To study the prevalence of cervical spine subluxations and predictive factors for atlantoaxial subluxations (including anterior atlantoaxial subluxation and atlantoaxial impaction, i.e., vertical subluxation) in patients with rheumatoid arthritis (RA) who were treated early and continuously with disease-modifying antirheumatic drugs for 8–13 years.
Radiographs of the cervical spine were obtained in 103 of 110 patients (the 110 surviving patients of the original 135-patient cohort) at their 8–13-year followup visits. The prevalence of cervical spine subluxations was determined. Demographic variables and the first 5-year serial data concerning disease course were analyzed in a logistic regression model to find predictive factors for atlantoaxial subluxations.
Atlantoaxial subluxations were found in 14 patients (14%), and 5 patients (5%) had subaxial subluxations. Older age at baseline, greater disease activity during the first 5 years, and early erosiveness in peripheral joints predicted the development of atlantoaxial subluxations. Patients who had ≥10% of the maximum possible radiographic damage (by Larsen score) in peripheral joints at 5 years were 15.9 times more likely to develop atlantoaxial subluxations at 8–13 years than patients whose peripheral joint damage remained <10% of the maximum.
Compared with historical control RA cohorts, a lower prevalence of cervical spine destruction was found in the present group of patients. Rapid erosiveness in peripheral joints was the best predictor for atlantoaxial subluxations. Extensive erosiveness in peripheral joints should alert rheumatologists to the possible development of atlantoaxial subluxations in patients with RA.
Cervical spine disorders are common consequences of severe rheumatoid arthritis (RA) (1). The most common characteristics of the cervical spine disorders in RA are horizontal and vertical subluxations of the atlantoaxial joint, i.e., anterior atlantoaxial subluxation (aAAS) and atlantoaxial impaction (AAI) (2). In addition, rheumatoid inflammation of the facet joints below the second cervical vertebra may result in subaxial subluxation (SAS), which can, however, be difficult to distinguish from subluxation caused by degeneration of the spine (3). We have previously reported that up to 42% of patients treated with conventional therapy exhibit aAAS, AAI, or SAS after 20 years of RA (4). We also recently showed that aggressive treatment of RA with a combination of disease-modifying antirheumatic drugs (DMARDs) retards or prevents early development of atlantoaxial subluxations (5). However, very little is known of the long-term prevalence of RA-related cervical spine changes in patients who are treated actively with DMARDs from the time of diagnosis.
The rheumatoid disorders of the cervical spine are known to be associated with the extent of erosions in the peripheral joints and the HLA status, as well as with the presence of rheumatoid factor (RF) and rheumatoid nodules (6–10). Continuous disease activity predicts the development of erosions in peripheral joints (11, 12). However, no formal studies have been conducted to determine whether continuous disease activity predicts cervical spine disorders in RA.
In the present study, we evaluated the 8–13- year–point prevalence of cervical spine subluxations in patients with RA who were treated with DMARDs according to the “sawtooth” strategy (13) from the early stages of the disease. In addition, we studied predictive factors for the development of atlantoaxial subluxations in these patients.
PATIENTS AND METHODS
Initially, a total of 135 patients with recent-onset RA were recruited into two separate cohorts. The first cohort of 58 patients was collected in 1983–1985 to study the erosiveness of RA (14). The second cohort of 77 patients was enrolled in 1988–1989 to investigate the efficacy and tolerability of sulfasalazine in early RA (15). All patients were subsequently enrolled in a longitudinal followup study to evaluate the benefits of early and continuous therapy with DMARDs (the “sawtooth” strategy ) on the long-term outcomes (16, 17). The treatments of these patients with DMARDs are described in more detail elsewhere (18). At enrollment, all patients met the American Rheumatism Association (ARA) 1958 criteria for definite or classic RA (19) as well as the American College of Rheumatology (formerly, the ARA) 1987 revised criteria for RA (20). All patients were clinically assessed every 6 months for the first 2 years and at least yearly thereafter. Twenty-five patients had died (21). Therefore, in this report we present the cervical spine findings in 103 of 110 patients who had radiographs of the cervical spine obtained in 1997.
