asymmetry scores predict diabetes will be an interesting area of research, given that the general understanding of the genetic and developmental bases of FA has been somewhat illusive.^38,39 Initially, symmetry was thought to indicate the genetic ability of an organism to undergo identical development on both sides of the body in response to environmental stress.^40-42 Results from laboratory studies examining stress and FA, however, have been inconsistent.³⁵ Morris et al²² hypothesized that the inconsistencies in the relationships detected between FA and stress are due to differences across genotypes. This hypothesis suggests that there is variation both within and across populations in a genetic threshold, which determines when environmental stress is sufficient to shift the allocation of resources toward growth and away form accurate development. Recent empirical research supports this hypothesis.⁴³ We suggest that the differences we detected in asymmetry between individuals with T2DM and controls represent genetic differences in the responses of the genotypes to environmental stressors during development. Gaining understanding of these early steps in the pathophysiology of diabetes could be extremely impactful.

Asymmetry scores for finger IV were the most predictive of T2DM in this population, suggesting that the time period during gestation when the prints on these fingers are forming may be a critical time stage of development for increasing risk of developing diabetes. Given that prints on little finger (finger V) are the last to form at 16-17 weeks, our results suggest that the critical time is around 14-16 weeks. There were differences between the finger pair we detected as most predictive, and the pairs of fingers that were most predictive in a previous study.²⁴ In this previous study, finger pair V was most predictive for males and finger pair II for females. Ravindranath et al²⁴ did not control for age in their analyses, however the age of participants in their study ranged from 38-82 years, and therefore most closely matched our ROC analyses of individuals over 40 year of age. We did not assess the ability of asymmetry of finger pairs to predict T2DM in males and females separately. However, we did include gender in our multivariate models, and did not detect any significant influences of gender in the ability to predict T2DM for any of the finger pairs. One possible explanation for the differences in finger pairs that were predictive across the 2 studies could be population differences in the genotypes leading to T2DM (current study primarily Caucasians; previous study Indian).²⁴ Of the 70 or more distinct genomic regions that have been found to be association with T2DM, some examples of heterogeneity in effects across ethnic groups have been observed,^44-46 including an Indian population.⁴⁷ However, a larger sample size with additional statistical power is needed to address this question.

One of the most surprising results from our study was the association between asymmetry scores and T1DM, as well as the overall higher asymmetry for the prints of individuals with T1DM as compared to individuals with T2DM. The scores for AUC suggest that asymmetry scores for finger pairs IV and V would be discriminative for T1DM. While T1DM and T2DM are considered quite different diseases, they are both multifactorial diseases with a complex interaction between predisposing genetic and environmental factors. In addition, an increased frequency of T2DM diabetes has been detected in families with T1DM diabetes, which could suggest a common genetic susceptibility.^48-50 Finally, an increasing proportion of diabetic patients with diabetes may have both T1DM and T2DM processes contributing to their diabetic phenotype.⁵¹ Our results support studies suggesting that individuals predisposed to T1DM and T2DM are both experiencing environmental influences during gestation that increase risk of developing these diseases.

Conclusions

Our results suggest that fingerprint asymmetry could be developed into a valuable tool for predicting risk of T2DM and T1DM, and that wavelet analysis is a method that can be used to assess asymmetry in fingerprints. The advantage of fingerprints scored using wavelet-based methods over genetic testing, is that it can indicate gestational environment and would be much less expensive. The cost is important, given recent reports that both risk-aware and risk-unaware individuals were interested in genetic testing, but identified the need for low-cost tests.⁵² We suggest a more comprehensive analysis of fingerprint asymmetry as a predicator of both T2DM and T1DM risk, scoring asymmetry with wavelet analysis and comparing to the predictive ability of genetics alone, is warranted.

Footnotes

Acknowledgements

We thank Eric Landman, Melinda Ruberg, Anna Saltman, Casey Shubrook, Elizabeth Snow, Chris Swyers, and Kavya Yellermraju for assisting with collection of the fingerprints.

Abbreviations

AUC, area under the receiver operating characteristic curve; FA, fluctuating asymmetry; ROC, receiver operating characteristic; T1DM, type 1 diabetes mellitus; T2DM, type 2 diabetes mellitus.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The Ohio University Baker Fund provided financial support.

References

Engelgau

Narayan

KMV

Herman

. Screening for type 2 diabetes. Diabetes Care. 2000;23:1563-1580.

Harris

Eastman

. Early detection of undiagnosed diabetes mellitus: a US perspective. Diabetes Metab Res Rev. 2000;16:230-236.

