Mortality in early inflammatory polyarthritis: Cardiovascular mortality is increased in seropositive patients

Abstract

Objective

To determine the degree and causes of any excess mortality observed during the early years of inflammatory polyarthritis (IP).

Methods

Between 1990 and 1994, a total of 1,236 patients were registered with the Norfolk Arthritis Register, a primary care–based inception cohort. All patients were tracked on the National Health Service Central Register for notification of death. The vital status of each patient was determined as of December 31, 1999. Causes of death were coded according to the International Classification of Diseases, Ninth Revision. Expected death rates were calculated using annual death rates for the Norfolk population. Standardized mortality ratios (SMRs) were calculated for all IP patients and for the subgroups of patients who did and did not satisfy the American College of Rheumatology (ACR) 1987 criteria for rheumatoid arthritis (RA) at baseline, as well as for the subgroups who were and were not rheumatoid factor (RF) positive at baseline.

Results

By December 31, 1999, 160 patients (13%; 79 women and 81 men) had died. The median duration of followup in the entire cohort was 6.9 years. Mortality rates were not significantly increased in the entire group of patients with IP or in the subgroup who met the ACR 1987 criteria for RA at baseline. In contrast, RF-positive patients had an increased rate of death from all causes (SMR in men 1.51, in women 1.41). Cardiovascular disease was the most common cause of death. The majority of the excess mortality in the RF-positive patients could be attributed to cardiovascular causes (SMR in men 1.34, in women 2.02).

Conclusion

Excess mortality in the early years of IP is confined to patients who are seropositive for RF. While excess cardiovascular mortality has been described in patients with established RA, this is the first report of premature death from heart disease in the early years of IP.

Rheumatoid arthritis (RA) is a chronic inflammatory disease of unknown etiology that may extend over several decades of a patient's life. As well as reducing the quality of the patient's life, RA may reduce life expectancy. Most studies of mortality in RA patients (1–15) have found increased mortality rates compared with the general population (1–10, 12, 14, 15). However, two recent studies of inception cohorts (11, 13) did not detect any increase in mortality rates during the early years of RA.

There is increasing interest in examining the link between death from cardiovascular causes and RA (2, 8). Inflammation is an intrinsic part of RA, and it has recently been recognized that inflammation also plays an important role in atherosclerosis (16). This shared pathogenesis may explain the link between RA and cardiovascular disease (CVD) (17). Previous studies in patients with well-established disease have shown that deaths from cardiovascular causes are increased. There is, however, little evidence to show whether this predominance of excess deaths due to CVD extends back into the early years of RA.

In RA, age and male sex (1–3, 9, 11, 13, 14) are consistent predictors of increased mortality, as expected from their role in the general population. Rheumatoid factor (RF) positivity has been found to predict mortality in a number of studies (1, 3, 9), but the influence of other disease-related factors is unclear. The disparity in results is due to variation in the recording of clinical variables, the method of analysis, methods of patient recruitment, study design, and length of followup.

In order to monitor mortality in the early years of disease it is essential that patients are followed from near the time of disease onset. This can best be achieved by the use of a community-based cohort, which avoids the selection bias inherent in studies that recruit patients from clinics or hospitals. The Norfolk Arthritis Register (NOAR) (18), which was established in 1989, is the largest population-based incidence register of patients with early inflammatory polyarthritis (IP) in Europe, and is therefore well placed to examine mortality in the early years of disease. The aim of the current analysis was to determine whether mortality rates are increased in the early years of IP compared with the general population, and to examine mortality rates in the subgroup of patients who are seropositive at an early stage in their disease course.

PATIENTS AND METHODS

The Norfolk Arthritis Register.

NOAR is a primary care–based inception cohort of patients with early IP. Details of the registry have been published elsewhere (18). Briefly, the registry covers the area of the former Norwich Health Authority, with a population of approximately half a million. Primary care providers notified NOAR of patients who were age 16 years or older at symptom onset, had swelling of at least 2 joints that had persisted for at least 4 weeks, and had symptom onset after January 1, 1989. A parallel notification system operated from hospitals within the catchment area.

Patient cohort.

Between 1990 and 1994, a total of 1,362 patients were referred to NOAR. Of these patients, 126 were subsequently diagnosed by a hospital consultant as having a condition other than RA, IP, psoriatic arthritis, or postviral arthritis that accounted for their symptoms. These 126 patients were therefore excluded from further analysis.

