Predictors of Pain for Patients With Early Inflammatory Polyarthritis


Address correspondence to Patricia L. Dobkin, PhD, McGill University Health Centre, 687 Pine Avenue West, V Building, Montreal, Quebec, Canada H3A 1A1. E-mail:



To identify predictors of pain at 1 year in patients with early inflammatory polyarthritis (EIP).


Using a prospective design, patients were examined by a rheumatologist and completed questionnaires at baseline and at 1 year after symptom onset. Separate regression analyses were run for pain intensity, sensory pain, and affective pain. Age and sex were adjusted in cross-sectional and longitudinal analyses; baseline potential predictors consisted of measures for corresponding pain values and disease activity, depression, coping scores, medication use, rheumatoid arthritis criteria being met, and duration of symptoms.


A total of 211 patients were enrolled in the study (mean ± SD age 58.8 ± 14.2 years, 63% women). There were significant improvements at 1 year for disease activity, instrumental coping, emotional coping, depression, and all 3 pain measures. At baseline, disease activity and depression were positively associated with all types of pain; in addition, instrumental coping was positively associated with sensory pain and palliative coping was positively associated with affective pain. At 1 year, pain intensity was predicted by baseline pain intensity, duration of symptoms, use of disease-modifying antirheumatic drugs (DMARDs), and emotional coping. Sensory pain was predicted by baseline sensory pain and DMARD use. Affective pain was predicted by baseline affective pain, DMARD use, and emotional coping.


The majority of treated EIP patients can expect improvements in clinical and psychosocial variables over the first year of their illness. Emotional coping at baseline may contribute to pain in the future, and therefore it may be useful for patients to learn other means of dealing with this chronic disease.


Pain is a significant problem for many patients with inflammatory polyarthritis (IP) and rheumatoid arthritis (RA) [1]. Most often it is pain that initially motivates them to seek medical care [2]. Ødegård et al tracked pain in a relatively large sample of recently diagnosed RA patients for 10 years and noted that 30% still experienced clinically important levels of pain (≥40 mm on a visual analog scale [VAS]) [3]. The aims of current treatment for arthritis are to control inflammation, regulate the immune system, and prevent joint damage. Clinicians frequently observe that although disease activity is reduced and progression is prevented, pain does not diminish accordingly [4]. Similarly, patients who have comparable evidence for disease activity or functional impairment may react differently [3]; one patient may continue to work, whereas another may reduce hours, or retire early. For example, arthritis patients with psychological distress are more likely to experience and be disabled by pain [5, 6]. Lee et al [7] found that even when in remission, 12% of patients with RA report significant levels of pain that are associated with global functioning, fatigue, sleep problems, and poor self-efficacy for dealing with the disease. Little empirical work has been devoted to this topic in early IP (EIP). Due to the multidimensional pain experience, biologic, psychological, and social factors influence the development, duration, and intensity of pain [8]. Therefore, to understand pain in EIP, one needs to monitor clinical and psychosocial variables over time [7].

Depression is the most prevalent mental health problem in patients with chronic pain. It reduces pain tolerance and threshold, increases analgesic requirements, and is associated with poor adherence to therapy [9, 10]. A cross-sectional study of patients with recently diagnosed EIP found that 23% of patients screened positive for depressive symptoms [11]. These patients showed evidence of significantly worse disease severity, disability, and pain; moreover, arthritis patients with depression engaged in more emotional coping (see below), had less self-efficacy for pain management and other arthritis-related symptoms, and had less social support than patients who did not screen positive for depressive symptoms.

Coping is a general concept used to describe cognitive, emotional, and behavioral reactions aimed at reducing the impact of stressors. Illness and living with chronic pain are stressful events, and the various strategies people use to reduce concomitant distress and/or symptoms associated with it have been studied in RA [12]. Ramjeet et al [13] reviewed the literature on the relationship between coping and psychological and physical adjustment in RA; of the 31 studies examined, only a few included patients with recent diagnoses. Most studies examined the impact of coping on psychological outcomes, such as depression. One cross-sectional study [14] of patients with recent-onset RA (<1 year) found that patients with higher levels of pain and disability reported more catastrophizing, less positive affect, and more negative affect. Another study [15] in which patients kept daily diaries found that relaxation coping predicted less pain the following day. Nonetheless, in general, little is known about how coping influences pain in the short and long term for patients who are just starting to live with their illness.

