Disease Progression and Treatment Responses in a Prospective DMARD-naïve Seropositive Early Rheumatoid Arthritis Cohort: Does Gender Matter?

Abstract

Objective

To assess gender differences in disease characteristics and treatment responses over time in a DMARD-naïve seropositive early rheumatoid arthritis (RA) cohort.

Methods

DMARD-naïve, seropositive early RA (<14 months) patients with polyarticular disease were recruited by the Western Consortium of Practicing Rheumatologists. Each patient was examined at study entry, after 6 and 12 months, and yearly thereafter. Clinical and demographic data were collected. We investigated gender differences in baseline disease characteristics and treatment using Chi-square, Mann-Whitney and t tests. We used generalized estimating equations (GEE) models for repeated measures to examine whether the rate of change of specific disease outcomes during the first 2 years after DMARD initiation were significantly influenced by gender.

Results

At baseline, men (n=67) and women (n=225) had similar disease activity and radiographic damage; men, however, had significantly worse erosion, while women had worse joint space narrowing. Despite similar treatment, women had worse disease progression over the 2-year follow up, as assessed by trends in DAS28ESR4, physician global scores and tender joint counts. In the GEE model, gender was significantly associated with the rate of change of DAS28ESR4 scores (p=0.009), though not being independently associated with disease activity. Self-reported measures (HAQ-DI, patient global scores, fatigue, pain) were worse among women at baseline and throughout the study period. Men were more likely to achieve remission.

Conclusion

At baseline, men and women had similar disease activity and joint damage. Responses to treatment over time were, however, better among men in this pre-biologic era; women had worse progression despite similar treatment.

Introduction

Gender differences in rheumatoid arthritis (RA) have been described at multiple levels. Men and women with RA have been found to differ in incidence of the disease [1], extent of disability [2], arthritis-related pain [3], load of genetic risk factors [4, 5] and even mortality [6]. Yet, the mechanisms by which these differences arise, including the possible involvement of sex hormones [7, 8] are as yet unclear. In the last few years, however, there has been a renewed interest in gender differences in RA within the rheumatology community, with a focus on disease outcomes and responses to treatment.

Studies of gender differences in disease outcomes and responses to treatment in RA have used a variety of study designs as well as patient populations that vary widely in disease duration and severity, ranging from prospective early RA cohorts [9–11] to cross-sectional patient populations with long-standing disease [3, 5, 12, 13]. A number of recent studies have reported poorer RA outcomes among women in terms of symptoms, disease activity and functional capacity, both in early [14, 2] and late RA [3, 12, 15], although the early RA studies also reported no gender differences in severity in early stages of the disease. In contrast, other studies of long-standing RA have found men to be more likely to have severity markers, such as higher levels of anti-citrullinated peptide antibodies (ACPA, formerly anti-CCP) and the HLA-DRB1 shared epitope [4, 5], and to have worse outcomes [13]. It is still unclear whether RA occurs in more severe forms among women or men, and whether the disease progresses differently in each gender. Furthermore, there is now emerging evidence that men may be more likely than women to achieve remission, in response to traditional as well as biologic DMARDs [11, 16–18]. There is currently much debate about whether the observed differences are intrinsic to the disease or not. It has been suggested that the gender differences may be accounted for by differences in duration of the disease between men and women [11, 19], or may just reflect a gender bias in the reporting of symptoms [12].

Given these conflicting findings and views regarding whether and how men and women with RA differ in their disease characteristics and responses to treatment, detailed examination of prospective early RA cohorts may be better suited to explore these issues, whereas studies of established RA of variable duration may be limited by their cross-sectional design. Thus, in order to determine whether gender differences, if they do exist, are present in the early stages of the disease or appear in later stages, we have investigated such differences in a prospective DMARD-naïve early RA (<14 months duration) cohort at baseline [20]; we have also examined differences in disease progression and treatment outcomes between men and women in the longitudinal setting, in this fairly homogeneous cohort limited to patients with positive rheumatoid factor (RF) tests and active polyarticular arthritis.