Tabák

Jokela

Akbaraly

Brunner

Kivimäki

Witte

. Trajectories of glycaemia, insulin sensitivity, and insulin secretion before diagnosis of type 2 diabetes: an analysis from the Whitehall II study. Lancet. 2009;373(9682):2215-2221.

Diabetes Prevention Program (DPP) Research Group. The Diabetes Prevention Program (DPP): description of lifestyle intervention. Diabetes Care. 2002;25(12):2165-2171.

Laaksonen

Lindström

Lakka

. Physical activity in the prevention of T2DM: the Finnish diabetes prevention study. Diabetes. 2005;54:158-165.

Shubrook

Johnson

. An osteopathic approach to type 2 diabetes mellitus. J Am Osteopath Assoc. 2011;111:531-537.

Lindström

Tuomilehto

. The Diabetes Risk Score: a practical tool to predict type 2 diabetes risk. Diabetes Care. 2003;26:725-731.

Sanghera

Blackett

. Type 2 diabetes genetics: beyond GWAS. J Diabetes Metab. 2012;3(198):6948.

Drong

Lindgren

McCarthy

. The genetic and epigenetic basis of type 2 diabetes and obesity. Clin Pharmacol Ther. 2012;92:707-715.

10.

Arrieta

Criado

Martinez

Lobato

Gil

Lostao

. Fluctuating dermatoglyphic asymmetry: genetic and prenatal influences. Annals Hum Biol. 1993;20:557-563.

11.

Clausen

Borch-Johnsen

Pedersen

. Relation between birth weight and the insulin sensitivity index in a population sample of 331 young, healthy Caucasians. Am J Epidemiol. 1997;146:23-31.

12.

Dabelea

Pettitit

. Intrauterine diabetic environment confers risks for type 2 diabetes mellitus and obesity in the offspring, in addition to genetic susceptibility. J Pediatr Endocrinol Metab. 2001;14:1085-1091.

13.

Pechenkina

Benfer

Vershoubskaya

Kozlov

. Genetic and environmental influence on the asymmetry of dermatoglyphic traits. Am J Phys Anthropol. 2000;111:531-543.

14.

Kahn

Ravindranath

Valdez

Narayan

. Fingerprint ridge-count difference between adjacent fingertips (dR45) predicts upper-body tissue distribution: evidence for early gestational programming. Am J Epidemiol. 2001;153(4):338-344.

15.

Wertheim

Maceo

. The critical stage of friction ridge and pattern formation. J Forensic Identification. 2002;52:35.

16.

Cummins

Midlo

. Fingerprints, Palms and Soles: An Introduction to Dermatoglyphics. New York, NY: Dover; 1976.

17.

Kobyliansky

Bejerano

Katznelson

MBM

Malkin

. Relationship between genetic anomalies of different levels and deviations in dermatoglyphic traits. Stud Hist Anthropol. 2006;4:61-121.

18.

Kahn

Graff

Stein

Lumey

. A fingerprint marker from early gestation associated with diabetes in middle age: the Dutch Hunger Winter Families Study. Int J Epidemiol. 2009;38(1):101-109.

19.

Nilesh

Rakate Zambare Balbhim

. Fingertip patterns: a diagnostic tool to predict diabetes mellitus. Nat J Med Dent Res. 2014:46.

20.

Yohannes

Alebie

Assefa

. Dermatoglyphics in type 2 diabetes with implications on gene linkage or early developmental noise: past perspectives. Current Trends Future Prospects. 2015;3(1D):297-305.

21.

Van Valen

. A study of fluctuating asymmetry. Evolution. 1962;16:125-142.

22.

Morris

Rios-Cardenas

Lyons

Tudor

Bono

. Fluctuating asymmetry indicates optimization of growth rate over developmental stability. Functional Ecol. 2012;26:723-731.

23.

Weisensee

. Assessing the relationship between fluctuating asymmetry and cause of death in skeletal remains: a test of the developmental origins of health and disease hypothesis. Am J Hum Biol. 2013;25(3):411-417.

24.

Ravindranath

Joseph

Bosco

Rajangam

Balasubramanyam

. Fluctuating asymmetry in dermatoglyphics of non-insulin-dependent diabetes mellitus in Bangalore-based population. Indian J Hum Genet. 2005;11: 149.

25.

Leon

. Fetal growth and adult disease. Eur J Clin Nutr Suppl. 1998;1:72-78.

26.

Atkinson

Eisenbarth

. Type 1 diabetes: new perspectives on disease pathogenesis and treatment. Lancet. 2001;358:221-229.

27.

