Introduction

Osteoarthritis is the most common form of arthritis among the elderly. It affects a large proportion of people aged over 50 years of age and its incidence increases with age [1–5]. In recent years, more and more attention has been focused on hand osteoarthritis (HOA), which is one of the most frequently affected sites [4].

The estimated prevalence of radiographic HOA in patients older than 65 ranges from 70 to 99% in women and 64 to 78% in men; the most frequently involved joints are the distal interphalangeal (DIP) and proximal interphalangeal (PIP), followed by metacarpophalangeal (MCP) joints [6].

Variable percentages of HOA patients may be affected by an erosive form of the disease which has a great impact on their daily life and patients affected by erosive HOA usually have more severe radiographic disease and higher functional impairment [7–14].

While interphalangeal (IP) joint involvement has been well documented [8, 10, 14], MCP and radiocarpal (RC) joint involvement has received less attention and is therefore less well-characterized.

The aim of our study was to evaluate MCP and RC joint radiographic involvement and its relationship with erosive disease; we also evaluated the progression of structural damage in MCP joints in a longitudinal study.

Patients and methods

We enrolled 368 patients with HOA. Enrolment was carried out by three secondary/tertiary level Rheumatology Centers (with special interest in HOA) between May 2004 and May 2006. Inclusion and exclusion criteria have already been reported [15]. In brief, patients with pain and/or bony enlargements in finger joints were selected and subjected to hand X-ray evaluation. None of the subjects had pain or swelling in MCP joints in order to exclude concomitant potential inflammatory joint diseases. Diagnosis of HOA was made according to ACR criteria before hand X-ray performance [16].

Patients with chronic inflammatory joint diseases, psoriasis, or other autoimmune disorders such as connective tissue diseases, uveitis, and inflammatory bowel diseases were excluded. In addition, patients with Calcium Pyrophosphate Deposition Disease (CPPD) and hemochromatosis or with history of hand joint fractures were excluded.

Patients were divided into three groups based on the presence of radiographic central erosions in their IP joints: group 0, no central erosion seen in either hand; group 1, only one joint with a central erosion detected; and group 2, two or more joints with central erosions found. Demographic data (age, age at disease onset, disease duration, and Body Mass Index (BMI)) were obtained and recorded, as previously described [10].

In addition to the cross sectional study, we performed a longitudinal study on 44 selected patients (39 females, 88.6%) whose X-rays were available after 3.9 years (mean ± sd = 3.96 ± 0.31) from the first exam. Radiographic changes in the OA patients were compared to the radiographic changes detected in nine normal controls, who were randomly selected from people attending our outpatient clinic for minor, non-specific complaints without finger joint pain and/or tenderness and no finger nodes, who had hand X-rays performed twice, 4 years apart (mean ± sd = 3.88 ± 0.38).

Posteroanterior X-ray examinations of both hands were performed in all patients, and the radiological involvement of the MCP and RC joints was evaluated.

Radiographs were scored by consensus opinion (jointwise at the same time) by two experienced readers (OA, CC) utilizing the Kellgren & Lawrence (K/L) scoring systems as described elsewhere [10, 17, 18]. Due to the absence of MCP and RC joints both in the OARSI atlas [19] and the Verbruggen [14] scoring systems, and in order to obtain a more detailed evaluation of OA radiographic features in MCP joints, we also separately scored Osteophytes (OST) and Joint Space narrowing (JSN) features according to the standard OARSI definitions already validated for IP joints, as well as the presence of Marginal Erosions (ME).

A total of 4416 joints were evaluated: 1548 in group 0, 492 in group 1, and 2376 in group 2. In the follow-up study, we evaluated 528 joints from HOA patients and 108 from NCs twice.

Written informed consent was obtained from the patients and the NCs as was approved by the Ethics Committee of our hospital.

Statistical analysis

All the statistical analysis was performed using SPSS v.19.0 (IBM Corp., Armonk, NY, USA).

All continuous data were expressed in terms of mean and standard deviation of the mean; the categorical data were expressed as frequencies and percentages. The estimated values adjusted for within-patient effects, age, sex, and BMI, were expressed as expected value and 95% confidence interval (CI). The Kolmogorov-Smirnov test was performed to test the normality of continuous variables. Levene’s test was used to test homoscedasticity. At the patient level, we used the ANOVA test followed by the Scheffè post hoc pairwise comparison to assess differences in continuous, normally distributed and homoscedastic data among the three groups; the Kruskal-Wallis test followed by the Mann-Whitney test with the Bonferroni correction was used for multiple comparisons for not normally distributed or heteroscedastic data, the Chi square test evaluated by Exact Methods for small samples to investigate the relationships between categorical variables and the three groups, and the Jonckheere-Terpstra test evaluated by Exact Methods for small samples to investigate the relationships between ordinal variables and the three groups.

