http://orcid.org/0000-0001-8537-1510
Figures
Abstract
Introduction
In patients with psoriatic arthritis (PsA) an increased cardiovascular (CV) risk has been observed. Recently, a EULAR taskforce suggested to use a multiplication by the factor of 1.5 of CV risk algorithms in patients with inflammatory arthritis. This study aims to evaluate the performance of five original and adapted according to EULAR recommendations CV risk algorithms in PsA: SCORE, CUORE, Framingham Risk Score (FRS), QRISK2, and Reynold’s Risk Score (RRS).
Methods
Prospectively collected data from two Italian cohorts were used. Discriminatory ability for CV risk prediction was evaluated by the area under the ROC curves. Calibration between predicted and observed events was assessed by Hosmer-Lemeshow (HL) tests. Sensibility and specificity were calculated for low-to-intermediate and intermediate-to-high risk cut-offs.
Results
One hundred fifty-five patients were enrolled with an observation of 1550 patient/years. Area under the ROC were 0.7679 (95% CI 0.64768 to 0.88812), 0.864 (95% CI 0.79675 to 0.93278), 0.7575 (95% CI 0.65784 to 0.85708), 0.8660 (95% CI 0.79428 to 0.93772), and 0.7183 (95% CI 0.57795 to 0.85862) for SCORE, CUORE, FRS, QRSIK2, and RRS, respectively. HL tests demonstrated poor model fit (p<0.05) for SCORE, CUORE, and RRS. Discriminative ability and calibration were not improved by adaption of the algorithms according to EULAR recommendations. Up to 80% of CV events occurred in patients at “low risk” and up to 93% of CV events in patients at “low-intermediate risk”.
Citation: Navarini L, Margiotta DPE, Caso F, Currado D, Tasso M, Angeletti S, et al. (2018) Performances of five risk algorithms in predicting cardiovascular events in patients with Psoriatic Arthritis: An Italian bicentric study. PLoS ONE 13(10): e0205506. https://doi.org/10.1371/journal.pone.0205506
Editor: Concetta Zito, University of Messina, ITALY
Received: June 22, 2018; Accepted: September 26, 2018; Published: October 11, 2018
Copyright: © 2018 Navarini et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All data generated or analyzed during this study are included in this article and all the results can be found in the body of the paper.
Funding: The authors received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Psoriatic arthritis (PsA) is an inflammatory arthritis associated with personal or familial history of psoriasis. PsA is characterized by heterogeneous clinical presentation, with possible axial and/or peripheral joint involvement, enthesitis and dactylitis [1].
In PsA, as well as in other forms of inflammatory arthritis, an increased cardiovascular risk (CV) has been observed. Increased prevalence of traditional CV risk factors, chronic inflammation and potential adverse effects of drugs, especially corticosteroids and nonsteroidal anti-inflammatory drugs (NSAIDs), contribute to CV comorbidity in PsA [2–4]. Notably, chronic inflammation can affect global CV risk on multiple points of view. The expression of a more atherogenic lipid profile, the forced reduction of physical activity, and the increased production of cytokines involved in plaque formation, such as interleukin (IL)-1, IL-17, and tumor necrosis factor (TNF)-alpha, seem to boost CV risk of patients [2]. Accelerated atherosclerosis seems to play a pivotal role predisposing patients with PsA to the development of CV and cerebrovascular events [5].
Therefore, the identification of high CV risk patients with psoriatic arthritis is particularly important, in order to implement preventive strategies, like lifestyle changes and pharmacological interventions. A large number of CV risk algorithms have been proposed over time [6]. The performance of algorithms for calculating CV risk in patients with another inflammatory, like rheumatoid arthritis (RA), is a largely debated subject. In particular, Arts and coworkers [7] assessed the performance of four CV risk algorithms in establishing the risk of fatal and non-fatal CV events in European patients with RA. The results show that different scores (Framingham, SCORE, Reynold’s Risk Score, QRISK2) tend to underestimate CV risk in patients with RA; the risk observed exceeds that predicted; the different scores appear slightly calibrated for RA patients [8]. In addition, Ernst and coauthors demonstrated that the Framingham Risk Score (FRS) underestimates CV risk in patients with psoriatic arthritis [9].