Patients and Methods

Patients

RA patients were recruited between January 1, 1993 and April 1, 2002, as a joint effort of the Western Consortium of Practicing Rheumatologists, through 29 recruitment centers in the western region of the US and Mexico, including 4 university medical centers, and 25 community practices as described elsewhere [20]. To enter the observational study, patients had to: (a) satisfy the 1987 ACR criteria for RA [21], (b) be within 14 months of symptom onset, (c) have no prior treatment with a disease-modifying anti-rheumatic drug (DMARD), (d) have positive titers for rheumatoid factor (RF) antibodies (≥ 40 IU/ml), and (e) have at least 6 swollen joints (of 66) and at least 9 tender joints (of 68). Approval for the study was obtained from local ethics committees and all participating patients provided informed consent.

Data collection

Each patient was examined by their rheumatologist at the time of entry in the study (baseline), after 6 and 12 months, and yearly thereafter. The rheumatologists collected demographic and clinical data, including joint counts for 68 tender and 66 swollen joints, and physician global assessment at each of these time-points, as well as data on utilization and outcomes for all DMARDs, biologics and combination therapies used while enrolled in the study. Joint counts for 28 tender (TJC28) and swollen (SJC28) joints were derived from the recorded 66/68 joint counts. Radiographs of hands/wrists and forefeet were obtained. Patients also completed self-report questionnaires, providing data on their Health Assessment Questionnaire Disability Index (HAQ-DI), arthritis-related pain on a visual analog scale (VAS), fatigue VAS, global health and depression scores (Center for Epidemiological Studies Depression Scale, CESD). Blood samples drawn at baseline and at each follow up were used to evaluate levels of RF, C-reactive protein (CRP), and erythrocyte sedimentation rate (ESR). ACPA titers were measured only at baseline from frozen samples of a subset of subjects at Specialty Laboratories.

Assessment of Disease Activity

DAS28 scores were computed using (a) ESR:

DAS28ESR4 = [0.56*sqrt(TJC28) + 0.28*sqrt(SJC28) + 0.70*ln(ESR) + 0.014*(global)] and (b) CRP:

DAS28CRP4 = 0.56*sqrt(TJC28) + 0.28*sqrt(SJC28) + 0.36*ln(CRP+1) +0.014*(global)+0.96

Patients were categorized, at baseline, as having mild (2.6<DAS28ESR4≤3.2), moderate (3.2<DAS28ESR4≤5.1) or severe (DAS28ESR4>5.1) RA disease activity using previously defined criteria based on the DAS28ESR4 [22].

Assessment of Radiographs

Radiographs for hands and feet were evaluated for RA joint damage by calculating the erosion and joint space narrowing (JSN) scores using the method developed by Sharp et al. [23, 24]. Total Sharp scores were calculated by adding erosion scores and JSN scores.

Improvement of RA Symptoms and Remission

Improvement of disease symptoms was determined based on the ACR criteria for 20% and 50% improvement from baseline [25, 26] at 6, 12 and 24 months. We did not use ACR 70 criteria since few patients met these criteria. Patients were considered to be in point remission if their DAS28ESR4 score was less than 2.6 [27].

Statistical analyses

1) Gender Differences at Baseline (cross-sectional)

Differences in disease features and treatment prescribed between men and women were examined at baseline, using Chi-square tests to compare distributions of categorical variables; t-tests were used for normally distributed continuous variables, and the Mann-Whitney test for those non-normally distributed. For continuous variables, if normally distributed, means and standard deviations (SD) are reported, while medians and 25^th and 75^th percentiles are reported if non-normally distributed.

2) Longitudinal Analyses

Generalized Estimating Equations (GEE) models for repeated measures, using an independent correlation structure, with robust estimation, were used to model differences in disease activity and disease severity measures as well as frequency of ACR improvement or DAS28 remission between men and women, from baseline to 24 months. For all outcomes examined, follow-up time and gender were included as the main explanatory variables in the GEE model, adjusting for age, baseline disease activity (DAS28ESR4) and DMARD prescribed at the baseline visit; an interaction term was also included in the model to evaluate whether the rate of change of the outcome over the 2 years of follow up was modified by gender. The GEE model with DAS28ESR4 as the outcome variable was tested with and without adjusting for RF titers, fatigue VAS, and CESD scores when evaluating the influence of gender and follow up time on disease activity over time. All analyses were repeated in the subset of patients who were prescribed DMARDs at the baseline visit to assess responses to treatment among men and women in terms of the outcomes described above. The ESR and CRP variables were analyzed using observed values as well as adjusted values obtained after gender-specific published formulae [28, 29] were applied to adjust for age.

