In this large, nationwide, population-based, 12-year, cohort study (PsA cases, n=33 026), the all-cause mortality risk in patients with PsA diagnosed in specialised rheumatology/internal medicine care in Sweden was significantly increased, although only by around 10%, compared with the general population, mainly driven by increased risks in women and patients with longer duration since PsA diagnosis. When adjusting for comorbidity status, the increased mortality risk disappeared, indicating that the excess comorbidity burden in PsA (whether causally related to PsO/PsA or not) is the driving force behind the elevated mortality. The cause of death distributions were similar between PsA cases and comparator-subjects, with CVD and malignancy being the leading causes of death in both groups. Predictors of mortality were also generally similar in both cohorts (lower socioeconomic status, general comorbidities, prior joint surgery), although IBD did not predict mortality in PsA and hip/knee replacement surgery was a stronger mortality predictor in PsA than in the general population.

Many of the contradictive, prior mortality estimates in PsA derive from hospital-based, single centre cohorts or smaller population-based studies (with fewer than 1000 PsA cases; online supplemental table S7). Considering that hospital-based studies are likely to capture patients with more severe disease compared with population-based cohorts, mortality estimates from such assessments may be expected to be inflated. However, even the results of these studies are inconsistent in observing an elevated mortality risk in PsA or not (online supplemental table S7). Turning to larger population-based assessments, two studies from Taiwan (9572 patients, 2001–2012, and 8795 patients, 2000–2012) demonstrated a standardised mortality ratio (SMR) of 1.47 (95% CI: 1.36 to 1.58) and HR of 1.52 (95% CI: 1.39 to 1.66), respectively.29 30 Furthermore, the SMR was also increased at 1.34 (95% CI: 1.16 to 1.52) in a Canadian study (15 430 patients, 1996–2016).33 On the other hand, large population-based studies from the UK (8706 patients, 1994–2010),27 Denmark (9817 patients, 1998–2014),36 Israel (5275 patients, 2003–2018)35 and Norway (18 700 patients, 2010–2017),18 did not demonstrate an increased mortality risk in PsA, with reported HR of 1.02 (95% CI: 0.92 to 1.12), stratified HR of 1.06 (no CI reported) and adjusted HR of 1.02 (95% CI: 0.90 to 1.15) and 1.06 (95% CI: 0.99 to 1.13), respectively. This diversity of results was also highlighted in a recent meta-analysis, wherein no increased all-cause mortality in the overall PsA population was found (relative risk 1.12 (95% CI: 0.96 to 1.30)), although the point estimate was very similar to that found in the present study (and an increased risk was indeed observed among female patients with PsA, see below).19 In the aforementioned Israeli cohort, the final, non-increased mortality estimate was adjusted for comorbidities, whereas their crude assessment did show an elevated HR in line with ours at 1.16 (95% CI: 1.04 to 1.29).35 Aiming to quantify the total impact of PsA and its associated comorbidities on mortality, in our main analyses we did not adjust our morality estimates for comorbidities, as we consider this to better illustrate the real burden of PsA. As expected, however, when including such adjustments in an additional analysis, the elevated mortality risk associated with PsA disappeared.

In women with PsA, we found the all-cause mortality risk relative to their sex-matched comparator-subjects to be higher than in male patients, although with lower absolute MR. This observation is in line with the above mentioned meta-analysis,19 as well as the recently published Norwegian study,18 both of which (similar to us) found a significantly increased mortality risk among female, but not male, patients with PsA. To establish the reasons behind this sex difference was beyond the scope of the current study. A higher impact of PsA-related comorbidities on mortality among women compared with men could, however, be hypothesised, and also receives some support from the observation that the percental differences in deaths due to CVD and infections between patients and comparator-subjects were marginally higher for women than men (table 3). The finding that the mortality risk in relation to age-matched and sex-matched comparator-subjects increases with longer duration since diagnosis is compatible with negative effects of long-term systemic inflammation, and may also reflect the accumulation of PsA-related comorbidities over time.

