Abnormal bone geometry at the metacarpal bone shaft of rheumatoid arthritis patients with maintained muscle–bone relationship




Metacarpal juxtaarticular bone is altered in rheumatoid arthritis (RA). However, a detailed analysis of disease-related geometric adaptations of the metacarpal shaft is missing. The aim of the present study was to assess the role of RA disease, forearm muscle cross-sectional area (CSA), age, and sex on bone geometry at the metacarpal shaft.


In 64 RA patients and 128 control subjects, geometric properties of the third metacarpal bone midshaft and forearm muscle CSA were measured by peripheral quantitative computed tomography (QCT). Linear models were performed for cortical CSA, total bone CSA, polar stress-strain index (SSI; a surrogate for a bone's resistance to bending and torsion), cortical thickness, and the metacarpal index (MI; cortical CSA/total CSA), with the explanatory variables muscle CSA, age, RA status, and sex.


Forearm muscle CSA was associated with cortical and total metacarpal CSA and polar SSI. RA group status was associated with all bone parameters except cortical CSA. There was a significant interaction between RA status and age, indicating that the RA group had a greater age-related decrease in cortical CSA, cortical thickness, and the MI.


Bone geometry of the metacarpal shaft is altered in RA patients compared to healthy controls. Whereas bone mass of the metacarpal shaft is adapted to forearm muscle mass, cortical thickness and the MI are reduced, but outer bone shaft circumference and the polar SSI increased in RA patients. These adaptations correspond to an enhanced aging pattern in RA patients.


In rheumatoid arthritis (RA), bone damage such as erosions, juxtaarticular osteopenia, as well as generalized osteopenia and osteoporosis, are well recognized (1, 2). Metacarpal bone loss in RA patients is commonly assessed on plain radiographs of the hands by one of 3 methods: 1) the combined cortical width (added thickness of the right and left cortical shell) measured at the second metacarpal bone shaft (3–5), 2) the metacarpal index (MI), defined as the ratio between the cortical width divided by the total bone width (6), and 3) digital x-ray radiogrammetry (DXR), a computer-automated analysis of hand radiographs (7) that measures the cortical thickness and bone width of the 3 middle metacarpal bones over a fixed-size region (8). The MI has been found to decrease with increasing RA disease activity or severity (9, 10), and the decrease of the MI over time was more pronounced in erosive than nonerosive disease (11). Detection of bone loss at the metacarpal shaft was among the earliest features of RA and precedes bone erosions (7, 12), highlighting the importance of quantifying metacarpal bone status.

We have recently developed a protocol for measuring metacarpal bone geometry and volumetric bone mineral density (BMD) by peripheral quantitative computed tomography (QCT) (13, 14). This protocol allows accurate measurements of volumetric BMD and cross-sectional geometry at the distal epiphysis and shaft of the third metacarpal bone. At the radius, the tibia, and the third metacarpal bone, we have found epiphyseal trabecular BMD and shaft cortical thickness to be reduced in female RA patients compared to healthy controls (13). While existing studies have shown that bone mass at the metacarpal bone shaft of RA patients is reduced (15), and its cross-sectional geometry different from healthy controls (13), it has not been assessed whether these differences may be due to the influence of factors other than disease activity, specifically muscle force or mass (16, 17), sex (18), and age (18, 19). The aim of the present study was to compose a detailed assessment of the cross-sectional bone geometry at the shaft of the third metacarpal bone in patients with RA compared to healthy controls relative to the main confounding factors for bone geometry. We measured bone shaft properties of the third metacarpal bone and the cross-sectional area (CSA) of the forearm muscles by means of peripheral QCT, and performed linear models for bone shaft parameters with regard to the explanatory variables forearm muscle CSA, age, sex, and RA status.


We conducted a cross-sectional cohort study in RA patients and healthy control subjects assessing the influence of various factors on bone geometry at the metacarpal bone shaft. The study protocol was approved by the Ethics Committee of the Canton of Bern.