In the present study, 3 aspects of pain, i.e., the intensity, sensory, and affective components, were tracked in a cohort of patients with EIP from the initial rheumatology visit to 1 year after symptoms began in order to identify predictors of pain. It was hypothesized that depression and emotional coping with illness would predict pain, but it was not known if this would be the case for all 3 aspects of pain. Moreover, given that the relationship between variables may differ when measured concurrently and longitudinally, we planned to examine both.

Significance & Innovations

  • At baseline, disease activity and depression were associated with all 3 aspects of pain (intensity, sensory, and affective).
  • At baseline, instrumental coping was associated with sensory pain and palliative coping was associated with affective pain.
  • One year later, baseline values for disease-modifying antirheumatic drug use and emotional coping were predictors of pain.
  • The majority of patients improve on disease activity, pain, and coping over time; men fare better than women.



Consecutive adult patients with synovitis affecting at least 3 joints for 1 to 12 months who were evaluated at the Centre Hospitalier Universitaire de Sherbrooke (CHUS) were invited to participate in a longitudinal observational study [16, 17]. Patients were excluded if they had bacterial or crystal-induced arthritis, a defined connective tissue disease, or systemic vasculitis according to the 1987 American College of Rheumatology (ACR) criteria [18]. Consenting patients were regularly followed by rheumatologists and treated using the current approach of early and intensive treatment with disease-modifying antirheumatic drugs (DMARDs) [19, 20]. DMARD treatment was individualized with the objective of sustained remission. The Ethics Review Board of the CHUS approved the study.

Baseline medical visit

A rheumatologist completed a physical examination, including 68 tender and 66 swollen joint count assessments. A study coordinator employed a structured interview at the inclusion visit and at the followup visit. The time of onset of arthritis was assumed to be the week or month during which joint symptoms/signs had appeared or, in patients with previous musculoskeletal symptoms (e.g., osteoarthritis), the time when signs or symptoms of IP appeared.

Variables assessed at the initial medical visit included demographics, tender and swollen joint counts, duration of morning stiffness, medication use at and between each visit, modified Health Assessment Questionnaire (M-HAQ) score [21], serum C-reactive protein (CRP) level (upper limit of normal 8.0 mg/liter), serum IgM rheumatoid factor (RapiTex RF, Dade Behring; positive threshold set at 40 IU/ml), measure of anti-Sa as described [16], and anti–cyclic citrullinated peptide 2 antibodies (QuantaLite, Inova Diagnostics; positive when >20 units/ml).

The Disease Activity Score in 28 joints (DAS28) was calculated using the appropriate formula ( using 4 variables (tender and swollen joint counts on 28 joints, global disease activity measured on a 100-mm VAS, and CRP level) [22].


The overall intensity of pain was derived from the Short Form of the McGill Pain Questionnaire (MPQ-SF) [23]. A 0–10 VAS, with 0 corresponding to “no pain” and 10 corresponding to “worst possible pain,” was employed to assess pain at the time of questionnaire completion. The sensory and affective aspects of pain were derived from a list of adjectives describing pain (11 items for sensory pain, e.g., throbbing, aching, hot burning, tender; and 4 items for affective pain, e.g., tiring/exhausting, fearful); each descriptor was rated on a 4-point intensity scale (range 0–3, where 0 = none and 3 = severe). The psychometric properties of this questionnaire have been examined by Grafton et al and the VAS showed good test–retest reliability in patients with osteoarthritis [24]. Jensen et al [25], in an examination of the reliability and validity of chronic pain measures, concluded that single pain intensity ratings have sufficient psychometric strengths to be used in chronic pain research.

Putative predictors of pain


The DAS28 is a composite of 4 variables: tender and swollen joint counts (a semiobjective measure by the physician), CRP levels, and global disease activity VAS (the patient's subjective evaluation).