The cohort utilized in this analysis comprised the remaining 1,236 patients. A subset of patients with IP met the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) 1987 criteria for RA (19), although this proportion varied depending on the point in the disease course at which the criteria were applied (20).

Baseline data.

At the baseline (or, registration) assessment, patients were visited at home by a research nurse who performed a structured interview and clinical examination. Data recorded included age at symptom onset, sex, time from symptom onset to presentation to NOAR, and history of smoking. In addition, a variety of clinical data were recorded, including the number of swollen and tender joints (range 0–49), the duration of early morning stiffness (on the day of the assessment), and the presence of rheumatoid nodules. A blood sample was also taken for RF testing by a latex agglutination technique. A titer of ≥1:40 was considered positive.

Completion of a Health Assessment Questionnaire (HAQ), modified for use in British patients (21), formed part of the baseline assessment. The HAQ comprised 20 questions covering 8 areas of daily living; scores ranged from 0 (no disability) to 3 (severe disability).

Notification of death.

In the UK, 98% of residents are registered with a National Health Service (NHS) general practitioner. The NHS Central Register (NHSCR) is a computerized registry of the records of all NHS patients. Access to this registry for patients residing in England and Wales is obtained via the Office for National Statistics (ONS).

All patients in this cohort were tracked on the NHSCR so that NOAR would be notified in the event of their death, and their cases were followed until December 31, 1999. Death certificates of patients who died before January 1, 1991, were obtained from central records. Causes of death were coded by the ONS, using the International Classification of Diseases, Ninth Revision (ICD-9) (22). The underlying cause of death on the death certificate was extracted for all patients and grouped according to ICD-9 chapter.

If a patient emigrated or was no longer registered with a general practitioner, the ONS notified NOAR of the date of this “embarkation.” The patient's vital status beyond this time point cannot be ascertained, and therefore, data for any patients who “embarked” from the NHSCR were deemed censored at the time of embarkation.

Statistical analysis.

Standardized mortality ratios (SMRs). Age- and sex-specific mortality rates for the Norfolk population for the years 1990–1998 were obtained from the ONS. No population mortality data for 1999 were available at the time of analysis and, therefore, the mortality rates for 1998 were substituted. Expected death rates in the NOAR cohort were calculated using these annual Norfolk population data.

Observed and expected numbers of deaths were compared using SMRs, which were calculated for all patients with IP, and then separately for all men and all women with IP. SMRs were also calculated for the subgroups of patients who did or did not satisfy the ACR 1987 classification criteria for RA (19) at baseline, as well as for patients who were or were not RF positive at baseline. SMRs were calculated for all causes of death, as well as for deaths from neoplasms (ICD-9 codes 140–239), cardiovascular causes (ICD-9 codes 390–459), and respiratory causes (ICD-9 codes 460–519).

Predictors of mortality. The following variables recorded at baseline were assessed univariately as predictors of mortality using Cox proportional hazards regression (23): age at symptom onset (in 10-year age bands), sex, delay to presentation (tertiles), smoking status at baseline (never smoked, current smoker, ex-smoker who stopped <15 years prior to registration, or ex-smoker who stopped ≥15 years prior to registration), number of joints with active IP (tertiles), early morning stiffness (≥60 minutes), HAQ score ≥1.50, RF (positive/negative), and nodules (absent/present).

Baseline variables that were significant at P ≤ 0.20 on univariate analysis were entered into a multivariate model. This model was simplified using the likelihood ratio test (24). Variables that were significant at P < 0.10 on multivariate analysis were retained. Previously excluded variables (P < 0.20 on univariate analysis) were added to the multivariate model to determine whether they contributed significantly. Those that were significant at P < 0.10 were retained. The proportional hazards assumption of the final multivariate model was checked.

Statistical analyses were conducted using Excel 97 (Microsoft, Redmond, WA), SPSS for Windows (25) and Stata (26).

RESULTS

Characteristics of the study cohort.

The study cohort comprised 814 women (66%) and 422 men (34%). The median age at symptom onset was 53 years (interquartile range [IQR] 41–66). Eighty-nine percent of the patients (n = 1,106) presented to NOAR within 2 years of symptom onset. The median delay from symptom onset to registration by NOAR was 6.2 months (IQR 3.1–12.7). At baseline, 575 patients (47%) satisfied the ACR 1987 classification criteria for RA. RF was measured in 522 patients in the RA subgroup, and 45% were found to be seropositive at baseline. Table 1 shows the baseline variables for the whole IP cohort and the RA-positive and RF-positive subgroups.

