Article Text
PDF
Abstract
Background Radiological damage is an important outcome measure in rheumatoid arthritis (RA), both for research and clinical purposes. Depending on the setting, both hands and feet are radiographed, or only a part of these. It is unknown whether radiographing part of the four extremities gives comparable information to radiographing both hands and feet. This study therefore aimed to compare the radiological information obtained both when evaluating single time point radiographs and progression over time, in early and advanced RA.
Methods 6261 sets of hands and feet x-rays of 2193 RA patients from Leiden, Groningen (both from The Netherlands) and North America were studied. Correlations between joint damage at different regions were compared (unilateral vs bilateral and hands vs feet). Analyses were done at single time points (cross-sectional) and for progression over time (longitudinal), both for continuous severity measures (Sharp/van der Heijde score; SHS) and binomial measures of erosiveness.
Results When studying single time points, the severity of joint damage (SHS) is highly correlated between left and right, but weakly correlated between hands and feet. Correlation coefficients were higher in advanced than early RA. These findings were comparable in the three datasets. When evaluating erosiveness using only unilateral x-rays or hands without feet, 19.3% and 24.0–40.4% are incorrectly classified as non-erosiveness. Similarly, when evaluating disease progression by imaging only unilateral x-rays or only hand x-rays, progression would have been missed in 11.6–16.2% and 21.2–31.0% of patients.
Conclusion Performing x-rays of both hands and feet yields additive information compared with imaging only a part of these.
Statistics from Altmetric.com
Joint damage is a hallmark of rheumatoid arthritis (RA) that is commonly measured using x-rays of hands and feet. For scientific research purposes, x-rays provide an important outcome measure on the severity of RA. It is reflective of the cumulative burden of inflammation over time and can be quantified using validated scoring methods. In clinical practice x-rays are taken to evaluate treatment efficacy, because radiological joint damage may progress both in the presence and absence of low disease activity scores.
In some hospitals radiographic protocols are such that complete sets of hands and feet are always portrayed, whereas in other hospitals only the regions with most complaints are depicted. In scientific studies it is common that bilateral extremities are radiographed, although this can concern both hands and feet1,–,3 or only both hands.4 If there is sufficient proof that radiographing only one extremity is as informative as radiographing both hands and feet, this would imply that radiographing in daily practice can be done with less exposure to radiation and less cost. In addition, this issue is relevant for scientific studies when patients are included that have radiographs made of only part of the extremities and the question arises as to whether missing radiological data can be imputed.
Scientific data on the advantages of radiographing all hands and feet compared with radiographing only a part (either the most painful site or only the upper extremities) are, to the best of our knowledge, lacking. The results of recent MRI studies increased our interest in this issue. MRI evaluations are generally performed on one hand and/or one foot and seldom on bilateral extremities because of practical concerns. Current data indicated that there are no major differences in joint damage between dominant and non-dominant hands measured with MRI,5 suggesting that imaging of a unilateral extremity may be sufficient. This led us to perform the present study in which we aimed to compare the radiological information that is obtained when all hands and feet, or only a part of these are radiographed. We studied this with regard to single time point radiographing as well as with regard to evaluations of the progression of joint damage over time, for continuous and binominal radiological outcome measures, and in early and advanced RA.
Patients and methods
Study population
Three datasets consisting of 6261 sets of hand and feet x-rays belonging to 2193 adult RA patients in early and advanced disease stages were studied. RA was defined according to the 1987 American College of Rheumatology criteria.6
Leiden early arthritis clinic cohort
Six hundred and seventy-eight early RA patients originating from the western part of The Netherlands, who were consecutively included in the Leiden Early Arthritis Clinic (EAC) cohort between 1993 and 2006, were studied. Patients had less than 2 years of symptoms. The median symptom duration at inclusion was 4.3 months. Mean age was 56.6 years and 457 (67.4%) were women. Hands and feet x-rays were taken at baseline and on yearly follow-up visits for 7 years. All x-rays were chronologically scored by one experienced reader using the Sharp/van der Heijde score (SHS).7 The intraclass correlation coefficient (ICC) within the reader was 0.91. The mean total SHS was 8.7 and the mean hands SHS was 5.74 at the first visit.
