https://orcid.org/0000-0002-3645-7059
Abstract
Whether metformin and its adenosine 5‘monophosphate-activated protein kinase (AMPK) activation protect from psoriasis risk is unconcluded. We investigated the effect of AMPK, a pharmacological target of metformin, on the risk of psoriasis and its comorbidities and mortality among participants in the UK Biobank (UKB).
To avoid immortal time biases in pharmacoepidemiologic studies, Mendelian randomization was used to infer the AMPK pathway-dependent effects. The cut-off age for distinguishing early-onset/late-onset psoriasis (EOP/LOP) was set at 60 years, based on the incident psoriasis peak in UKB. A genetic instrument comprising 44 single-nucleotide polymorphisms associated with glycated haemoglobin (HbA1c), serving as a proxy for AMPK genetic risk score (negatively associated with AMPK activation), was employed as previously reported in the literature. Log-binomial models were used to estimate the effect size of AMPK regarding relative risk (RR) and 95% CI.
A total of 407 159 participants were analysed, including 9126 EOP and 3324 LOP. The AMPK genetic risk score was associated with a 12.4% increase in the risk of LOP in men (RR = 1.124, 95% CI: 1.022–1.236). This association was not significant for EOP or women. AMPK genetic risk score exhibited an elevated risk of ischemic heart disease (RR = 1.217, 95% CI 1.062–1.395) in male psoriasis patients.
AMPK activation may protect against LOPs and associated ischemic heart disease in men. A sex-specific, comorbidity-targeted intervention for psoriasis is needed.
The beneficial effect of metformin on psoriasis has been discussed but unconcluded.
We added new epidemiological evidence highlighting the potential benefits of metformin in psoriasis through its activation of AMPK. Our findings indicate that AMPK activation is associated with an increased risk of late-onset psoriasis and cardiovascular comorbidities in men.
Further research about the AMPK signaling pathway in psoriasis pathogenesis with different sex/onset-age is suggested. A sex-specific, comorbidity-targeted intervention for psoriasis is needed.
Introduction
Psoriasis, an inflammatory skin disorder with a genetic predisposition and immune-related characteristics, affects ∼125 million people worldwide and imposes a significant disease burden [1–4]. Moreover, there is evidence of an increasing prevalence of psoriasis [5]. Previous studies have indicated that immune imbalance due to specific genetic susceptibility factors contributes greatly to the development of psoriasis [6, 7]. However, the genetic risk of psoriasis is inversely correlated with age at onset of psoriasis. With the older the onset age of psoriasis, the role of genetic predisposition involved in its aetiology is less [8]. This updated evidence helped to explain the solid reported association of HLA-Cw6 with juvenile-onset psoriasis rather than with late-onset patients [9]. Interestingly, most well-established psoriasis susceptibility loci are not strongly associated with psoriasis onset after the age of 60 years [10, 11], which coincides with a second peak of psoriasis occurrence [10, 12]. Emerging aetiological studies have demonstrated the interplay between metabolic disorders and immune dysfunction in the development of psoriasis, particularly in cases with chronic inflammatory comorbidities and those that are initiated in later stages of life [13]. When stratified by age of onset, diabetes, dyslipidaemia and major cardiovascular events were more frequent in subjects with late-onset, implying the metabolic-driven aetiology may underlie the initiation of psoriasis in the geriatric population [14, 15].
Adenosine 5’monophosphate-activated protein kinase (AMPK) is an enzymatic complex known for its regulatory role in energy metabolism and the immune homeostasis network [16]. AMPK has been reported to engage in the regulation of immune-related diseases such as psoriasis, osteoarthritis, inflammatory bowel disease and atherosclerosis [17]. Activation of AMPK has been shown to increase autophagy, thereby eliminating inflammatory factors and subsequently alleviating psoriasis [18]. Multiple studies have indicated that the activation of AMPK exhibits variations based on sex, with more pronounced biological effects observed in males, despite the diverse disease phenotypes examined [19, 20]. Metformin, a well-known activator of AMPK through increased phosphorylation of its catalytic α-subunit, has been widely used as a first-line treatment for managing hyperglycemia in patients with type 2 diabetes for several decades [21]. The anti-inflammatory effects of metformin are mainly based on AMPK activation and inhibition of mTOR pathways [22]. Increasing evidence suggests that metformin offers numerous benefits beyond glycemic control, including its anti-inflammatory effects and ability to maintain homeostasis [23–25].
