To determine the relationship between the index to ring finger (2D:4D) length ratio and the risk of knee and hip osteoarthritis (OA).
To determine the relationship between the index to ring finger (2D:4D) length ratio and the risk of knee and hip osteoarthritis (OA).
We conducted a case–control study, in which cases with persistent symptoms and radiographic evidence of knee or hip OA were compared with controls with no symptoms and no radiographic evidence of knee or hip OA. Hand radiographs were visually classified as type 1 (index finger longer than the ring finger), type 2 (index finger equal to the ring finger), or type 3 (index finger shorter than the ring finger). The 2D:4D phalangeal and metacarpal length ratios were measured separately. The odds ratio (OR) and 95% confidence interval (95% CI) were calculated and adjusted for possible confounding factors using a logistic regression model.
Of 2,049 cases, 1,013 had radiographic evidence of knee OA and 995 had hip OA. Of 1,123 controls, 836 had no knee OA and 1,050 had no hip OA. The type 3 finger pattern was associated with knee OA (OR 1.94, 95% CI 1.54–2.44), and the risk was greater in women (OR 3.05, 95% CI 2.08–4.47) than in men (OR 1.45, 95% CI 1.08–1.95). There was a dose-response relationship between both 2D:4D phalangeal and metacarpal length ratios and the risk of knee OA. The risk of hip OA was inconsistent.
Compared with types 1 and 2, the type 3 “male” pattern 2D:4D length ratio is associated with OA, especially knee OA. The risk is independent of other major OA risk factors.
The index to ring finger length ratio (2D:4D) is a trait that is sexually differentiated in a variety of species (1, 2). In humans, males typically have shorter second (index) digits (2D) compared with fourth (ring finger) digits (4D), whereas in women the fingers are more equal in length (3). Smaller 2D:4D length ratios have been associated with higher prenatal testosterone levels, higher sperm counts, and lower estrogen concentrations (3–7). Reduction in this ratio has been shown to have a number of associations, ranging from sexual ability (8, 9), physical and athletic ability (10–14), and “masculine” facial shape (15–17) to offspring sex ratio (6, 18) and performance in examinations (19). It has been considered a masculine surrogate marker and a risk factor for autism (20), myocardial infarction (21), and human immunodeficiency virus in men (22). However, it has not been examined as a possible risk factor for osteoarthritis (OA), a condition that can be related both biomechanically to physical activity and hormonally to estrogen deficiency (23). We therefore undertook this case–control study to examine whether smaller 2D:4D length ratio is associated with knee and hip OA, and whether any such association is independent of other established risk factors for OA.
Participants were men and women recruited originally from 2002 to 2006 in Nottingham into a case–control study designed primarily to investigate gene–environment interaction in large-joint OA (the Genetics of Osteoarthritis and Lifestyle Study). Approval to assemble this cohort and to examine radiographs for OA research purposes was obtained from the Nottingham Research Ethics Committee.
Cases were recruited from hospital orthopedic surgery lists (current and for the previous 5 years) and from a rheumatology clinic. Individuals who had clinically significant symptomatic OA of the hips or knees were eligible and were invited to participate in the study. Subjects were excluded if they had any other major arthropathy (e.g., rheumatoid arthritis), childhood joint disease (e.g., Legg-Calvé-Perthes disease, slipped femoral epiphysis, hip dysplasia), total joint replacement (TJR), avascular necrosis of the femoral neck or distal condyle, or terminal illness. Controls were recruited from hospital lists of people who had undergone intravenous urography (IVU) within the last 5 years. Individuals who had no radiographic evidence of hip OA on their screening IVU film and who had no hip or knee symptoms and none of the exclusions listed above for cases were invited to take part in the study. Cases and controls were further characterized by clinical and radiographic assessments. The presence of interphalangeal (IP) nodes was determined clinically using established methods (24, 25); nodal change was defined as Heberden's and/or Bouchard's nodes that affected at least 2 rays of each hand.
Radiographs of the knees and pelvis were obtained to ascertain the diagnosis, as described below. The presurgical radiographs of cases who had undergone TJR were copied using Hipax Dicom digitizing software (Hipax, Vorstetten, Germany) to enable scoring. If cases and controls had not undergone TJR, new radiographs were obtained, unless radiographs of the same joint (using the same standardized method) had been obtained not more than 2 years previously. All participants had radiographs of both hands, both knees, and the pelvis examined.