Table 1. Baseline variables for the entire IP cohort and the RA-positive and RF-positive subgroups*

Baseline variables	Entire IP cohort (n = 1,236)	RA-positive subgroup (n = 575)	RF-positive subgroup (n = 320)
* IP = inflammatory polyarthritis; RA = rheumatoid arthritis; RF = rheumatoid factor; IQR = interquartile range; HAQ = Health Assessment Questionnaire. † RF was measured in 1,085 patients. ‡ Smoking history was obtained from 1,234 patients.
Sex, no. (%) female	814 (66)	393 (68)	186 (58)
RA, no. (%)	575 (47)	–	235 (73)
RF titer ≥1:40, no. (%)†	320 (29)	235 (45)	–
Nodules, no. (%)	93 (8)	74 (13)	51 (16)
Smoking history, no. (%)‡
Never smoked	389 (32)	185 (32)	74 (23)
Current smoker	334 (27)	144 (25)	113 (35)
Ex-smoker	511 (41)	241 (43)	132 (42)
Age at symptom onset, median (IQR)	53 (41–66)	57 (45–68)	57 (48–68)
Delay to presentation, median (IQR) months	6.2 (3.1–12.7)	6.6 (3.3–13.1)	6.5 (3.1–12.6)
Swollen and tender joint count, median (IQR)	3 (0–8)	7 (3–14)	4 (1–9)
HAQ score, median (IQR)	0.75 (0.25–1.375)	1.125 (0.5–1.75)	0.75 (0.375–1.625)

Among the entire cohort of 1,236 patients, 1,085 were tested for RF, of whom 320 (29%) were classified as being RF positive. The baseline demographics and arthritis profiles of the 151 patients who did not have RF measured were very similar to those of the patients who did have RF measured. The two groups also had similar rates of mortality (data not shown). However, there was evidence of a difference between men and women. RF was measured in 91% of the male patients, compared with 86% of the female patients (risk difference 5% [95% confidence interval (95% CI) 2, 9]).

Traditional nonsteroidal antiinflammatory drugs (NSAIDs) were taken by 75% of the patients (n = 927) at baseline. During the course of followup, 652 patients (53%) were treated with disease-modifying antirheumatic drugs (DMARDs) and/or steroids. Sulfasalazine was the first drug prescribed for 52% (n = 340) of those treated, with steroids being used as the first drug in a further 28% (n = 182). Combination therapy and use of methotrexate were fairly rare (3% and 7%, respectively).

Followup and main causes of death.

The median followup time for the cohort was 6.9 years (IQR 5.6–8.2). By December 31, 1999, 160 patients (13%; 79 women and 81 men) had died. The death certificate for 1 woman was unobtainable, and this patient was excluded from further analysis. Cardiovascular disease was the most common cause of death in women (47%) as well as in men (33%) (Table 2). Neoplasms also ranked high as a common cause of death (23% of women and 33% of men). Only 27 death certificates (17%) mentioned RA (ICD-9 code 714.0) as a contributory factor, and RA was given as the main cause of death on only 5 death certificates (3%).

Table 2. Main cause of death in the NOAR cohort*

Cause of death	ICD-9 code	Deaths from all causes
		No. (%) of women (n = 78)	No. (%) of men (n = 81)
* The death certificate for 1 woman was unobtainable; this patient was therefore excluded from further analysis. NOAR = Norfolk Arthritis Register; ICD-9 = International Classification of Diseases, Ninth Revision.
Infectious and parasitic diseases	001–139	1 (1.3)	5 (6.2)
Neoplasms	140–239	18 (23.1)	27 (33.3)
Mental disorders	290–319	0 (0.0)	1 (1.2)
Diseases of the nervous system and sense organs	320–389	1 (1.3)	1 (1.2)
Diseases of the circulatory system	390–459	37 (47.4)	27 (33.3)
Diseases of the respiratory system	460–519	6 (7.7)	14 (17.3)
Diseases of the digestive system	520–579	5 (6.4)	1 (1.2)
Diseases of the genitourinary system	580–629	2 (2.6)	0 (0.0)
Diseases of the musculoskeletal system and connective tissue	710–739	6 (7.7)	3 (3.7)
Other	780–999	2 (2.6)	2 (2.5)

Among women, 40% of deaths from malignancy were attributed to lung cancer, which is nearly double the percentage in men (19%) (Table 3). In men, deaths from cancer were predominantly due to carcinoma of the esophagus (n = 5), stomach (n = 3), colon (n = 3), and associated sites (n = 4). The site of malignancy was not specified for 22% of the women and 7% of the men.

