Clinical, laboratory, and genetic markers for the development or presence of psoriatic arthritis in psoriasis patients: a systematic review

Twenty to thirty percent of psoriasis (Pso) patients will develop psoriatic arthritis (PsA). Detection of Pso patients that are (at risk for) developing PsA is essential to prevent structural damage. We conducted a systematic search of five bibliographic databases, up to May 2020. We searched for studies assessing markers (clinical, laboratory, genetic) associated with the development or presence of PsA in Pso patients. Study selection and quality assessment of the included studies was performed, followed by a qualitative best evidence synthesis to determine the level of evidence for a marker and its association with concomitant/developing PsA in Pso. Overall, 259 possible markers were identified in 119 studies that met the inclusion criteria. Laboratory markers related to inflammation and bone metabolism reached a strong level of evidence for the association (not prediction) of PsA in Pso. Only CXCL10 showed strong evidence for a positive predictive value for PsA in Pso. The importance of timely detecting PsA in a Pso population, and finding more (bio)markers contributing to early detection, remains high. Supplementary Information The online version contains supplementary material available at 10.1186/s13075-021-02545-4.


Introduction
Psoriatic arthritis (PsA) is an immune-mediated inflammatory disease affecting joints and entheses and is strongly associated with psoriasis (Pso). Twenty to thirty percent of Pso patients will develop PsA, with an average lag time between Pso and PsA of 10 years [1,2]. This lag time creates a unique opportunity to identify patients with an increased risk for (developing) PsA. The (timely) recognition of concomitant PsA, or ideally early prediction, is important, because untreated PsA can lead to irreversible joint damage [3,4]. Treatment of arthritis leads to an improvement of both function and quality of life [5]. However, patients with Pso are mostly seen by physicians (e.g., dermatologists) who are not trained in recognizing early signs of arthritis. Identifying markers for PsA in patients with Pso can optimize screening to detect the onset of PsA as early as possible.
Current screening strategies mostly use questionnaires based on clinical characteristics to detect PsA [6,7]. Both characteristics of Pso as well as environmental factors may be relevant variables for PsA screening [8][9][10]. Next to clinical characteristics, extensive research has been done on genetic markers, in both HLA (human leukocyte antigen) and non-HLA regions [10][11][12]. Likewise, there are laboratory markers involved in inflammation pathways who might be able to help detect PsA in Pso patients [13,14]. However, most research focuses on the differentiation between Pso and/or PsA on one side and healthy controls on the other side. To our knowledge, no comprehensive overview has been made to summarize the evidence for these clinical, genetic and laboratory markers.
Therefore, we conducted a systematic review to identify possible markers for the onset of PsA in a Pso population, with the purpose of providing a comprehensive summary of the available markers for PsA in Pso.

Protocol
The protocol was designed according to the Preferred Reporting Items for Systematic review and Meta-Analysis [15] and registered in Prospero (CRD42018093982).

Search strategy
Five bibliographic databases (PubMed, EMBASE, Web of Science, Medline and Cochrane) were searched for studies from January 1, 1990, up to April 29, 2020. Search terms compromised keywords involving study population, study design, and etiology (supplementary table 1). In addition, reference lists of included articles were used for cross-reference checking.

Study selection
Studies were screened for eligibility based on title and abstract by two independent reviewers (MM, JV for laboratory and genetic studies; MM and TH for clinical studies). Potentially relevant papers were assessed in full text (MM, TH). Any disagreement was resolved by consensus or by discussion with a third reviewer (JR, MW, JV). Studies were excluded based on the following criteria: (1) < 10 patients per group (Pso and PsA, respectively), (2) age of patients < 18 years, (3) no statistical comparison between Pso and PsA, and (4) languages other than English, German, or Dutch. We primarily focused on studies with a longitudinal design, meaning that the marker was present before the presentation of PsA. A very low number of longitudinal studies was available for laboratory studies (n = 2), and none for genetic studies. To not miss potential relevant markers in these two categories, we also included genetic and laboratory studies with a cross-sectional design (i.e., marker was present at the same time as PsA) as a "second best" option. While these might not be useful to identify predictors for development of PsA, they could provide information about possible markers for concomitant PsA.