Some established in vitro experiments have demonstrated the anti-inflammatory and anti-proliferative effects of metformin in inhibiting the growth of activated keratinocytes via AMPK [26, 27]. Consequently, the potential benefits of metformin in psoriasis have been investigated [27]. However, two large-scale pharmacoepidemiologic studies examining the effects of metformin on psoriasis have yielded inconsistent conclusions [28, 29]. Due to inherent time-related and confounding bias in pharmacoepidemiologic studies, the reported effects of metformin on psoriasis may be limited by bias and hence remains unconcluded. While some relevant trials have shown the benefits of metformin in psoriasis patients, these trials were not sufficiently large or of long duration to provide definitive evidence [30, 31], such that one study reported a mean difference in severity score reduction of 2.2 (P = 0.001) while the other reported a mean difference in per cent change of 14.3% (P = 0.215). Moreover, existing studies have primarily focused on the therapeutic efficacy of metformin in treating psoriasis itself, leaving a knowledge gap regarding the potential protective effects of metformin on long-term comorbidities associated with psoriasis.
To address these limitations, we employed a Mendelian randomization (MR) approach, which minimizes biases by leveraging randomly allocated genetic variants at conception. MR studies have become increasingly utilized to infer the health effects of medications [32], such as the use of PCSK9/HMGCR variants as proxies for the effects of statins [33, 34] and the use of AMPK variants to represent the effect of metformin on the risks of cardiovascular diseases and cancer [35]. In our study, we used genetic variants of AMPK associated with glycated haemoglobin (HbA1c) to construct AMPK genetic risk score, serving as an inverse proxy for AMPK activation. This approach we used, employed in established pharmaco-epidemiological studies to infer the health effects of certain medications [35], provides additional evidence regarding the effects of AMPK activation and metformin use on the risk of psoriasis and its comorbidities.
Materials and methods
Study design and participants
This is a two-sample MR analysis, using AMPK as a target of metformin to infer the AMPK pathway-dependent effects of metformin on psoriasis and its comorbidities.
The UK Biobank recruited >500 000 participants aged 40–69 years between 2006 and 2010 in the UK. The UK Biobank received ethical approval from the North West Multicentre Research Ethics Committee (11/NW/0382), and all participants provided informed consent. No ethics approval was acquired for the analyses using summary statistics. The contributing studies to the consortium received ethical approval from their specific institutional review boards, and consent was obtained from all participants. Participants completed questionnaires covering self-reported diseases, underwent assessments and interviews, and provided biological samples. A blood sample was collected and tested at the central processing laboratory in Stockport, within 24 h of collection. HbA1c was determined by high-performance liquid chromatography on Bio-Rad Variant II Turbo analysers (Bio-Rad Laboratories, Hercules, CA, USA). Longitudinal follow-up via record linkage to all health service encounters is conducted. Hospital inpatient data used ICD-9 and ICD-10 codes.
Genotyping was undertaken with two similar arrays, the UK Biobank Lung Exome Variant Evaluation (BiLEVE) Axiom array (49 979 participants) and the UK Biobank Axiom array (438 398 participants). Genotype imputation was based on the reference panel combining the UK10K haplotype and the Haplotype Reference Consortium reference panels. To reduce confounding by latent population structure, we restricted the analysis to genetically verified white British participants and further excluded participants with a mismatch between genetic and reported sex or missing data on genotype. We used genotype and phenotype data from the UK Biobank provided in March 2022.
