The worldwide prevalence of psoriatic arthritis—a systematic review and meta-analysis

Abstract

Objectives

Previous attempts to pool prevalence studies in PsA have failed to take account of important methodological differences between studies that may have created biased estimates. The aim of this review is to estimate the prevalence of PsA within the adult general population worldwide, considering potential differences between population-based and health administrative studies separately.

Methods

Four electronic databases were systematically searched for articles reporting the prevalence of PsA. Data were pooled to generate worldwide prevalence estimates. Where sufficient data were available, results were summarized by continent.

Results

Thirty studies were identified, with half from Europe (n = 15). Thirteen population-based studies were identified comprising >92 000 adults, plus 17 studies (>180 million adults) based on health administrative data. The worldwide prevalence of PsA was 112 per 100 000 adults. The prevalence of PsA estimated using population-based studies was 113 per 100 000 with continent-specific estimates of 207 (Europe), 64 (North America) and 37 (Asia) per 100 000. Health administrative studies gave a global prevalence of 109 per 100 000 with continent-specific prevalence of 175 (Europe), 147 (North America), 78 (Asia) and 17 (South America).

Conclusion

This review compiles currently available estimates of PsA prevalence in the general population into global and continent-based estimates and considers important study design characteristics. There is wide variability between continents, and data in some geographical areas are sparse, but available evidence suggests that PsA is more common in Europe and North America compared with Asia and South America, and current best estimates suggest a global prevalence of 112 per 100 000 adults.

Rheumatology key messages.

• The global prevalence of PsA is 112 per 100 000 adults.

• There is higher prevalence in Europe/North America versus Asia/South America, although data are sparse in many regions.

• Understanding the burden of PsA aids planning and securing resources for its treatment and management.

Introduction

Psoriatic arthritis (PsA) is a complex chronic inflammatory arthritis that often occurs in association with psoriasis. Most cases become symptomatic from the third decade, men and women are equally affected and the highest peak of incidence is observed among persons aged 30–60 years [1, 2].

The prevalence of PsA has been investigated in many epidemiological studies using various approaches. Several meta-analyses have attempted to summarize existing studies, and estimates have varied nearly 20-fold geographically from 0.01% in Asia to 0.19% in Europe [3], and some have reported the prevalence worldwide to be 0.13% [4]. While some authors have focused on the general population, others have examined subgroups of the population of interest. A recent meta-analysis reported the PsA prevalence within the adult psoriasis population to be 21.6%, although these investigators pooled data from different study designs with different inclusion criteria [5]. Others have shown that the prevalence of rheumatologist-diagnosed PsA among persons with psoriasis to be 17.1% [6]. These differences may be explained by differences in methodology and inclusion criteria [7]. For example, if the psoriasis cohort is one with moderate to severe skin disease, participants may be at higher risk of developing PsA, and this in turn will inflate prevalence estimates. Furthermore, some studies use alarmingly simple methods, where case ascertainment is based primarily on self-reported symptoms and without rigorous diagnostic criteria or specialized rheumatology assessment.

Other methodological concerns include the lack of distinction between different study designs when including studies for meta-analyses. Population-based studies and health administrative data studies have distinct advantages and disadvantages. For many years population-based studies have been considered the gold standard as they collect information from a representative sample of the entire population, providing precise estimates of disease prevalence using standardized or known diagnostic criteria. In contrast, health administrative data studies use records from sources like insurance companies, which might not fully capture the disease’s scope due to reliance on healthcare-seeking behaviour and diagnostic coding accuracy. Pooling data from population-based and health administration data requires careful consideration due to methodological and data quality differences, which may introduce important potential biases.

To date, there are no meta-analyses that estimate the global prevalence of PsA taking account of these issues, an omission that is both important and surprising. Thus, the aims of this study are (i) to estimate the prevalence of PsA worldwide in the adult general population, accounting for methodological differences between the studies, and (ii) where possible, to provide prevalence estimates within continents.