Data extraction
Data extracted included study design, patient characteristics, markers, and outcome. Extraction was performed by two reviewers, with 10% overlap to check extraction quality (MM, TH).

Assessment of risk bias
Risk bias was assessed using the Newcastle Ottawa Scale for case-control and cohort studies [16]. This tool comprises three domains: selection, comparability, and outcome/exposure. A study was considered of "good" quality when it had a minimum of 3 stars in the selection domain, 1 star in the comparability domain, and 2 stars in the outcome/exposure domain. "Fair" quality was given when a study had a minimum of 2 stars in the selection, 1 star in the compatibility, and 2 stars in the outcome/exposure domain [17]. If a study failed to meet these standards, it was considered to be of "poor" quality. Risk of bias assessment was performed by two reviewers (MM, TH) independently. Any disagreement was resolved by consensus or by discussion with a third reviewer (JR, MW, JV).

Best evidence synthesis
For the best evidence synthesis (BES), we included markers that either showed a significant difference between Pso and PsA in at least one study or markers that showed no significant results in at least two studies (i.e., we excluded markers who were only investigated once and showed no association). Markers were grouped into overarching categories (see Tables 1, 2 and 3). In addition, for markers presented as a categorical variable, we used the data of the most extreme level. For example, in the study from Love et al., body mass index (BMI) was categorized into four levels: < 25 (normal), 25-30, 30-35, > 35 kg/m 2 [33]. For the best evidence synthesis, we looked at the highest level (i.e., BMI > 35 kg/m 2 ) compared to reference level (i.e., BMI < 25 kg/m 2 ). We then assessed the consistency of the results within and across studies. If within a study, a marker was represented in multiple nonhierarchical conceptually similar constructs, we considered the result consistent if ≥ 75% of the constructs pointed in the same direction. Otherwise, we considered the result for that marker "mixed." For example, one study looked at fracture, any trauma, and trauma leading to medical care [21]. Because two of these were not predictive of PsA, and one was, we considered this study to have "mixed results" with respect to the marker "trauma." If across multiple studies, < 75% of studies were in agreement with each other, we considered this "conflicting evidence." If ≥ 75% of studies were in agreement, we applied the evidence grading according to Sackett [17]. Because only a small minority of the included studies were of "good" quality, we adapted the Sackett best evidence synthesis as follows: strong evidence in case of two or  [18,19] Strong evidence of no association Diarrhea 2x no association [18,20] 1x no association [21] Strong evidence of no association Infection requiring antibiotics 1x positive association [20] 1x no association [18] Conflicting evidence Uveitis 1x positive association [18] Moderate evidence of positive association Disease characteristics (general) (worsening) Fatigue 1x positive association [22] Moderate evidence of positive association Worsening function 1x positive association [22] Moderate evidence of positive association Younger age at Pso onset 2x positive association [23,24] 1x no association [25] Conflicting evidence

Medication
Corticosteroids use 1x positive association [19] Moderate evidence of positive association Influenza vaccination 1x no association [20] 1x no association [21] Moderate evidence of more studies with good or fair quality, moderate evidence in case of two or more studies with low quality or one study of good or fair quality, and limited evidence in case of one study with low quality. In case of two or more good/fair quality studies, the results of the poor quality studies were not taken into account for the BES. The heterogeneity of the markers and statistics precluded a quantitative meta-analysis.

