Cervical spine involvement in rheumatoid arthritis over time: results from a meta-analysis

Introduction Complications in rheumatoid arthritis (RA) seem less common than they were years ago. The prevalence and progression of anterior atlantoaxial subluxations (aAASs), vertical subluxations (VSs), subaxial subluxations (SASs), and associated cervical myelopathy in RA over the past 50 years were determined. Methods A literature search was performed by using Medline-OVID/EMBASE, PubMed, and Scopus (from 1960 to June 21, 2014). Prevalence studies were included if the sample size was at least 100 or the prevalence/progression of cervical subluxations was reported. Study quality was assessed by using the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist. Prevalence of cervical subluxations was calculated for each study. Student’s t test and meta-regression were used to evaluate for significance. Results In total, 12,249 citations were identified and 59 studies were included. The prevalence of aAAS decreased from 36% (95% confidence interval (CI) 30% to 42%) before the 1980s to 24% (95% CI 13% to 36%) in the 2000s (P = 0.04). The overall prevalence rates were 11% (95% CI 10% to 19%) for VS, 13% (95% CI 12% to 20%) for SAS, and 5% (95% CI 3% to 9%) for cervical myelopathy, and there were no significant temporal changes. Rates of progression of aAAS, VS, and SAS were 4, 6, and 3 lesions per 100 patients per year, respectively. The incidence of new or progressive cervical myelopathy was 2 cases per 100 patients with known cervical subluxations per year. Conclusions Since the 1960s, only aAAS has decreased dramatically. It is still more than twice as common as VS or SAS. No temporal changes in the development of cervical myelopathy in affected patients with RA were noted. The progression rates of cervical subluxations and myelopathy were unchanged over time.


Introduction
Rheumatoid arthritis (RA) is a chronic systemic inflammatory disease that primarily affects the joints. Although inflammatory arthritis of the small joints in the hands and feet is a common clinical manifestation, the cervical spine can also be affected. In fact, cases surfaced as early as 1890, when A. E. Garrod reported 178 patients with cervical spine involvement in a series of 500 patients with RA [1]. Cervical spine involvement can take many forms. Reported findings include vertebral endplate erosions, spinous process erosions, and apophyseal joint changes, such as osteoporosis, blurring, and fusion [2,3]. The most characteristic lesions are subluxations [4,5].
The atlas-axis-cervical vertebrae 1 and 2 (C1 and C2)-articulation is one of the prime disease targets. The erosive pannus formation at this joint often leads to bony destruction and laxity in the surrounding ligamentous complex, especially the transverse ligament that aligns that atlas and axis. The subsequent loss of anchoring structures results in atlantoaxial subluxation (AAS) [6]. The subluxation can be anterior, posterior, lateral, and rotatory. The anterior atlantoaxial subluxation (aAAS) is the most common subtype; the reported prevalence ranges from 10% to 55% [2,3,[7][8][9]. This wide range may be related to the variation in the radiographic definition of aAAS. Separation of the anterior odontoid peg (dens) from the anterior ring of atlas (altanto-dental interval) of greater than 2.5 mm to 5 mm has been considered pathological [2,[10][11][12]. Posterior AASs are less frequent and often due to fracture of the dens [13]. Lateral subluxations are considered when lateral masses of C1 are displaced laterally more than 2 mm in comparison with C2. They can lead to head tilt and rotational deformities [5].
Another form of subluxation is the vertical subluxation of the axis (VS), also known as altantoaxial impaction, cranial settling, superior migration of the odontoid, or psuedobasilar invagination. It is secondary to the destruction of occipitoatlantal and atlantoaxial joints and surrounding soft tissues [5]. Not surprisingly, many studies have shown that AAS and the destruction of C1-C2 articulation tend to predate the development of VS [14][15][16].
Lastly, subaxial subluxation (SAS) is associated with the damage to the facet joints, interspinous ligaments, and intervertebral discs. It can happen at one or multiple levels, leading to a step-ladder deformity and kyphosis [5]. SAS is defined by a misalignment of more than 3 mm between the anterior bodies of two consecutive vertebrae, although a threshold of 1 mm has also been found in studies [4,17]. Neck subluxations can develop over the course of the disease process, and there is an increased risk of developing neurological complications. Tinnitus, vertigo, and visual changes associated with vertebrobasilar insufficiency or brainstem compression can arise [6]. Patients may experience myelopathic symptoms. Once severe neurological complications develop, the prognosis can be grave; 1-year mortality can be as high as 50% if the condition is left untreated [18].
It appears that RA complications such as rheumatoid nodules, corneal melts, vasculitis, and C-spine subluxations are less frequent than they were years ago. Given the improved RA management over the past 50 years, our objectives were to quantify and evaluate temporal changes in the prevalence and progression of various RA-related cervical subluxations and cervical myelopathy.
Identified titles/abstracts were reviewed, and full reports were obtained if appropriate. Studies were considered if they provided data on the prevalence or progression of any cervical subluxations in patients with RA. Studies reporting prevalence required a sample size of at least 100 patients with RA. Another inclusion criterion was publication in English. Case reports, case series of fewer than 100 patients for the estimate of prevalence, and review articles were excluded. Cervical spine involvement of RA patients who were followed over time was included for the rate of progression analyses. Additional articles were retrieved by hand by searching relevant references.