Söderström

Åman

Hjern

. Being born in Sweden increases the risk for type 1 diabetes-a study of migration of children to Sweden as a natural experiment. Acta Paediatrica. 2012;101(1):73-77.

28.

Ziegler

Mathies

Ziegelmayer

. Dermatoglyphics in type 1 diabetes mellitus. Diabetic Med. 1993;10(8):720-724.

29.

Medland

Park

Loesch

Martin

. Ridgecounter: a program for obtaining semi-automated finger ridge counts. Annal Hum Biol. 2007;34:504-517.

30.

Jain

Prabhakar

Hong

Pankanti

. Filterbank-based fingerprint matching. IEEE Trans Image Process. 2000;9(5):846-859.

31.

Tico

Immonen

Ramo

Kuosmanen

Saarinen

. Fingerprint recognition using wavelet features. ISCAS. 2001;2:21-24.

32.

Bharkad

Kokare

. Hartley transform based fingerprint matching. J Inf Processing Sys. 2012;8:85-100.

33.

Dale

Joshi

. Fingerprint matching using transform features. In: TENCON 2008 IEEE Region 10 Conference. New York, NY: IEEE; 2008:1-5.

34.

Logue

Walker

Colhoun

. Scottish Diabetes Research Network Epidemiology Group. Do men develop type 2 diabetes at lower body mass indices than women? Diabetologia. 2011;54:3003-3006.

35.

Heikes

Eddy

Arondekar

Schlessinger

. Diabetes Risk Calculator: a simple tool for detecting undiagnosed diabetes and pre-diabetes. Diabetes Care. 2008;31:1040-1045

36.

Wang

Larson

Vasan

. Metabolite profiles and the risk of developing diabetes. Nat Med. 2011;17(4):448-453.

37.

Van Hoek

Dehghan

Witteman

. Predicting type 2 diabetes based on polymorphisms from genome-wide association studies a population-based study. Diabetes. 2008;57:3122-3128.

38.

Bjorksten

Fowler

Pomiankowski

. What does sexual trait FA tell us about stress? Trends Ecol Evol. 2000;15(4):163-166.

39.

Lens

Dongen

Kark

Matthysen

. Fluctuating asymmetry as an indicator of fitness: can we bridge the gap between studies? Biol Rev. 2002;77(1):27-38.

40.

Palmer

Strobeck

. Fluctuating asymmetry and developmental stability: heritability of observable variation vs. heritability of inferred cause. J Evol Biol. 1997;10(1):39-49.

41.

Møller

. Symmetry, size and stress. Proc R Soc London Ser B. 2000;263:423-1427.

42.

Polak

Starmer

. The quantitative genetics of fluctuating asymmetry. Evolution. 2001;55(3):498-511.

43.

Polak

Hooker

Tyler

. Consistent positive co-variation between fluctuating asymmetry and sexual trait size: a challenge to the developmental instability-sexual selection hypothesis. Symmetry. 2015;7(2):976-993.

44.

Cho

Chen

. DIAGRAM Consortium; MuTHER Consortium. Meta-analysis of genome-wide association studies identifies eight new loci for type 2 diabetes in east Asians. Nat Genet. 2012;44:67-72.

45.

Guo

Hanson

Traurig

. TCF7L2 is not a major susceptibility gene for type 2 diabetes in Pima Indians: analysis of 3,501 individuals. Diabetes. 2007;56:3082-3088.

46.

Hanson

Rong

Kobes

. Role of established type 2 diabetes-susceptibility genetic variants in a high prevalence American Indian population. Diabetes. 2015;64:2646-2657.

47.

Saxena

Saleheen

Been

. Genome-wide association study identifies a novel locus contributing to type 2 diabetes susceptibility in Sikhs of Punjabi origin from India. Diabetes. 2013;62(5):1746-1755.

48.

Ramachandran

Snehalatha

Premila

Mohan

Viswanathan

. Familial aggregation in type 1 (insulin-dependent) diabetes mellitus: a study from south India. Diabet Med. 1990;7:876-879.

49.

Erbey

Kuller

Becker

Orchard

. The association between a family history of type 2 diabetes and coronary artery disease in a type 1 diabetes population. Diabetes Care. 1998;21:610-614.

50.

Tuomi

. Type 1 and type 2 diabetes: what do they have in common? Diabetes. 2005;54(2):S40-S45.

51.

Teupe

Bergis

. Epidemiological evidence for “double diabetes.” Lancet. 1991;337:361-362.

52.

de Groot

Wessel

. Genetic testing and type 2 diabetes risk awareness. Diabetes Educ. 2014;40:427-433.