Instrumental variable
A weighted AMPK genetic risk score was created for each participant in the UK Biobank participant based on the strength of the association of genetic variants in the relevant gene regions with HbA1c in the Meta-Analyses of Glucose and Insulin-related traits Consortium, a genome-wide association study (GWAS) of HbA1c, with validation in the UK Biobank [36]. The meta-analyses included up to 159 940 individuals from 82 cohorts of European, African, East Asian and South Asian ancestry, with a mean age ranging from 52.2–60.9 and a proportion of females ranging from 44.4% to 66.1%. According to a previous study, we selected genetic variants within one megabase pair downstream and upstream of genes that encode AMPK subunits and selected low-linkage disequilibrium (r2<0.3) variants associated with HbA1c at a significance level of 0.05 in MAGIC, restricted to participants of European ancestry [35]. A total of 44 variants were used to construct the AMPK genetic risk score. The weighted AMPK genetic risk score was then calculated for each participant by summing the number of HbA1c-related alleles that a participant inherited at each variant included in the AMPK genetic risk score, weighted by the effect of that variant on HbA1c%. Because a positive effect size of a single nucleotide polymorphism (SNP) signifies an increase in HbA1c and vice versa, a higher AMPK genetic risk score is indicative of genetically higher HbA1c levels, which could be considered a proxy of less possibility of metformin use (Supplementary Fig. S1, available at Rheumatology online).
This score was dichotomized by median to mimic metformin use (AMPK under median indicates genetically lower HbA1c) and non-use (AMPK above-median indicates genetically higher HbA1c). As a result, a higher AMPK genetic risk score indicates a genetically higher HbA1c level and may serve as a proxy for less metformin use [34, 35, 37]. This dichotomization strategy allows us to evaluate the relationship between AMPK genetic risk scores and metformin use. Additionally, we utilized the continuous AMPK genetic risk score to assess the potential dose-response relationship.
Outcome variables
The outcomes were psoriasis and its several elderly-onset comorbidities with great disease burden. Cardiovascular and neurodegenerative comorbidities, including stroke, ischemic heart disease and dementia were selected due to their natural geriatric incident peak. Each disease outcome was defined based on self-reported medical conditions at baseline, and subsequent primary and secondary diagnoses of hospital episodes and death (ICD-9: 696 for psoriasis, 430/431/432/433/434 for stroke, 410/411 for ischemic heart disease, 290/294.1 for dementia, 332 for Parkinson’s disease; ICD-10: L40 for psoriasis, I60/I61/I62/I63/I64 for stroke, I20.0/I21/I24 for ischemic heart disease, F00/F01/F02/F03/F051/G30/G31.1/G31.8 for dementia, G20 for Parkinson’s disease). Specifically, psoriasis was further categorized into early- and late-onset by the onset age of 60 years according to the previously reported bi-peak of incidence of psoriasis in the UK Biobank [12].
Statistical analysis
To assess the assumption of independence of the genetic instruments (AMPK groups) from potential confounders, we assessed the association of AMPK groups with covariates (age at recruitment, education level, body mass index, smoking status, alcohol drinking status, and Townsend deprivation index) using χ2 tests or Student t-tests. To demonstrate that AMPK had the expected effect on HbA1c, we assessed the differences in HbA1c and random glucose between each group using t-tests. We assessed the association of genetically predicted increase in HbA1c% instrumented by AMPK variants with risks of EOP and LOP psoriasis using log-binomial regression models to obtain relative risks (RRs) and 95% confidence interval (CIs) as effect size. Cubic splines were used to evaluate the dose-response relationship between AMPK score and psoriasis. The interaction effect between AMPK genetic risk score and sex was tested, and further models were stratified by sex and adjusted for age at recruitment, genotyping array and the first 20 principal components of genetic ancestry. We also investigated the association of AMPK genetic risk score with cardiovascular and neurodegenerative comorbidities in patients with psoriasis. All analyses were performed using R software, version 3.5.1 (R Foundation for Statistical Computing, Vienna, Austria). A two-tailed P-value <0.05 was considered statistically significant. For multiple comparisons (EOP vs LOP, multiple outcomes), false discovery rate (FDR)-adjusted P values were also calculated.