Cervical spine radiographs.
Lateral-view cervical spine radiographs (during flexion and extension) were taken using a 150-cm tube-to-plane distance. A diagnosis of aAAS was made if the distance between the anterior aspect of the dens and the posterior aspect of the anterior arch of atlas was >3 mm during flexion. Lateral-view radiographs obtained during flexion were used for evaluation of AAI. AAI was diagnosed using the Sakaguchi-Kauppi (S-K) method, which has been developed in particular for screening purposes and is used for evaluating the position of the atlas in relation to axis (22). The S-K method divides the condition into 4 grades (grade I represents normal and grades II–IV represent abnormal). Since aAAS and AAI are typical cervical spine changes in RA, they are referred to as atlantoaxial subluxations in the present report. SAS was diagnosed if a vertebra had moved more than 3 mm in relation to the next vertebra when measured from the posterior line of the vertebral bodies.
Clinical examination, performed every 6 months during the first 2 years and yearly thereafter, included laboratory tests (erythrocyte sedimentation rate [ESR], C-reactive protein [CRP] and hemoglobin levels, and RF [at least at baseline]), swollen joint counts (SJCs; 28-joint count) (23), radiographs of hands and feet, and patient questionnaires. To get an estimate of overall disease activity during the first 5 years of disease, the area under the curve (AUC) was calculated for ESR, CRP level, and SJC. Unfortunately, during the first 5 years, different questionnaires were used in the two cohorts. Therefore, the questionnaire data could not be included in the analyses.
Radiographs of hands and feet were obtained in posterior–anterior projection at the beginning of the followup and once a year thereafter. The Larsen score of 0–100 was applied to grade structural damage of the wrists, the first-through-fifth metacarpophalangeal joints, and the second-through-fifth metatarsophalangeal joints at baseline and at 1, 3, and 5 years (24–26). All radiographs from a given patient were evaluated at the same time according to a method we have previously described (17), in chronological order and without information regarding the clinical status of the cervical spine.
Data analyses were performed with SPSS 11.0 software (SPSS, Chicago, IL). The baseline differences between the groups (patients who developed atlantoaxial subluxations later versus those who did not) were computed by two-way analysis of variance for continuous variables and by logistic regression for noncontinuous variables, adjusted for age and sex when appropriate.
A logistic regression model was used to explore predictive factors for atlantoaxial subluxations. The AUC values over 5 years for ESR, CRP level, and SJC as well as the Larsen scores (as a continuous variable) at baseline and at 1, 3, and 5 years were entered separately into a model, adjusted for age and sex. All these variables were statistically significant in the model, except for the Larsen score at baseline. A multivariate logistic regression model was then used with the AUC values over 5 years for ESR, CRP level, and SJC as well as the Larsen score at 5 years as covariates. The 5-year Larsen score as a continuous variable appeared to be the only statistically significant risk factor for atlantoaxial subluxations in a multivariate analysis. Therefore, in order to get a clinically relevant interpretation of the results of the regression model, a dichotomized 5-year Larsen score (damage ≥10% of maximum versus <10% of maximum) was used in the final model, adjusted for age and sex. The study was approved by the ethics committee of Jyväskylä Central Hospital.
Atlantoaxial subluxations were found in 14 patients (14%), 9 of whom had aAAS only, 4 of whom had AAI only, and 1 of whom had a combination of aAAS and AAI (Table 1). Two of these patients had undergone surgery because of severe aAAS. In all other patients, the atlantoaxial subluxations were less severe, i.e., aAAS ≤7 mm and AAI S-K grade II–III.
Table 1. Prevalence of cervical spine subluxations in 103 patients with rheumatoid arthritis 8–13 years after disease onset*
SAS was detected in 5 patients, including 2 who had SAS in two levels of cervical spine and 2 others with a combination of AAI and SAS. The extent of SAS was <5 mm in all patients.
