Survival in polymyalgia rheumatica and temporal arteritis: a study of 398 cases and matched population controls

Abstract

Objective. To estimate survival in polymyalgia rheumatica (PMR) and temporal arteritis (TA).

Methods. The present study encompassed 338 incident cases who were diagnosed at the Department of Rheumatology during the period 1987–1997 and 60 cases diagnosed in the same period but admitted to hospital for reasons other than PMR or TA. The 398 patients were each assigned four age‐ and sex‐matched controls from the same population and mortality ascertained.

Results. Among the 338 incident cases, there were 69 deaths compared with 360 deaths among their 1352 controls. The mortality was thus 28% lower in cases than in controls [relative risk (RR)=0.72, 95% confidence interval (CI) 0.55–0.95]. The 274 incident cases with pure PMR had increased survival compared with controls (RR=0.70, 95% CI 0.52–0.95), whilst among the 64 incident TA patients and their controls, no difference in mortality was found (RR=1.2, 95% CI 0.55–2.74). Patients diagnosed at other departments and their controls had the same mortality. In the incident cases, the mean initial dose of prednisolone, the mean maintenance dose of prednisolone, the mean initial erythrocyte sedimentation rate and C‐reactive protein and frequency of peripheral arthritis did not differ between survivors and those dying during the observation period.

Conclusion. The study showed increased survival in patients with PMR compared with controls, whilst mortality in TA equalled that of controls. There was no association between use of corticosteroids and level of disease activity and death. The increased survival in PMR might be explained by improved medical surveillance.

Polymyalgia rheumatica, Temporal arteritis, Mortality, Survival.

Polymyalgia rheumatica (PMR) and temporal arteritis (TA) are usually regarded as two clinical variants of the same disease. However, inflammation of arteries is the typical lesion in TA, whereas in PMR, musculoskeletal symptoms predominate and synovial tissues may represent primary antigenic targets. Consequently, peripheral arthritis is seen in about one‐third in PMR, while in TA such clinical manifestations appear rather rare [1].

PMR and TA may thus be more different than traditionally regarded, and the clinical and pathological discrepancies may result in different morbidity and mortality. Although several studies on mortality and survival in PMR and TA have emerged [2–16], very few surveys have separated TA and PMR [7, 8], and important information regarding possible prognostic differences may therefore have been concealed. There is also a lack of prospective studies with appropriately matched controls which appear necessary to determine morbidity and mortality.

The results of previous studies on mortality in PMR and TA have been inconclusive, some estimating increased survival [9], others showing increased mortality [3, 5, 7, 10, 11], while rates of mortality and survival similar to those of the general population have been obtained in some investigations [2, 6, 12]. The present study encompassed 338 incident cases who were followed during the period 1987–1997 and 60 cases diagnosed during the same period but having been admitted to hospital for causes other than PMR and TA. The 398 patients were each assigned four age‐ and sex‐matched controls and mortality ascertained.

Materials and methods

Geographics

The study included patients from the county of Aust Agder only. The county is located on the coast of south Norway, and its population reached 100 000 in 1993. Less than 10% of the working population are employed in heavy manual labour. The rate of unemployment during the study period was on average 3%. More than 95% of the population are Caucasian of Nordic origin (Norsemen). The county has only one hospital and no private practising rheumatologists.

Prior to the start of the investigation in April 1987, all general physicians and hospital doctors were informed of the study and asked to refer patients suspected of developing PMR and/or TA to the Department of Rheumatology before initiating drug administration. The methods and design of the study have been described in detail in previous publications [17, 18].

Patients

In total, 398 patients had onset of PMR or TA during the study period. A total of 338 cases were incident cases diagnosed by us. In addition, 60 cases were found by reviewing all hospital records with a diagnosis of PMR or TA given during the study period. The review was carried out after the study period, and only cases diagnosed with PMR or TA between 1 April 1987 and 30 December 1997 were included. Each patient was assigned four sex‐matched controls from the general population of Aust Agder. The controls were born in the same year and month as the index case. The control patients had to be alive at the time of onset of disease in the index case. Thus, cases and controls were the same age at the time of study entry. The controls were selected at random from the National Population Registry. This registry contains names, dates of birth and death and present address.

For PMR, the criteria suggested by Bird et al. [19] were used. In TA, the American College of Rheumatology (ACR) criteria were applied [20], but only cases with a biopsy‐proven diagnosis were included. Other diagnoses were carefully excluded. The patients were classified as PMR, TA or PMR and TA, the latter patient group meeting diagnostic criteria of both PMR and TA [19, 20]. Due to missing descriptions of temporal artery biopsies in the hospital records, three females diagnosed at other departments could not be definitely classified (denoted unclassifiable cases).