Methods

Study selection

A protocol for the systematic review was published on PROSPERO (registration number: CRD42021255494) [8]. The systematic search was developed in consultation with a medical librarian and conducted from database inception to November 2023 in the MEDLINE, EMBASE, CINAHL and Web of Science databases. A predefined search strategy was used and adapted for each database with a combination of medical subject heading (MeSH) terms and text words including terms related to PsA and prevalence: (‘Arthritis, Psoriatic’ [MeSH] OR ‘Psoriatic Arthritis’ OR (‘Psoriasis’ AND ‘Arthritis’) OR ‘Spondyloarthropathy(ies)’ [MeSH] OR ‘Spondylarthropathy(ies)’ OR ‘Psoriatic disease(s)’) AND (‘Prevalence’ [MeSH] OR ‘Incidence [MeSH] OR ‘Occurrence’). The detailed search syntax per database can be viewed in Supplementary Table S1, available at Rheumatology online.

After removing duplicates, titles were screened for eligibility by one reviewer (S.L.). Abstracts were screened by the same reviewer, and a second reviewer (G.T.J.) checked independently 10% of the decisions. All relevant full texts were then retrieved, screened and assessed for eligibility by S.L. and uncertainties were resolved in discussion with G.T.J. and G.J.M. In addition to the electronic searches, the reference lists of eligible papers were manually searched.

Inclusion criteria

Full papers were included if they identified PsA using clinical diagnosis or internationally recognized classification criteria, used a sampling frame that could be considered population-based, and reported on the prevalence of PsA or provided sufficient information to calculate prevalence. If multiple papers were available using data from the same data source, the larger study was included—otherwise no restriction on date was applied.

Exclusion criteria

Studies that assessed the prevalence of PsA in specific population sub-groups (such as psoriasis or indigenous populations) were excluded, as were conference papers and/or letters, as they were unlikely to provide sufficient data to assess eligibility. In addition, studies were excluded that were based on a self-reported diagnosis alone without physician confirmation.

Quality assessment

The quality of all included studies was assessed using the Joanna Briggs Institute’s (JBI) critical appraisal checklist for studies reporting prevalence data [9].

Data extraction

The following data were extracted from each study: geographical location, year of study, study period, study design/data source, size of study population, number of PsA cases, published prevalence, 95% CI, age, distribution of gender, PsA diagnostic/classification criteria. When several annual prevalence estimates were presented, the most recent year up to and including 2019 was extracted to avoid estimates likely to have been distorted by the COVID-19 pandemic.

Where required data were not available, the corresponding authors were contacted to request missing data and studies were included if these were provided.

Health administrative data were defined as insurance data, electronic medical records, hospital- or clinic-based records, and/or disease registries [10].

Statistical analysis

Due to theoretical differences between population-based vs health administrative data, global pooled estimates were calculated separately, and continent-wide estimates were computed using inverse variance weighted random effects models (DerSimonian–Laird method) [11] with logit transformed proportions for variance stabilization [12, 13]. Variability due to between-study heterogeneity was presented using the I² statistic [14]. All effect sizes and pooled estimates were presented per 100 000 with 95% CI. All meta-analyses were performed in R using the ‘meta’ and ‘metafor’ packages [15].

Results

Study selection and main characteristics

In total 9243 papers were identified after removal of duplicates. Title and abstract screening led to 87 papers reaching the full-text screening stage (Fig. 1). Authors from 16 publications were contacted, 14 responded, of whom seven provided the required information to be included in this study. Nine studies were excluded from the analysis for the following reasons: (i) the authors no longer had access to the data required for inclusion, (ii) the provided data did not meet the criteria for inclusion, or (iii) there was no response from authors. A total of 30 studies were included, conducted between 1982 and 2020 and covering over 180 million adults from 24 countries. Fifteen studies were from Europe, seven from Asia, six from North America and two from South America. Thirteen studies used a population-based study approach and 17 studies used health administrative data (Table 1). Overall, based on the JBI assessment tool, study quality was good (Supplementary Table S2, available at Rheumatology online).

Figure 1.

Flow diagram for study selection during literature review. In accordance with the PRISMA reporting guidelines [16]

Table 1.