Study characteristics
The characteristics of the included studies are listed in supplementary table 2. All clinical studies had a Methotrexate use 2x no association [18,19] Strong evidence of no association Retinoid use 1x positive association [18] Moderate evidence of positive association Rubella vaccination 1x no association [20] 1x positive association [21] Conflicting evidence Tetanus vaccination 1x no association [20] 1x no association [21] Moderate evidence of no association

Miscellaneous
Anti-ADAMTS-L5 IgG antibodies 1x positive association [95] Limited evidence of positive association Anti-LL37 antibodies 1x positive association [95] 1x mixed results [82] Conflicting evidence longitudinal design. Two laboratory studies had a longitudinal design and 67 had a cross-sectional design. All of the genetic studies had a cross-sectional design. Based on the criteria described in the best evidence synthesis, 259 markers were selected for further description (clinical 51, laboratory 137, genetic 71), of which 104 were described in multiple studies (clinical 32, laboratory 36, genetic 36). All markers are shown in supplementary tables 3, 4, 5.

Quality assessment
Of the included studies, 19 studies were qualified as good quality, 11 studies were qualified as fair quality, and 89 studies were qualified as poor quality. Quality assessment of the included studies is shown in supplementary tables 6 and 7.

Best evidence synthesis
Qualitative best evidence synthesis is depicted separately for clinical, laboratory, and genetic studies in Tables 1, 2  and 3. With respect to predictive markers for PsA in Pso, we report the markers for which there was at least a moderate level of evidence, or which were investigated in more than one study. With respect to markers associated with the presence of PsA in Pso, we report only the markers which were investigated in more than one study. An overview of the most promising findings is also shown in Fig. 2.
Only one of the 14 markers which were investigated more than once showed moderate evidence of a negative association with the presence of PsA in Pso: a lower ratio of OPG to receptor activator of nuclear factor kappa-B ligand (RANKL) was associated with the presence of PsA in Pso [50,56].

Genetic markers Strong level of evidence
There were no genetic markers which reached a strong level of evidence for a positive, negative, or no association with the presence of PsA.

Moderate level of evidence
Moderate evidence was available for 30 of 71 investigated genetic markers. Twenty-two of those 31 have been investigated in more than one study.

