Introduction

Rheumatoid arthritis (RA) patients have premature mortality, frequently attributed to cardiovascular diseases (CVDs), especially coronary heart disease (CHD) [1]. Despite new effective RA medication, the mortality gap between RA patients and the general population has become even wider during the most recent years [2]. Only few studies report changes over time in cause-specific mortality. These studies have shown coronary [3] and RA-related [4, 5] deaths to decline.

Causes of deaths (CoDs) are likely to be inaccurate without autopsy, the gold standard, in at least one third of cases [6]. However, autopsy rates in Europe and the US have declined from 60% in the 1960s to 10% or less nowadays [6].

Our aim was to investigate in patients treated for RA at a single hospital CoDs in 1971–1991. We also aimed to clarify contribution of autopsy to the CoDs.

Materials and methods

Every RA patient treated for RA in 1971–1991 at Kivelä Municipal Hospital, responsible for primary to tertiary RA treatment in Helsinki, was identified by an identification code from the National Hospital Discharge Register. Of the 3,186 RA patients, 960 died in 1971–1991. RA patients were treated by rheumatologists, who routinely used the international RA classification criteria and started disease-modifying antirheumatic drugs (DMARDs) in the hospital ward. Because the identification code came into regular use in the 1970s, we included RA patients deceased from 1971. During the period studied, autopsy rate declined (Table 1; p value for linear trend = 0.010). Therefore, the cutoff year was set at 1991. Statistics Finland provided CoD information.

Table 1 Characteristics of patients
Full size table

We report multiple CoDs including, if not stated otherwise, immediate, underlying, contributory, and intervening antecedent CoD from death certificates. CoDs based on autopsy when performed.

Because majority of infections classified by the International Classification of Diseases (ICD) are recorded under the affected organ system, some infections may remain undiscovered. To study this, we analysed underlying CoDs classified by the ICD (eighth revision <1979; ninth revision ≥1979) and by a modified grouping recording infections in one category, used by us [7]. Our grouping records cardiac, vascular, and cerebrovascular diseases in CVDs and other categories as those classified by the ICD, but infections excluded.

The Ethics Committee of Helsinki University Central Hospital provided favourable opinion on the study protocol.

Statistical analysis

We used NCSS and SPSS statistical software systems. The χ 2 or Fisher's exact probability tests were used to compare frequencies. Comparisons of continuous variables were executed by Student's t test or Welch's test. Statistical significance for hypotheses of linearity was evaluated by analysis of variance or Cochran–Armitage test. Logistic regression analysis was adopted to get age- and sex-adjusted probabilities.

Results

Of the 960 patients, 53% were autopsied (medicolegal autopsy in 11%; Table 1). Natural deaths occurred for 925 patients (97%). Over the 21 years, mean age at death increased from 65 to 73 years (p value for linear trend < 0.001).

The leading CoD was CVD (Table 2). If RA was the CoD, mean ± standard deviation age at death was 67 ± 10 years in males and 72 ± 9 years in females (p < 0.001). The most frequent RA-related complication was reactive amyloidosis (9% of patients). Other such complications were very uncommon.

Table 2 Causes of death in patients with rheumatoid arthritis
Full size table

Infection appeared as an underlying CoD less frequently by the ICD (1%) than by our grouping (8%). Respiratory disorders were more common by the ICD (8%) than by our grouping (3%). Other CoDs showed no major differences.

Males died more frequently of malignancies (19% vs 11%; p = 0.004), lung cancer (10% vs 3%; p < 0.001), and respiratory diseases (10% vs 4%; p < 0.001), but females died more frequently of RA (50% vs 31%; p < 0.001). Between the genders, amyloidosis (7% of males vs 10% of females; p = 0.196) or other CoDs showed no significant differences.

Over 1971–1991, RA (Fig. 1; p value for linearity < 0.001) and renal (p value for linearity = 0.010) deaths declined, whereas CVDs, CHD, and infections showed no trend.

Fig. 1
figure 1

Sex- and age-adjusted deaths caused by rheumatoid arthritis with 95% confidence intervals. Multiple causes of death

Full size image

Compared to nonautopsied patients, CHD and gastrointestinal (GI) disorders were more frequent, and RA and endocrinologic and renal diseases were less frequent in autopsied patients (Table 3). Between patients with and without autopsy, CVDs showed no significant difference in an analysis of underlying CoD (45% vs 47%; p = 0.539) and multiple causes (both 61%). Autopsied patients (3%) died more frequently of pericarditis or myocarditis than nonautopsied patients (0.2%; p < 0.001). Between these patient groups, no other RA-associated complication showed significant differences. However, all the patients dying of medullar compression of the cervical spine were in the autopsy group (three patients).

Table 3 Causes of death in rheumatoid arthritis patients with and without autopsy
Full size table

Sudden unexpected death occurred in 33 (3%) patients, of whom nine (27%) were autopsied. The cause of sudden death was CVD in all the autopsied patients and in all but one of nonautopsied.

Discussion

Our patients represent a wide clinical spectrum of RA treated by rheumatologists at a single hospital. The rheumatologists routinely used the international RA classification criteria and started DMARDs in hospital ward. Thus, RA diagnoses are likely to be accurate, and the study has included all RA patients treated in 1971–1991. Here, we report 21-year mortality data of these patients.