Separate radiographs of the right and left hands were obtained. The participant was seated adjacent to the x-ray table, with the forearm and hand flat and prone on the table with no lateral angulation at the wrist. The hand was centered on the cassette with fingers slightly spread apart but flat. The x-ray beam was centered on the third metacarpophalangeal joint. Images were obtained using a small focal point and a detail cassette. Exposures and distances were as follows: 48 KV, 3.2 mAs, and 90-cm focus-to-film distance. Each film was scanned and written to CD using the Hipax 4.2 x-ray image processor. CD images were read using the Hipax Private Health disk image viewer, which enables straight-line measurements to an accuracy of 0.01 mm (manufacturer precision).
The 2D:4D length ratio was determined on the hand radiographs using 2 methods: a visual classification and the measured 2D:4D length ratio. The visual classification consisted of classifying each hand according to whether the index finger was visually longer (type 1), equal to (type 2), or shorter than the ring finger (type 3). The measured 2D:4D length ratio is the length (mm) from the midpoint of the base of the proximal phalanx to the midpoint of the tip of the distal phalanx, and the length (mm) from the midpoint of the base to the midpoint of the tip of the metacarpal bone. The 2D:4D length ratio was calculated separately for phalanges and metacarpal bones, as well as for the combined finger lengths. All measurements were performed by 1 observer (JR) and were entered directly into a Microsoft Access database (Microsoft, Redmond, WA). If no measurement was possible, the reason for this was recorded.
Posterior-anterior weight-bearing knee radiographs were obtained with the SynaFlexer positioning frame (Synarc, San Francisco, CA) (26), with feet externally rotated 10 degrees, and knees and thighs touching the vertical platform anteriorly and the x-ray beam angled 10 degrees caudally. Skyline 30-degree views of both patellofemoral compartments were obtained with the participant seated, and the beam was angled from feet to knees. Anteroposterior views of the pelvis were obtained with the participant supine and the feet internally rotated 10 degrees.
Knee and pelvis radiographs were all graded for features of OA by a single investigator (SD) (not the same individual as the hand radiograph observer). For knees, the following were scored separately in each compartment (medial tibiofemoral, lateral tibiofemoral, and patellofemoral): osteophytes (graded from 0 to 5) and joint space narrowing [JSN] (graded from −1 to 5), using the Nottingham logically derived line drawing atlas (27). In this system, −1 = wider than normal joint space, 0 = normal joint space (derived from a community sample), and 5 = bone on bone, with each grade being separated by the same interval (20% of the normal width) and using separate atlases for men and women (because of sex differences in normal width) (27). Knee OA was defined radiographically as a JSN score of ≥2 plus ≥1 osteophyte in any knee compartment. Tibiofemoral and patellofemoral joints were examined separately. Subjects who had only flexed or straight lateral knee radiographs obtained before total knee replacement were scored using an equivalent Kellgren/Lawrence (K/L) score (28) of 0–4 for the patellofemoral compartment; radiographic knee OA was defined as a K/L score of ≥3.
For hips, the following were recorded: actual measurement (mm) of minimum joint space (using Hipax Dicom software), site of maximum narrowing (superolateral, superoinferior, supero indeterminate, axial, medial, concentric, unable to score), osteophytes (graded 0–3) in the acetabulum, femoral head, and femoral neck, presence and location (acetabulum, femoral head) of subchondral cysts and sclerosis, and an overall Croft score (graded 0–5) (29). Hip OA was defined as a Croft score of ≥3 (i.e., 2 or more of the following: osteophytes, narrowing, subchondral sclerosis, cysts).
Reproducibility of finger length measures was examined in a random sample of 30 participants, whose radiographs were read under blinded conditions by the same observer 3 times during the study period (beginning, middle, and end). Reproducibility of OA changes in knees and hips was determined in a random sample of 20 participants, whose radiographs were read under blinded conditions twice by the observer halfway through the reading process, with an interval of >2 weeks between the 2 readings. The intraclass correlation coefficient (ICC) and weighted kappa were used to present agreement for continuous and ordinal data, respectively.
