The risk of osteopenia/osteoporosis and psoriatic disease: A systematic review

Abstract Background Psoriasis (Ps) is a multisystem inflammatory disease associated with several comorbidities; however, its effect on bone health remains uncertain. This systematic review aimed to evaluate the risks of osteopenia (OPe) and osteoporosis (OP) in psoriasis. Methods A systematic search was performed for published studies evaluating cutaneous Ps and psoriatic arthritis (PsA) compared with healthy control groups utilizing a validated bone mineral density (BMD) assessment score. Meta‐analysis was performed using a random‐effects model; pooled estimates and their confidence intervals (CIs) were calculated. For analysis, Ps and PsA groups were combined due to the small number of studies. Results Twenty‐one studies were included for final analysis; three Ps only, 15 PsA and three both. There was a significant difference between psoriatic disease (combination Ps and PsA group) compared with controls relating to an association with OP/OPe, with an overall odds ratio (OR) of 1.71 (95% CI 1.07–2.74: p‐value = 0.026). The Ps group had significantly lower BMD than the control group at both the lumbar spine and femoral neck (mean difference −0.04; 95% CI −0.090 to 0.002 and −0.03; 95% CI −0.059 to 0.003 respectively). Conclusion Putative risks of OPe and OP in both Ps and PsA are supported but not confirmed. Significant heterogeneity of reported data limits definitive conclusions in this meta‐analysis. This review contributes to the further understanding of Ps as a multisystem disease and future management of potential comorbidities, but highlights key gaps in the literature. Further studies addressing standardised OP reporting, specific disease group characteristics comparing Ps with PsA, patient characteristics and medication use, are required in order to make more certain conclusions with greater clinical impact.

density (BMD) assessment by dual-energy X-ray absorptiometry (DXA), with a T score ≤−2.5 and between −1.0 and −2.5 standard deviations (SD), indicating OP and osteopenia (OPe) respectively. 3,4 Chronic inflammation is a vital component in OP pathogenesis. OP and Ps have many overlapping pathways, including raised systemic inflammatory cytokines, namely interleukin-6, interferon-γ and tumour necrosis factor-α. 1,2 Further, psoriatic arthritis (PsA) shares a similar T helper type 1 and 17 driven inflammatory mechanism with rheumatoid arthritis. Rheumatoid arthritis is known to be associated with OP, suggesting that Ps may also potentially be associated with OP. 5 In addition to sharing molecular similarities, there are clear potential risk factors for OP in patients with psoriatic disease. Physical inactivity, compounded by arthralgia or joint dysfunction, and lifestyle factors including smoking or alcohol may contribute. 1 Systemic corticosteroid treatment is a known risk factor for OP. Topical corticosteroids (TCS) are suggested to be an independent risk factor for OP, particularly the use of potent or very potent TCS in high cumulative doses. 6 Clinical guidelines for Ps recommend the long-term use of TCS, potentially involving application over large surface areas for prolonged periods. Other Ps treatments may also affect bone remodelling, including nonsteroidal anti-inflammatory drugs (NSAIDs), diseasemodifying anti-rheumatic drugs (DMARDs) including methotrexate (MTX), azathioprine (AZA) and ciclosporin (CsA), or biologic agents.
Current guidelines on managing comorbidities in Ps do not address bone health. Published studies evaluating the association between Ps and OP have produced conflicting results. This systematic review aimed to evaluate the risks of OPe/OP in Ps and PsA patients to better guide management.

| Search strategy and selection criteria
This systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement. 7 The electronic databases PubMed and Embase were screened from their inception through to 31 May 2021, using the keywords "psoriasis" OR "psoriatic" AND "osteoporosis" OR "osteopenia" OR "bone" OR "fracture", to identify studies evaluating associations between Ps and OP. The search was restricted to human subjects, publications in English and controlled trials. A detailed search strategy of electronic databases can be found in Tables S1 and S2. References of included studies were reviewed, as were systematic reviews on similar topics.
Studies that satisfied the following criteria were eligible for inclusion: case-control, cohort, or crosssectional studies with original data and more than 10 patients; study population with Ps or PsA; control groups without psoriatic disease; and reported outcomes of effect estimates of OPe/OP or BMD T-score. Studies not reporting the primary outcome were excluded.

| Data extraction and outcomes
Search results were compiled into a single Endnote file, with duplicates removed. Two investigators screened titles and abstracts (AS,AKU) independently, with What is already known about this topic?
� Psoriasis (Ps) and osteoporosis (OP) are both chronic inflammatory conditions with complex pathogenesis. � Several patients with psoriasis and psoriatic arthritis have risk factors for osteoporosis; long term topical corticosteroid use, treatments impacting bone remodelling including non-steroidal anti-inflammatory drugs, and lifestyle factors. � Current guidelines on managing comorbidities of psoriasis do not address bone health and previous studies evaluating the association between Ps and OP have produced conflicting results.