Groningen cohort
The second set of radiological data were obtained from 261 RA patients from the northern part of The Netherlands (Groningen). These patients were radiographed between 1965 and 2010. The mean age at diagnosis was 45.1 years and 177 (67.8%) were women. Hands and feet x-rays were available over a follow-up duration of at most 25 years, with a mean number of x-rays per patient of 3.1 (with a maximum of eight x-rays per patient). The x-rays were scored chronologically by one of two readers using the SHS. ICC within readers were both greater than 0.90 and between readers 0.96. The mean total SHS was 12.0 and the mean hands SHS was 6.24 at the first visit.
North American dataset
One thousand two hundred and fifty-four RA patients deriving from different North American centres, who were radiographed between 1975 and 2011 were studied. x-Rays were made of hands but not feet. The mean age at the time of the diagnosis was 48.9 years; 981 patients (78.2%) were women. x-Rays were available over a follow-up duration of at most 25 years with a mean number of x-rays per patient of 2.0 (with a maximum of eight x-rays per patient). The x-rays were scored with known time order by one experienced reader; the within-reader ICC was 0.99. The mean hands SHS at the first visit was 14.9.
Study design
This study was divided into two parts. First, we compared the severity of joint damage (expressed by the SHS) at different regions, both when evaluating single time points cross- sectionally and when evaluating the progression of joint damage over time. This was done to explore how well the quantity of radiological damage at one site was correlated with that at the other site.
Second, it was studied whether classifying a patient as having erosive disease was different when both hands and feet x-rays were studied or only a part of them. Also here, x-rays obtained at a single time point were evaluated cross-sectionally and progression of erosiveness was evaluated longitudinally. Erosiveness was defined as having two or more erosive joints in the cross-sectional analysis. This cut-off was chosen based on data showing that the presence of two or more erosive joints is more predictive of RA than the presence of one erosive joint.8 In the longitudinal analyses, progression was expressed as a binomial outcome. Progression of erosiveness was defined as an increase in the erosion score of 2 points or more in a 1-year interval.9 Differences in the classification of erosiveness if both hands and feet or a part of these are radiographed were evaluated. As sensitivity analyses, other cut-offs for progression were also studied.
Statistical analyses
The severity of joint damage at different regions on x-rays of single time points was compared using Pearson correlation coefficients. This was done for the individual follow-up years in the Leiden RA patients. As there was no yearly follow-up schedule in the Groningen and North American datasets, follow-up years were combined into ‘year strata’, to include a larger number of patients per time point. Here x-rays taken at years 1–4, years 5–9 and years 10–25 were combined. x-Rays taken at years 10–25 in the Leiden EAC were also combined. As a high correlation coefficient does not by definition mean high agreement, the level of agreement was tested by plotting data in Bland–Altman plots.10
Analyses on the progression of the severity of joint damage were performed in Leiden EAC only, as it contained sufficient radiological information for the subsequent years of follow-up.
The presence of erosiveness at different sites was compared by cross-tabulations. Patients who would be classified as having erosive disease when evaluating x-rays at all sides but classified as non-erosive when evaluating single locations were labelled as misclassified patients. Similar analyses were done with regards to the progression of erosiveness.
All analyses were performed using SPSS version 17.0. p Values less than 0.05 were considered significant.
Results
Comparisons on severity of joint damage
First it was evaluated how well the severity of radiological damage (expressed by the SHS) at one site was correlated with radiological damage at other sites. This was done both when radiological joint damage at single time points was studied and when progression over time was evaluated.
Comparison at single time points
When evaluating the SHS of the Leiden RA patients, it was observed that correlation coefficients were rather high between left and right comparisons (table 1). In addition, with increasing follow-up duration the correlation coefficients increased. However, the correlation coefficients of comparisons of hands with feet were moderate at all points in time (table 1). Analyses in the Groningen and North American datasets also showed that the level of joint damage at the left and right sides were fairly comparable, although the correlation coefficients in Groningen did not increase during follow-up. Similar to the Leiden data, in the Groningen data, the correlation of the severity of joint damage between hands and feet was moderate (table 2).
Correlations between the severity of joint damage (according to the SHS method) at different sites and different disease durations in Leiden RA patients
Correlations between the severity of joint damage (according to the SHS method) at different sites and different disease durations in Groningen and North American RA patients
As the total SHS is composed of information on the severity of erosions and joint space narrowing, analyses were repeated for the total erosion score and total joint space narrowing score separately, leading to findings comparable to those of the total SHS (tables 1 and 2).