The baseline demographic and clinical variables in the 14 patients with atlantoaxial subluxations compared with those in the other 89 patients are shown in Table 2. The patients who later developed atlantoaxial subluxation were older than the other patients. Further, values of the disease activity measures at baseline were higher in patients who would later develop atlantoaxial subluxations than in those who would not, but the differences did not reach statistical significance. Comparatively higher disease activity values were also found at the time of radiographic evaluation of the cervical spine in patients who developed atlantoaxial subluxations (Table 2). Interestingly, the duration of RA (at the 8–13-year followup) was not associated with the occurrence of atlantoaxial subluxations.
Table 2. Demographic and clinical characteristics at baseline, and clinical characteristics at radiographic evaluation of the cervical spine at 8–13-year followup visit, of patients who did and those who did not develop atlantoaxial subluxations*
Patients with atlantoaxial subluxations (n = 14)
Patients without atlantoaxial subluxations (n = 89)
Except where indicated otherwise, values are the mean (median). ESR = erythrocyte sedimentation rate; CRP = C-reactive protein; RF = rheumatoid factor; HAQ = Health Assessment Questionnaire; VAS = visual analog scale.
Two-way analysis of variance was performed for continuous variables, and logistic regression was performed for noncontinuous variables, adjusted for age and sex when appropriate.
Duration of symptoms before diagnosis, months
Formal education ≥9 years, %
Disease activity measures
CRP level, mg/liter
No. of swollen joints, 28-joint count
Peripheral joint Larsen score, 0–100
At 8–13-year followup
Duration of disease, years
Disease activity measures
RF positive, %
No. of swollen joints, 28-joint count
No. of tender joints, 28-joint count
HAQ, 0–3 scale
Pain, 0–100-mm VAS
Global health, 0–100-mm VAS
Peripheral joint Larsen score, 0–100
In a univariate logistic regression model with adjustment for age and sex, the AUC values for the disease activity measures ESR, CRP level, and SJC showed them to be statistically significant risk factors for atlantoaxial subluxations, as were the Larsen scores at 1, 3, and 5 years (Table 3). In a multivariate model including the AUC values over 5 years for ESR, CRP level, and SJC as well as the Larsen score at 5 years, the Larsen score was the only statistically significant predictor for atlantoaxial subluxations. Patients with a Larsen score ≥10% of the maximum at 5 years were 15.9 times more likely (95% confidence interval [95% CI] 3.38–74.7) to develop atlantoaxial subluxations at 8–13 years than patients whose peripheral damage remained <10% of the maximum (Table 4).
Table 3. Odds ratios (ORs) and 95% confidence intervals (95% CIs) for atlantoaxial subluxations, by univariate and multivariate logistic regression analysis*
OR (95% CI)
OR (95% CI)
All analyses are adjusted for age and sex. AUC = area under the curve; SJC = swollen joint count (see Table 2 for other definitions).
Increasing age in years
Men vs. women
For CRP level
At 1 year
At 3 years
At 5 years
Table 4. ORs for atlantoaxial subluxations, by logistic regression analysis in patients with peripheral joint damage ≥10% of the maximum possible Larsen score at 5 years*
The proportion of patients who subsequently developed atlantoaxial subluxations, according to Larsen score at 1, 3, and 5 years, is illustrated in Figure 1. Four of the 7 patients (57%) whose Larsen score at 1 year was ≥10% of the maximum developed atlantoaxial subluxations later. At the 3- and 5-year visits, 31–32% of the patients whose Larsen score was ≥10% of the maximum would later develop atlantoaxial subluxations, while only 8% and 4% of those whose Larsen score remained <10% of the maximum at 3 years and 5 years, respectively, developed atlantoaxial subluxations during the 8–13-year followup.
The present study confirms the previous observation that atlantoaxial subluxations of the cervical spine are associated with the extent of erosions in the peripheral joints (7, 8, 10, 27–29). However, our report is the first to present an estimation of the risk ratio for atlantoaxial subluxations, based on the radiographic scores in the peripheral joints during the first years of RA. In fact, patients who developed ≥10% of the theoretical maximum damage in their wrist, hand, and foot joints in the first 5 years were 15.9 times more likely to develop atlantoaxial subluxations than those whose peripheral damage remained <10% of the maximum.