Study characteristics and prevalence of PsA by population-based and health administrative data studies, country and continent

Reference	Country	Cases	Study size	Prevalence per 100 000 (95% CI)a	Age-standardized prevalence (95% CI)b	Diagnostic method/case definition for PsA
Population-based studies
Europe
Saraux et al. 2005 [17]	France	12	9395	127.7 (72.6, 224.8)	190 (80, 350)	Telephone interview + confirmation by clinician via interview or clinical examination and fulfilment of ESSG criteria
Trontzas et al. 2005 [18]	Greece	14	8740	160.2 (94.9, 270.3)	170 (100, 240)	Face-to-face interview + confirmation by clinician via clinical examination and fulfilment of ESSG criteria
De Angelis et al. 2007 [19]	Italy	9	2155	417.6 (217.4, 800.7)	—	Postal questionnaire + confirmation by clinician via clinical examination and fulfilment of ESSG criteria
Anagnostopoulos et al. 2010 [20]	Greece	6	1705	351.9 (158.2, 781.0)	—	Postal questionnaire + confirmation by clinician via clinical examination and fulfilment of CASPAR criteria
Çakır et al. 2012 [21]	Turkey	9	17 835	50.5 (26.3, 97.0)	—	Face-to-face interview + confirmation by clinician via clinical examination and fulfilment of ESSG criteria
Rodrigues et al. 2019 [22]	Portugal	20	10 661	187.6 (121.1, 290.6)	—	Face-to-face interview + confirmation by clinician via clinical examination
Pérez et al. 2020 [23]	Spain	27	4916	549.2 (376.9, 799.7)	580 (380, 870)	Telephone interview + confirmation by clinician via interview or clinical examination and fulfilment of CASPAR criteria
Asia
Hoa et al. 2003 [24]	Vietnam	1	2119	47.2 (6.6, 334.2)	—	Face-to-face COPCORD questionnaire + confirmation by clinician via clinical examination
Al-Awadhi et al. 2004 [25]	Kuwait	1	7670	13.0 (1.8, 92.5)	—	Face-to-face COPCORD questionnaire + confirmation by clinician via clinical examination
Joshi et al. 2009 [26]	India	5	8145	61.4 (25.6, 147.4)	40 (20, 90)	Face-to-face COPCORD questionnaire + confirmation by clinician via clinical examination
Li et al. 2012 [27]	China	2	10 556	18.9 (4.7, 75.7)	—	Face-to-face interview + confirmation by clinician via clinical examination and fulfilment of CASPAR criteria
North America
Alvarez-Nemegyei et al. 2011 [28]	Mexico	1	3915	25.5 (3.6, 181.1)	—	Face-to-face COPCORD questionnaire + confirmation by clinician via clinical examination
Rodriguez-Amado et al. 2011 [29]	Mexico	4	4713	84.9 (31.9, 225.9)	—	Face-to-face COPCORD questionnaire + confirmation by clinician via clinical examination
Health administrative data studies
Europe
Alamanos et al. 2003 [30]	Greece	221	488 435	45.2 (39.7, 51.6)	56.6 (49.9, 63.2)	Systematic recording system for autoimmune rheumatic diseases and fulfilment of ESSG criteria
Hanova et al. 2010 [31]	Czech Republic	96	154 374	62.2 (50.9, 70.6)	49.1 (39.5, 60.4)	Medical records for diagnosis by clinicians and fulfilment of Vasey and Espinoza criteria
Tekin et al. 2019 [32]	Denmark	10 577	4 689 946	225.5 (221.3, 229.9)	—	ICD-10 codes
Grellmann et al. 2021 [33]	Germany	2877	971 910c	296.0 (285.4, 307.0)	—	ICD-10 codes
Pina Vegas et al. 2021 [34]	France	63 598	67 250 000c	94.6 (93.8, 95.3)	—	Read/SNOMED codes
Kerola et al. 2023 [35]	Norway	18 896	4 094 425	461.5 (445.0, 468.1)	—	ICD-10 codes
Scott et al. 2022 [2]	England	33 636d	11 719 961c	287.0 (284.0, 290.1)	303 (300, 306)	Systematic recording system for autoimmune rheumatic diseases and fulfilment of ESSG criteria
Exarchou et al. 2023 [36]	Sweden	29 359	7 998 644	367.0 (326.9, 371.3)	—	Medical records for diagnosis by clinicians and fulfilment of Vasey and Espinoza criteria
Asia
Eder et al. 2018 [37]	Israel	4490	2 931 199	153.2 (148.8, 157.7)	137 (134, 141)	ICD-9 codes
Wei et al. 2018 [38]	Taiwan	336	840 193	40.0 (35.9, 44.5)	37.75 (33.54, 41.97)	ICD-9 codes
North America
Asgari et al. 2013 [39]	USA (Northern California)	2941	4 299 700c	68.4 (66.0, 70.9)	12.6 (11.6, 13.7)	ICD-9 codes and fulfilment of CAPSAR criteria if diagnosis not entered by rheumatologist
Eder et al. 2019 [40]	Canada	18 655	10 774 802	173.1 (170.7, 175.6)	149 (147, 151)	ICD-9/10 codes
Karmacharya et al. 2021 [41]	USA (Minnesota, Olmsted County)	200	117 323c	170.5 (148.4, 195.8)	181.8 (156.5, 207.1)	ICD-9/10 codes and fulfilment of CAPSAR criteria
Ogdie et al. 2021 [42]	USA	37 150	16 180 180d	229.6 (227.3, 231.9)	—	ICD-9/10 codes
South America
Soriano et al. 2011 [43]	Argentina	65	88 112	73.8 (57.9, 94.1)	—	ICD-9/10 codes + ICPC + SNOMED codes and fulfilment of CASPAR criteria
Júnior et al. 2019 [44]	Brazil	2	60 413	3.3 (0.8, 13.2)	—	ICD-10
Fernández-Ávila et al. 2023 [45]	Colombia	6433	47 651 852e	13.5 (13.2, 13.8)	—	ICD-10