3) Sensitivity analyses

We assessed gender differences in loss to follow up, and repeated the analyses in the subset of patients with complete DAS28ESR4 data at all time points.

All statistical analyses were performed using the Stata software package (Release 10, College Station, Texas, USA).

Results

Gender Differences at Baseline

A total of 292 patients, consisting of 225 women and 67 men, with seropositive early RA were included at baseline. The main gender differences are summarized in Table 1. As expected from the selection criteria, at study entry, all patients had recent onset RA with a mean duration of 6.2 months (SD: 3.5 months), and men and women had equally active disease, as assessed by the DAS28ESR4. Overall, the patients had moderate (men: 13%, women: 15%) or severe disease activity (men: 87%, women: 84%), with only 1 woman having mild disease activity. Physician global assessment scores were also similar between genders at baseline, as were the total Sharp scores. Compared to women, men had significantly higher erosion scores, swollen joint counts (SJC28), RF titers and frequencies of nodules; they were also significantly more likely to have ever smoked. Men tended to have later mean age at onset, slightly longer duration of RA symptoms, and were more likely to have radiographic erosions, higher tender joint counts (TJC28) and higher CRP levels and ACPA titers than women, although these differences were not statistically significant. In contrast, women had significantly worse function (HAQ-DI) and significantly higher patient global scores, fatigue, and depression scores (CESD). They had worse arthritis-related pain and joint space narrowing scores than men, though these were not statistically significant.

Table 1 Disease features at baseline.

Table 1 shows a summary of the disease features and medications prescribed at baseline in the cohort of 292 RA patients, and in men and women separately. In the case of continuous variables, mean values and standard deviations (SD) are reported for those variables that were normally distributed; else, medians and inter-quartile range (IQR) are given. Proportions are shown for categorical variables. P values are also shown for differences derived from Student’s t-tests, Mann-Whitney tests and Chi-squared tests.