Various comparisons were made between cases and controls. The prevalence of a definite or probable type 3 pattern on either hand was calculated. Because the 2D:4D length ratio was symmetric between the right and left hand in this population (data to be published elsewhere), the mean 2D:4D length ratio of the right and left hand was calculated for each participant. The data were categorized into tertiles for subject-based analysis, in which tertile 1 had the smallest 2D:4D length ratio (i.e., longer ring finger) and tertile 3 had the greatest 2D:4D length ratio. For each case–control comparison, the t-test was used for continuous data and the chi-square test was used for dichotomous data. The odds ratio (OR) and 95% confidence interval (95% CI) were used to estimate the relative risk of OA. The analyses were undertaken separately for knee OA, hip OA, and other types of OA, in which controls were redefined according to the target joint OA regardless of other joint involvement. However, other joint involvement was adjusted, together with other confounding factors, using a logistic regression model.
Dose-response relationships between tertiles of the 2D:4D length ratio and the risk of OA were examined. Possible confounding factors were adjusted, including age; sex; body mass index (BMI); bone mineral density (BMD), which was estimated by a dual x-ray absorptiometry (DXA) scan of the calcaneus using the Apollo 501A00Z DXA (Norland Medical Systems, Trumbull, CT) and the age-adjusted Z score; previous significant joint injury (self-reported, dichotomized as yes or no); regular energetic activities (defined as >20 minutes/time and >3 times/week, which caused sweating, breathlessness, or increased pulse for those between the ages of 20 and 40); male hormone surrogates, such as the age at onset of male pattern baldness, the earliest decade when definite baldness appeared (age 20–29 years = 3, 40–49 years = 2, 60–69 years = 1), shaving frequency during the ages of 20–29 years (1 = did not shave, 2 = shaved every other day, 3 = shaved less than every other day, 4 = shaved once a day, 5 = shaved twice a day), and adult acne (i.e., acne during the ages of 30–39 years); female hormone surrogates, such as the age at onset of menopause (<40 years = 7, 40–49 years = 6, 50–59 years = 5, 60–69 years = 4, 70–79 years = 3, ≥80 years = 2, not yet = 1); IP nodes; and other large-joint OA. Analyses were performed using SPSS software, version 13 (SPSS, Chicago, IL). P values less than 0.05 (2-tailed) were considered significant.
In total, 3,475 cases and 3,441 controls were contacted for this study. Of 3,180 eligible cases, 2,168 agreed to participate and 2,049 completed the clinical and radiographic assessments (response rate 64%). Of 2,022 eligible controls, 1,312 provided consent and 1,123 completed the assessment (response rate 55%). The total study population was 3,172 before the radiographic assessment, including 1,042 index knee OA cases, 1,007 index hip OA cases, and 1,123 controls. After radiographic examination, 1,013 cases were confirmed as having knee OA (81% had undergone TJR) and 995 cases were ascertained as having hip OA (93% had undergone TJR). Of 1,123 controls, 836 had no radiographic evidence of knee OA and 1,050 had no radiographic evidence of hip OA; these controls were compared with knee and hip OA cases separately. All radiographically assessed subjects were included in the analyses (Table 1). Compared with controls, cases with knee or hip OA were older, heavier, had higher bone density, and more frequently had IP nodes.
|Knee OA||Hip OA|
|Cases (n = 1,013)||Controls (n = 836)||P||Cases (n = 995)||Controls (n = 1,050)||P|
|Age, mean ± SD years||68.4 ± 7.3||62.9 ± 8.4||<0.001||67.7 ± 7.1||64.1 ± 8.4||<0.001|
|No. of women/men||488/525||399/437||0.852||499/496||500/550||0.269|
|BMI, mean ± SD kg/m2||31.3 ± 5.3||27.2 ± 4.4||<0.001||29.3 ± 5.2||27.5 ± 4.7||<0.001|
|BMD, mean ± SD Z score||1.01 ± 1.25||0.59 ± 1.21||<0.001||0.95 ± 1.26||0.68 ± 1.24||<0.001|
|Subjects with IP nodes, no. (%)||338 (33.4)||89 (10.6)||<0.001||282 (28.3)||138 (13.1)||<0.001|
|Visually assessed finger pattern type, no. (%)|
|Indeterminate||106 (10.5)||17 (2.0)||29 (2.9)||22 (2.1)|
|2D:4D phalangeal ratio, mean ± SD‡||0.912 ± 0.023||0.917 ± 0.022||<0.001||0.914 ± 0.023||0.916 ± 0.023||0.083|
|2D:4D metacarpal ratio, mean ± SD§||1.154 ± 0.031||1.155 ± 0.032||0.326||1.154 ± 0.032||1.155 ± 0.033||0.243|
The ICC for reproducibility of the radiographic measurement of finger length ranged from 0.95 to 0.99, and the weighted kappa for reproducibility of the visual classification of hand types ranged from 0.58 to 0.78. The weighted kappa for reproducibility of the radiographic OA scores ranged from 0.60 to 1.00, depending on the joint and the individual feature assessed.