^aPrevalence calculated based on available data.

^bAs reported in the publication. Adjusted to per 100,000 to provide comparison with the crude estimate.

^cData provided by the author(s).

^dData provided in online supplementary data of the publication.

^eSelf-calculated (back calculation of the denominator). CASPAR: Classification Criteria for Psoriatic Arthritis; COPCORD: Community Oriented Program for the Control of Rheumatic Disease; ESSG: European Spondylarthropathy Study Group criteria; ICD: International Classification of Disease; ICPC: International Classification of Primary Care; SNOMED: Systematized Nomenclature of Medicine Clinical Terms.

Population-Based studies

Based on data from 13 population-based studies, the worldwide prevalence was calculated to be 113 PsA cases per 100 000 (95% CI: 64, 198) (Fig. 2).

Figure 2.

Worldwide prevalence of PsA (population-based studies)

Seven studies reported prevalence of PsA within Europe with sample sizes varying between 1705 and 17 835 [17–23]. The individual reported prevalence per 100 000 varied 10-fold from 51 (26–97) to 549 (377–800) [17, 23] and the pooled prevalence was 207 per 100 000 (113–379). All studies used either a face-to-face, telephone or postal questionnaires to identify potential cases who were then clinically examined by a rheumatologist. Four studies applied the European Spondylarthropathy Study Group (ESSG) criteria to classify cases [17–19, 21], two studies used the Classification Criteria for Psoriatic Arthritis (CASPAR) [20, 23], and one classified PsA cases based on clinical diagnosis alone [22].

The prevalence of PsA in Asia was reported by four studies, from Vietnam, Kuwait, India and China [24–27]. Three identified cases through the Community Oriented Program for the Control of Rheumatic Diseases (COPCORD) questionnaire, which was performed door-to-door, followed by a clinical examination by a rheumatologist [24–26]. The sample sizes ranged from 2119 to 8145 and the reported prevalence per 100 000 varied between 13 (2–93) and 61 (26–147). The fourth study used a bespoke screening questionnaire followed by a rheumatologist’s clinical examination of positive questionnaire responders using the CASPAR criteria to identify PsA cases [27]. The study reported 10 556 participants, and a prevalence estimate of 19 per 100 000 (5–76). Across the four studies the pooled prevalence estimate was 38 per 100 000 (18–76).