Discussion
In this review, we summarized the available evidence for possible markers for the onset or presence of PsA in a Pso patient population in a systematic way. Thereby, we provide an update and addition to a recent narrative review regarding this subject by Scher et al. [10]. When looking at clinical markers, we found only strong evidence for markers which were not associated with the development of PsA. Regarding laboratory markers, there was strong evidence for the predictive value of (a change in) CXCL10 serum titers [27,60]. There was also strong evidence for the association with (but not prediction of) PsA of several markers related to bone metabolism [49][50][51][52][53][54][55] and inflammation [27, 43, 44, 47, 49, 53, 54, 56, 58, 61-64, 66, 70-75, 130]. With respect to genetic markers, we found no markers which reached a strong level of evidence for the association with PsA.
In line with previous beliefs on possible clinical risk factors [10,140], we found moderate evidence for a positive association of gluteal fold lesions [25] and nail pitting for the onset of PsA [18]. However, for nail involvement in general (e.g., distal onycholysis, oil drop phenomenon and crumbling), there was strong evidence of no association [18,19,25,27]. Therefore, this relationship seemed to be restricted to this specific nail feature.
Notably, we found conflicting evidence for the predictive value of obesity [18,20,22,27,29,33,34] and psoriasis severity [18,20,22,25,27,28] for the development of PsA in Pso patients. These studies may also be prone to bias because patients with severe Pso differ from patients with mild Pso in several aspects. For instance, when looking at Pso severity in particular, one can argue that more severe skin involvement is treated more intensively, thereby possibly suppressing concomitant arthritis. These kinds of bias may be the reason why these frequently reported markers reach conflicting evidence when all the studies are taking into account in a systematic way.
When looking at BMI at one unspecified timepoint, this marker shows conflicting evidence for a relationship with the development of PsA. In three out of five high/fair quality studies, there was no association [18,22,27], while two out of five showed a positive association [29,33]. Even when taking into account that the before mentioned three studies are performed in a partially overlapping cohort, this marker does not reach the 75% agreement level we consider necessary for a conclusive result. Therefore, BMI at any unspecified timepoint may not be specific enough for prediction of PsA. Interestingly, more specified markers of weight and body composure (e.g., recent weight gain, BMI at younger age or abdominal adiposity) showed a positive association with the development of PsA in Pso but were only investigated in one study of poor quality [34]. Increasing the evidence in a more detailed way may be more valid and relevant.
The association of trauma and psoriatic arthritis was theorized to be due to a deep Koebner phenomenon [140]. This phenomenon is comparable to the wellknown Koebner phenomenon in the skin, where trauma can cause the appearance of new skin lesions. The theory on the deep Koebner phenomenon is based on a study of Thorarensen et al., who used diagnostic codes to establish two comparable cohorts (Pso with and without PsA) [35]. However, when forming cohorts in this way, there is a higher risk of misclassification in either cohort. This study is in disagreement with two other papers with higher diagnostic certainty [19,20]. Therefore, we concluded that there is currently strong evidence that physical trauma is not associated with a higher rate of PsA in Pso patients.
The relationship between smoking and PsA development has been described previously as the "smoking paradox" [31]. This entails the fact that smoking appears to be a risk factor for PsA when looking at the general population, but this association disappears when only looking at psoriasis patients. This paradox may be explained by collider bias: bias resulting from correcting for a variable which is a common effect of the exposure and outcome [10]. In our review, we found conflicting evidence for an effect of (current) smoking [18,20,28,29,31,32]. However, due to this collider bias, it is hard to determine if smoking leads to additional risk for the development of PsA in a Pso population, unrelated to its effect on the development of Pso. Studies focusing on a change in smoking status after the development of Pso may shed a light on this enigma, as suggested by Nguyen [31].
With regard to laboratory markers, only CXCL10 was studied longitudinally. This cytokine was described in two good/fair quality studies; both found an association between CXCL10 and PsA. Pso patients who developed PsA had a higher CXCL10 serum level at baseline [27]. It was also shown that during the evolution to arthritis the serum level of CXCL10 diminished: a larger negative change was associated with a higher risk of PsA [60]. The reason why CXCL10 levels decreased towards the development of PsA is still unknown. One hypothesis could be that the psoriasis patient group with a high level of CXCL10 is more prone to develop arthritis due to its chemoattractant properties on CXCR3 + CD4 + and CD8 + T cells [141]. In the evolution towards clinical manifest PsA, locally produced CXCL10 might get depleted by these infiltrating and locally expanding inflammatory cells, subsequently lowering circulating CXCL10 levels over time. However, since these two studies were published by the same research group, results may be based on (partially) overlapping patient groups. Therefore, the predicting value of CXCL10 should be interpreted cautiously.