Of note, while we do observe an increased all-cause mortality in the overall PsA population, the risk-increase of around 10% is lower than that previously reported for RA (estimated excess risk 24–126%) or AS (38–64%).15–20 This may also partly explain the conflicting results of prior PsA studies, some of which may have been underpowered to detect a risk-increase of this size. Interestingly, prior studies have found a more clearly elevated mortality risk in PsO than in PsA.27 36 While this could be due to different effects of skin and joint inflammation, it may also be hypothesised that one mechanism underlying such results could be that patients with mild PsO more often than PsA may not seek medical care at all, or are managed exclusively in primary care, rendering them not to be included in mortality assessments. Anyhow, in a recent meta-analysis, the all-cause mortality in PsO (of any severity) was increased in the same general range as that observed for PsA in the present study (pooled relative risk 1.21 (95% CI: 1.14 to 1.28)), whereas it was higher in severe PsO.37

To estimate cause-specific mortality was beyond the aims of the present study. However, apart from the minor sex-differences alluded to above, the cause of death distributions were overall similar between PsA cases and comparator-subjects. In accordance with a number of previous publications,18 29 33 35 36 deaths due to CVD and malignancy were the leading causes in both groups.

Within our PsA population, disease severity, as reflected by previous joint surgery, was a significant predictor of all-cause mortality. Of particular interest, prior hip and/or knee replacement surgery was a stronger mortality predictor in PsA than in comparator-subjects. Prior studies regarding the predictive roles of disease activity and severity on mortality in PsA are limited. In two earlier studies, higher disease activity and elevated acute phase reactants were, however, associated with an increased risk of death in PsA.28 32 General comorbidities were strong predictors of mortality in both PsA and comparator-subjects, whereas extra-musculoskeletal manifestations (ie, IBD, uveitis) did not predict increased mortality in PsA, despite IBD being related to mortality in the population. Our results also indicate that the relative contribution of some conditions, in particular anxiety/depression, to mortality may be greater in the general population than in patients with PsA, where disease-specific factors also play a role. Finally, lower socioeconomic status, approximated by shorter education, was a significant non-differential predictor of mortality among PsA cases and comparator-subjects, in accordance with prior assessments in both PsA and the general population.32 35 47

The large size (the largest to date on this topic) and population-based approach of our study, including patients with PsA with variable disease severity, is expected to provide estimates close to the true MR. The register-based methodology allowed nationwide identification of PsA cases and comparator-subjects from the general population in a uniform and systematic manner. This enabled direct comparisons with internal comparator-subjects, which has previously been shown to yield better mortality approximations than external comparisons (SMRs).48

PsA cases managed exclusively in primary care, as well as a minority subset of cases managed only by private caregivers, are not captured by the NPR and thus not included in the current study. The proportion of individuals with only primary care health contacts for PsA has been estimated at 8.5% and 26.7%, respectively, in two assessments from southern Sweden prior to 2010.38 49 In the first of these studies, however, only 18–24% of the PsA cases with a diagnosis deriving exclusively from primary care were found to fulfil established PsA classification criteria.38 The proportion of PsA cases exclusively followed in private specialised care units that do not report to the NPR (which most nowadays do) is expected to differ regionally, depending on the availability of private caregivers. Patients followed at such units might, however, occasionally also consult public rheumatology departments. Considering that both of these groups of patients may have a milder disease than those included in our study, the observed mortality risk could be somewhat overestimated, although the exact size of such overestimation is unknown.

Furthermore, the validity of the PsA diagnoses in the NPR may be a potential limitation. However, the PsA case definition used for the present assessment has shown high validity, with a positive predictive value for the fulfilment of established PsA classification criteria of 86%.46 Misclassification of PsA diagnoses in the NPR is thus unlikely to have had a substantial impact on our mortality estimates, which is further supported by the similar results in the sensitivity analysis in which 20% of the PsA cases were randomly replaced by one of their own matched comparator-subjects. Misclassification of the main outcome is not a major concern because of virtually complete ascertainment, but misclassification of the cause of death is likely to be present despite the reliable data deriving from the Cause of Death Register. However, this misclassification is possibly non-differential for the PsA cases and comparator-subjects and unlikely to have significantly flawed our results.

Finally, the nationwide register sources used for the current study unfortunately do not provide sufficient coverage regarding lifestyle factors, disease characteristics and clinical variables that would allow for adjustment or stratification for important variables such as smoking, other risk factors for CVD not captured by ICD-codes, disease severity or PsA phenotype. Nor do we know the time of symptom onset, and thus used time since PsA diagnosis as surrogate marker for disease duration. In light of the clear risks of confounding by indication in relation to comorbidities, as well as difficulties to distinguish impact of treatments from that of disease activity, attempting to assess potential relationships between anti-rheumatic treatment exposures (non-steroidal anti-inflammatory drugs, glucocorticosteroids, DMARDs) and mortality in PsA would require a dedicated study in its own right, and was thus beyond the scope of the present manuscript.