Disease features (Baseline)		Sample sizes *			All patients	Men	Women	p-value
		All	Men	Women
Age at onset (yrs), mean±S.D.		291	66	225	49.6±13.3	51.3±12.3	49.1±13.6	0.20
Disease duration (months), mean±S.D.		292	67	225	6.2±3.5	6.9±3.7	6.1±3.5	0.10
Presence of erosions (% (n))		259	59	200	78.8 (204)	84.8 (50)	77.0 (154)	0.20
Erosion score, median (IQR)		259	59	200	0.8 (0.3–2.5)	1.3 (0.5–3.7)	0.7 (0.2–2.0)	0.007
Joint space narrowing score, median (IQR)		258	59	199	2.0 (0.5–5.6)	1.8 (0.5–4.3)	2.3 (0.5–5.8)	0.25
Total Sharp score, median (IQR)		259	59	200	3.6 (1.3–8.0)	3.6 (1.5–8.8)	3.6 (1.3–8.0)	0.74
Presence of nodules (% (n))		280	64	216	13.9 (39)	21.9 (14)	11.6 (25)	0.04
Tender joint count, mean±S.D.		279	64	215	13.7±7.5	14.8±7.8	13.4±7.4	0.20
Swollen joint count, mean±S.D.		279	64	215	13.1±7.1	14.9±6.9	12.6±7.1	0.02
CRP (mg/dl), median (IQR)		290	67	223	1.4 (0.5–3.4)	1.7 (0.5–3.9)	1.3 (0.5–3.0)	0.40
ESR, median (IQR)		290	67	223	37 (23–55)	35 (22–51)	38 (25–55)	0.66
DAS28ESR4, mean±S.D.		267	61	206	6.2±1.1	6.3±1.1	6.2±1.2	0.45
DAS28CRP4, mean±S.D.		267	61	206	5.1±1.1	5.2±1.1	5.0±1.1	0.48
Physician global VAS, mean±S.D.		278	64	214	49.2±21.0	49.1±23.2	49.2±20.4	0.99
HAQ-DI score, mean±S.D.		252	58	194	1.2±0.7	1.0±0.7	1.3±0.7	0.003
Patient global VAS, mean±S.D.		280	63	217	56.2±27.3	50.0±27.4	58.0±27.1	0.04
Pain VAS, mean±S.D.		213	44	169	60.4±27.1	53.8±25.9	62.1±27.3	0.07
Fatigue VAS, mean±S.D.		213	44	169	52.0±24.6	41.4±25.8	54.8±23.6	0.003
CESD score, median (IQR)		217	49	168	13 (7–22)	7 (4–16)	14 (8–23)	0.001
Ever smoked (% (n))		226	55	171	63.3 (143)	83.6 (46)	56.7 (97)	<0.0005
RF titer (IU/ml), median (IQR)		281	60	221	211 (88–463)	257 (164–505)	187 (79–459)	0.03
Low titer (12–50 IU/ml), (% (n))		265	57	208	10.9 (29)	3.5 (2)	13.0 (27)	0.02
Medium titer (51–100 IU/ml), (% (n))		265	57	208	13.6 (36)	7.0 (4)	15.4 (32)
High titer (>100 IU/ml), (% (n))		265	57	208	75.5 (200)	89.5 (51)	71.6 (149)
ACPA titer (units/ml), median (IQR)		123	29	94	249 (129–288)	273 (210–290)	235 (124–284)	0.10
Low titer (20–49 units/ml), (% (n))		107	25	82	1.9 (2)	0	2.4 (2)	0.54
Medium titer (50–99 units/ml), (% (n))		107	25	82	7.5 (8)	4.0 (1)	8.5 (7)
High titer (≥100 units/ml), (% (n))		107	25	82	90.7 (97)	96.0 (24)	89.0 (73)
DMARD** prescribed at baseline visit (% (n))		292	67	225	75.7 (221)	71.6 (48)	76.9 (173)	0.38
Methotrexate (monotherapy/combination therapy) (% (n))		292	67	225	49 (143)	46.2 (31)	49.8 (112)	-
Hydroxychloroquine (monotherapy) (% (n))		292	67	225	15.8 (46)	10.5 (7)	17.3 (39)	-
Sulfasalazine (monotherapy) (% (n))		292	67	225	7.5 (22)	13.4 (9)	5.8 (13)	-

^*These sample sizes denote number of patients on whom data were available for each variable

^**Methotrexate, sulfasalazine, hydroxychloroquine or gold therapy

Treatment Initiated at the Baseline Visit and First DMARD

Treatment regimens prescribed at the baseline visit are summarized in Table 1. A DMARD (methotrexate, sulfasalazine, hydroxychloroquine or gold) was prescribed at the baseline visit for 76% (221 of 292) patients, i.e. 72% of men and 77% of women (p=0.38). In terms of first DMARD used for each patient, including those started after the "baseline" visit, methotrexate, as monotherapy or in combination with other DMARDs, was the first DMARD in 172 patients (37 men and 135 women), i.e. 55% of men and 60% of women (p=0.49), which is consistent with expected practice patterns in this pre-biologic era.

Changes in Disease Outcomes over Time

(a) Disease Activity

During the first 2 years after baseline, mean DAS28ESR4 scores showed more improvement in men compared to women (Figure 1a). Similar trends were observed for DAS28CRP4 scores (Figure 1b). In the GEE model, after adjusting for age, RF titers, fatigue VAS, depression score (CESD), baseline DAS28ESR4 and baseline DMARD, gender was found to significantly influence the rate of change of the DAS28ESR4 (p=0.009), but was not independently associated with this outcome (p=0.18) (table 2). Significant predictors for DAS28ESR4 over the 2-year follow up included baseline DAS28ESR4 (p<0.0005), follow up time (p<0.0005), fatigue VAS (p<0.0005), RF titer (p=0.006) and depression score (p=0.007). As shown in Figure 1c and e, tender joint counts and physician global assessment showed similar trajectories to the DAS28ESR4, with more improvement among men over time. Swollen joint counts, on the other hand, were significantly higher among men at baseline, but were quite similar in both genders thereafter (Figure 1d). Gender was a significant predictor for swollen joint counts (p=0.03), but not for tender joint counts and physician global scores. Gender did not seem to influence the rate of change of these outcomes (swollen and tender joint counts and physician global scores) over time.