The two population-based studies in North America were both from Mexico [28, 29], with a combined sample size of 8628. Both studies used the COPCORD questionnaire followed by a clinical examination of positive responders to confirm the PsA diagnosis based on rheumatologist expert opinion. The pooled prevalence estimate was 64 per 100 000 (23–174).

Health administrative data studies

Based on data from 17 health administrative data studies, the worldwide prevalence, was estimated to be 109 per 100 000 (95% CI: 75–158) (Fig. 3).

Figure 3.

Worldwide prevalence of PsA (health administrative data studies)

Eight studies reported the prevalence of PsA using European data [2, 30–36, 46, 47], with sample sizes varying from 154 374 to over 67 million, and prevalence estimates ranging per 100 000 between 45 (40–52) and 462 (455–468), resulting in a pooled prevalence of 175 per 100 000 (105–293). For the identification of PsA cases, most studies used ICD-10 codes within different health administrative data sources, i.e. insurance/healthcare databases, medical records and rheumatology/autoimmune specific registries, [32–35, 46].

Two studies were included from Asia (Israel and Taiwan) [37, 38], and the sample sizes varied from 840 193 to over 2 million with prevalence estimates of 153 per 100 000 (149–158) and 40 per 100 000 (36–45), respectively. There were major methodological differences between studies: in the study from Taiwan [38], PsA cases were identified by examining health insurance records using ICD-9 codes whereas the study from Israel [37] identified the PsA cases through an ICD-9 algorithm applied to a health services database. The pooled prevalence was 78 per 100 000 (21–292).

One study from Canada and three from the USA were included [39–42] with sample sizes ranging up to >16 million. Individual prevalence estimates per 100 000 varied from 68 (66–71) to 230 (227–232). One American study [39] identified cases by applying case-finding algorithms using ICD-9 codes coded by rheumatologists or non-rheumatologists in health maintenance organization records. Coding by non-rheumatologists was confirmed by a rheumatologist using CASPAR criteria. The remaining studies used either or both of ICD-9 and ICD-10 codes to identify cases reported in health administrative data records. The fulfilment of the CASPAR criteria was ascertained retrospectively in one of these studies [41]. Pooling all four studies resulted in an estimate of 147 per 100 000 (100–215).

Three studies were identified to represent the South American continent (Argentina, Brazil and Colombia) with samples sizes of 88 112 and over 47 million, retrospectively [43–45]. There was a 4-fold difference in prevalence per 100 000 between the Argentine (3, 1–13) and the Colombian study (13.5, 13.2–13.8) and more than a 20-fold difference in prevalence between the Argentine and the Brazilian study (74, 58–94). The pooled estimate was 17 per 100 000 (4–70). The studies identified PsA cases through different screening of medical records: the Argentine study used ICD-9/10 and SNOMED-CT (Systematized Nomenclature of Medicine-Clinical Terms) codes and potential patient records were reviewed by two rheumatologists for compliance with CASPAR criteria whereas the studies from Brazil and Colombia solely relied on ICD-10.

Global prevalence

Despite theoretical differences, prevalence estimates based on population-based, or health administrative data studies were similar, 113 and 109, respectively, per 100 000 adults. Based on all 30 included studies, the pooled global prevalence of PsA was 112 per 100 000 (95% CI: 83, 151) (Supplementary Fig. S1, available at Rheumatology online). Combining both study designs the prevalence estimate was 188 (128–289), 48 (20–115), 133 (93–191) and 17 (4–70) per 100 000 for Europe, Asia, North and South America, respectively.

Discussion

This systematic review and meta-analysis has shown that the prevalence of PsA worldwide is 112 per 100 000 adults. Further, we have shown that minimal differences exist between population-based studies (113 per 100 000) and those using health administrative data (109 per 100 000). Evidence suggests that PsA is more prevalent in Europe and North America compared with Asia and South America, although outside Europe data are sparse. Of note, no eligible studies were identified from Africa or Oceania.