The distribution of the different finger patterns and the mean values of measured 2D:4D length ratios are given in Table 1. The number of subjects available for each analysis varied, depending on the study group and the characteristic assessed. While a larger proportion of subjects could not be assessed for the visual classification (10.5%) and the 2D:4D phalangeal measure (27.6%) in the knee OA group, the majority could be assessed for the 2D:4D metacarpal measure (>97%) regardless of study group (Table 1).
Compared with other finger pattern types (types 1 and 2), the visual type 3 finger pattern was associated with an increased risk of OA involving any knee compartment, the hip, or “generalized” nodal OA (IP nodes plus OA at the knee and/or hip). The risk of knee OA in participants with type 3 finger patterns was nearly twice that for participants without this pattern (Table 2). Women with the type 3 finger pattern had a greater risk of knee OA (OR 3.05, 95% CI 2.08–4.47) than men (OR 1.45, 95% CI 1.08–1.95) (Figure 1). There was no significant difference between unilateral risk (OR 2.47, 95% CI 1.65–3.72) and bilateral risk (OR 1.78, 95% CI 1.39–2.29) of knee OA should a subject have the type 3 finger pattern. The association was independent of other joint OA and risk factors, as evidenced by the fact that significance was retained after adjustment for OA of other joints and other risk factors (Table 2).
|Frequency of type 3 finger pattern||OR (95% CI)|
|Any||550/1,013||341/836||1.72 (1.43–2.08)||1.94 (1.54–2.44)|
|Tibiofemoral||507/924||392/951||1.73 (1.44–2.08)||1.89 (1.50–2.37)|
|Patellofemoral||252/498||414/974||1.39 (1.12–1.72)||1.73 (1.31–2.27)|
|Hip OA||512/995||455/1,050||1.39 (1.17–1.65)||1.37 (1.13–1.67)|
|IP nodes||371/780||1,198/2,392||0.90 (0.77–1.06)||0.99 (0.83–1.18)|
|IP nodes + knee and/or hip OA||302/620||290/708||1.37 (1.10–1.70)||1.69 (1.33–2.16)|
The smaller the 2D:4D phalangeal ratio, the greater the risk of tibiofemoral joint knee OA (P for trend = 0.024) (Table 3). Such a dose-response relationship was not found with patellofemoral joint knee OA (P for trend = 0.531) or hip OA (P for trend = 0.341), but was observed with the combination of IP nodes plus knee and/or hip OA (P < 0.001) (Table 3). A similar dose-response pattern was observed for the 2D:4D metacarpal bone ratio (Table 4). However, isolated IP nodes were associated only with the 2D:4D phalangeal ratio and not with the metacarpal bone ratio, and some marginal differences were seen with the combinations (Tables 3 and 4).
|Adjusted OR (95% CI)||P for trend†|
|Tertile 3 (reference)||Tertile 2||Tertile 1|
|Any||1||1.01 (0.74–1.38)||1.36 (0.99–1.87)||0.046|
|Tibiofemoral||1||1.01 (0.74–1.37)||1.41 (1.03–1.94)||0.024|
|Patellofemoral||1||0.88 (0.61–1.28)||1.08 (0.73–1.59)||0.531|
|Hip OA||1||1.05 (0.82–1.34)||1.21 (0.92–1.57)||0.341|
|IP nodes||1||1.44 (1.12–1.86)||1.55 (1.19–2.03)||<0.001|
|IP nodes + knee and/or hip OA||1||1.49 (1.05–2.12)||2.02 (1.40–2.90)||<0.001|
|Adjusted OR (95% CI)||P for trend|
|Tertile 3 (reference)||Tertile 2||Tertile 1|
|Any||1||1.11 (0.85–1.46)||1.36 (1.03–1.79)||0.038|
|Tibiofemoral||1||1.14 (0.87–1.49)||1.43 (1.09–1.88)||0.013|
|Patellofemoral||1||1.09 (0.79–1.50)||1.19 (0.86–1.64)||0.322|
|Hip OA||1||1.11 (0.88–1.41)||1.19 (0.95–1.50)||0.138|
|IP nodes||1||1.15 (0.93–1.43)||1.16 (0.94–1.43)||0.131|
|IP nodes + knee and/or hip OA||1||1.17 (0.88–1.55)||1.41 (1.05–1.91)||0.013|
The type 3 finger pattern was associated with sex in that men were >2.5 times more likely than women to have this pattern (OR 2.68, 95% CI 2.29–3.14). Type 3 finger pattern was also associated with a female estrogen deficiency surrogate of earlier onset of menopause (OR 1.18, 95% CI 1.01–1.36). However, it was not associated with physical activities (OR 1.02, 95% CI 0.87–1.19) or male hormone surrogates such as shaving frequency (OR 0.90, 95% CI 0.80–1.01), adult acne (OR 1.25, 95% CI 0.72–2.15), or age at onset of baldness (OR 1.03, 95% CI 0.91–1.18). Similar results were obtained with the measured 2D:4D length ratios (data not shown).