MR analysis for HbA1c and psoriasis
To test whether the hypothesis that the observed effect of AMPK is due to HbA1c, we assessed the association of genetically elevated HbA1c with psoriasis. We downloaded the summary data, including effect sizes (regression coefficients, β), s.e. and P values from https://gwas.mrcieu.ac.uk/. We used the R package ‘TwoSampleMR’ according to the guideline (https://mrcieu.github.io/TwoSampleMR). We used the inverse variance weighted (IVW) method as the primary approach and other algorithms as supplements. We then tested pleiotropy using MR-Egger regression. For HbA1c, the data from the MAGIC were gathered as the exposure variables (ID: ieu-b-103, N = 46 368). For psoriasis, we extracted the results from the studies by the Neale laboratory, with 3871 cases and 333 288 controls, to generate the outcome variable (ID: ukb-a-100). After removing linkage disequilibrium, 11 genetic variants were selected from the exposure datasheet and incorporated into the outcome. We tested the pleiotropy and heterogeneity using the MR-Egger method. The leave-one-out sensitivity analysis that removes one single nucleotide polymorphism (SNP) at a time was conducted to test the robustness of the result.
Results
Participant characteristics
A total of 407 159 participants (54.1% women) were included in the main analysis. There were 9126 and 3324 cases of EOP and LOP, respectively. In Table 1, HbA1c was significantly higher in the high AMPK genetic risk score group, and the proportions of diabetes were significantly higher in the high AMPK genetic risk score group in men. No other significant differences in baseline characteristics between the two groups were identified.
Baseline characteristics of participants in the UK Biobank by sex and AMPK score
| Female (n = 220 168) | Male (n = 186 991) | |||||
|---|---|---|---|---|---|---|
| Baseline characteristics | AMPK score < median | AMPK score ≥ median | P | AMPK score < median | AMPK score ≥ median | P |
| Age at recruitment (years), mean (s.d.) | 56.7 (7.9) | 56.7 (7.9) | 0.776 | 57.2 (8.1) | 57.1 (8.1) | 0.384 |
| College/university degree, n (%) | 32 397 (29.3) | 32 173 (29.3) | 0.784 | 30 158 (32.3) | 30 188 (32.2) | 0.453 |
| Townsend deprivation index, mean (s.d.) | –1.58 (2.88) | –1.58 (2.89) | 0.182 | –1.52 (2.99) | –1.52 (2.99) | 0.969 |
| Smoker, n (%) | 44 594 (40.4) | 44 470 (40.5) | 0.651 | 47 221 (50.7) | 47 900 (51.1) | 0.070 |
| Alcohol drinker, n (%) | 101 449 (91.9) | 101 026 (92.0) | 0.453 | 88 507 (95.0) | 89 208 (95.1) | 0.077 |
| BMI (kg/m2), mean (s.d.) | 27.0 (5.1) | 27.0 (5.1) | 0.697 | 27.8 (4.2) | 27.8 (4.2) | 0.496 |
| HbA1c (mmol/mol), mean (s.d.) | 35.5 (5.6) | 35.8 (5.7) | <0.001 | 36.1 (7.1) | 36.5 (7.5) | <0.001 |
| HbA1c (%), mean (s.d.) | 5.40 (0.51) | 5.43 (0.52) | <0.001 | 5.45 (0.65) | 5.49 (0.68) | <0.001 |