Figure 1.

(a)–(k): The observed mean values for different outcomes, including the ACR core set measures, in men (solid line) and women (dashed line) over the 2-year follow up. The bars indicate the 95% confidence intervals at each point. The numbers of men and women with available data at each time point, for each outcome, are shown below each graph.

Table 2. Longitudinal analysis of predictors of DAS28ESR4 scores over time using a Generalized Estimating Equations (GEE) model.

Regression coefficients, 95% confidence intervals (CI) and p values are shown for the GEE model with DAS28ESR4 as the outcome variable. All covariates adjusted for in the model are listed as independent variables.

Outcome variable in GEE model	Independent variables	Regression coefficient (β) (95% CI)	p value
Disease activity score (DAS28ESR4)	Follow up time	−0.09 (−0.12, −0.07)	<0.0005
	Gender¶	−0.23 (−0.57, 0.11)	0.18
	Gender*Follow up time (interaction)	0.04 (0.009, 0.06)	0.009
	Age	0.003 (−0.006, 0.01)	0.57
	RF	0.0006 (0.0002, 0.001)	0.006
	Fatigue VAS	0.02 (0.01, 0.02)	<0.0005
	DAS28ESR4 at baseline	0.50 (0.39, 0.61)	<0.0005
	CESD	0.02 (0.004, 0.03)	0.007
	DMARD prescribed at baseline	−0.20 (−0.54, 0.15)	0.27

^¶The referent group for gender was male

(b) Self-reported Measures

Measures of disease activity, functional capacity and quality of life obtained by patient self-report, i.e. patient global scores, HAQ-DI scores, fatigue VAS, pain VAS, and CESD scores, were all higher among women at baseline (table 1). Although they improved in both genders over time, they remained higher among women (Figure 1f, g and h; data not shown for pain VAS and CESD). Accordingly, female gender was significantly associated with these measures in the GEE models where these were the outcome variable. Only pain VAS was not significantly influenced by gender although it remained higher among women over time. The mean difference, from the GEE models, in each of these measures between women and men was as follows: patient global VAS scores: 9.3 (p=0.001; 95% CI: 3.6, 15.0), HAQ-DI: 0.30 (p<0.0005; 95% CI: 0.14, 0.46), fatigue VAS: 10.2 (p=0.002; 95% CI: 3.9, 16.5), and CESD scores: 4.6 (p=0.003; 95% CI: 1.6, 7.5). However, gender did not influence the rate of change of these self-reported measures during the 2-year follow up.

(c) Acute-phase Reactants and RF

A different trend was observed for ESR, CRP and RF. Although levels of these measures decreased significantly in men and women during the first 6 months after baseline, and continued to decrease among men, mean levels among women increased after 6 months (Figure i, j and k). However, gender was not significantly associated with these measures (ESR: p=0.8; CRP: p=0.8; RF: p=0.1), or their rate of change. Similar results were obtained for ESR and CRP after gender-specific published formulae were applied to adjust for age.

(d) Radiographic Damage

As shown in Figure 2, radiographic damage, as assessed by total Sharp scores, increased significantly over time in both men and women, with mean scores being consistently higher among women. Interestingly, throughout the study period, men had worse erosion scores, and women had worse JSN scores. In the GEE models with JSN scores as outcome, female gender was a significant predictor for higher JSN scores (p=0.005), but did not influence the rate of change of this outcome. The JSN scores were also significantly influenced by age (p<0.0005). Interestingly, after stratifying by gender, age was found to be a significant predictor of JSN scores only among women (p<0.0005), whereas among men, there was no association with age (p=0.28).

Figure 2.

The changes in radiographic damage, as assessed by erosion scores, joint space narrowing scores and total Sharp scores, are shown over the 2-year follow up.

(e) ACR Improvement and DAS28 Remission

In the GEE models, gender did not significantly influence ACR20 or ACR50 improvement or DAS28 remission. Nonetheless, there was a non-significant trend for increased proportions of men satisfying the criteria for ACR20 improvement and DAS28 remission throughout the 24-month follow up, as shown in Figure 3a and c. This trend was, however, not observed for ACR50 improvement (Figure 3b).