At the joint level, the Generalized Linear Model for repeated measures with a Gamma distribution adjusted for within-patient, sex, age, and BMI was utilized to assess changes in follow-up analyses. The Generalized Linear Model with a Gamma distribution (continuous variables) and a Poisson distribution (dichotomic variables) adjusted for within-patient, sex, age, and BMI was utilized to assess differences among the groups.

The intraclass correlation coefficient (ICC) with a 95% CI was performed to assess the intrareader reliability for the K/L score. Cohen’s kappa with a 95% CI was used to assess intrareader reliability for the scoring of each feature in each single joint using the OARSI score. Values of 0.8 were considered as excellent [20].

For all tests p < 0.05 was considered significant.

Intrareader variability was obtained on examination of 20 randomly selected radiographs and showed excellent results (total K/L in each patient ICC = 0.994, 95% CI 0.992–0.995).

Furthermore, intrareader reliability for each item on the radiographic score showed excellent results [(osteophytes Cohen-Kappa (k) = 0.98, 95% CI (0.97–0.99); joint space narrowing k = 0.95, 95% CI (0.92–0.97); erosions k = 0.90, 95% CI (0.84–0.96)].

Results

On the basis of the presence of central erosions, patients were sub-grouped as follows: group 0, 129 (35%) patients; group 1, 41 (11.2%) patients; and group 2, 198 (53.8%) patients.

Significant differences in age at disease onset were found between patients in groups 0 and 2 and between patients in groups 1 and 2, whereas we did not find significant differences between patients in group 0 and patients in group 1 (Table 1). Similarly, significant differences in age were found between patients in groups 0 and 1, between patients in groups 0 and 2, and also between patients in groups 1 and 2 (Table 1).

Table 1 Patients demographics
Full size table

No differences were found in gender and BMI distribution among the three groups (Table 1).

Cross-sectional study

Among the three groups, only a very low number of joints showed K/L values ≥2 (Fig. 1a) 42/1290 (3.3%) in group 0, 10/410 (2.4%) in group 1, and 36/1980 (1.8%) in group 2.

Fig. 1
figure 1

a Metacarpophalangeal joint with Kellgren & Lawrence score ≥2. b Metacarpophalangeal joint with marginal erosions

Full size image

Low mean score values were obtained for all the radiographic features. Only the JSN score showed a significant increase from groups 0 to 2, although when adjusted for age, sex, and BMI statistical significance was not confirmed (Table 2).

Table 2 Mean per patient values of radiographic features of MCP joints
Full size table

No significant differences were found when performing “per joint” evaluations of the severity of radiographic damage among the ten MCP joints (Supplementary Data—Table 1S).

Marginal erosions (ME-Fig. 1b) were rarely found with percentages of positivity ranging from 3% (2nd right MCP) to 11% (4th right MCP)—mean percentage 6.7%. No preferential localization of ME was found (metacarpal heads vs basal phalanges) (Table 2).

No RC joint with a K/L score ≥2 was observed. Very low scores were obtained for all the items utilized, without any differences among the three patient groups. (Table 2S/Supplementary Data)

Longitudinal study

To evaluate the progression of radiographic damage, 44 HOA patients (Females 39 (88.6%)) and nine NCs were re-evaluated after a period of about 4 years of follow-up. Normal controls showed similar demographic characteristics: gender 7 females (77.8%), age (years, mean ± sd) 66.8 ± 9.0 and BMI (mean ± sd) 24.9 ± 3.4.

Worsening of erosive disease in IP joints was found. The number of patients in groups 0 and 1 decreased, and the number of patients in group 2 increased accordingly (Table 3).

Table 3 Progression of erosive disease at follow-up
Full size table

A total number of 690 joints were evaluated both at baseline and at follow-up—582 in the OA patient groups and 108 in NCs.

In HOA patients, both MCP and wrist per patient radiographic scores had worsened at follow-up, the highest differences being detectable in MCP joints (Table 4).

Table 4 Per joint-group features of radiographic follow-up in HOA patients
Full size table

The increase of MCP joints with K/L ≥2 did not reach statistical significance (baseline 4/440 (0.9%) vs follow-up 8/440 (1.8%), p = 0.245).

In NCs, no significant progression in radiographic scores was found (Table 5).

Table 5 Per joint-group features of radiographic follow-up in NCs
Full size table

Discussion

To our knowledge, no extensive reports concerning MCP joint radiographic features have been published with the aim of evaluating their radiographic involvement in HOA, especially when differentiating between erosive and non erosive disease.