What does this study add?
� This systematic review is the first evaluating bone mineral density (BMD) and psoriatic disease exclusively in studies with healthy control comparison groups and using outcomes of validated bone quality assessment measures. We demonstrate the possible risks of osteopenia and osteoporosis in psoriatic disease. � Given inconsistencies, bias, and heterogeneity of reporting in the existing published literature, more extensive longitudinal studies with standardized BMD reporting measures and subgroup analyses, are needed to better characterize the risk of low BMD in psoriasis and medication risk factors.
discrepancies resolved by a third-party reviewer (CBB), to identify articles for full qualitative review. If the same dataset appeared in different reports, the earliest study was included. The primary outcome of this systematic review was to define the association between psoriatic disease and OPe/OP. The main outcomes of interest were the incidence and prevalence data of OPe and OP and absolute values of BMD as an ancillary marker. Secondary outcomes included data on the effect of concurrent psoriatic treatments.

| Assessment of bias
Bias was assessed for each study included in the systematic review by two independent reviewers (AS, AKU). A modified Newcastle Ottawa Scale (NOS) was applied to each study to assess for bias in the selection and comparability of study groups, and ascertainment of outcomes of interest 8 (Tables S3-S5). A score of seven or higher was considered high quality.

| Data synthesis and statistical analysis
The prevalence/incidence data, and BMD for the psoriatic disease cohort and control group were presented. The estimated odds ratios (OR) and 95% confidence intervals (CI) were calculated for the primary binary outcomes (risk estimates of OPe/OP). For continuous outcomes (absolute value of BMD) the mean differences (MD) were calculated. Due to the heterogeneity in T-score reporting, quantitative analysis was not possible. Given the small number of studies evaluating Ps alone in comparison to a combination of Ps and PsA, separate subgroup analysis was not performed due to the variability of small groups.
Pooled prevalence estimates (expressed as OR) and BMD across studies were generated using a random-effects model, and heterogeneity between studies was evaluated using a χ 2 test on Cochran's Q quantified with the I 2 statistic. A p-value of <0.05 was used as a cut-off to indicate significant heterogeneity, whereas, on I 2 statistic, limits of 30%, 50% and 75% indicated low, medium and high levels of heterogeneity, respectively. 9 A random-effects framework was utilized given the variation in population and characteristics across the studies. Meta-analysis was performed with R version 4.1.0 and package Meta generating a forest plot showing fixed and random-effects models, using Mantel-Haenszel for OR and inverse variance for BMD. Inter-study heterogeneity was assessed using DerSimonian and Laird variance estimator and evaluated using a χ 2 test on Cochran's Q test and quantified with the I 2 statistic.

| Effect measures
Publication bias was assessed with the aid of funnel plots. Influence analysis was performed to evaluate the impact of each study on the overall effect by omitting them individually. The publication bias was quantified using Eggers's test; this is considered statistically significant when p < 0.05.

| RESULTS
From an initial selection of 165 references, 79 were excluded after reading the title or abstract as duplicates or irrelevant, and 37 after reading the article; 21 articles were retained. Screening references did not yield any additional studies. The selection process is further detailed in Figure 1.