Table 3. Deaths from neoplasms in the NOAR cohort*

Cause of death	ICD-9 code	Deaths from neoplasms
		No. (%) of women (n = 18)	No. (%) of men (n = 27)
* NOAR = Norfolk Arthritis Register; ICD-9 = International Classification of Diseases, Ninth Revision.
Malignant neoplasm
Digestive organs and peritoneum	150–159	2 (11.1)	15 (55.6)
Bronchus and lung	162.9	7 (38.9)	5 (18.5)
Breast	174.9	3 (16.7)	–
Ovary	183.0	1 (5.6)	–
Prostate	185	–	1 (3.7)
Bladder	188.9	0 (0)	2 (7.4)
Brain	191	0 (0)	1 (3.7)
Site unspecified	199	4 (22.2)	2 (7.4)
Non-Hodgkin's lymphoma	200	1 (5.6)	0 (0)
Myeloid leukemia	205	0 (0)	1 (3.7)

In women, 5 of the 6 deaths attributed to respiratory causes were due to pneumonia (ICD-9 codes 480–487). However, pneumonia was given as the cause of death in only 4 of the 14 men who died of respiratory illnesses. The remaining deaths from respiratory disease in men were attributed to chronic obstructive pulmonary disease (ICD-9 codes 490–496) (n = 7) or fibrosing alveolitis (n = 3).

Standardized mortality ratios.

Rates of death from all causes were not significantly elevated in women (Table 4) or in men (Table 5) with IP. Stratifying by RA status did not alter these findings (Tables 4 and 5). In contrast, death from all causes was increased in RF-positive patients (in women, SMR 1.41 [95% CI 0.93, 2.05]; in men, SMR 1.51 [95% CI 1.06, 2.08]). There was also an increased risk of death from cardiovascular causes in RF-positive patients compared with the general population (Tables 4 and 5). In addition, the risk of death from respiratory disease was increased in RF-positive men (SMR 2.66 [95% CI 1.14, 5.23]).

Table 4. Standardized mortality ratios (SMRs) in women*

Group	Cause of death
	All causes				Neoplasms (ICD-9 codes 140–239)				Cardiovascular disease (ICD-9 codes 390–459)				Respiratory disease (ICD-9 codes 460–519)
	Obs	Exp	SMR	95% CI	Obs	Exp	SMR	95% CI	Obs	Exp	SMR	95% CI	Obs	Exp	SMR	95% CI
* Values are the observed (Obs) and expected (Exp; based on Norfolk population data) number of deaths, with SMRs and 95% confidence intervals (95% CIs). The rheumatoid arthritis (RA) subgroups either met (positive) or did not meet (negative) the American College of Rheumatology (formerly, the American Rheumatism Association) 1987 criteria for the classification of RA at baseline. The rheumatoid factor (RF) subgroups either had (positive) or did not have (negative) RF titers ≥1:40 at baseline.
All women with IP	78	77.5	1.01	0.80, 1.26	18	22.0	0.82	0.49, 1.30	37	32.2	1.15	0.81, 1.59	6	9.3	0.64	0.24, 1.40
RA
Positive	43	43.5	0.99	0.71, 1.33	10	12.1	0.83	0.40, 1.52	20	18.2	1.10	0.67, 1.69	3	5.3	0.56	0.11, 1.65
Negative	35	34.0	1.03	0.72, 1.43	8	9.8	0.81	0.35, 1.60	17	13.9	1.22	0.71, 1.96	3	4.0	0.75	0.15, 2.19
RF
Positive	27	19.1	1.41	0.93, 2.05	5	5.6	0.90	0.29, 2.09	16	7.9	2.02	1.15, 3.28	2	2.2	0.89	0.10, 3.21
Negative	36	44.9	0.80	0.56, 1.11	10	12.9	0.77	0.37, 1.42	15	18.5	0.81	0.45, 1.34	4	5.3	0.75	0.20, 1.92