The Center for Epidemiologic Studies Depression Scale (CES-D) is a 20-item scale that was designed to detect depression in the general population [26-28]; it is also useful in clinical and psychiatric settings. This scale asks patients to rate the frequency of depressive symptoms in the past week from 0–3, where 0 = rarely or none of the time and 3 = most or all of the time. It has good psychometric properties, including reliability and concurrent and discriminate validity. Scores range from 0–60, with higher scores indicating greater depression. A CES-D cutoff score of 16 is used to screen for depression in the general population. However, recent studies have shown that a higher cutoff score of 19 would be more appropriate for use specifically in rheumatic disease populations to increase specificity [28]. The Cronbach's alpha coefficient (internal consistency) for this sample was 0.89.


The Coping with Health Injuries and Problems Scale (CHIP) is a 32-item self-report measure of coping with health problems developed by Endler and Parker [29]. Coping strategies are grouped into 4 subscales: 1) instrumental coping, which involves task-oriented behaviors, such as seeking health information or advice; 2) palliative coping, which involves self-care, such as resting; 3) emotional coping, which involves rumination about the health problem and wishing it would improve; and 4) distraction coping, which involves engaging in behaviors that allow the individual not to dwell on the health problem, such as taking part in pleasant experiences either alone or with others. The CHIP has been found to have acceptable internal consistency as well as very good test–retest reliability scores, and is valid for patients with chronic musculoskeletal pain [30]. The Cronbach's alpha coefficients (internal consistency) for this sample were 0.73, 0.77, 0.79, and 0.86 for distraction, palliative, instrumental, and emotional coping, respectively.

Statistical analyses

Descriptive statistics, including proportions, means, and SDs, were used to characterize the study population. Paired t-tests and McNemar's tests for continuous and binary variables, respectively, were used to compare patients' clinical characteristics measured at baseline with their characteristics 1 year later. To account for 18 comparisons, we used a Bonferroni-corrected criterion of P values less than 0.05/18 = 0.0028 for statistical significance. Missing values were given a value of 0. In sensitivity analyses, calculations were repeated after imputation of missing data of the mean value of the available answers.

To better understand how the selected clinical variables correlate with current pain and/or predict future pain scores, we carried out 2 types of multivariable analyses. First, cross-sectional analyses were employed to identify clinical correlates of pain scores at the baseline visit. Second, longitudinal data were analyzed to assess potential predictors, measured at the baseline assessment, of pain 1 year later.

In cross-sectional analyses, we fit 3 separate multivariable linear regression models to identify statistically significant correlates of baseline values of pain intensity, sensory pain, and affective pain. For each of these regression analyses, 10 baseline clinical variables were hypothesized, based on the clinical expertise of the investigators, to be potential correlates of the baseline pain scores, and were initially included as independent variables. These variables consisted of duration of symptoms (in months), a binary indicator of whether or not the patient met at least 4 ACR criteria for RA [18], binary indicators of taking DMARDs and corticosteroids, disease activity measured by the DAS28, screening positive for depression as measured by the CES-D, and 4 coping scores as measured by CHIP subscales: distraction, palliative, instrumental, and emotional coping. In addition, 2 sociodemographic variables, i.e., age at recruitment and sex, were considered potential confounders, and therefore were included in all multivariable models. We relied on forward selection and backward elimination procedures to achieve parsimonious multivariable models. These procedures retained only those variables among the 10 clinical variables that met the criterion for entering the final model (during forward selection procedure) or the criterion for remaining in the final model (during backward deletion procedure), while forcing the 2 sociodemographic variables into the model. To reduce the risk of residual confounding, we included in the final model the marginally nonsignificant clinical variables, and therefore set both criteria at P values less than 0.1 for a 2-tailed model-based t-test.

We then fit 3 separate multivariable linear models to longitudinal data in order to identify significant predictors for pain intensity, sensory pain, and affective pain 1 year later. For each of these regression analyses, the aforementioned 10 clinical variables were hypothesized as potential predictors, while age and sex were considered potential confounders and, therefore, adjusted for in all of the models. In addition, to account for the regression to the mean, the baseline values of the respective pain scores were forced into the multivariable models. Similar to cross-sectional analyses, forward selection and backward elimination procedures were used to select statistically significant or marginally nonsignificant variables (P values less than 0.1) among the 10 potential predictors into the final, parsimonious multivariable models. All analyses used SAS statistical software, version 9.2.