| Study population and characteristics
Overall, 21 studies (Table 1) compared Ps patients with a control group, specifically Ps (n = 3), PsA (n = 15) and both, as separate comparison groups (n = 3). Hereon 'psoriatic disease' refers to Ps and PsA groups combined. Two further studies included additional comparison groups: pre-and post-menopausal females, and men. In all studies, the diagnosis of Ps was made clinically in the Dermatology outpatient setting, and the diagnosis of PsA predominantly by Classification criteria for Psoriatic Arthritis (CASPAR) diagnostic criteria or the International Classification of Diseases (ICD) for large national databases. Cases were enroled from various settings (specialist Dermatology/Rheumatology outpatient [n = 11], inpatient Dermatology ward [n = 1], large national databases [n = 6]), with sample sizes reflecting this and a total of 267 415 overall cases included. Control groups comprised of healthy individuals and were primarily age and gender-matched (n = 15). The modified NOS appraisal scale deemed study quality as high 7 (n = 5); intermediate 4-6 (n = 14) and low 3 points (n = 2); noting that cross-sectional studies may score a maximum of 6 points (n = 13).
The mean participant age lay between 35 and 66 years; there was a female-predominant sex distribution in most of the studies (n = 13) with two studies excluding males entirely. The mean BMI was between 24.9 and 30.3. Menopausal status was recorded (n = 10), excluded (n = 2) and otherwise not stated (n = 11). SCHAUER ET AL.

| Pooled effects on the primary outcome
The association between OPe/OP and Ps using effect measures was reported in 14 studies, and these were pooled for meta-analysis ( Figure 2). The psoriatic disease group differed significantly from the control group, with an overall OR of 1.71 (95% CI 1.07-2.74: pvalue = 0.026). Only three studies did not show an increased risk of low BMD.
The gold standard for reporting BMD is with Tscores, and values were reported in 10 different studies, with specific measurements at LS (n = 9), TH (n = 3), FN (n = 8) (Table 3). Due to missing variables required for meta-analysis, this was not performed.

| Ps treatment and OP/OPe
Medication use was recorded in 15 studies; specifically included TCS (n = 1), systemic corticosteroid (n = 8), NSAID (n = 3), DMARD (n = 10), Biologic (n = 7); antiosteoporotic (n = 2) and medications excluded (n = 4). Regarding the secondary outcome measure evaluating the possible effect of psoriatic medications on OP, nine studies performed subgroup analyses adjusting for medications. They were included for qualitative analysis, with three studies demonstrating a possible F I G U R E 1 Prisma protocol for systematic review negative impact of systemic treatment on bone mineralization (Table 4).

| Reporting biases and certainty of the evidence
The risk of bias was moderate to high in most studies (Table 1). Inclusion and exclusion criteria meant that baseline characteristics of the psoriatic disease group were highly variable to begin, and the selection of nonpsoriatic controls was often poorly defined. Overall, it is likely that there were too few studies in the metanalyses to draw definitive conclusions, but there were significant differences between Ps and control groups at face value. There was a very high level of heterogeneity with studies included in OR calculation (I 2 99.7%; 95% CI 99.7-99.8; Q = 4876.06). Further influence analysis resulted in a non-significant outcome; heterogeneity remained high even with excluding studies from analysis, and a formal test indicated publication bias.
In the meta-analysis of areal BMD, omitting individual studies altered whether the difference in observed BMD was significant or not for LS and TH parameters. Heterogeneity was very high, particularly for LS, indicating substantial intra-study difference and thereby questioning the validity of pooling data to obtain an overall effect. Further, there were not enough data to draw a definitive conclusion regarding publication bias, but visual inspection of relevant charts indicated that this was occurring.