Table 5. Standardized mortality ratios (SMRs) in men*

Group	Cause of death
	All causes				Neoplasms (ICD-9 codes 140–239)				Cardiovascular disease (ICD-9 codes 390–459)				Respiratory disease (ICD-9 codes 460–519)
	Obs	Exp	SMR	95% CI	Obs	Exp	SMR	95% CI	Obs	Exp	SMR	95% CI	Obs	Exp	SMR	95% CI
* Values are the observed (Obs) and expected (Exp; based on Norfolk population data) number of deaths, with SMRs and 95% confidence intervals (95% CIs). The rheumatoid arthritis (RA) subgroups either met (positive) or did not meet (negative) the American College of Rheumatology (formerly, the American Rheumatism Association) 1987 criteria for the classification of RA at baseline. The rheumatoid factor (RF) subgroups either had (positive) or did not have (negative) RF titers ≥1:40 at baseline.
All men with IP	81	71.7	1.13	0.90, 1.40	27	20.2	1.34	0.88, 1.95	27	32.5	0.83	0.55, 1.21	14	9.2	1.53	0.84, 2.57
RA
Positive	41	37.8	1.08	0.78, 1.47	17	10.4	1.63	0.95, 2.61	12	17.2	0.70	0.36, 1.22	7	5.03	1.39	0.56, 2.87
Negative	40	33.9	1.18	0.84, 1.61	10	9.8	1.03	0.49, 1.89	15	15.3	0.98	0.55, 1.62	7	4.1	1.69	0.68, 3.49
RF
Positive	37	24.6	1.51	1.06, 2.08	10	7.1	1.41	0.68, 2.60	15	11.2	1.34	0.75, 2.20	8	3.0	2.66	1.14, 5.23
Negative	38	42.7	0.89	0.63, 1.22	16	11.9	1.35	0.77, 2.19	11	19.3	0.57	0.28, 1.02	4	5.6	0.72	0.19, 1.84

Mortality rates were not significantly increased in the small group of patients who were seropositive and were assigned RA status at baseline (in women, SMR 1.31 [95% CI 0.81, 2.01]; in men, SMR 1.31 [95% CI 0.85, 1.93]). This may reflect the small number of patients in this subgroup.

There were 8 excess deaths (calculated by subtracting expected deaths from observed deaths) among RF-positive women, all of which were due to CVD. When stratified by age at death, women younger than 65 years had a 3-fold increased risk of dying from CVD compared with women in the general population (Table 6), although the wide 95% CI precluded definitive interpretation. When the stratification was repeated using a cutoff of 75 years of age in women, this younger excess cardiovascular mortality was confirmed (SMR 3.09 [95% CI 1.33, 6.09]). In RF-positive men, there were 12 excess deaths, 4 of which were due to CVD. All 4 of these deaths in men occurred prior to the age of 65 years (Table 6).

Table 6. Mortality rates in seropositive patients, by sex, age at death, and cause of death*

Group, age at death	Cause of death
	All causes				Cardiovascular causes (ICD-9 codes 390–459)
	Obs	Exp	SMR	95% CI	Obs	Exp	SMR	95% CI
* Values are the observed (Obs) and expected (Exp; based on Norfolk population data) number of deaths, with standardized mortality ratios (SMRs) and 95% confidence intervals (95% CIs). The rheumatoid factor (RF) subgroups either had (positive) or did not have (negative) RF titers ≥1:40 at baseline. ICD-9 = International Classification of Diseases, Ninth Revision.
RF-positive women
<65 years	2	2.7	0.74	0.08, 2.68	2	0.60	3.31	0.37, 11.9
≥65 years	25	16.4	1.52	0.98, 2.25	14	7.3	1.91	1.04, 3.21
<75 years	13	7.8	1.65	0.88, 2.84	8	2.6	3.09	1.33, 6.09
≥75 years	14	11.3	1.24	0.68, 2.08	8	5.3	1.50	0.65, 2.95
RF-positive men
<65 years	9	2.9	3.14	1.43, 5.96	5	1.1	4.37	1.41, 10.2
≥65 years	28	21.7	1.29	0.86, 1.86	10	10.1	0.99	0.47, 1.82
<75 years	18	9.7	1.86	1.10, 2.94	8	4.3	1.85	0.80, 3.65
≥75 years	19	14.9	1.28	0.77, 1.99	7	6.9	1.01	0.41, 2.08

Patients who satisfied classification criteria for RA at baseline or who were RF positive were significantly older at symptom onset than those who did not satisfy criteria for RA or who were RF negative. Therefore, the SMRs calculated for these patient subgroups are not directly comparable.