Patients' characteristics at baseline and 1 year later

A total of 248 patients were recruited between August 2006 and July 2011. Of these, 11 did not complete the pain questionnaires, 18 dropped out and 1 died before the first followup, and 7 had not yet reached the 12-month benchmark by March 1, 2012. We therefore report on 211 patients who completed the pain assessments at both time periods. Missing values represented 3.9% and 4.5% of the answers to the pain questionnaires at baseline and 1 year, respectively. Table 1 summarizes and compares the patients' characteristics at baseline and 1 year and reports the statistical significance of the differences between the mean values for continuous variables, or the frequency distributions for categorical variables, at the 2 time points. None of the baseline characteristics were statistically different between the included and excluded patients. Sixty-three percent of the 211 patients were women. At baseline, their mean age was 59 years and more than half (58.8%) were taking DMARDs, while less than half (40.8%) were taking corticosteroids. One year later, the majority of patients were receiving DMARDs (90.7%). The mean baseline pain scores were 48.9 (maximum of 100) for intensity, 8.6 (maximum of 33) for sensory pain, and 3.1 (maximum of 12) for affective pain. On average,all of these pain scores decreased significantly, as did the disease activity and depression (Table 1). At baseline, RA patients had significantly higher disease burden and higher pain levels than undifferentiated arthritis patients; these differences disappeared at 1 year (data not shown). After 1 year, the mean values of instrumental and emotional coping scores decreased significantly, whereas distraction and palliative coping scores did not change systematically. In addition, the M-HAQ score decreased significantly after 1 year.

Table 1. Comparison of selected characteristics measured at baseline and 1 yeara
 Baseline1 year2-tailed Pb
Total NValueTotal NValue
  1. aACR = American College of Rheumatology; RA = rheumatoid arthritis; DMARD = disease-modifying antirheumatic drug; anti–CCP-2 = antibodies to cyclic citrullinated peptide 2; RF = rheumatoid factor; M-HAQ = modified Health Assessment Questionnaire; DAS28-CRP = Disease Activity Score in 28 joints using the C-reactive protein level.
  2. bBy paired t-test or McNemar's test.
Female sex, %21163.3 
Age, mean ± SD years21158.8 ± 14.2 
Duration of symptoms, mean ± SD months2114.51 ± 2.84 
Fulfilling ≥4 1987 ACR criteria for RA, %20977.920480.90.609
DMARD use, %21158.820490.7< 0.0001
Corticosteroid use, %21140.820442.21.000
Anti–CCP-2 positive, %19240.318536.80.458
Anti-Sa positive, %14217.01117.20.004
RF positive, %20933.220226.20.012
Any antibody positive (anti-Sa and/or anti–CCP-2 and/or RF), %19548.320442.70.118
M-HAQ score, mean ± SD2080.79 ± 0.612090.40 ± 0.46< 0.0001
DAS28-CRP, mean ± SD2104.8 ± 1.42063.1 ± 1.3< 0.0001
Remission rate (DAS28 <2.6), %2105.220639.3< 0.0001
Depression score, mean ± SD21019.6 ± 11.321014.2 ± 10.1< 0.0001
Coping scores, mean ± SD     
Distraction coping21024.0 ± 6.620924.6 ± 7.30.318
Palliative coping21023.7 ± 6.220922.9 ± 6.20.056
Instrumental coping21031.2 ± 5.920929.0 ± 6.6< 0.0001
Emotional coping21025.8 ± 7.920922.5 ± 8.3< 0.0001
Pain scores, mean ± SD     
Pain intensity20848.9 ± 27.220433.9 ± 26.9< 0.0001
Sensory pain2088.6 ± 7.62075.8 ± 6.8< 0.0001
Affective pain2063.1 ± 3.52032.0 ± 2.9< 0.0001

Correlates of pain scores at baseline

Table 2 shows the results of 3 separate multivariable linear regression analyses for baseline values of pain intensity, sensory pain, and affective pain. For each of these outcomes, both forward selection and backward deletion procedures selected the same final models, and Table 2 shows the estimated associations only for the variables included in the corresponding final model. Among the 10 initially considered baseline characteristics, higher baseline disease activity and higher depression scores were independently associated with significantly higher baseline pain intensity (Table 2). Similar associations were also found to be statistically significant for baseline sensory pain and baseline affective pain (Table 2). For sensory pain, in addition to baseline disease activity score and depression score, higher baseline instrumental coping was associated with significantly higher baseline sensory pain. For affective pain, higher baseline palliative coping was significantly associated with higher baseline affective pain, in addition to baseline disease activity and depression.