Data collection
Information from the studies was extracted by one reviewer (TZ) and ambiguities were resolved with the other reviewer (JP). A standard data extraction form was used to extract the following information: Prevalence studies: year of publication, author, location of study, study design, patient population, sample size, number of females, mean age, mean disease duration, and proportion with any cervical changes, aAAS, posterior AAS, lateral AAS, rotatory AAS, VAS, SAS, cervical myelopathy, brainstem compression/vertebral artery involvement, and their respective definitions.
Progression studies: year of publication, author, location of study, study design, patient population, sample size, number of females, mean age, mean disease duration, years of follow-up, baseline, and rate of progression of aAAS, VS, SAS, cervical myelopathy and number of progressive/new cervical myelopathies. If studies included patients with RA and other rheumatic diseases, only data pertaining to the RA cohort were extracted.

Quality assessment
Each study was assessed by using the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist [19], which is a 22-item checklist relating to the title and abstract of the article (item 1), background and objectives (items 2 and 3), methods (items 4 to 12), results (items 13 to 17), discussion (items 18 to 21), and funding (item 22). The purpose of STROBE is not to give a quality score but to ensure clear presentation of reporting.

Statistical analysis
Proportions were pooled with a random effects model [20]. Forest plots were created to estimate prevalence with a 95% confidence interval (CI). The I 2 statistic was used to quantify the magnitude of heterogeneity (mild, 0% to 30%; moderate, 31% to 50%; high, more than 50%). Tau-squared was the square root of the betweenstudy variance, and P value was for Cochrane's Q, the classic measure of heterogeneity. Owing to significant heterogeneity, Forest plots and pooled estimates were not reported. Student's t test was performed to evaluate regional differences in aAAS. Meta-regression through a random effects model was also used to explore the study-level associations between the prevalence of various C-spine complications and year of publication. A P value of less than 0.05 was considered statistically significant. Publication bias was determined by using funnel plots.

Search results
The literature search identified 12,249 citations with 412 duplicates. Titles or abstracts (if available) were screened for eligibility, yielding 126 citations for full text review. Twelve studies were case reports. Twenty-nine studies were editorials or reviews. Eleven studies did not report on the rate or progression of cervical complications. Eleven studies examined patients with severe neck complaints or suspected/known cervical instabilities. Six excluded studies examined the prevalence of cervical instabilities in RA but had a sample size of less than 100. Two articles were identified by hand-searching key references. In total, 67 articles were excluded, leaving 59 studies to be included in our meta-analysis ( Figure 1).
Across all studies, the median number of items fulfilled on the STROBE checklist was 16 (range of 11 to 19) and reporting was improved in more recent studies. Characteristics of the included studies are shown in Tables 1 and 2.

Prevalence of any rheumatoid arthritis-related changes, cervical subluxations, and cervical myelopathy
Twelve studies reported the prevalence of any RA-related cervical changes, not limited to subluxations. Out of 3,559 patients, 1,597 (45%) were found to have one or more of those radiographic changes. There were no significant changes over time of publication (P = 0.71).
Twenty-three studies with 5,106 patients with RA reported on cervical myelopathy. Two hundred seventy-one patients (prevalence of 5%) had observed neurological deficits corresponding to Ranawat Cervical Myelopathy Classification II and above (95% CI 3% to 9%) [67]. We did not observe a significant change over time (P = 0.22). In terms of serious neurological complications, nine patients (0.2%) were observed to develop paraplegia. Ten additional patients (0.2%) had symptoms of possible brainstem involvement, including nausea, vertigo, or drop attacks.