We analysed multiple CoDs to get comprehensive data on determinants of RA mortality. An analysis of solely underlying CoD may underestimate the contribution of RA to mortality [5]. Reactive amyloidosis may remain undetected in an analysis of underlying CoD, because, according to the World Health Organization (WHO), it should not be recorded as the underlying cause. Furthermore, CoD analysis according to the ICD may leave some infections unrecognised, as observed here. We, therefore, analysed multiple CoDs according to our classification recording infections in one category.

Similar to one population-based study over 1950–1981 [4], RA-related deaths declined here. The explanation may be that RA is becoming milder or that RA treatment has been improving [8]. Close to our findings, another population-based study over 1970–2002 reported RA-related deaths to decline from 1970 to the 1990s [5]. However, RA deaths in that study started to increase, particularly in the oldest patients, in the mid-1990s. One explanation suggested was the implementation of aggressive treatment on fragile elderly patients. Both these studies reported more RA-related deaths in females, but when RA was a CoD, males tended to die younger [4, 5]; both findings similar to ours. RA, therefore, may be more aggressive in males.

Parallel to one 11-year follow-up study [9], half of our nonautopsied cases died of RA, a higher proportion than for our autopsied cases (38%). This higher proportion could be explained by referral bias. Unclear CoD most likely has been the indication for autopsy referral here. This may have affected our study populations. A clinician may have considered an autopsy unnecessary if a patient had severe RA. Similar explanation may be suggested for our nonautopsied patients with severe diabetes or end-stage renal failure. Compared with autopsied patients, nonautopsied patients had more unspecific CoDs, such as heart failure and renal failure, which, in nonautopsied patients, constituted the majority of renal deaths. However, renal failure merely represents a common end-stage for various renal disorders. The same is true for heart failure. Furthermore, the clinicians here did not refer a patient with sudden death for autopsy, as might be expected. Studies have shown similar discrepancy rate in CoD diagnoses between cases where the clinician would have requested an autopsy to those where they would not [6]. Despite intensive modern clinical investigations, autopsies continue to reveal major premortem diagnostic errors in one third of cases [10].

Parallel to other findings [11–14], the leading CoD here was CVD. Studies have reported CVD to appear as an underlying CoD in 39–53% of RA patients [11, 12, 14], figures close to ours (46%). Active RA plays a significant role in development of atherosclerosis [1]. Although RA-related deaths declined, no decline occurred in cardiovascular deaths here. Several potential factors may contribute. One factor may be that nonaspirin nonsteroidal anti-inflammatory drugs (NSAIDs), associated with increased cardiovascular risk [15], were introduced in Finland mainly in the 1970s [16]. Another factor may be that the most effective RA medication to reduce cardiovascular risk, i.e., antitumour necrosis factor (TNF) therapy [1], was not used during the study period. Glucocorticoids, used frequently in RA, may promote CVDs, but this is controversial [11, 17]. Furthermore, we can speculate that our RA cases may have gained suboptimal benefit of improvement in diagnostic techniques and treatment of CHD, because RA patients less likely than those without RA report symptoms of angina and more likely experience unrecognised myocardial infarction [18].

Similar proportions of autopsied and nonautopsied patients died of CVDs, but the autopsied more frequently died of CHD. One explanation for the higher proportion of CHD in our autopsied cases may be that those dying of CHD were more frequently referred for autopsy. Another explanation may be that CHD in our nonautopsied cases underwent undetected. This is supported by several observations: (1) RA patients are susceptible to silent coronary events [18]; (2) coronary deaths in autopsied RA patients are, compared to non-RA cases, diagnosed less accurately [7]; (3) myocardial infarction often remains undiscovered before autopsy [6]; and (4) in autopsied subjects, even if the major disease category of CVD is correct, the specific type of CVD may be incorrect [19].

As shown earlier [1], our cases frequently died of infection. Parallel to others' findings [9, 14], the most frequent infectious CoD here was respiratory infection, which has been shown to go often undetected before autopsy [6]. We, however, found no significant difference between the patient groups in infectious deaths. This may indicate that our patients' infections were diagnosed accurately. Because anti-TNF therapy was not in use during the period studied, our patients may have been less susceptible to opportunistic infections than are RA patients nowadays.

Between the patient groups, deaths caused by malignancies showed no significant differences. Because of this similarity, we may also speculate that those CoDs had been determined accurately, as described previously [19]. Compared to females, the males in our study died more frequently of pulmonary cancer, in parallel with findings in Finnish population during 1955–1992 [20].

GI deaths in our autopsied patients, compared to nonautopsied, were threefold higher. One explanation may be that our RA patients dying of GI disease were more often referred for autopsy. Another explanation may be that GI deaths remained undetected during lifetime. Before autopsy, GI disorders [19], especially GI haemorrhage [21], have often been overlooked. Our RA cases most likely had been taking nonselective NSAIDs—risk factors for serious and frequently asymptomatic GI complications [22]. In our nonautopsied cases on nonselective NSAIDs, GI deaths may have, thus, remained undiscovered.

In conclusion, our findings indicate that RA patients die most frequently of CVDs, RA, and infections. Deaths caused by RA declined within 21 years. Our findings may indicate that the specific type of CVD such as CHD may be overlooked during lifetime.