Next it was questioned whether the high correlations in the left and right comparisons may have been influenced by the magnitude of the SHS. To evaluate this, the agreement between left and right total SHS was portrayed with Bland–Altman plots for all three datasets (see supplementary figures S1–S3, available online only). Good agreements were observed, indicating that the correlation between right and left total SHS was highly independent of the severity of joint damage.
Comparison of the severity of progression of joint damage
It was evaluated whether using only one hand or foot, two hands, two feet or both hands and feet provided comparable information to monitor the progression of joint damage. The progression of joint damage was measured quantitatively using the total SHS.
In Leiden RA patients two intervals were studied, progression during the first year of follow-up, and progression between year 6 and year 7. Correlation coefficients were determined (table 3).These were fairly high for left and right comparisons, but lower for hands and feet comparisons. Also here, Bland–Altman plots were made for left and right comparisons (see supplementary figure S4, available online only), showing that the correlation coefficients were independent of the magnitude of the SHS scores.
Correlations between the severity of progression of joint damage at different sites at different intervals in Leiden RA patients
Progression analyses could not be performed in the Groningen and North American datasets because too few patients had radiological data at pairs of subsequent follow-up years.
Comparison of the presence of erosiveness
In the clinical setting, continuous measures for joint damage are seldom used and radiographs are often reported on the presence of erosiveness. Therefore, we next evaluated to what extent identification of erosiveness was different when x-rays of all hands and feet were considered compared with considering a part of these x-rays.
Comparison at single time points
First, x-rays at individual time points were evaluated. Erosiveness was defined as having two or more joints with erosions. When studying the baseline x-rays of the Leiden RA patients, 67.4% of the patients had erosive disease. When only the hands were studied, 54.9% of the patients had erosive disease. Evaluating one hand instead of both hands would lead to incorrectly classifying a patients as having non-erosive disease in 19.3% of cases. This proportion was also 19.3% when the x-rays at year 7 were studied. When x-rays of both hands were made but not of the feet, 40% and 24% of the patients were incorrectly classified as having non-erosive disease at baseline and 7 years of follow-up, respectively (table 4).
Comparison of the frequency of erosive disease at hands and feet in Leiden RA patients at baseline and after 7 years of follow-up
In the Groningen and North American datasets, similar results were obtained (see supplementary tables S1 and S2, available online only). For instance, when evaluating the x-rays made during the first years of the disease, 19.8% and 23.2% of patients, respectively, would incorrectly be classified as non-erosive if only one of the two hands was evaluated. In addition, evaluating hands without feet in the Groningen set of patients would result in misclassification of 37.1% of the patients.
Comparison of the severity of progression of joint damage
Progression of joint damage can be expressed quantitatively using scoring methods such as the SHS method. Although this quantitative outcome measure is more informative than qualitative outcome measures,11 progression can also be expressed binomially. Here patients were identified as having progressive disease in case the erosion score increased for 2 points or more in a 1-year interval. When studying the first 1-year interval in the Leiden RA patients, it was observed that 25.7% of patients had progression in the right hand, 24.0% in the left hand and 38.6% in both hands (table 5). If only one hand was used to monitor progression for both hands, 16.2% of patients would be misclassified as having non-progressive disease. If x-rays of both hands without feet were used to monitor progression, 31.0% of patients would be misclassified as having non-progressive disease. When studying the interval between years 6 and 7, it was observed that smaller percentages of patients progressed; 11.6% of patients progressed in the right hand, 10.6% in the left hand and 21.2% in both hands. If, for instance, only one hand was used to monitor progression in both hands in this disease phase, 14.8% of patients would be misclassified as non-progressors. Similarly, monitoring both hands without feet would lead to such misclassification in 21.2% of patients.
Comparison of the frequency of progression of erosiveness at hands and feet in Leiden RA patients at baseline and after 7 years of follow-up
As sensitivity analyses, analyses were repeated when defining progression in erosiveness as an increase in erosion score of 3 or more or 1 or more in a 1-year interval. This gave comparable results (see supplementary table S3, available online only).