Our study demonstrates that

  • MCP joints are involved in the OA process albeit in a milder form. No preferential location among the ten MCP joints was observed

  • JSN may be the only feature related to erosiveness

  • Wrist involvement can be considered negligible or absent

  • A significant worsening of OA scores in follow-up compared to NCs is found

In a population-based study, Kalichman and coworkers demonstrated that MCP are the first joints to undergo OA changes until the sixth decade of life, then DIP joints become the most affected [4]. In another population-based study, Dahaghin et al. showed MCP joint involvement (K/L ≥2) in 8.2% of joints affected by HOA [8]. More recent papers have confirmed that MCP joint involvement is present but less frequent than DIP and PIP joint involvement. [21, 22]. A higher MCP OA prevalence was shown in another population-based study by Haugen et al. from the Framingham Osteoarthritis study cohort, where the observed MCP involvement was of 11.9% in men and 6.8% in women; furthermore, wrist involvement was demonstrated more frequently in men than in women. Interestingly, signs of previous wrist fractures were present in 17.5% of wrist OA [7].

We studied an HOA population diagnosed on the basis of ACR criteria. When evaluating radiographic features, we have to remember that a patient can be considered affected by osteoarthritis when at least one joint shows a K/L score ≥2, and we found that only a small minority of MCP joints can be classified as having osteoarthritis.

Our study showed low mean score values for all radiographic features. It is also worth noting that ME were rarely found and without correlation with bone erosions in IP joints. In DIP and PIP joints, higher percentages of ME were demonstrated (between 18 and 77%)

[10]. On the other hand, we have to mention a recent paper published by Punzi et al. on erosive osteoarthritis. We may need to consider the potential localisation of such erosive features in joints other than IP joints of the hands, such as thumb base and facet joints as well as MCP joints [23].

Since we did not perform a population-based study, we cannot compare our results with the previously mentioned papers. In our series, only a small minority of patients were male. On the other hand, the previously recognized higher prevalence of MCP OA in males was described in populations which were not screened for disorders which are commonly associated with MCP OA (i.e., hemochromatosis and CPPD). These disorders were exclusion criteria in our population enrollment.

Since IP joint structural damage has been demonstrated to be more severe in patients with erosive disease [10], we also asked whether MCP joint structural damage was more severe in the erosive patients. We found that only JSN and not OST score or bone erosion presence were correlated to erosiveness. It is worth remembering that in inflammatory arthritis, JSN and not bone erosions is correlated to clinical burden. [24, 25] Likewise, we might hypothesize that a similar process could be involved in OA. Our data demonstrate that JSN, more than OST or ME, might be correlated with erosiveness and accordingly with a more severe disease, suggesting a role for cartilage pathology in disease progression.

In our study, wrist involvement was minimal with no significant score values, and therefore it might be considered negligible/absent. The only recently published paper dealing with wrist involvement in HOA [7] also found minor wrist involvement and a correlation with a history of fractures. None of our patients had a history of wrist fractures.

In order to obtain a more complete picture of MCP joint OA, 44 patients from our series were re-evaluated after a period of 3.9 years of follow-up, together with nine NCs. A significant worsening of OA scores in follow-up compared to baseline data was seen. It is noteworthy that NCs did not show a similar worsening, thus suggesting that aging per se is not responsible for structural damage progression.

Nonetheless, the progression of radiographic scores was mild, even if significant, and it did not entail a significant increase in the joints with a K/L ≥2. It is conceivable that a more prolonged follow-up would be associated with a larger number of new joints reaching a K/L score ≥2.

No significant progression was seen in RC joints, thus confirming almost no wrist joint involvement in HOA.

Our study has some limitations. It was not a population-based study, it only evaluated patients with an established diagnosis of HOA. The follow-up study involved a low number of patients and an even lower number of NCs, and the follow-up time was limited. Again we did not have clinical data on the Australian/Canadian Osteoarthritis Hand Index (AUSCAN) or Dreiser’s algo-functional finger index scores or about patients’ professions which might help in making correlations with MCP involvement. However, it is the first follow-up study on MCP joints. As regards the radiographic scores, the K/L system is well utilized for IP as well as for MCP joints; on the other hand, to date, no other paper has scored the single items JSN and OST in MCP joints. In order to do that, we applied the OARSI Atlas definitions validated for IP joints to MCP joints. Therefore, our results should be interpreted with some caution.

A strong point of our paper is the careful selection of the patients, which allowed us to study a large case series of subjects without concurrent joint disorders.

In conclusion, MCP involvement in HOA is mild, but progressive, while wrist involvement is negligible. The role of MCP involvement in causing pain and function disability should be addressed in future studies.