Consecutive RA patients fulfilling the American College of Rheumatology criteria (20), seen in the Department of Rheumatology, Inselspital Bern, were recruited. Previous therapy (including bisphosphonate and glucocorticoids), disease duration, erosiveness of disease as assessed by total the Ratingen score (21), as well as serologic status were obtained. For the control group, we recruited healthy volunteers by locally distributed flyers and advertisement on the hospital internal web. Inclusion criteria were, for both groups, age 20–90 years. Exclusion criteria were, for both groups, bone metabolic diseases, hyper/hypoparathyroidism, hyper/hypothyroidism, chronic renal insufficiency, cancer, pregnancy, lactation, and drug addiction on the basis of medical history and questionnaires for osteoporosis risk factors. For the control group, established osteoporosis and previous or present bisphosphonate therapy were also exclusion criteria. All of the patients and volunteers gave written informed consent.

Metacarpal bone measurements.

Measurements were performed with a Stratec XCT 3000 scanner (Stratec Medical). This peripheral QCT apparatus measures attenuation of radiographs, which are linearly transformed into hydroxyapatite densities. Unlike some other peripheral QCT scanners, the Stratec XCT 3000 is calibrated with respect to water, which is set at 60 mg hydroxyapatite so that fat results in 0 mg hydroxyapatite (22). Hydroxyapatite equivalent densities are automatically calculated from the attenuation coefficients by employing the manufacturer's phantom, which itself is calibrated with respect to the European Forearm Phantom (QRM) (22).

The length of the third metacarpal bone was palpated and measured from base to head by measuring tape to the nearest 5 mm. A scout view was performed of the head of the third ossa metacarpalia and the reference line was placed at the distal end of the bone. Scans were performed at 50% of total bone length measured from the distal bone end. Slice thickness was 2.2 mm, scanning speed was set at 15 mm/second, and voxel size at 0.3 mm edge length. The threshold for the periosteal surface was set at 280 mg/cm3 and from this, bone mineral content, total bone CSA (including the bone marrow space), and the polar bone stress-strain index (SSI; in mm3) were calculated. The polar SSI is a measure for diaphyseal bone resistance to bending and torsion (23, 24) and can be used as a surrogate for bone strength (25, 26). Cortical bone was selected by the threshold 710 mg/cm3, and from this, cortical CSA (excluding the bone marrow space) and cortical BMD were calculated. Cortical thickness was calculated from total CSA and cortical CSA based on the assumption that the bone shaft is cylindrical.

Muscle size assessment.

Muscle CSA of the forearm was measured as a surrogate for the muscle forces acting on the metacarpal bones (27–29). For this purpose, a peripheral QCT scan was performed at the lower arm at 66% of total ulnar length measured from ulnar styloid. Slice thickness was 2.2 mm and voxel size was set at 0.5 mm, with a scanning speed of 20 mm/second. Muscle CSA was determined by selecting the area with a lower threshold of 40 mg/cm3 and an upper threshold of 280 mg/cm3 hydroxyapatite density after smoothing the image.

Statistical analysis.

Subject characteristics were compared by nonparametric Mann-Whitney tests. Linear models for metacarpal bone midshaft cortical CSA, total CSA, cortical wall thickness, polar SSI, and the MI were calculated for all of the subjects together. The following explanatory variables were entered into the model: forearm muscle CSA, age, sex, and RA status, as well as all 2-way cross terms (interactions).

In order to center the data, mean muscle CSA was subtracted from all muscle CSA data and mean age was subtracted from age data. Statistical analyses were performed using SPSS, version 17.0. Due to the large number of computed P values, statistical significance was set at 0.01.


Subject parameters.

A total of 64 consecutive RA patients and 128 control subjects were recruited for the present study. The RA group consisted of 49 women and 15 men, whereas the control group included 98 women and 30 men. Group means and SDs as well as P values of nonparametric tests of subject characteristics are shown in Table 1. Subject characteristics were comparable between the RA and control groups; however, male RA patients tended to be 3.1 cm shorter than control males (P = 0.09).

Table 1. Subject anthropometric data*
ParameterRA patientsReference groupMann-Whitney P
  • *

    Values are the mean ± SD (used for better interpretability of data) unless otherwise indicated. RA = rheumatoid arthritis; CSA = cross-sectional area.

  • P values of nonparametric tests are shown because some of the parameters in the smaller male groups were not normally distributed.