The incident cases were regularly examined by a rheumatologist and followed until permanent disease remission and drug withdrawal or death. To calculate mortality rates, all 398 patients were followed to emigration, death or the end of the study period (31 December 1997). The mean observation time for the group was 64 months (0–150.5 months). Information about deaths during the follow‐up was provided by Statistics Norway. The study was approved by the Regional Committee for Medical Ethics.

Statistics

The χ² test, Mann–Whitney test, Student's t‐test and Fisher's exact test were used to compare characteristics between the patients and the control group. In order to assess the association between the two groups and mortality, stratified Cox proportional hazard regression analysis was used. The stratification made sure that the matched patient–control sets were controlled for. Kaplan–Meyer survival curves showed the graphical association between the patients and the control group. P values less than 0.05 indicated significant results. All statistical analyses were carried out using the Epi Info (version 6) package and the SAS system [21].

Results

Incident cases

Deaths

There were 69 deaths during the observation period among the 338 incident cases compared with 360 deaths among 1352 controls (Table 1). Taking into consideration the length of follow‐up, the relative risk (RR) for the incident cases was 0.72 (P=0.02) [95% confidence interval (CI) 0.55–0.95].

Table 1.

Incident cases: number of deaths among patients with PMR and TA and their controls

	Patients			Controls
	Number	Deaths	Number	Deaths	RRa	95% CI
PMR
Females	181	37	724	185	0.74	0.51−1.07
Males	93	19	372	113	0.65	0.39−1.07
All PMR	274	56	1096	298	0.70	0.52−0.95
TA
Females	30	6	120	20	1.37	0.54−3.51
Males	12	2	48	10	0.94	0.20−4.44
All TA	42	8	168	30	1.23	0.55−2.74
TA and PMR
Females	19	4	76	27	0.48	0.16−1.40
Males	3	1	12	5	0.76	0.08−6.89
All TA/PMR	22	5	88	32	0.52	0.20−1.36
All cases with TAb	64	13	256	62	0.82	0.44−1.52
All patients	338	69	1352	360	0.72	0.55−0.95

^aAdjusted for the matched patient–control sets.

^bIncludes all patients from group II (TA) and from group III (TA with PMR).

Mortality by disease group

Patients with pure PMR had an increased survival compared with controls (RR=0.70, P=0.02) (Table 1). In pure TA, survival was not significantly different from that of controls (RR=1.22) (Table 1). There were no significant differences between cases and controls for the other disease categories or between males and females within the disease categories.

Survival and time of death

Figure 1 shows the calculated Kaplan–Meyer survival curves for men and women with PMR and their controls. The major differences in survival between patients and controls appeared after a disease duration of approximately 2–3 yr.

Figure 2 shows the Kaplan–Meyer survival curves for patients with TA and their controls. No significant differences between patients and controls were found, in particular no trend towards increased survival among patients appeared.

Fig. 1.

Survival in patients with PMR and their controls.

Fig. 2.

Survival in patients with TA and their controls.

Therapy, disease activity and peripheral arthritis

The mean initial dose of prednisolone in dying patients was not significantly different from that of survivors (30.3 and 27.7 mg, respectively). Similarly, the mean maintenance dose of daily prednisolone during the first year of observation was almost identical in survivors and dying patients (6.2 and 6.3 mg, respectively). The mean initial erythrocyte sedimentation rate (ESR) was 72.5 mm among survivors compared with 80.8 mm among dying cases (not significant). The mean initial C‐reactive protein (CRP) was 65.3 g/dl in survivors compared with 80.4 g/dl among those who died during the observation period (not significant). Peripheral arthritis was present in 27.6 and 25.0% among survivors and dying patients, respectively (not significant). Adjustment for age and gender did not change these conclusions (data not shown).

Other cases of PMR and TA

The mortality rate of 60 cases found in other departments did not differ significantly from that of their controls (RR=0.80, 95% CI 0.42–1.50). There were, however, few cases in each diagnostic category which precluded further analysis. If all 315 cases of pure PMR were included in one analysis, we confirmed the lower mortality of PMR patients initially found in incident cases only (RR=0.73, P=0.03). Similarly, if all cases of pure TA were merged into one analysis, we confirmed that TA patients and their controls had the same mortality (RR=0.95).