Table 2. Correlates of pain at baselinea
Baseline characteristicsPain intensitySensory painAffective pain
  1. aValues are the regression coefficient (95% confidence interval), shown only for those among potential correlates that were included in the respective reduced model.– = a given variable was not included in the reduced model, i.e., was not an independent, statistically significant correlate of a given outcome; ACR = American College of Rheumatology; RA = rheumatoid arthritis; DMARD = disease-modifying antirheumatic drug; DAS28 = Disease Activity Score in 28 joints.
  2. bP < 0.001 for the 2-sided t-test based on the multivariable linear model with adjustment for all other correlates included in the respective final model (as shown in the corresponding column).
  3. cP > 0.001 and < 0.01 for the 2-sided t-test based on the multivariable linear model with adjustment for all other correlates included in the respective final model (as shown in the corresponding column).
  4. dP > 0.01 and < 0.05 for the 2-sided t-test based on the multivariable linear model with adjustment for all other correlates included in the respective final model (as shown in the corresponding column).
Age at recruitment (1-year increase)−0.09 (−0.31, 0.13)−0.04 (−0.10, 0.03)−0.02 (−0.05, 0.01)
Sex (male vs. female)−2.58 (−8.97, 3.81)−1.34 (−3.25, 0.56)−0.31 (−1.16, 0.55)
Duration of symptoms (1-month increase)
Fulfilling ≥4 1987 ACR criteria for RA
Corticosteroid use
DAS289.21 (6.86, 11.55)b1.47 (0.77, 2.16)b0.53 (0.21, 0.84)c
Depression score0.65 (0.37, 0.93)b0.23 (0.15, 0.31)b0.13 (0.10, 0.17)b
Coping scores   
Palliative0.08 (0.01, 0.15)d
Instrumental0.20 (0.04, 0.36)d

Predictors of pain scores at 1 year

Table 3 shows the results of the 3 final multivariable linear models, selected for predicting the values of the 3 pain scores at 1 year. For each of these outcomes, both forward selection and backward deletion procedures selected the same final models. Imputing for missing data did not modify the selected variables. Table 3 shows that, as expected, for each of the 3 pain scores, the higher baseline score was associated with a significantly higher value of the corresponding score at 1 year. Among patients with the same baseline pain intensity, compared to women, men on average had lower pain intensity at 1 year, while patients with a longer duration of symptoms and those who took DMARDs at baseline reported higher pain intensity at 1 year (Table 3). In other words, female sex, longer duration of symptoms, and taking DMARDs at baseline predicted significantly smaller improvements (i.e., reductions) in pain intensity at 1 year. Higher emotional coping also significantly predicted higher pain intensity score, indicating that patients who engaged in this type of coping showed less improvement during the following year.

Table 3. Baseline predictors of pain at 1 yeara
Baseline characteristicsPain intensitySensory painAffective pain
  1. aValues are the regression coefficient (95% confidence interval), shown only for those among potential predictors that were included in the respective reduced model.– = a given variable was not included in the reduced model, i.e., was not an independent, statistically significant predictor of a given outcome; ACR = American College of Rheumatology; RA = rheumatoid arthritis; DMARD = disease-modifying antirheumatic drug.
  2. bP < 0.001 for the 2-sided t-test based on the multivariable linear model with adjustment for all other predictors included in the respective final model (as shown in the corresponding column).
  3. cP > 0.01 and < 0.05 for the 2-sided t-test based on the multivariable linear model with adjustment for all other predictors included in the respective final model (as shown in the corresponding column).
  4. dP > 0.001 and < 0.01 for the 2-sided t-test based on the multivariable linear model with adjustment for all other predictors included in the respective final model (as shown in the corresponding column).
Pain intensity0.28 (0.14, 0.42)b
Sensory pain0.26 (0.13, 0.38)b
Affective pain0.25 (0.14, 0.37)b
Age at recruitment (1-year increase)−0.02 (−0.27, 0.23)−0.01 (−0.07, 0.05)0.005 (−0.02, 0.03)
Sex (male vs. female)−8.54 (−15.74, −1.34)c−1.31 (−3.17, 0.56)−0.48 (−1.27, 0.31)
Duration of symptoms (1-month increase)1.53 (0.25, 2.82)c
Fulfilling ≥4 1987 ACR criteria for RA
DMARD use9.02 (2.02, 16.02)c2.59 (0.80, 4.38)d0.75 (−0.009, 1.50)
Corticosteroid use
Disease activity score
Depression score
Coping scores   
Emotional0.49 (0.02, 0.95)c0.10 (−0.02, 0.23)0.06 (0.0032, 0.11)c