 Age, years55.18 ± 11.5254.02 ± 12.760.492
 Height, cm163.61 ± 6.07164.85 ± 5.640.163
 Weight, kg67.37 ± 13.7263.50 ± 9.900.111
 Forearm muscle CSA, cm223.96 ± 3.9825.31 ± 3.520.117
 Age, years57.40 ± 8.2355.23 ± 11.640.596
 Height, cm175.07 ± 6.82178.20 ± 5.570.093
 Weight, kg89.07 ± 13.5785.33 ± 6.560.267
 Forearm muscle CSA, cm238.95 ± 8.2642.03 ± 5.140.135

Mean ± SD disease duration in the RA group was 11.3 ± 9.1 years (11.0 ± 9.0 years in women and 12.6 ± 9.7 years in men), with a range from 1–40 years. Seventy-one percent of all patients (70% in women and 73% in men) were erosive, 69% (68% in women and 73% in men) were positive for rheumatoid factor, and 89% (85% in women and 100% in men) were positive for anti–cyclic citrullinated peptide antibodies. Anti–tumor necrosis factor therapy was previously given to 61% of women and 53% of men (median duration was 3 months for both sexes). Bisphosphonates had been administered to 33% of female and 20% of male patients. Seventy-five percent of patients (72% in women and 87% in men) had been receiving corticosteroid therapy during the year previous to peripheral QCT measurement.

Linear models.

Our models including only 4 explanatory variables and their interaction terms explained the major part of the bone variables' variance, specifically between 52% and 72% (Table 2), meaning that, for example, only 28% of the variance in total CSA could not be explained by forearm muscle CSA, age, sex, and RA status. Table 2 shows the intercepts and the slopes of all explanatory variables in the models. Positive values for RA status indicate that RA patients had a greater mean value than controls in this particular bone parameter when controlled for the other explanatory variables. Parameter estimates for sex, age, and forearm muscle CSA indicate the slopes in the control group. RA group status was a significant explanatory variable for all assessed bone parameters except cortical CSA, meaning that even after controlling for age, sex, and forearm muscle bulk, total CSA and polar SSI were greater in the RA group and cortical thickness and MI were smaller in the RA group compared to the control group. Cortical and total CSA as well as polar SSI increased significantly with increasing forearm muscle CSA (Table 2 and Figures 1A–C) in our model, controlled for age, sex, and RA group status.

Table 2. Linear models for metacarpal bone shaft parameters in all subjects*
 Cortical CSATotal CSACortical thicknessPolar SSIMI
  • *

    Values are the parameter estimates/slopes (P) unless otherwise indicated. For age, forearm muscle CSA, and sex, the slopes of the control group are shown. Age was defined as age − mean age and muscle CSA was defined as muscle CSA − mean muscle CSA in order to center the data. CSA = cross-sectional area; SSI = stress-strain index; MI = metacarpal index; RA = rheumatoid arthritis.

  • Significant explanatory variables (P ≤ 0.01).

  • Shows the variance explained by the model.

Intercepts41.04 (0.000)61.38 (0.000)1.89 (0.000)116.79 (0.000)0.68 (0.000)
RA group status−4.59 (0.063)10.44 (0.002)−0.40 (0.003)32.69 (0.001)−0.15 (0.000)
Age−0.04 (0.731)0.21 (0.210)−0.01 (0.465)0.46 (0.378)−0.00 (0.171)
Forearm muscle CSA0.65 (0.000)1.06 (0.000)0.01 (0.155)3.22 (0.000)−0.00 (0.564)
Sex−1.60 (0.523)−0.70 (0.833)−0.07 (0.599)−1.12 (0.912)−0.03 (0.516)
Group × age−0.22 (0.001)0.13 (0.158)−0.01 (0.000)0.38 (0.171)−0.01 (0.000)
Group × forearm muscle CSA0.16 (0.324)−0.35 (0.102)0.01 (0.130)−1.15 (0.079)0.01 (0.027)
Age × forearm muscle CSA0.00 (0.793)0.00 (0.922)−0.00 (0.956)0.01 (0.839)0.00 (0.534)
Group × sex3.52 (0.262)−7.88 (0.058)0.28 (0.101)−25.93 (0.043)0.11 (0.044)
Sex × forearm muscle CSA0.30 (0.082)0.07 (0.745)0.02 (0.047)−0.03 (0.965)0.01 (0.046)
Sex × age−0.01 (0.971)0.01 (0.947)−0.00 (0.635)0.15 (0.812)0.00 (0.917)
Figure 1.

Fits from separate linear regressions for rheumatoid arthritis (RA) patients and healthy controls displaying the relationship between forearm muscle cross-sectional area (CSA) and A, metacarpal shaft cortical CSA, B, total CSA, and C, polar stress-strain index (SSI). R2 values for the regressions of the RA group and the control group are also indicated. P values for all regressions were <0.0005.