Predictors of mortality.

Preliminary analysis demonstrated that the effect of a number of clinical predictors differed between the sexes (data not shown), and therefore, multivariate predictors of mortality were identified separately for men and women (Tables 7 and 8). In all models, age was a strong predictor of mortality, as was expected.

Table 7. Multivariate predictors of mortality in women*

Baseline variable	All women with IP		RA-positive subgroup		RF-positive subgroup
	No.	HR (95% CI)	No.	HR (95% CI)	No.	HR (95% CI)
* The multivariate Cox regression model developed for all women with inflammatory polyarthritis (IP) did not satisfy the proportional hazards assumption of the model; therefore the effect of rheumatoid factor (RF) was stratified by followup time in order to meet the assumption. The hazards ratios (HRs) for RF are interpreted according to a followup of <4 years versus ≥4 years. Among women who survived for at least 4 years, those who were RF positive were not more likely to die compared with those who were RF negative. RA = rheumatoid arthritis; 95% CI = 95% confidence interval; HAQ = Health Assessment Questionnaire.
All causes of death
Age at symptom onset (10-year age bands)	650	3.02 (2.32, 3.94)	327	3.34 (2.33, 4.79)	176	2.95 (1.96, 4.43)
HAQ score <1.50	473	1.00	205	1.00
HAQ score ≥1.50	177	1.68 (0.98, 2.91)	122	1.96 (0.93, 4.15)
Cardiovascular causes of death
Age at symptom onset (10-year age bands)	650	4.00 (2.63, 6.08)	327	4.76 (2.72, 8.34)	176	3.33 (1.92, 5.77)
RF negative			187	1.00
RF positive			140	2.41 (0.87, 6.67)
With followup <4 years		10.4 (2.25, 48.3)
With followup ≥4 years		0.87 (0.27, 2.72)

Table 8. Multivariate predictors of mortality in men*

Baseline variable	All men with IP		RA-positive subgroup		RF-positive subgroup
	No.	HR (95% CI)	No.	HR (95% CI)	No.	HR (95% CI)
* IP = inflammatory polyarthritis; RA = rheumatoid arthritis; RF = rheumatoid factor; HR = hazards ratio; 95% CI = 95% confidence interval.
All causes of death
Age at symptom onset (10-year age bands)	372	2.53 (2.00, 3.20)	160	2.45 (1.66, 3.62)	128	2.30 (1.62, 3.26)
Nodules absent	346	1.00	139	1.00	111	1.00
Nodules present	26	3.37 (1.85, 6.14)	21	4.20 (1.85, 9.54)	17	4.49 (2.09, 9.67)
RF negative	244	1.00	77	1.00
RF positive	128	1.64 (1.03, 2.61)	83	1.87 (0.93, 3.78)
Cardiovascular causes of death
Age at symptom onset (10-year age bands)	372	2.32 (1.57, 3.43)	160	2.06 (1.04, 4.11)	128	1.95 (1.16, 3.27)
Nodules absent	346	1.00	139	1.00	111	1.00
Nodules present	26	5.38 (2.18, 13.2)	21	11.1 (2.90, 42.6)	17	6.59 (2.11, 20.6)
RF negative	244	1.00
RF positive	128	2.09 (0.94, 4.67)

Age at symptom onset and severe disability (HAQ score ≥1.50 at baseline) were identified as multivariate predictors of death from any cause in the entire group of women with IP as well as in the subgroup who satisfied classification criteria for RA at baseline. The rate of death among women who had a HAQ score ≥1.50 at baseline was between 68% and 96% higher than that in women with a HAQ score <1.50 at baseline. For RF-positive women, the rate of death again increased 3-fold for each 10-year increase in age, although no other predictors were identified on multivariate analysis for this subgroup (Table 7).

Among women, RF was a strong predictor of death from cardiovascular causes in both the entire group with IP and the RA subgroup (Table 7). In the entire group of women with IP, the initial multivariate model did not satisfy the proportional hazards assumption of the Cox model, and therefore, the effect of RF was stratified by followup time. For this group, RF had the strongest effect in the initial stages of followup. For example, in the first 4 years of followup, the rate of death from cardiovascular causes was 10 times higher in women who were RF positive than in women who were RF negative. However, in women who had survived for ≥4 years, there was no increase in the rate of death according to RF status. Again, in the subgroup of women who were RF positive, age was the only variable identified as being predictive of mortality.