The review included studies that varied in terms of PsA classification, but also, crucially, in terms of study design. Despite methodological differences, results were actually similar between population-based and health administrative data studies, so we considered it appropriate to combine a single global PsA prevalence estimate.

During the literature review the search results were confined to papers reporting a prevalence estimate within the adult general population. Studies that reported a PsA prevalence estimate using a population-based approach but including children and/or adolescents were excluded unless data for adults could be separately extracted. For example, recently published papers from Poland [48] and Taiwan [49] reported the PsA prevalence to be 73 and 77 per 100 000, respectively, including children and/or adolescents. The authors of the Polish study estimated that if they had focused solely on the adult population for the calculation of PsA prevalence, it would increase by a factor of 1.15–1.20 (as the Polish population aged 0–17 years is about 18%), but no further explanation is given. It is noteworthy, however, that these estimates fall within the ranges of the included studies but could not be included. In addition, the search identified one study from Japan [50] that investigated the prevalence of PsA. However, the study’s methods of case identification are qualitatively different from all other included studies and indeed it is not clear whether all PsA cases would have been identified from the denominator’s population, artificially lowering the prevalence estimate and introducing a significant source of heterogeneity into the meta-analysis. Additionally, nine studies had to be excluded because the original authors (i) reported that they no longer had access to the necessary data, (ii) provided data that led to ineligibility, or (iii) were non-responsive to data requests.

One strength of the search strategy is that it was developed to capture the wide spectrum of where and how PsA estimates might have been reported. For example, as PsA is part of the SpA family, the search strategy was designed to identify both PsA-specific prevalence studies and papers that present prevalence estimates for SpA and its subtypes separately. Furthermore, this review focused on the importance of distinguishing between methodological differences when summarizing epidemiological studies on the prevalence of PsA. It demonstrates that it is appropriate to summarize a single estimate incorporating different study designs if a robust approach is applied to the included studies.

A previous attempt to collate prevalence information was made by Scotti et al. [4] who estimated the PsA prevalence to be 0.13% worldwide. Another attempt was made by Stolwijk et al. [3] who reported the prevalence of PsA in Europe to be 0.19%, and in the Middle East, 0.01%. Differences between these estimates and the current study may reflect the fact that the current study includes several more recently published studies. Previous reviews included studies with more diverse methods to identify PsA cases whereas the current review applied stricter inclusion and exclusion criteria in this regard to make a methodologically more robust estimation of the prevalence of PsA. It is of note, however, that the prevalence estimate is remarkably similar.

Previous research demonstrated that prevalence estimates differ significantly depending on the target population investigated and diagnostic criteria used. When investigating the prevalence of PsA within psoriasis cohorts, results have shown much higher prevalence estimates compared with investigations within the general population [5]. As well as being one step removed from the general population, one challenge is related to diagnostic accuracy and misclassification. The diagnosis of PsA can be complex and requires clinical assessments, including physical examinations, patient-reported symptoms, and imaging or laboratory tests. In some studies, the reliance on clinical examination to determine the presence of PsA may have introduced a bias, because of the risk of dropout of individuals asked to attend for clinical assessment. This will have served to underestimate the prevalence of PsA. In contrast, in other studies, within psoriasis cohorts, misclassification could lead to the overestimation of PsA.

Within the present work 13/30 studies used a population-based survey approach, and the remainder utilized health administrative data to identify cases of PsA. From an epidemiological perspective population-based surveys are the gold standard to determine prevalence. While this study design uses the consistent application of one of the recognized classification criteria used for PsA, the most essential matter with this study design is that it requires active patient participation, which is costly and time consuming. Therefore, the utilization of health administrative data to estimate the prevalence of PsA within large study populations is attractive because it is associated with lower costs and less time, and algorithms and analytical models can be easily reused and adapted. However, there are disadvantages. For example, using data from statutory/private insurance databases to calculate PsA prevalence in countries where not every citizen has equal access to the health care system may systematically exclude citizens who are uninsured or who may be insured by private insurers [33, 42]. Also, PsA patients who do not seek medical advice or who are undiagnosed will not be captured with approaches using health administrative data [37, 46]. Moreover, the accuracy of the algorithm is dependent on the accuracy of coding in practice. The estimate of PsA in Europe was generally higher within population-based studies than within health administrative data, highlighting the potential effect on prevalence estimates.