There was a significant interaction between RA group status and age for all of the parameters except total CSA and polar SSI, indicating that there was a significantly steeper age-related decline of cortical CSA, cortical thickness, and the MI in the RA group than the control group. This is indicated by the interaction term group × age in Table 2 and shown graphically in Figures 2A and B. The interaction term of group × forearm muscle CSA in Table 2 shows nonsignificant values for all bone parameters, meaning that the slopes between forearm muscle CSA and bone parameter was the same in both groups (Figures 1A–C). The intercepts or distance between the two regression lines were significant for all bone parameters except metacarpal shaft cortical CSA (as indicated by the significant parameter estimates for RA group status in Table 2), with the RA group having a greater total CSA, a greater polar SSI, but a thinner cortical thickness and a lower MI for a given forearm muscle CSA than the control group (Figures 1A–C).

Figure 2.

Fits from separate linear regressions for rheumatoid arthritis (RA) patients (and healthy controls) displaying the relationship between A, age and metacarpal thickness and B, the metacarpal index. R2 values for the regressions of the RA group and the control group are also indicated. P values for all regressions were <0.0005.


We found the muscle–bone relationship (cortical CSA/muscle CSA) at the inflammation-prone site of the metacarpal bone shaft to be maintained in RA patients. However, the cross-sectional geometry of the metacarpal shaft was altered in RA patients in that the cortex was thinner but the bone outer circumference was greater in RA patients compared to healthy controls. The consequence of this “outward drift” is improved bone strength (polar SSI). These geometric changes are consistent with the morphologic alterations found with aging, although they seem to be enhanced in the RA group.

To date, only a few studies have explored the muscle–bone relationship in adult RA patients. One study found femoral neck areal BMD to be related to quadriceps strength in female RA patients (30). In postmenopausal women with early RA, grip strength has been found to be reduced compared to healthy controls and also to osteoporotic women (31). Furthermore, in the RA group, grip strength was correlated with total body areal BMD (31). These studies corroborate our results of the largely maintained muscle–bone relationship in RA patients. We found metacarpal cortical CSA to be slightly but insignificantly reduced in RA patients when corrected for forearm muscle CSA (Table 2 and Figure 1A). We measured forearm muscle CSA as a surrogate for muscle forces acting on the metacarpal bones rather than grip strength because we expected it to be less dependent on the highly variable pain in these patients. In fact, forearm muscle CSA has been found to be a better predictor of radius bone strength than grip force (32).

While bone mass (approximated here by cortical CSA) was largely adapted to muscle mass (approximated here by muscle CSA), this smaller (by trend) bone mass was distributed further away from the central axis of the metacarpal bone shaft, resulting in a thinner cortex and a larger outer bone circumference. An intact muscle–bone relationship but altered bone shaft geometry was also found in patients with juvenile idiopathic arthritis (JIA) (33, 34). These studies also found a thinner cortex but greater outer bone circumference and polar SSI at the radial shaft measured by peripheral QCT.

Some of the thinning of the cortex and increase in outer bone circumference may have been purely related to aging. Age-related thinning of the cortex and an increase in the outer bone circumference has been documented in healthy populations at the shafts of the tibia (35, 36), radius (37), and metacarpal bone (18, 38). From our data, a mean age-related increase in metacarpal total CSA of approximately 1 mm2 per decade in healthy controls can be derived (data not shown). This is comparable to the increase of approximately 0.1 mm per decade found in a normative data study for metacarpal bone width in men (18). However, we found significantly steeper slopes for age-related decreases in cortical CSA, cortical thickness, and the MI in the RA compared to the control group (Table 2 and Figures 2A and B). A possible explanation may be an inflammation-driven endosteal resorption–periosteal bone formation drift. Accordingly, studies in other inflammatory disease populations such as JIA (34) experimentally inducing inflammatory knee arthritis (39, 40), adjuvant-induced arthritis (41), and collagen-induced arthritis (42) have documented inflammation-driven endosteal resorption. Periosteal bone formation has been found in human (34) and animal studies (40, 41), with one of these studies observing that periosteal bone formation occurred after the endosteal resorption, when the inflammation had already subsided (41). This suggests that periosteal bone formation may be a compensatory mechanism to restore bone strength. While inflammation seems to be the likely culprit for the endosteal bone resorption, our study cannot identify whether this process is primary or secondary to inflammation. When inflammation is accompanied by pain, disuse may be a consequence of pain and therefore, bone resorption may be a consequence of disuse rather than inflammation as such. This question needs to be addressed in future longitudinal studies. Likewise, a different bone morphologic genotype is unlikely to be responsible for the greater outer metacarpal shaft diameter at a concomitantly rather smaller stature and smaller forearm muscle CSA (Table 1) in the RA group. In a recent study, we found total CSA at the tibial shaft not to be increased in RA patients compared to healthy controls (13).