| Random glucose (mmol/L), mean (s.d.) | 5.06 (1.03) | 5.07 (1.05) | 0.071 | 5.17 (1.36) | 5.19 (1.37) | 0.093 |
| Diabetes, n (%) | 1992 (1.8) | 2095 (1.9) | 0.073 | 3340 (3.6) | 3626 (3.9) | 0.001 |
| Female (n = 220 168) | Male (n = 186 991) | |||||
|---|---|---|---|---|---|---|
| Baseline characteristics | AMPK score < median | AMPK score ≥ median | P | AMPK score < median | AMPK score ≥ median | P |
| Age at recruitment (years), mean (s.d.) | 56.7 (7.9) | 56.7 (7.9) | 0.776 | 57.2 (8.1) | 57.1 (8.1) | 0.384 |
| College/university degree, n (%) | 32 397 (29.3) | 32 173 (29.3) | 0.784 | 30 158 (32.3) | 30 188 (32.2) | 0.453 |
| Townsend deprivation index, mean (s.d.) | –1.58 (2.88) | –1.58 (2.89) | 0.182 | –1.52 (2.99) | –1.52 (2.99) | 0.969 |
| Smoker, n (%) | 44 594 (40.4) | 44 470 (40.5) | 0.651 | 47 221 (50.7) | 47 900 (51.1) | 0.070 |
| Alcohol drinker, n (%) | 101 449 (91.9) | 101 026 (92.0) | 0.453 | 88 507 (95.0) | 89 208 (95.1) | 0.077 |
| BMI (kg/m2), mean (s.d.) | 27.0 (5.1) | 27.0 (5.1) | 0.697 | 27.8 (4.2) | 27.8 (4.2) | 0.496 |
| HbA1c (mmol/mol), mean (s.d.) | 35.5 (5.6) | 35.8 (5.7) | <0.001 | 36.1 (7.1) | 36.5 (7.5) | <0.001 |
| HbA1c (%), mean (s.d.) | 5.40 (0.51) | 5.43 (0.52) | <0.001 | 5.45 (0.65) | 5.49 (0.68) | <0.001 |
| Random glucose (mmol/L), mean (s.d.) | 5.06 (1.03) | 5.07 (1.05) | 0.071 | 5.17 (1.36) | 5.19 (1.37) | 0.093 |
| Diabetes, n (%) | 1992 (1.8) | 2095 (1.9) | 0.073 | 3340 (3.6) | 3626 (3.9) | 0.001 |
HbA1c: glycated haemoglobin.
Baseline characteristics of participants in the UK Biobank by sex and AMPK score
| Female (n = 220 168) | Male (n = 186 991) | |||||
|---|---|---|---|---|---|---|
| Baseline characteristics | AMPK score < median | AMPK score ≥ median | P | AMPK score < median | AMPK score ≥ median | P |
| Age at recruitment (years), mean (s.d.) | 56.7 (7.9) | 56.7 (7.9) | 0.776 | 57.2 (8.1) | 57.1 (8.1) | 0.384 |
| College/university degree, n (%) | 32 397 (29.3) | 32 173 (29.3) | 0.784 | 30 158 (32.3) | 30 188 (32.2) | 0.453 |
| Townsend deprivation index, mean (s.d.) | –1.58 (2.88) | –1.58 (2.89) | 0.182 | –1.52 (2.99) | –1.52 (2.99) | 0.969 |
| Smoker, n (%) | 44 594 (40.4) | 44 470 (40.5) | 0.651 | 47 221 (50.7) | 47 900 (51.1) | 0.070 |
| Alcohol drinker, n (%) | 101 449 (91.9) | 101 026 (92.0) | 0.453 | 88 507 (95.0) | 89 208 (95.1) | 0.077 |
| BMI (kg/m2), mean (s.d.) | 27.0 (5.1) | 27.0 (5.1) | 0.697 | 27.8 (4.2) | 27.8 (4.2) | 0.496 |
| HbA1c (mmol/mol), mean (s.d.) | 35.5 (5.6) | 35.8 (5.7) | <0.001 | 36.1 (7.1) | 36.5 (7.5) | <0.001 |
| HbA1c (%), mean (s.d.) | 5.40 (0.51) | 5.43 (0.52) | <0.001 | 5.45 (0.65) | 5.49 (0.68) | <0.001 |
| Random glucose (mmol/L), mean (s.d.) | 5.06 (1.03) | 5.07 (1.05) | 0.071 | 5.17 (1.36) | 5.19 (1.37) | 0.093 |
| Diabetes, n (%) | 1992 (1.8) | 2095 (1.9) | 0.073 | 3340 (3.6) | 3626 (3.9) | 0.001 |
| Female (n = 220 168) | Male (n = 186 991) | |||||
|---|---|---|---|---|---|---|