With regard to cross-sectional studies, and markers that may indicate the presence of PsA in Pso patients, we found strong evidence for a positive association with PsA in Pso for markers of inflammation and bone. CRP is a well-known, widely used inflammatory marker. We found strong evidence that the CRP level in PsA patients was higher than in patients with Pso only [27, 43, 44, 47, 49, 53, 54, 56, 64, 66, 70-75, 124, 130]. We argue that the co-appearance of joint inflammation is responsible for this observation. However, we found no articles which studied the level of CRP before the start of PsA in Pso. Therefore, it is unknown whether it can be used as a predictive marker. Also, a clear CRP cutoff value for the presence of PsA (and therefore, specificity and sensitivity) is lacking.
Other markers for which strong evidence of a positive association with the development of PsA in Pso exist were IL-6, MMP3, and OPG. IL-6 is widely regarded as a marker for systematic inflammation and an important contributor to the production of CRP by the liver. MMP3 and OPG are associated with bone metabolism; one of the hallmark signs of PsA is new bone formation [142]. Also, untreated arthritis can lead to irreversible erosions [4]. Therefore, it is not surprising that MMP and OPG showed an association with the presence of PsA in our review. In line with CRP, the predictive value of these markers is unknown, because longitudinal studies are not performed yet.
Laboratory markers for cardiovascular disease are studied extensively in psoriatic disease [42,46,62,64,71,72,76,77,130]. From these findings, we can conclude with strong evidence that these levels do not differ between psoriasis patients with and without arthritis. This is in contrast to a recent review which showed that the prevalence of cardiovascular comorbidities is higher in patients with PsA when compared to Pso [143]. This suggests that there are additional factors (e.g., systemic inflammation) that play a role in cardiovascular morbidity in PsA.
With respect to genetic markers, we focus here on the most important HLA-markers for Pso and PsA, and the IL-12 -IL-23 -IL-17 axis. The most important genetic marker for psoriasis is HLA-C*06, also known as PSOR1 [144]. This marker is responsible for up to 50% of Pso heritability in the healthy population. It is associated with type-I (early onset) psoriasis, as well as a guttate phenotype [145]. Interestingly, our review shows that, when looking within the population of Pso patients, patients with the HLA-C*06 marker were less likely to also have PsA. Despite multiple studies investigating this marker, high-quality studies are needed to confirm the robustness of the negative relationship between HLA-C*06 and the onset of PsA.
We found a moderate level of evidence for the presence of concomitant PsA in Pso for HLA-B*27, known for its high prevalence (90%) in ankylosing spondylitis (AS) [146]. In other diseases of the spondyloarthritis spectrum, the presence of HLA-B*27 is still higher than in the general population, but less than in AS. Our review showed that the presence of HLA-B*27 was higher in the Pso patients who developed arthritis than in the Pso patients who did not. This could indicate that HLA-B*27 may be able to differentiate between Pso patients who do or do not have PsA, which is also considered a part of the spondyloarthritis spectrum.
When looking at the IL-17/IL-23 axis from a genetic viewpoint, there was moderate evidence that there are no SNPs in the IL23 gene for which the presence differs significantly between PsA and Pso patients [109,111,114,115,147]. We found limited evidence that the presence of rs79877597 in the IL17 gene was more common in PsA versus Pso patients [118]. With regard to the common IL-12/IL-23 pathway, there was moderate evidence regarding several SNPs in the IL12 gene [148]. We found that the presence of one SNP in IL12 (rs2082412) was lower in PsA versus Pso patients, while other SNPs in this gene showed no difference [109,111,114,115]. While the IL-17/IL-23 axis may be important for the development of psoriatic disease in the general population, these results may indicate that it is of limited importance in the development of PsA in Pso.
The strengths of this study include the extensiveness and systematic way of the search with respect to markers for PsA in patient cohorts with Pso, subsequentially providing a comprehensive overview of the available evidence. Also, the intertwining of clinical, laboratory, and genetic markers in a systematic way is unique. By conducting a best evidence synthesis, taking the study quality into account, we made a qualitative overview of the extensive data.
The limitations of this systematic review are mostly due to the limitations of the included studies. Since there were (almost) no prospective/longitudinal studies looking at genetic and laboratory markers, we could only summarize the level of evidence with regard to the relationship between laboratory and genetic markers with the presence of PsA in patients with Pso (i.e., only one predictive factor could be identified). The level of evidence was limited by a paucity of high or fair quality studies. Mostly, this was because of a lack of appropriate definition of patient and control groups, in addition to not adjusting for possible confounders.

Conclusion
This comprehensive systematic review on clinical, laboratory, and genetic markers for PsA in patients with Pso revealed that a useful set of markers is not established yet. There were no clinical or genetic markers with strong evidence which could predict the development of PsA in Pso cohorts. There was strong evidence that laboratory markers related to bone metabolism and inflammation were associated with the presence of PsA. Promising is CXCL10, which reached a strong level of evidence for predicting development of PsA in a Pso population [27,60]. The importance of timely detecting PsA in a Pso population, and finding more (bio)markers contributing to early detection, remains high.