The clinical significance of our results does not only relate to metacarpal bone assessment by peripheral QCT, but also more commonly applied methods such as dual x-ray absorptiometry, DXR, or other numerical assessments of radiographs. These 2-dimensional imaging methods cannot quantify cross-sectional bone geometry, but they can approximate cortical thickness and total bone width from the anteroposterior views and use these measures to calculate the MI (cortical bone width/total bone width) and areal BMD (8). Any measured areal BMD is directly dependent on bone shaft geometry, i.e., the same bone mass with the same volumetric BMD distributed further away from the central bone shaft axis will result in a lower areal BMD. Therefore, areal BMD at the metacarpal shaft is forcibly lower in RA patients compared to controls because of their altered shaft geometry. It is noteworthy that the reduced MI in our female and male RA group was partly due to a reduced cortical thickness (−14% in women and −17% in men) (Table 2), but also due to an enlarged total CSA (5% in women and 6% in men) (Table 2).

The clinical implications of our detailed assessment of metacarpal bone shaft geometry in RA patients are that therapies that maintain or restore forearm muscle mass should be encouraged because metacarpal shaft bone mass is proportional to it. This may be achieved by an optimal pain management that will allow the patient to fully use the affected hand/arm. However, it remains to be elucidated whether the reduced appendicular lean mass in RA patients is a consequence of disuse or is cytokine driven (43). Due to their adverse effect on muscle mass (44) and bone structure (45, 46), steroids should be replaced, if possible. Furthermore, our results may help to explain why metacarpal shaft fractures are not a common complication in RA patients. Periosteal bone formation increases the bone's resistance against bending and torsion (a thinner cortex has to be further away from the bone's central axis to achieve the same polar or areal moment of inertia), so periosteal bone formation drift may indeed be a compensatory mechanism to avoid bone failure. However, despite the fact the polar SSI was improved, bone strength in our study population may not have been truly improved compared to the control population because, for example, the risk for buckling increases when the cortex becomes thinner.

Limitations of the present study are the fact that only 15 men with RA were included in the present study. A larger group of men may have shown significant interaction terms between sex and forearm muscle or sex and age. However, the main findings of a normal muscle–bone relationship and abnormal bone geometry at the metacarpal shaft of RA patients would not be changed. Second, there was a strong correlation between muscle CSA and sex in the models; however, similar results were found when sexes were analyzed separately. Third, it should be mentioned that the model fits were slightly improved if quadratic terms were added in the models for cortical thickness and the MI. Since the improvements were only marginal, we decided to omit the quadratic terms in favor of interpretability of the linear models. Finally, some of the variability in the metacarpal bone shaft geometric data of our RA group can most probably be ascribed to varying treatments with biologic agents, bisphosphonates, and glucocorticoids. However, our study population reflects a typical clinical population and to single out the effects of different treatments would require a much larger study population.

In conclusion, the muscle–bone relationship at the third metacarpal bone was found to be normal in RA patients, but metacarpal bone shaft geometry was altered in RA patients compared to healthy controls. The metacarpal cortex was thinner and the outer bone circumference was larger in the RA patients, corresponding to an enhanced aging pattern. The consequence of these geometric alterations was that bone strength at the metacarpal shaft was increased in RA patients.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Eser 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 conception and design. Eser, Aeberli, Villiger.

Acquisition of data. Eser, Aeberli, Widmer, Möller.

Analysis and interpretation of data. Eser, Aeberli, Möller, Villiger.


We would like to thank all of the participating patients and volunteers. We thank Ms Jeannette Colosio for conducting some of the peripheral QCT measurements. Dominic Schuhmacher from the Institute for Mathematical Statistics of the University of Bern advised us with statistical analyses.