| Baseline characteristics | AMPK score < median | AMPK score ≥ median | P | AMPK score < median | AMPK score ≥ median | P |
| Age at recruitment (years), mean (s.d.) | 56.7 (7.9) | 56.7 (7.9) | 0.776 | 57.2 (8.1) | 57.1 (8.1) | 0.384 |
| College/university degree, n (%) | 32 397 (29.3) | 32 173 (29.3) | 0.784 | 30 158 (32.3) | 30 188 (32.2) | 0.453 |
| Townsend deprivation index, mean (s.d.) | –1.58 (2.88) | –1.58 (2.89) | 0.182 | –1.52 (2.99) | –1.52 (2.99) | 0.969 |
| Smoker, n (%) | 44 594 (40.4) | 44 470 (40.5) | 0.651 | 47 221 (50.7) | 47 900 (51.1) | 0.070 |
| Alcohol drinker, n (%) | 101 449 (91.9) | 101 026 (92.0) | 0.453 | 88 507 (95.0) | 89 208 (95.1) | 0.077 |
| BMI (kg/m2), mean (s.d.) | 27.0 (5.1) | 27.0 (5.1) | 0.697 | 27.8 (4.2) | 27.8 (4.2) | 0.496 |
| HbA1c (mmol/mol), mean (s.d.) | 35.5 (5.6) | 35.8 (5.7) | <0.001 | 36.1 (7.1) | 36.5 (7.5) | <0.001 |
| HbA1c (%), mean (s.d.) | 5.40 (0.51) | 5.43 (0.52) | <0.001 | 5.45 (0.65) | 5.49 (0.68) | <0.001 |
| Random glucose (mmol/L), mean (s.d.) | 5.06 (1.03) | 5.07 (1.05) | 0.071 | 5.17 (1.36) | 5.19 (1.37) | 0.093 |
| Diabetes, n (%) | 1992 (1.8) | 2095 (1.9) | 0.073 | 3340 (3.6) | 3626 (3.9) | 0.001 |
HbA1c: glycated haemoglobin.
Association of AMPK score with psoriasis
Compared with participants with a low AMPK genetic risk score (under the median, favouring lower HbA1c), participants with a high AMPK genetic risk score (above the median, favouring higher HbA1c) had a similar risk of psoriasis (RR = 1.010, 95% CI 0.975–1.046, P = 0.578). We identified a significant interaction effect between AMPK genetic risk score and sex on LOP (Pinteraction < 0.001), such that men with a high AMPK genetic risk score had an additional 12.4% risk of LOP [RR = 1.124, 95% CI 1.022–1.236, P = 0.016, false discovery rate (FDR)-adjusted P = 0.048] but not of EOP (Fig. 1). We further investigated the dose-response relationship between AMPK genetic risk score and LOP and identified a significant non-linear association in men rather than women, such that the effect of AMPK was significant in high quantiles (above median) but not in low quantiles (Fig. 2).
Association of AMPK scores with psoriasis. FDR: false discovery rate; RR: relative risk
Dose-response relationship between AMPK score and the risk of psoriasis by sex
Association of HbA1c with psoriasis using two-sample Mendelian randomization
As shown in Fig. 3A, the IVW method, in conjunction with other methods, suggested that the genetically elevated HbA1c was associated with an increased risk of psoriasis (βIVW = 0.008, P = 4 × 10−4). Tests for pleiotropy (P = 0.712) and heterogeneity (P = 0.541) showed negative results. The leave-one-out sensitivity analysis showed stable results (Fig. 3B). The reverse MR analysis treating psoriasis as the exposure and HbA1c as the outcome demonstrated no significant association (P = 0.187), indicating the unidirectionality of the relationship.