As was true for the women, increasing age at symptom onset was a strong predictor of mortality in men (Table 8). Among men, rheumatoid nodules and RF positivity were consistently identified on multivariate analysis as predictors of mortality from any cause and mortality from CVD. At all times during the study, the rate of death from any cause among men with nodules was 3–4 times that among men who did not have nodules at the time of the baseline assessment. The presence of rheumatoid nodules was more strongly associated with death from cardiovascular causes. Men who were RF positive at baseline died at approximately twice the rate of RF-negative men throughout the study.

DISCUSSION

We examined the mortality experience of a large inception cohort of patients with early IP. There were 159 deaths during 8,423 person-years of followup (excluding 1 woman whose death certificate was unobtainable). Mortality rates among patients who were seropositive at baseline were increased compared with the local population. No excess mortality was detected in the entire group of patients with IP or in the subgroup of patients who satisfied the ACR 1987 criteria for RA at baseline.

Women who were seropositive at baseline were twice as likely to die of cardiovascular causes as were women in the general population. Most of the excess cardiovascular mortality among men occurred before age 65. Respiratory deaths were increased 3-fold in seropositive men. These deaths occurred at older ages (≥65 years) and most (50%) were attributed to chronic obstructive pulmonary disease. Eighty-five percent of this subgroup had a history of smoking, which may have contributed to the development of the obstructive lung disease. However, due to the small numbers, it was not possible to identify predictors of respiratory death.

The increased cardiovascular mortality rates in seropositive patients in our inception cohort is consistent with the findings of previous studies of patients with established rheumatoid disease (2, 3). Interestingly, seropositivity in the absence of arthritis has also been shown to be associated with increased rates of cardiovascular mortality (27).

Most previous studies of mortality in RA have been restricted to patients who satisfy the ACR 1987 criteria for RA (19). Given that one of the criteria for RA is being RF positive, it is likely that RF is a strong contributor to the increased mortality rates seen in previous studies of RA patients. We know that the likelihood of being classifiable as having RA or being RF positive increases with followup (20) and that in an established RA cohort, RA status and RF positivity correlate closely. If mortality rates are elevated in RF-positive patients, one would expect to find increased mortality rates in RA patients. However, in our inception cohort, fewer than half of the patients in the RA subgroup were RF positive at baseline. Therefore, the results of this study suggest that baseline measurement of RF is better than RA status at predicting mortality in the early years of the disease process.

When interpreting the results of this study, we have to consider that a number of potential predictors of mortality were investigated. The possibility that the association between mortality and RF is a false-positive result cannot be discounted. However, the fact that this association was observed in both men and women when the two groups were analyzed separately increases the probability this is a genuine association.

Our results appear to conflict with those of 2 recent studies of RA patients with disease onset during the 1980s and 1990s, which did not show any increase in mortality rates during the early years of disease (11, 13). A high percentage of patients in both of those cohorts were seropositive. In the study by Kroot et al (13), 91% of the patients received DMARD therapy during the first year, while 62% of the cohort studied by Lindqvist and Eberhardt (11) were treated with DMARDs. In contrast, only 41% of our cohort had taken DMARDs and/or steroids during the first year of followup. The improved prognosis observed by Lindqvist and by Kroot (11, 13) may be the result of more aggressive early treatment of rheumatoid disease. This hypothesis is compatible with the newly discovered role of inflammation in the pathogenesis of atherosclerosis (16).

Over recent years, it has become clear that ischemic heart disease and RA share some common features. It has been suggested that atherosclerosis represents a low-grade inflammatory condition associated with elevated levels of C-reactive protein (CRP), cytokines, and fibrinogen. Elevation of CRP levels has been shown in several studies to increase the future risk of cardiovascular events (28–31). Other markers of inflammation also shown to predict cardiovascular events include interleukin-6, soluble intercellular adhesion molecule type 1, serum amyloid A, and fibrinogen (30, 32). Of all these markers of inflammation, CRP remains the strongest independent predictor of future cardiovascular events in apparently healthy populations. The temporal relationship between atherosclerosis and elevated CRP production is not known. It is possible that the inflammation that occurs in RA may accelerate the atherosclerotic process in these patients.