High levels of uncertainty as well as high between-study heterogeneity was observed within the present meta-analyses. This was partly to be expected due to various factors such as variation of prevalence due to time and geographical area [51]. Another explanation could be the observed differences of included studies with regard to the classification and case ascertainment criteria applied to identify PsA cases. In recent decades, several classification and case ascertainment methods as well as coding opportunities for PsA have been developed and implemented into practice [52–59]. One disadvantage is that the different criteria address different requirements as well as domains leading to high heterogeneity within the included studies. For example, over 10 different approaches were reported to identify PsA cases within the 30 included studies of this review. Some studies, for example, have used the CASPAR classification criteria to identify cases. Typically, classification criteria are developed to be used to identify a homogeneous group of patients for clinical study in an already-diagnosed population. Therefore, if employed correctly, it does mean that any prevalence estimate based on classification criteria might underestimate the true prevalence in that population. However, prevalence estimates from the studies that apply CASPAR as part of their case ascertainment are in line with estimates from other studies, suggesting that this is not a major concern in the current review. Unfortunately, the available data within this review were not sufficient to investigate this in a statistically meaningful sub-group analysis [51]. To enhance the comparability between the included studies, it would have been desirable to calculate the prevalence of PsA using reported age-standardized estimates. However, as not all studies consistently reported age categories, age-standardized estimates, or provided information on reference populations for standardization, it was not feasible to carry out the analysis.

Interestingly, among the included health administrative studies, which investigated the prevalence of PsA over time, an increase in prevalence has been observed over the last decades. For example, Eder et al. [37] showed a doubling of PsA prevalence cases from 0.071% in 2006 to 0.153% in 2015 in Israel. Scott et al. [2] reported from 2004 to 2020 a similar increase in the PsA prevalence in England. Rather than a true increase in PsA prevalence per se, it is possible that increased awareness and knowledge among healthcare professionals may lead to a higher proportion of actual PsA cases being diagnosed. Additionally, more accurate coding practices could be one reason for the observed increase in prevalence. Another possible reason for the increasing number of cases may be that there is an increased interest in screening psoriasis patients for PsA in dermatology clinics to detect them early enough and to prevent worse long-term outcomes.

Conclusion

To our knowledge this is the first systematic approach to assess the global prevalence of PsA with particular attention to the methodological differences of the included studies. Although some differences exist between population-based epidemiological studies and those utilizing health administrative data, the pooled prevalence of PsA, globally, is 112 per 100 000 adults, with clear geographical differences: estimates for Europe, Asia, North America and South America are 188, 48, 133 and 17 per 100 000, respectively. Robust estimates of prevalence are crucial for healthcare planning and resource allocation. Where local data exists, and prevalence estimates diverge from what might be expected, investigators may wish to determine why this is the case. Is there an unmet burden of PsA, hitherto unidentified? Or can a local excess prevalence help inform aetiology? However, there are also areas of the world with no good quality prevalence estimates and there may be merit of additional studies in these areas if it is believed, locally, that the global or continent estimates may not be applicable. This is for local health providers and/or government to determine.

Supplementary material

Supplementary material is available at Rheumatology online.

Data availability

The data underlying this article are available in the article and in the supplementary files.

Author contributions

For detailed information, please see the Contributor Roles Taxonomy (CRediT) Statement [60] in Supplementary Table S3, available at Rheumatology online.

Funding

S.L. is part-funded by the University of Aberdeen through the Elphinstone Scholarship. No specific funding was received from any bodies in the public, commercial or not-for-profit sectors to carry out the work described in this article.

Disclosure statement: The authors have declared no conflicts of interest.