Two-sample Mendelian randomisation analysis for HbA1c as exposure and psoriasis as the outcome. (A) Mendelian randomisation analysis using different methods. (B) Leave-one-out sensitivity analysis. MR: Mendelian randomization; SNP: single nucleotide polymorphism
Association of AMPK score with comorbidities of psoriasis
We further removed participants with no psoriasis in baseline and follow-up. Psoriasis patients with a high AMPK genetic risk score had an increased risk of ischemic heart disease (RR = 1.137, 95% CI 1.013–1.278, P = 0.030, FDR-adjusted P = 0.150). AMPK genetic risk score was significantly associated with a higher risk of ischemic heart disease (RR = 1.217, 95% CI 1.062–1.395, P = 0.005, FDR-adjusted P = 0.025) but not other comorbidities in men. In contrast, the associations were not significant in women (Fig. 4).
Association of AMPK scores with cardiovascular and neurodegenerative comorbidities of psoriasis. FDR: false discovery rate; RR: relative risk
Discussion
To the best of our knowledge, this is the first MR study to ascertain the effects of metformin, instrumented by AMPK variants, on psoriasis and its long-term comorbidities, including cardiovascular and neurodegenerative diseases. We propose that AMPK activation has a protective role in reducing the risk of incident psoriasis and associated ischemic heart diseases in late-onset male patients. We added genetic evidence on the putative therapeutic effect of metformin for psoriasis patients via AMPK pathways.
There were two retrospective studies examining the effects of metformin on the risk of psoriasis among diabetes patients. Brauchli et al. reported that current use of metformin (≥15 prescriptions) was associated with a reduced psoriasis risk in the General Practice Research Database [29], while this association was only significant among frequent users of metformin (≥90 days of prescriptions per year) in a Taiwan database [28]. However, our study, where the start of ‘exposure’ is at birth, effectively reduced this bias. By using a design more robust to avoid time-related and confounding biases inherited in traditional pharmacoepidemiologic studies, our study adds powerful evidence to the ongoing discussion regarding the effects of metformin on psoriasis risk.
Because the anti-inflammatory effects of metformin are mainly based on AMPK activation and inhibition of mTOR pathways [22], AMPK genetic variants, also known as the main therapeutic signal pathway of metformin, have been extensively studied and utilized in measuring the heritable risk of developing a particular disease [36, 38, 39]. These genetic variants provide a reliable surrogate for assessing the effects of metformin. Similar strategies have been employed in previous studies, utilizing random allocation of genetic variants to infer the health effects of medications, which were demonstrated to be good surrogate predictors. For example, 3-hydroxy-3-methyl-glutaryl-coenzyme A reductase (HMGCR) variants have been used to represent the effects of statins, and HbA1c% genetic variants have served as an opposite proxy for AMPK activation to mimic the effects of metformin [35, 36]. Based on 44 SNPs in constructing AMPK genetic variants, we found that AMPK genetic risk score was associated with a higher risk of LOP in men rather than women. In contrast, no effect was observed between AMPK genetic risk score and early-onset psoriasis. The onset-age theory in psoriasis has been a topic of discussion for several decades. It was initially proposed by Henseler et al. in the 1980s [40], suggesting the existence of two distinct forms of psoriasis based on the age at onset. Psoriasis, which was initiated in later life, initially defined by onset age over 40 years, had a less severe clinical course and a more continuous evolution [41]. Building upon this concept, Kwon et al. proposed that the later incident peak of psoriasis, which occurs around the age of 60 years, has special clinical features and comorbidities [42]. Iskandar et al. further confirmed this hypothesis in a meta-analysis of 90 studies from 22 countries [12]. Our study focused on the bi-peak incidence pattern of psoriasis as previously reported in the UK Biobank, with a specific later incident peak at around 60 years of age. By investigating the characteristics of late-onset psoriasis, particularly in psoriasis initiated over 60 years, our findings align with existing evidence and further reinforce the validity of the age-onset theory. It is important to note that our study included participants from the UK Biobank who were all aged over 40 years, thus limiting our ability to test our findings against a more generalized late-onset psoriasis cut-off at 40 years old.