Another common feature of both diseases is dyslipidemia. The level of disease activity in RA seems to affect the type of dyslipidemia (33). Active inflammatory disease leads to a lowering of the levels of total cholesterol and high-density lipoprotein cholesterol, which leads to an unfavorable ratio of high-density lipoprotein to low-density lipoprotein. This dyslipidemia may predispose the RA patient to the development of CVD. If uncontrolled inflammatory joint disease leads to a dyslipidemia that favors atherosclerosis, this would partly explain the excess cardiovascular mortality seen in the early years of followup in our study.

Other possible etiologic agents include elevated levels of homocysteine. In one study (34), elevated homocysteine levels moderately increased the risk of cardiovascular events in postmenopausal women, but other studies have failed to show this effect. Elevated homocysteine levels have been found in RA patients, and DMARDs such as methotrexate or sulfasalazine may increase the risk of hyperhomocysteinemia through their antifolate effects (35–37).

Other drugs used to treat RA may influence the risk of developing CVD. Although one might expect corticosteroids to increase the risk of cardiovascular events by accelerating the atherosclerotic process, Wållberg-Jonsson et al (38) found that prolonged steroid use was actually associated with a lowering of the risk of cardiovascular events in a seropositive RA cohort. We did not find that steroid use at baseline predicted death from all causes or death from cardiovascular causes in this study. The use of traditional NSAIDs may, because of their antiplatelet effects, be protective against CVD events. NSAID use in this study was not associated with death. There have been suggestions that the use of selective cyclooxygenase 2 (COX-2) inhibitors may be associated with an increased risk of CVD events (39). These drugs were not widely available during the period of our study. It will be interesting to see if the incidence of CVD events increases as the use of selective COX-2 inhibitors becomes more prevalent in the treatment of RA.

Smoking is one of the classic risk factors for CVD in the general population. However, smoking was not identified as a predictor of mortality within our cohort when examined as categories of smoking status or as years of smoking (data not shown). Wållberg-Jonsson et al (2) also found that smoking did not predict cardiovascular events in a cohort of seropositive RA patients. Like us, those investigators did not have detailed “pack-year” smoking histories for their patients, and it may be that heavy cigarette smoking is more strongly associated with outcome. It is possible that the limitation in the way we recorded smoking history has led to a false-negative observation that smoking was not associated with mortality in our study. Wolfe et al (1) found an association between death from cardiovascular causes and the number of packs of cigarettes smoked per day (relative risk 1.5 per pack, 95% CI 1.2, 1.8). Wolfe (40) has also described a linear relationship between years of smoking and RF and nodule formation. In our study, multivariate Cox regression identified RF as a predictor of cardiovascular death in both men and women. In addition, in men, nodules were a very strong predictor of mortality from all causes and from CVD.

There is a complicated relationship between smoking, RF positivity, disease severity, and death from cardiovascular causes that is not yet fully understood. Previous studies have identified a link between smoking and RF status. As expected, seropositive patients were more likely to be smokers compared with seronegative patients. However, on univariate analysis, smoking was not identified as a predictor of death from all causes or from cardiovascular causes. This suggests that RF positivity was not simply acting as a marker for smoking status.

Classic risk factors for CVD do not seem to be increased in the RA population (41, 42). It is not known whether targeted reduction of cardiovascular risk factors would lead to a reduction in cardiovascular risk in RA patients similar to that seen in the general population (43). The identification of seropositive patients as “high-risk” patients may justify aggressive DMARD therapy in this group. This may lead to a reduction in cardiovascular mortality rates through maximum suppression of inflammation in these patients. These areas warrant further investigation.

In summary, this is the first report of increased cardiovascular mortality rates in the early years of seropositive inflammatory arthritis. Excess cardiovascular mortality appears to be strongly linked to early seropositivity in patients with IP. Seropositive RA may, like diabetes, act as an independent risk factor for CVD. The ability to identify patients at increased risk of death from cardiovascular causes is obviously of interest to the clinician. It remains to be proven whether targeted cardiovascular risk factor modification, aggressive suppression of the inflammatory rheumatoid disease, or a combination of the two would actually reduce cardiovascular mortality rates in the seropositive RA patient.

Acknowledgements

We would like to thank the general practitioners and hospital consultants within the catchment region for their continued support of NOAR. We would also like to thank the NOAR staff based at Aylsham for their continued hard work. In addition, we would like to acknowledge the assistance of the Office for National Statistics in providing access to the NHSCR.