Figure 3.

(a)–(c): The changes in response to treatment, as assessed by ACR20 and ACR50 improvement and DAS28 remission are shown over the 2-year follow up.

Responses to treatment in men and women

When the analyses were limited to the subset of 221 patients (48 men, 173 women) who were prescribed DMARDs at the baseline visit, similar results were obtained as those described above, with women showing worse progression, and men showing better improvement in the different outcome measures over time (data not shown).

Sensitivity Analyses

A total of 15 men and 62 women were lost to follow up during the 2-year follow up: 4 men, 19 women were lost to follow up by 6 months; 3 men, 13 women by 12 months; 8 men, 30 women by 24 months. Among the patients who remained in the study, data items for random variables were missing for various time points in random patients. Repeating the analyses using a subset of 106 patients (26 men and 80 women) who had complete DAS28ESR4 data from all visits also yielded results similar to those described above for the entire cohort (Figure 4).

Figure 4.

(a)–(j): The observed mean values for different outcomes, including the ACR core set measures, for 106 patients (26 men (solid line) and 80 women (dashed line)) with complete DAS28ESR4 data at all time points over the 2-year follow up. The bars indicate the 95% confidence intervals at each point.

Discussion

The present study is the first to report the longitudinal analysis of a prospective DMARD-naïve seropositive early RA cohort assessing how gender influences disease activity, functional disability, and radiographic outcomes, as well as subsequent treatment responses, over time, adjusting for within-patient correlation of data [30]. Previous studies examining gender differences in RA have mostly used ordinary linear regression and correlation methods to examine the influence of baseline characteristics on disease outcomes after a specified follow up time, thus not taking into account within-subject correlation. Our results show that, in this DMARD-naïve, seropositive early RA cohort with active polyarticular arthritis, men and women had equally active disease and radiographic joint damage at baseline, although women had higher scores for self-reported measures. Similar proportions of men and women were prescribed DMARDs at baseline, or had methotrexate as their first DMARD during the course of the study. Over time, however, even among patients who initiated DMARDs at baseline, disease progression was worse among women, while men showed better responses to treatment. Our results are consistent with the existing evidence that, in early stages of the disease, both men and women have similar disease activity, as shown in a cohort in the Netherlands [9], the Swedish BARFOT cohort [10, 2], and a cohort in Greece [31], all with RA duration less than 12 months at study entry. In contrast, cross-sectional studies of long-standing RA have reported women having more active disease than men [12, 16]. This difference is most likely due to different progression rates of RA in men and women. As demonstrated in a number of early RA cohorts [9, 14, 2], after approximately 2 years, disease activity tends to be significantly worse among women compared to men, as we also observed in the present study. Thus, all currently available evidence point to similar disease activity in the early stages of the disease, followed by a worse disease course among women over time.

It is unclear why women have higher disease activity than men after the disease has progressed for over a year. As seen from the GEE results, in this early RA cohort, the rate of change of disease activity scores was significantly influenced by gender; interestingly, follow up time was a significant predictor – a covariate often not taken into account. In examining components of the DAS28ESR4, it appears that the main differences between men and women were in ESR, with minor differences in tender joint counts and patient global scores. However, even after adjusting the ESR by the standard correction factor for gender and age [28], the reported gender differences in the DAS28ESR4 at 2 years remained present. Furthermore, the trends in improvement of both the DAS28CRP4 and DAS28ESR4 scores were significantly better among men in our cohort, suggesting that gender differences in ESR do not explain the observed differences in disease activity, as previously suggested [18]. It has also been argued that observed differences in disease activity are not due to differences in the disease per se, but arise as a result of gender differences in reporting of disease activity measures [12, 17]. Although the women in our cohort reported worse scores for pain, function and global health compared to men, this pattern of reporting was consistent throughout the study, including at baseline when there was no gender difference in disease activity. Based on the equation to calculate the DAS28, patient global health scores contribute little to the overall DAS28, while pain VAS and HAQ-DI do not contribute at all, suggesting that differences in these self-reported measures do not account for the increasing difference in disease activity between men and women over time. Furthermore, measures assessed by the physician, i.e. physician global assessment and tender joint counts, also followed the same trends over time as the disease activity scores, with the improvement rates among women starting to slow down after 6 months. Interestingly, an increase in the levels of inflammation markers, i.e. ESR and CRP, as well as in RF levels was observed among women after 6 months, when disease activity scores started to diverge between men and women. It is possible that women are not as responsive to anti-inflammatory medication as men, which might explain the short-lived amelioration in levels of inflammatory markers. This increase in inflammation markers could be related to the increase in tender joint counts, patient global scores, as well as physician global assessment, observed after 6 months, and eventually higher disease activity among women. Increased patient global scores and joint tenderness among women were also observed in the BARFOT cohort, and attributed to increased pain perception in women [2]. It is unclear, however, why swollen joint counts were the same in men and women throughout the study duration if there was increased inflammation among women after 6 months.