Discussion

The incident cases included in the present prospective study were based on patients referred to hospital by general practitioners. Prior to the investigations, all physicians in the county were asked to refer all patients suspected of having PMR or TA to the Department of Rheumatology. They were also asked not to initiate treatment with corticosteroids. As there are no private practising rheumatologists in the county and only one hospital, the patient population studied appears representative of the majority of PMR and TA developing during the study period. Nevertheless, the review of hospital records after the inclusion of prospective cases showed that in spite of serious efforts to include all incident cases in the county, some patients were diagnosed and treated by other hospital units or general practitioners. However, as cases later found in the hospital records were also included, the only cases possibly missed in the present study could be patients treated by general practitioners exclusively. These patients most probably have pure PMR and no coexistent disease or corticosteroid‐induced toxicity, hence without the need for hospital admittance. An earlier study of incidence rates estimated very high numbers of pure PMR [18], strongly indicating that the number of cases not referred to the hospital in the present study was very small. The study of TA was, however, limited to biopsy‐proven cases and a possible disadvantage of the present study was thus that comparisons with other surveys on TA following strictly the ACR criteria may be difficult. Finally, the controls were randomly selected and then matched for sex, age and place of residence. We thus believe that the cases studied are representative of the total population of PMR and TA developing during the observation period, and that the reported survival reflects the true survival in the patient population.

Mortality in PMR and TA has been the subject of a number of investigations [2–16], the majority focusing on TA only. We are aware of only two studies analysing mortality in PMR separately, one finding reduced survival [7], the other increased survival [9]. Three surveys of patients with PMR, TA and PMR/TA all found improved survival among patients compared with the general population [13–15]. Most studies were retrospective and based on cases admitted to hospital. Survival among patients was unanimously compared with that of the total population in the country of study. Thus, due to different study designs, it is difficult to compare the results of the present study with previous observations. The present investigation was prospective, analysed TA and PMR separately and included sex‐ and age‐matched controls from the same county to avoid differences in social, cultural and environmental background between cases and controls.

Among the cases with PMR only, survival was significantly increased compared with controls. In one previous study on pure PMR, survival was increased, but in males only [9]. We did not find any difference between males and females, and the estimated increased survival reached statistical significance only when males and females were analysed as one group. An increased survival can in general be explained by factors specifically related to the disease process, but we are unable to find any plausible explanation as to why the pathogenesis of PMR should lead to increased survival. Similarly, the therapy employed, in almost all cases being oral corticosteroids, cannot provide an acceptable explanation for increased survival among patients. One possible explanation would be improved medical surveillance as our patients were followed regularly from onset of disease to disease remission or death.

It is worth noting that the lower mortality in PMR patients was found in all PMR patients registered in the hospital and in the cases of PMR followed by the Department of Rheumatology from disease onset.

In the present study, no statistically significant difference in survival between patients with TA and their controls emerged. Our results regarding TA thus clearly contrast the increased survival estimated among PMR patients. Previous studies on mortality in TA have revealed diverging results, some with reduced survival [3, 4, 10, 11], others with similar survival in cases and controls [2, 5, 12]. In series of patients with both PMR and TA, some have found increased survival [13–15], while one study [6] found similar mortality in cases and controls. Although the actual patient population of TA was small, our results indicated that survival rates are similar in patients and controls when incident cases are selected from the total population. A potential bias may, however, have been the death of patients with TA prior to the events of diagnosis and initiation of treatment. Similar to all previous investigations, we have no means of evaluating such a possible selection bias.

In the present study, efforts were also made to look for possible predictors of survival. We suspected that patients with less severe disease, as expressed by moderate disease activity and low requirements for corticosteroids, were those with the best chance of survival. This was not the case. As maintenance doses of corticosteroids were guided by severity of symptoms only, the similar maintenance doses during the first observation year between survivors and dying patients do not suggest a relationship between disease severity and early death. Also, mean ESR and CRP at diagnosis were not significantly different among dying patients and survivors, supporting the lack of association between disease activity and survival. In a previous study of 284 patients with giant cell arteritis [5], an increased risk of dying from vascular disorders in the initial phase of the disease was found. After 4 months of disease, the risk equalled that of the general population. Based on the results of that study [5], one may suspect patients with high initial disease activity to be at particular risk of early death, but disease activity during the subsequent disease course does not influence survival. Further studies are clearly warranted to explore possible factors related to survival and death in PMR and TA. In a study in progress, causes of death among patients and controls will be analysed. Moreover, the incidence of cardiovascular and cerebrovascular disorders are at present also subject to investigation.

Correspondence to: J. T. Gran, Revmatologisk avdeling, Aust Agder Sentralsjukehus, 4809 Arendal, Norway.