The other interesting observation in our study is that the prevalence of cervical spine subluxations was low compared with historical data. In 1978, Rasker and Cosh (10) reported a 42% prevalence of aAAS and a 32% prevalence of AAI in patients who had had RA for 15 years. In a 9.5-year followup study by Winfield et al (29), cervical spine subluxations were found in 34% of RA patients. Furthermore, in the Heinola Follow-up Survey of RA, 42% of patients exhibited cervical spine subluxations at 20 years (4). In these studies, patients were treated with traditional DMARDs which were either rather ineffective (antimalarials) or often discontinued due to side effects (gold salts). In the present study, the patients were treated early and continuously with DMARDs, and the prevalence rates of aAAS, AAI, and SAS at 8–13 years were only 10%, 5%, and 5%, respectively. Comparisons with historical controls introduce several potential confounding factors. Nevertheless, we suggest it is possible that the extensive drug therapy with conventional DMARDs according to the “sawtooth” strategy (13, 18) played some role in preventing cervical spine changes, at least in some patients.
Structural damage at the atlantoaxial region is a potentially life-threatening condition. Severe rheumatoid destruction of the cervical spine and periodontoid pannus formation may cause compression of the spinal cord or brainstem, which may result in myelopathy or sudden death (30–32). Therefore, early recognition of patients who may develop or who already have atlantoaxial subluxations is important, in order to apply preventive strategies and early treatment. We have shown earlier that rheumatoid atlantoaxial subluxations can be retarded or even prevented with early and active therapy with DMARDs (5). Moreover, in patients with rheumatoid cervical spine disorders, active conservative treatment has been reported to relieve chronic neck pain (33).
Paimela et al (7) studied the influence of disease activity on the occurrence of cervical spine disorders as well as the functional capacity of patients with these disorders at the onset of RA and after 5 and 7 years of followup. Although patients with cervical spine disorders had higher CRP levels and ESRs and more swollen joints, reported more pain, and had worse self-reported functioning than patients without cervical spine disorders, the only statistically significant differences were observed in the CRP level at baseline and in pain on a visual analog scale at 5 years. The results of the present study are consistent with the observations of Paimela et al. At baseline, the patients who were later to develop atlantoaxial subluxations were older; however, they were comparable with the other patients with respect to demographic, clinical, and radiographic variables. Therefore, at the time of diagnosis of RA, it appears to be impossible to predict which patients are at increased risk of developing cervical spine disorders. In the present study, continuous high disease activity during the first years of RA (measured by the AUC values for ESR, CRP level, and SJC) was a statistically significant risk factor for later atlantoaxial subluxations. However, the extent of radiographic damage in the peripheral joints appeared to be the best predictor for atlantoaxial subluxations.
From a statistical standpoint, the low prevalence of atlantoaxial subluxations (only 14%) is one of the limitations of our study. Although the odds ratio (OR) of 15.9 for Larsen score at 5 years was impressive, the 95% CI of 3.38–74.7 was wide. A larger proportion of patients with atlantoaxial subluxations would have been statistically preferable, since the results of the statistical analysis would have been more reliable. In our study, due to the wide 95% CI, the OR estimate of 15.9 is robust. Nonetheless, even the lower limit of the 95% CI (3.38) indicates a substantial risk of atlantoaxial subluxations in patients who have ≥10% of the maximum possible destruction in peripheral joints.
We reported earlier that combination therapy with traditional DMARDs retarded the early development of atlantoaxial subluxations over 2 years (5). In the present study, we found a low long-term prevalence of cervical spine subluxations in patients who were treated early and continuously with traditional DMARDs. We conclude that rapid development of erosions in peripheral joints in patients with RA, regardless of the modality of their current treatment, should alert the rheumatologist to a high risk of cervical spine destruction in these patients.