Discussion
In the current study, we aimed to examine the difference in information content obtained when radiographing all hands and feet versus only a selection of these. To this end, radiological joint damage data of 6261 sets of x-rays of 2193 RA patients in early and advanced phases of the disease were studied. It was observed that both absolute SHS scores and progression in SHS scores were highly correlated between left and right side x-rays, but that the correlations between hands and feet were much weaker. When evaluating qualitative outcome measures of joint damage (erosiveness), it was demonstrated that using only unilateral or only hands x-rays resulted in approximately 20% and 40% of patients that are incorrectly classified as non-erosive, as they had erosions on the non-radiographed site. Together, these data suggest that feet x-rays importantly add to the information obtained by hand x-rays, and that when joint damage is expressed qualitatively, the percentage of patients with erosive disease is underestimated.
A strength of this study is that it includes a large number of patients and x-rays that are scored by experienced readers. In addition, findings are obtained from datasets of two different populations and different levels of severity, which adds to the generalisability of the results.
The SHS, measuring the severity of joint damage quantitatively, was highly correlated between the left and right side, both at single time points and when evaluating progression over time. The Bland–Altman plots indicated that there is actually good agreement, which is independent of the level of joint damage. The correlation coefficients even improved at increasing disease durations in the Leiden and North American RA patients. In the Groningen dataset this trend was not seen, presumably due to the already high correlation at baseline in this dataset. The symptom durations at the first visit of the Groningen and North American RA patients are not known. It is therefore unclear whether baseline visits represent the same disease stage in the different cohorts.
The above-mentioned findings are relevant for studies in which missing radiological data are imputed.12 Our findings may support imputing missing SHS hands or feet data with information of SHS scores of the radiographed contralateral hand or foot. In contrast, based on these data it is not advocated to impute missing feet data with those of hands. It is important to note that this conclusion concerns missing SHS data in research studies. However, the data in this study demonstrate that for monitoring purposes in the clinical setting, it is better to x-ray both hands and both feet.
Two previous studies showed that patients in an early phase of RA had more feet than hand joints involved.13 ,14 We observed a higher frequency of erosive joints in hands than in feet joints, although this may have been influenced by the fact that according to the SHS method, more hand than feet joints are evaluated (32 vs 12, respectively). Several studies have described the value of foot x-rays to diagnose RA,14,–,18 and a recent study also concluded that both erosions and joint space narrowing should be assessed and that there are no redundant joints.15 Nonetheless, to the best of our knowledge, the present study is the first to evaluate the correlation between hands and feet joint damage.
In the clinical setting, continuous outcome measures such as the SHS are not used and radiographs are generally reported in terms of erosiveness. In an attempt to study erosiveness, continuous radiological data were categorised. It was observed that, when expressing the data in this way, radiographing a part of the extremities leads to a loss of information compared with radiographing both hands and both feet. Inherent to categorisation is the question of which cut-off defined erosiveness best. In the cross-sectional analyses a cut-off of two or more erosive joints was chosen as the definition of erosive disease. This was done based on a previous study.8 In order to categorise the progression of erosiveness, an increase of two or more erosive scores was used, which was also in line with previous studies.19 Other cut-offs were also evaluated and these analyses did not result in different findings.
Left and right comparisons were made, disregarding information on dexterity or the side of most complaints. Unfortunately, such data were lacking. van der Heijde et al13 previously observed that there was no difference in the SHS score between dominant and non-dominant hands.
The result of a recent study on MRI indicated that there is no major difference between dominant and non-dominant hands.5 In that study joint involvement was evaluated with the RAMRIS score and thus quantitatively and not qualitatively. Our findings are partly in line with this as the total SHS between left and right hands was highly correlated.
In conclusion, evaluations on the presence/absence of erosive disease is preferably done by radiographing both hands and both feet because of the risk of false-negative classification. The severity of joint damage is highly correlated between left and right but not between hands and feet.
Acknowledgments
The authors would like to thank Dr P K Gregersen at the Feinstein Institute for Medical Research and Dr F Wolfe at the National Databank for Rheumatic Diseases for providing access to x-ray data on RA patients from the USA.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Web Only Data - This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
-
Funding The work of RK is supported by the Dutch Arthritis Association. The work of AHM is supported by The Netherlands Organisation for Health Research and Development. This research has been funded by the European Community Seventh Framework Program FP7 Health-F2-2008-223404 (Masterswitch).
-
Competing interests None.
-
Ethics approval Ethics approval was obtained from the local ethics committees of each dataset.
-
Provenance and peer review Not commissioned; externally peer reviewed.