Given the higher disease activity among women compared to men over time, it not surprising that men appeared to be more likely to satisfy the ACR criteria for 20% or 50% improvement, and to achieve DAS28 remission in our cohort. This is consistent with previous findings of increased remission among men, both in the early RA cohorts [11, 14, 18, 32] and in well-established RA [16, 17]. Although most studies have used the less stringent DAS28 remission criteria, maleness has been significantly associated with remission regardless of the remission criteria used [17]. In addition to these observational studies, a meta-analysis of data from randomized controlled trials comparing methotrexate to placebo or other DMARDs, has also suggested that women were less likely to respond to treatment than men [32].

Interestingly, in our dataset, men and women had similar radiographic joint damage calculated as total Sharp scores throughout the study, resulting from higher proportions of men having radiographic erosions and worse mean erosion scores, and women having worse JSN, as reported from other early RA cohorts [9, 11, 2]. Although the mean JSN scores increased in both men and women over 2 years, age was a significant predictor only among women. It is not clear what other factors are involved among men. It is also unclear why nodules were more prevalent among the men in our study. We and others have previously shown that in long-standing severe RA, men are more prone to nodules [5, 13], whereas other studies found no such gender differences [33]. Since the patients in our early RA cohort were DMARD-naïve at study entry, increased nodules in men cannot be explained as a side-effect of methotrexate use, but instead appears to be intrinsic to the disease process. Furthermore, men had higher RF titers than women even though all patients, irrespective of gender, had been selected to be seropositive for RF. We had previously reported such a gender bias in RF and ACPA in familial RA [5]. Higher RF titers may be explained in part by the significantly higher proportions of ever smokers among men since RF titers appear to be increased among smokers [34]. The ACPA titers among men in this early RA cohort, on the other hand, were only marginally higher than in women, suggesting that the difference we had previously observed [5] could be a feature of familial RA as opposed to sporadic RA.

The study has some limitations. First, the early RA cohort used was clinic-based and was selected for severe RA. It is thus not representative of population-based early inflammatory polyarthritis cohorts. However, men and women were selected using the same criteria; we are not aware of any gender biases in patient selection that may have affected our results. Second, the sample size was small, especially for male patients. Even though the GEE models use all available data, thus increasing the power to detect small associations, the results should be interpreted with caution until replicated in larger early RA datasets. Third, there was some loss to follow up and missing data items during the 24 months study period; however, we found no gender biases in the proportions of men and women lost to follow up or in the disease activity of those who remained. However, we cannot rule out the possibility of selection bias in the outcome measures of those who were lost to follow up. Fourth, we used the DAS28<2.6 criterion for remission, which reflects remission at one time point, rather than sustained remission, and is thus less stringent. Since only few patients satisfied the criteria for ACR remission, we felt that the DAS28 remission was appropriate, and it also allowed our results to be compared with previous reports, most of which have used DAS28 remission. Last, the GEE approach does not provide a measure of how well the model fits the data, and hence, the choice of a correlation structure is not always clear. We therefore compared different correlation structures and got similar results.

In summary, based on the results from our cohort, responses to treatment over time in early RA, as assessed by disease outcomes, ACR improvement and DAS28 remission, appeared to be better among men in the pre-biologic era. And although there were no gender differences at baseline in disease activity and radiographic damage, disease progression seemed to be worse among women.