The major statistically significant predictors of the 1-year values of sensory pain and affective pain (Table 3) were the baseline values of the corresponding pain scores. In addition, those who took DMARDs at baseline had statistically significant higher sensory pain and borderline significant (P = 0.0527) affective pain scores at 1 year. Higher baseline emotional coping scores predicted significantly higher sensory pain and higher affective pain at 1 year. In contrast to pain intensity, for sensory and affective pain, neither duration of symptoms nor sex was identified as a significant predictor of the corresponding pain scores at 1 year.


Our goal was to determine whether baseline psychosocial variables would predict persistence of pain 1 year after the onset of symptoms in patients with recent-onset IP. We sought to identify patients at risk for adverse pain outcomes. To do so, we studied pain and collected clinical and biologic parameters prospectively in a cohort of consecutive EIP patients treated according to the current guidelines aiming at remission. At baseline, disease severity and depression correlated significantly with all aspects of pain, and specific coping strategies significantly correlated with some aspects of pain, i.e., emotional coping with affective pain and instrumental coping with sensory pain.

As expected, in treated EIP patients at 1 year into the disease, a mean of 5 months after inclusion, disease activity decreased significantly and function improved. Similarly, as a group, patients experienced reductions in all aspects of pain. This finding is consistent with previous research pertaining to improvements in pain during the early phases of disease [4, 31]. Interestingly, neither baseline depression nor the DAS28 predicted pain at 1 year. Instead, baseline emotional coping was the single psychosocial variable predictive of pain at 1 year. Emotional coping, much like catastrophizing [10], may negatively influence pain through behavioral (e.g., low levels of physical activity), cognitive (e.g., increased attention to pain), and physiologic (e.g., impaired neuroendocrine function) pathways. This maladaptive coping strategy is correlated with psychological distress (e.g., depression and anxiety) [6]. Zyrianova et al [12] examined the complex relationship between coping and illness perception of pain in 68 patients with longstanding RA using a structural model. They found that greater pain was detected in patients with higher scores on passive coping (similar to emotional coping), more physical disability, and increased depression and anxiety. Illness perception mediated this relationship. In general, patients need to be flexible with regard to coping in order to reduce stressors inherent in chronic illness. Although the other coping strategies assessed with the CHIP may or may not be useful, depending on the situation, emotional coping is a type of reactivity that is not helpful at any stage of illness.

Because sex and increasing age were both expected to contribute positively to pain at all times, they were forced into the statistical models. Indeed, being a man was negatively associated with all pain measures at both time points, although not statistically so, except for pain intensity at 1 year. This finding is consistent with Barnabe et al, who conducted a systematic review of sex differences in pain in patients with inflammatory arthritis [32]. Surprisingly, age was negatively associated with pain modalities at baseline only. One would reason that younger individuals may initially react more intensely to pain and loss of function [6], but that later on, other individual variables (such as coping or depression) would matter more. This is noteworthy, since we previously observed that increasing age positively correlated with radiologic and functional severity in patients with EIP [16, 33].

Given that much prior work in this area was cross-sectional, we examined the data first in this way. We reasoned that the way people cope when first diagnosed may differ from a year later, when medications were stabilized and the initial reaction to having an illness was quelled. Similar to the literature, we observed that higher levels of both disease activity and depression were significantly associated with higher reports of all aspects of pain at baseline. Different coping styles were associated with 2 different aspects of pain: instrumental coping for pain intensity and palliative coping for affective pain. At the onset of illness, if pain intensity is high, one takes action such as seeking medical care, taking medications, and finding information. Also, when emotions are influencing pain (e.g., fear, sadness), patients experience fatigue and tend to rest more. This passive approach to pain management is common among patients with chronic pain.