This is the first study to examine the possible association between the 2D:4D length ratio and OA. We found a positive association with both knee OA and hip OA, which is stronger for tibiofemoral knee OA and for the combination of knee and/or hip OA plus IP nodes (i.e., “generalized nodal OA”). The association is dose-dependent and independent of other well-established risk factors for OA, such as age, sex, BMI, joint injury, and physical activity. Most of these established risk factors were replicated in this study apart from sex (due to matching), suggesting internal validity of the study (data not shown). The 2D:4D length ratio therefore appears to be a newly identified risk factor for the development of OA; specifically, women with the male pattern of 2D:4D length ratio (i.e., ring finger relatively longer than the index finger) are more likely to develop knee OA.
The mechanism that accounts for this association is unknown. Previous studies have shown the 2D:4D length ratio to associate with high standards of athletic achievement (10–14) and with high levels of testosterone and “male” characteristics (3–7). Increased activity and performance in physically demanding sports could contribute to the development of OA through repetitive joint trauma and injury, and higher testosterone levels could cause hormonal modulation of joint connective tissue metabolism, including chondrocyte, osteoblast, and osteoclast activity, and thus contribute to the development and/or progression of OA (30).
In our study, we found no supporting evidence to suggest that the 2D:4D length ratio operates through high levels of physical activity or via surrogates of male hormone levels such as shaving frequency, occurrence of adult acne, or male pattern baldness. However, since we estimated the degree of occupational and recreational activity by retrospective self-reporting and did not directly assess sporting ability or measure serum testosterone levels, as has been undertaken in other studies (3–7), it is possible that the recall bias of these measures may influence the results. Furthermore, we did not intentionally recruit a proportion of individuals who were elite athletes to specifically address this question. Nevertheless, consistent with other studies, we did confirm that the ratio relates to male sex (OR 2.68, 95% CI 2.29–3.14) and earlier onset of menopause in women, a surrogate outcome for estrogen deficiency, which suggests that our study population is not unrepresentative with respect to sex characteristics.
Apart from an indirect sex hormone effect mediated by biomechanical factors and effects on joint connective tissue metabolism, the 2D:4D length ratio could also associate with OA if it were a phenotypic marker for genetic variation in nonendocrine mechanisms. For example, the ratio has been associated with facial shape (17) and with an increased risk of heart disease (21), and such associations could relate to generalized regulatory mechanisms that influence structural connective tissue morphology and function (31). Certainly the UK female twin study (32) has shown strong heritability of the 2D:4D length ratio (OR 0.66, 95% CI 0.50–0.78). It has been proposed that the homeobox (Hox) genes are involved in the growth of bone, cartilage, and soft tissue of human fingers and toes (33). Hox genes are also associated with sex determination, the morphogenesis of the genitourinary system, fertility, and hematopoiesis, and hence with sex-dependent diseases (33). Further work is warranted to determine the responsible genes and the mechanism that leads to variation in 2D:4D length ratio and the link with predisposition to OA.
We used 3 different methods to assess the 2D:4D length ratio from radiographs: a direct visual comparison of the 2 finger ends, the measured ratio from the base of the proximal to the tip of the distal phalanx, and the measured ratio of the metacarpal lengths. Each method has its strengths and weaknesses. The visual assessment is simple and quick but particularly prone to bias from radial–ulnar malpositioning. Although the visual classification picks up the signals for most associations (Tables 2 and 3) and it relates to the measured ratios (to be reported elsewhere), caution must be taken when using this assessment, and supporting evidence from other measures such as the dose-response analysis is useful.