Metformin appears to have more benefits in men. Orchard et al. reported that metformin reduced the incidence of metabolic syndrome only in men in a large RCT [43]. The existence of a sex difference in the AMPK signal pathway has been proposed in previous literature [44]. Specifically, the AMPK pathway has been reported to exhibit higher activity in males, with significantly higher phosphorylation of AMPK observed in males compared with females after stimulation [19]. Our MR analysis suggested that the male-dominated efficacy of metformin could potentially be attributed to sex differences in the AMPK pathway. This finding is remarkably consistent with the speculation made by Glossmann and Reider [45], who suggested that a pilot trial combining metformin with methotrexate should first be conducted in male psoriasis patients. However, the underlying mechanism behind this sex difference in metformin's effects remains not yet fully understood.
Our study further suggests that AMPK activation can reduce subsequent comorbidities, specifically ischemic heart disease, only in men with psoriasis. While most previous studies reported the benefits of metformin on psoriasis [46], these studies primarily focused on analysing the clearance of psoriatic lesions rather than assessing the associated comorbidities, which cannot be overlooked in the lifelong management of psoriasis.
In all, our study suggested the necessity for subgrouping preventive and treatment strategies among the high-risk population and psoriasis patients. More sex-specific and metabolic-driven aetiology studies on psoriasis should be conducted.
Although our study is less affected by biases than previous observational studies, there are limitations. First, the conclusion of our study is restricted to participants of European ancestry. Second, while we found that AMPK activation by metformin may protect against psoriasis and its particular types of cardiovascular comorbidities, the estimates from this study cannot be directly inferred the health impacts of metformin, given the differences in exposure time [47]. However, RCTs often use short-term treatment, in contrast to the lifelong exposure estimated by MR analysis. Third, the AMPK genetic variants may only predict the effect of metformin which acts on the AMPK pathways [48, 49]. Besides, using a proxy for metformin by AMPK genetic variants assumed that the effect of genetically predicted AMPK activation in the body has an environmental equivalent effect, potentially leading to false-negative results. However, the use of SNPs associated with HbA1c, as an opposite proxy for AMPK activation to mimic metformin is a strategy adopted from several established MR studies [35]. More importantly, it is important to highlight that AMPK is one of the main targets of metformin's anti-inflammatory effects on psoriasis, as reported by in vitro psoriasis studies. Last, owing to the multiple comparisons for onset age, sex and comorbidity, we adjusted the P values using the FDR method. The results should be interpreted with caution for possible type 1 error caused by multiplicity.
Conclusion
In conclusion, AMPK activation, as the main target of metformin, may protect from incident LOP and their ischemic heart disease comorbidities in men rather than women. There is a clear need for sex-specific, comorbidity-targeted, metabolic-driven protective and therapeutic interventions for psoriasis.
Supplementary material
Supplementary material is available at Rheumatology online.
Data availability
This work has been conducted using the UK Biobank Resource (Application Number: 55242, 55257). The UK Biobank is an open-access resource, and bona fide researchers can apply to use the UK Biobank dataset by registering and applying at http://ukbiobank.ac.uk/register-apply/. Further information is available from the corresponding author upon request.
Funding
This work was supported by the National Natural Science Foundation of China (8210373), Huxiang Youth Talent Support Program of Hunan Province (2022RC1014) and Ministry of Industry and Information Technology of China (TC210804V).
Disclosure statement: The authors declare no conflict of interest.
Acknowledgements
We thank Matt Hodgson, Paul Flood, Laura Battersby, Jamie King and Charlotte Morrissey from UK Biobank Access Management Team, for helping us with the data preparation.
References
Author notes
X.C., H.L., and M.S. contributed equally.
Comments