At 1 year, our patients reported mean scores for the VAS levels of global pain that were similar to those reported by Ødegård et al (34.0 and 28.7, respectively), with similar levels of variability (26.9 and 24.4, respectively). Pain intensity, as well as sensory and affective aspects of pain, was predicted by baseline levels of the corresponding measures. Van der Heide et al [34] also found that the only predictor of pain 1 year following diagnosis was baseline pain in EIP patients; however, psychosocial variables were not assessed in that study. In our patients, emotional coping predicted higher levels of each type of pain (significantly for pain intensity and affective pain). Our finding is similar to Covic et al [35] for patients with established RA; the authors showed that physical disability, helplessness, and passive coping (i.e., catastrophizing, praying, hoping) predicted pain (using the pain subscale of the Arthritis Impact Measurement Scales) 1 year later. It is well known that patients with chronic pain who engage in this type of coping unintentionally amplify their pain [36]. They ruminate and engage in a self-perpetuating vicious cycle of worry, tension, and pain.

The observation that DMARD use, but not corticosteroid use, at baseline predicted pain at 1 year is puzzling. One may speculate that DMARD use reflects the fact that sicker patients are prescribed more medications (i.e., confounding by indication), whereas corticosteroid use may not be as much determined by patients' pain as by a physician's perceived disease severity and personal prescribing habits. Because inclusion of patients occurred between 1 and 12 months into the disease (mean 5 months), while the second assessment occurred at 12 months, the fact that duration of symptoms at baseline correlated with pain intensity at 1 year may reflect the slow mode of action of DMARD-based treatment of EIP; patients recruited later would have less time to improve. A longer duration of observation allowing full adjustment of the treatment to individual patients may help to disentangle the role of delayed diagnosis and of intrinsic disease severity in pain persistence over time.

Depression was positively correlated with all aspects of pain when the symptoms were first detected. This finding is consistent with cross-sectional study results [11]. Contrary to our hypothesis, depression at baseline was not an independent predictor for any aspect of pain at 1 year. This finding is consistent with Ødegård et al, who reported that pain was longitudinally explained by anxiety (not depression), disease activity (erythrocyte sedimentation rate), physical function, and female sex [3]. Our observations contradict those of a longitudinal study in Montreal, where high levels of depressive symptoms correlated with less improvement in pain in 180 early inflammatory arthritis patients 6 months after baseline assessment. That study differs from the current study because they used the MPQ-SF total pain score only, and they did not report on coping strategies.

As in all studies, there are both strengths and weaknesses herein. A positive aspect of this work is that it is based on a longitudinal cohort of consecutive EIP patients that enabled us to make causal inferences based on psychometrically sound measures. Missing data were few and imputing did not change the results. Another consideration is that the population studied was predominantly French Canadian living in a midsized city or in the suburban and rural area surrounding it; it is unknown if our results can be generalized to other groups. Moreover, we do not have information pertaining to comorbid conditions (e.g., fibromyalgia) that may have contributed to the results [4].

The clinical implications of our study are relatively straightforward. Patients can be reassured that the majority does well in the first year of illness, with reductions in pain and clinical and psychological symptoms over time. Nonetheless, given the unpredictable and sometimes uncontrollable nature of autoimmune-mediated arthritis, patients should be encouraged to be flexible in the type of coping they use such that it fits the nature of the stressor encountered and where they are in the trajectory of their illness. For example, when patients are first diagnosed, instrumental and palliative coping may be helpful. However, patients relying heavily on emotional coping may be identified as being at risk for pain persistence over time. Psychosocial interventions (e.g., Mindfulness-Based Stress Reduction [37]) could be offered to these at-risk patients in the early phases of the disease to teach effective coping skills.


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. Dobkin 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. Dobkin, Cossette, Boire.

Acquisition of data. Carrier, Boire.

Analysis and interpretation of data. Liu, Abrahamowicz, Carrier, de Brum-Fernandes, Cossette, Boire.


The authors would like to thank Ms Sehar Manji for preparation of the manuscript and the patients in the study for their participation.