The 2 measured ratios take longer to determine and yield somewhat differing results. For example, the dose-response relationship of the 2D:4D length ratio with OA was stronger for the phalangeal than for the metacarpal ratio, and the OR for presence of IP nodes was significant for the phalangeal but not the metacarpal ratio. Although the phalanges and metacarpal bones both contribute to variation in the 2D:4D length ratio and presumably reflect the same underlying constitutional characteristic, their relative contribution may differ. Furthermore, the phalangeal lengths are more susceptible to acquired alterations secondary to trauma and to IP OA (foreshortening due to bone and cartilage attrition, radial–ulnar deviation, or flexion deformity). The greater targeting of IP nodes for the index finger may even lead to systematic bias, in this case, foreshortening of the index finger biasing toward a lower value for 2D:4D length ratio. Indeed, in our analysis, isolated IP nodes were associated with the phalangeal but not the metacarpal ratio, but after removal of IP nodes from the analyses (adjustment) we found almost identical relationships between the 2 measured 2D:4D length ratios and OA (Tables 3 and 4). Therefore, care must be taken in interpreting the phalangeal ratio since it may overestimate the association with OA.
In contrast, the metacarpal ratio is less affected by trauma and OA and is applicable to almost all hand radiographs. Therefore it would appear to be the preferred radiographic measure, especially in older populations and in those who are especially at risk of IP nodes and hand OA (e.g., in genetic studies of generalized OA). The visual classification, however, may prove to be a useful tool, especially for clinical assessment with a carefully standardized hand position, and this merits further investigation.
There are several limitations to this study. First, we undertook a hospital-based case–control study of participants who were recruited primarily for a study of gene–environmental interaction in knee and hip OA. Although the sample size was large (n = 3,172), all participants with knee OA and hip OA had clinically severe OA and had been referred to secondary care for consideration of TJR (this was undertaken in the majority of cases). Therefore, these results may not be generalizable to the whole population of people with knee or hip OA.
Second, the controls were selected from hospital patients who had undergone intravenous urography; thus, although “matched” in being referred to a hospital for non-OA reasons, they were not a random sample from the general population. The associations found in this study therefore require confirmation in a community setting.
Third, of 1,007 patients with clinical hip OA, more than half of them (n = 540) also had radiographic knee OA. After the radiographic confirmation and exclusion of knee OA and IP nodes, only 352 patients had isolated hip OA in this group; this may reduce the power to detect a positive association. Also, a frequency-, age-, and sex-matched case–control study was attempted to reduce differences between age and sex, so the ORs for these 2 factors may be underestimated.
Fourth, we measured phalangeal and metacarpal lengths separately. The advantage of this is that it allowed us to examine the contribution of the 2 bone types to the 2D:4D length ratio independently. The disadvantage, however, is that the separation led to difficulty in dichotomizing the data for type 3 finger pattern (i.e., 2D:4D ratio ≤1), since all phalangeal ratios were <1 and all metacarpal ratios were >1 (Table 1). We therefore grouped the data into tertiles to estimate relative risk of tertile 1 or 2 compared with tertile 3. The possible bias in using the extreme reference is open to debate.
Fifth, because IP nodes were used as a surrogate for hand OA, the results for radiographically defined hand OA have yet to be confirmed. Finally, the hand radiographs were standardized primarily for assessment of OA, and if the 2D:4D length ratio became an important radiographic assessment, more detailed attention to finger positioning to minimize radial–ulnar deviation might permit more confident visual assessment of finger type in a greater proportion of participants.
In summary, the smaller 2D:4D length ratio assessed from hand radiographs is associated with an increased risk of knee OA, especially tibiofemoral OA. It also associates with the combination of IP nodes plus knee and/or hip OA. The association is independent of other established OA risk factors. It is best determined by measurement of metacarpal length, particularly in older populations. The underlying mechanism of the risk is unclear and merits future exploration.
Dr. Doherty had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study design. Zhang, Muir, M. Doherty.
Acquisition of data. Robertson, S. Doherty, Liu, Maciewicz, M. Doherty.
Analysis and interpretation of data. Zhang, Robertson, Maciewicz, Muir, M. Doherty.
Manuscript preparation. Zhang, Robertson, Maciewicz, Muir, M. Doherty.
Statistical analysis. Zhang.