Discussion
Cervical spine involvement in RA has long been recognized. Varying degrees of prevalence have been reported as reflected in our literature search. This can be explained by numerous factors, such as differences in sample size, disease duration, follow-up period, medications used, and radiological criteria. There may also be reporting bias in which once something is well described, fewer publications follow over time. To the best of our knowledge, this is the first study to systematically evaluate the prevalence and progression of cervical subluxations and cervical myelopathy in patients with RA.
The earliest form of cervical instability in patients with RA has been shown to be a reducible AAS [62]. AAS is the most common cervical subluxation in RA. (AAS is twice as common as VS or SAS, and other types are very uncommon.) No regional differences in aAAS were found between Asia, Europe, and North America. However, a significant temporal reduction in the prevalence of aAAS exists, decreasing by one third from the 1960s and 1970s to 2000. The could reflect the changes in RA management (earlier introduction of DMARDs with higher doses and more combinations), or changing epidemiology of RA with fewer complications compared to years ago, and or reporting bias where the complications that are well described are less likely to be published over time so the contemporary rates may not be fully reported (publication bias).
The effects of DMARDs were addressed specifically in a number of studies. For example, Neva and colleagues recruited 198 early RA patients who were randomly assigned to either combination DMARDs (sulfasalazine, methotrexate, and hydroxychloroquine) or a single DMARD with or without prednisone [43]. After 2 years of follow-up, aAAS was found in 3% of the patients. None of the patients who were receiving combination therapies had aAAS. Lending further support to the role of DMARDs in preventing cervical involvement, biological agents were examined in two studies. A cohort of 38 early RA patients who received more than 2 years of continuous infliximab treatment in addition to methotrexate and prednisone was followed for 4 years [64]. Only one of the patients developed de novo AAS. In another study, a group of 91 patients who were on infliximab, etanercept, or tocilizumab was followed for 4 years [65] and three developed AAS.
Whereas the prevalence of aAAS seemed to be on the decline with the introduction of DMARDs, that of SAS and VS did not. This could be related to the natural progression of cervical subluxations. Our results demonstrated that, once formed, 4 out of 100 aAAS lesions would have an increase in atlanto-dental interval per year. After a certain point, AAS can have the appearance of stabilization or even improvement in the atlanto-dental interval on x-ray, signaling the beginning of VS. For example, Rana followed 41 AAS patients over a period of 10 years. Ten percent of the patients had less AAS secondary to new development of VS and upward translocation of odontoid process [16]. Smith and colleagues reported a similar phenomenon in 13% of their cases [23]. SAS generally develops at a later stage and is often seen after development of VS [68]. The lack of temporal reduction in the prevalence of VS and SAS may therefore be because DMARDs are less effective in preventing progression of existing lesions or due to a very low prevalence resulting in insufficient power to detect a change. When biologics were used, if there was C-spine subluxation, 79% to 80% progressed further [64,65]. We suspect that, eventually, the prevalence of VS and SAS will decrease as fewer de novo AASs are being formed.
(See figure on previous page.) Figure 1 (A) Search strategy for articles and (B) Funnel plots to determine if there was publication bias. Funnel plot 1: Graph of sample size of individual study (x-axis) and its respective prevalence of anterior atlanto-axial subluxations (y-axis). Funnel plot 2: Graph of sample size of individual study (x-axis) and its respective prevalence of rheumatoid arthritis (RA)-related cervical changes (y-axis). Funnel plot 3: Graph of sample size of individual study (x-axis) and its respective prevalence of vertical atlanto-axial subluxations (y-axis). Funnel plot 4: Graph of sample size of individual study (x-axis) and its respective prevalence of subaxial subluxations (y-axis). Funnel plot 5: Graph of sample size of individual study (x-axis) and its respective prevalence of cervical myelopathy (y-axis). Although cervical subluxations are common as illustrated above, their clinical presentation is often asymptomatic or associated with neck pain. Neurological complications are less frequent and tend to occur later in the disease course [69]. In our study, we reported a 5% prevalence rate for observed neurological deficits in patients with RA on the basis of the Ranawat grading system. Class I patients have pain but no neural deficits and therefore were not counted for our study. Class II patients have subjective weakness with hyperreflexia and dyssthesias. Class III patients have objective weakness and long tract signs. Class IIIA patients are ambulatory whereas IIIB patients are non-ambulatory [67]. Serious complications were rare, however. Paraplegia was reported in 0.2% of patients with RA, and symptoms of possible brainstem involvement, such as vertigo and drop attacks, were found in another 0.2%. We did not observe a decrease in the already-rare prevalence or progression of cervical myelopathy over time.
This study has limitations. Non-English studies were excluded. Although some asymmetries existed in our funnel plots, we do not believe that this represented major publication bias. Our data had significant heterogeneity. Age, disease duration, RA activity and severity, and treatments such as DMARDs and corticosteroids were not assessed. Estimates were generated at a population level. Inclusion criteria for some case series were not described in detail, and so the generalizability of the findings is unknown. We did not present the overall meta-analysis results, because heterogeneity (as measured by the I 2 with respect to  between-study variation) was high, but we could explore trial-level differences contributing to the heterogeneity (such as year of publication) by using meta-regression, which quantifies the effect of a study characteristic (publication year) on the overall effect of heterogeneity in the rate of C-spine subluxations. The review may be limited by variation in study quality, especially with the earlier publications having lower STROBE scores (as is often seen when comparing older with more recent studies due to different standards). Lastly, another factor to consider when interpreting the study is that cervical complications tend to develop over time. Even though our review included data from recent studies, the long-term effects of modern DMARDs (type and how we use them) on development and progression of cervical instabilities have yet to be adequately studied, and it could take years before differences in complications can be identified as C-spine complications could take years to develop or be prevented.

Conclusions
Patients with RA are still at risk for cervical spine involvement. Over the past 50 years, we have found a significant decrease in the prevalence of aAAS, possibly due to improved disease control with modern DMARDs or changing epidemiology of RA. However, aAAS remains the most common lesion with a prevalence of 27%, more than twice as common as SAS or VS. The prevalence of cervical myelopathy did not change dramatically and remains at 5% among patients with RA. The rate of progression of cervical subluxations and myelopathy did not seem to be affected significantly by DMARDs.