| DISCUSSION
The association between Ps and osteoporotic disease is contested. Ps has been purported to promote bone demineralization due to various mechanisms, including chronic systemic inflammation, the use of anti-psoriatic medications and prolonged immobilization due to pain and joint dysfunction. This systematic review demonstrated a higher risk of OP/OPe with psoriatic disease by the odds ratio outcome and total BMD at the LS and FN, but not the TH. Furthermore, discrepancies were noted on qualitative analysis of T-scores across different anatomical sites, and these variable findings are in keeping with other related systematic reviews.
Chen et al., 30 found no significant association between psoriatic disease and the absolute value of BMD. Further, patients with Ps did not have a higher risk of OP than the controls; but, interestingly, did have an increased OR of sustaining fractures unrelated to lower BMD/OP. Similar findings were reported by Sepehri et al., 31 demonstrating an increased risk of fracture but not OP/OPe in psoriatic disease. Chandran et al., 32 showed a possible association between various factors and low BMD in PsA, but with inconsistent findings. This reflects the complex relationship between Ps, PsA and low BMD and the multifactorial nature of potential risk factors related to Ps and its treatment.
There are many established risk factors for low BMD, and several included studies accounted for these confounders performing sub-analyses. Females were found to have lower BMD 28 and higher fracture risk 23 ; conversely, another observed the association between Ps/OP observed only among males. 1 Post-menopause leads to lower BMD, 24 increased fragility fractures 27 and years since menopause determining BMD. 25,26 Vitamin D deficiency was associated with Ps 21 and increased risk of hip fractures. 26 Other studies have extrapolated that vitamin D deficiency plays a role in the pathogenesis of OP in psoriatic disease rather than psoriatic disease activity. 33 Ps susceptibility genes are associated with genetic loci for metabolic syndrome, type 2 diabetes mellitus F I G U R E 2 Pooled association between osteoporosis/osteopenia and psoriasis by measure of odds ratio SCHAUER ET AL. dermatitis 19 ). There was a higher risk of OP in Ps with obesity, but only in males 1 and, interestingly, in one study, higher BMI was found to be protective with a higher T-score. 21 Heightened systemic inflammation levels in Ps represent a proposed mechanism for the association with low BMD, with higher C-reactive protein (CRP) & erythrocyte sediment rate (ESR) in Ps. 28 Although in the context of other evidence, there is a question of uncaptured cumulation of the effect of repeated attacks on bone demineralization. 33 Other potential psoriatic disease-specific risk factors for low BMD include higher PASI associated with lower T-scores, 21 the presence of erosions 11 and an increased number of damaged joints. 13 In addition, increased disease duration was associated with low BMD, 17,26 noting that one study did not find a significant correlation, albeit these groups were subdivided by an arbitrary less than or more than a 9-year duration of arthritis. 13 The presence of axial disease was excluded in two studies 24,27 but not explored elsewhere, with evidence suggesting more severe bone loss in peripheral PsA than axial PsA. 34 Several Ps treatments have been shown to interfere with physiological bone metabolism, including corticosteroids, 35 MTX, 36 CsA. 37 In Ps there is a dichotomy; AZA has been associated with overall increased fracture risk 38 and glucocorticoids, 39 CsA, 38 and MTX 38,40 do not appear to impact bone health negatively. A large-scale observational study, drawing on UK Biobank genomic data and using Mendelian randomization to evaluate the causal effect, concluded that the effect of PsA on OP was not genetically determined but secondary to treatment with medication (MTX, CsA). 41 These results, however, must be interpreted with caution given methodological concerns, including high disease activity likely contributing to taking these medications in the first place. 42,43 In this review, most studies did not find any increased risk associated with Ps treatment; however, a few noted adverse outcomes associated with systemic corticosteroid 29 and DMARD use. 14,26 As there was no standard approach to grouping medications across papers, and sub-analysis for medications was only performed in a few-studies, meta-analysis was not performed. Biologics, as the newest therapeutic agents, were poorly represented, with at least one other study demonstrating no progression of catabolic and anabolic bone changes in the joints of patients with PsA with IL-17 inhibition. 44 Furthermore, emerging evidence suggests a beneficial effect of biologic treatment, that despite longer disease duration, there was better bone strength than patients receiving MTX or no DMARD, 45 an area which perhaps warrants further attention.

Study
Significant morbidity is associated with low BMD, with the degree of disability correlated to lower BMD. 13 A large UK population-based cohort study demonstrated that patients with Ps have an increased fracture incidence than the general population by up to 26%. 5 This is in keeping with our findings, with an increase in fragility fracture associated with psoriatic disease 16,26,27 bar one study. 20 Other outcome measures such as increased hospitalization, loss of quality of life, economic burden and mortality were not within the scope of included studies.
Summarizing the results of studies within this systematic review demonstrates the considerable breadth of this subject. It perhaps explains the observed high level of heterogeneity, with individual studies focussing on different aspects of the complex interplay between Ps, PsA, OP and OPe. While most studies recorded additional population characteristics and comorbidities, they failed to adjust for these variables in subsequent analyses. Secondly, the methodological quality was highly heterogeneous, with potential selection bias introduced to both cases and controls and deficits in the length of follow-up data. Notably, the largest study included 16 (which if omitted renders the results not significant) encompassed event-level records as opposed to patient-level records. This means that the same patient could potentially be included more than once if they visited ED on multiple occasions. Further, several studies utilized national claim databases and diagnostic codes (i.e., ICD), introducing potential bias by coding errors and diagnosis misclassification. Finally, retrieved articles included brief reports, which scored poorly on the modified NOS quality assessment tool.
To our knowledge, this is the first systematic review evaluating BMD and psoriatic disease exclusively in studies with healthy control comparison groups and using outcomes of validated bone quality assessment measures. While these are strengths, this review is hampered by the small number of evaluable studies and demonstrates a reasonably high level of reporting bias. Given the small number of included studies, Ps and PsA groups were combined for analysis, due to variability of smaller groups and subsequent inability to perform useful subgroup analysis. As a result, we are unable to make a distinction between cutaneous Ps and PsA as a potential risk factor for low BMD, and there is a paucity of data related to cutaneous Ps alone. It is plausible that PsA alone may have a greater effect on BMD than Ps, given associated comorbidities and potential more frequent treatment with medications such as systemic corticosteroids. However this needs to be further evaluated in future specific studies.
The different modalities of expressing BMD (i.e., areal BMD as g/cm 2 , T-score or Z-score measured at multiple skeletal sites) diluted collected data further, making it difficult to draw meaningful conclusions. The gold standard of reporting BMD is using T-scores, however due to the absence of studies reporting required SCHAUER ET AL. variables, including means, standard deviations, medians, IQR, meta-analysis was not performed. Chen et al. 30 performed analyses using absolute BMD measurement only, and Chandran et al. 32 also highlighted variable means of BMD reporting with subsequent limited analyses. Though it is the gold-standard for OP diagnosis according to the World Health Organisation (WHO) classification, DXA has inherent flaws, with high specificity but low sensitivity, and may not encapsulate the actual risk of fracture in psoriatic disease. 30,46 Other outcome measures used to evaluate bone health and quality included FRAX, an assessment algorithm used in the prediction of hip fracture and other osteoporotic fractures, and quantitative computed tomography (QCT), a three-dimensional quantification of volumetric BMD (g/cm 3 ) and bone microstructure. 26,29 Notwithstanding, DXA remains the most widely validated tool for assessing BMD.
Finally, clearly accounting for medication use to allow sub-analysis, or preferably designing a separate study evaluating specific individual treatment options would yield more results. Admittedly, this research topic is fraught with difficulty given polypharmacy and variation in personalized treatment regimens balanced against optimal disease control. Furthermore, disease severity and treatment are often closely interrelated, so it may be challenging to disentangle the role of chronic inflammation and systemic treatment on low BMD in severe Ps.
The limitations discussed highlight areas which require addressing in future studies to improve the quality of the literature. Further studies evaluating cutaneous psoriasis alone and the effect on OP are required, with studies being adequately powered to compare cutaneous psoriasis versus psoriatic arthritis, and OP. Importantly, standardized reporting methods of OP using T-scores with additional parameters to allow use in meta-analysis may yield more meaningful clinical conclusions. From the varied countries they were performed, the studies included had a diverse patient population. Ethnic background was sometimes specifically paired for controls 25 but no sub-analysis performed. Dheda et al. 13 suggested Indians may be more at risk of PsA and low BMD than other Oriental ethnic groups, however further targeted studies may help identify if certain ethnic groups are at higher risk.
To reduce the burden of osteoporotic fracture, WHO advocates using clinical risk factors in conjunction with BMD to deliver timely interventions to individuals at high risk. 46 While psoriatic disease as an independent risk factor of low BMD remains undetermined, using knowledge of established risk factors, patients with Ps generally have multiple and increased risk factors, placing them at higher risk of OP. 5 This review highlights the need to be cognisant of the risk of OP in association with Ps and consider early evaluation and prophylactic treatment.

| CONCLUSION
This systematic review evaluates BMD in psoriatic disease and treatment by summarising the published literature using healthy control comparison groups, synthesising existing evidence and highlighting deficits in current knowledge. It shows inconsistencies and a likelihood of bias in the published literature about low BMD prevalence and medication risk factors in Ps and PsA. This study highlights that Ps patients should be monitored for complex comorbidity, including OPe/OP. However, more extensive, longitudinal studies or Ps registries, including robust assessment of comorbidities and pharmacological treatment, together with more standardised BMD reporting measures, are needed to more accurately characterise the risk of low BMD in cutaneous Ps.