Recently, the European League Against Rheumatism (EULAR) recommended to adapt the general population risk algorithms with a multiplication by the factor of 1.5 in patients with rheumatoid arthritis and other inflammatory arthritis, except for QRISK2 which is characterized by a multiplication factor intrinsic to the algorithm for rheumatoid arthritis [10].
Only a few data evaluate the performance of different algorithms, both original and adapted according to EULAR indication, in predicting the actual CV risk in PsA. In this study, the performance of FRS, Systematic Coronary Risk Evaluation (SCORE), QRISK2, Reynolds risk score (RRS), and the Italian Progetto CUORE individual score is evaluated in a bicentric cohort of patients with psoriatic arthritis from Rome and Neaples (Italy).
Methods
A retrospective analysis of prospective collected data from PsA cohort of Unit of Immunorheumatology, Università Campus Bio-Medico di Roma, and Unit of Rheumatology, University of Neaples “Federico II”, has been made in November 2017. Patients without a personal history of CV disease (CVD) at baseline (November 2007) were consecutively included in this study. At baseline, all the patients fulfilled the Classification Criteria for Psoriatic Arthritis (CASPAR) [11].
Ethics committee of Università Campus Bio-Medico di Roma approved the study, which complied with the Declaration of Helsinki.
Baseline characteristics extracted from the cohort database: were age (years), gender (male/female), weight (kg), height (cm), CRP (mg/l), disease activity score (DAS) 28-joints, axial joint arthritis (Y/N), peripheral joint arthritis (Y/N), enthesitis (Y/N), dactylitis (Y/N), psoriasis area severity index (PASI), history of inflammatory bowel disease (Y/N), history of uveitis (Y/N), family history of CVD (Y/N), smoking status (Y/N/previous), hypertension (Y/N), systolic blood pressure (mmHg), total cholesterol (mg/dl), high-density-lipoprotein (HDL) cholesterol (mg/dl), use of statins (Y/N), use of antihypertensive medication (Y/N), diabetes mellitus (Y/N), atrial fibrillation (Y/N), chronic kidney disease stage IV-V (Y/N), angina or heart attack in a 1st degree relative <60 years (Y/N).
The primary outcome was the first CV event (fatal and non-fatal), as report by electronic patient files. CV events were considered: sudden cardiac death, coronary artery diseases (CAD) (stable and unstable angina pectoris, myocardial infarction), cerebral vascular accident (CVA), transient ischemic attack (TIA), peripheral artery disease (PAD) and heart failure.
The 10-year general FRS for CVD [12], QRISK2 [13], CUORE [14] and RRS [15, 16] were calculated using already published algorithms. SCORE risk charts for low risk countries were used [17]. The items included in each CV risk algorithm are reported in Table 1. Cut-off values that mark the difference between low-to-intermediate risk and intermediate-to-high risk were 10% and 20% respectively, except for SCORE in which cut-off values were 1% and 5%.
At baseline, data from medical record were used to calculate individual risk for CV within 10 years for all five algorithms. The discriminatory ability of all five algorithms was evaluated using the area under the receiver operating characteristic (ROC) curve, which is similar to the concordance-statistic (c-statistic). Calibration was assessed by comparing the agreement between observed and predicted number of CV events in groups stratified in deciles, sextiles, or septiles of the predicted risk, as appropriate, using Hosmer-Lemeshow (HL) test. Fisher’s exact test has been used for analysis of contingency table, while Mann-Whitney test has been used to compare ranks. The sample size, calculated setting a type I error to 0.05, the test power to 0.8, the AUC of the adapted test to 0.7, was 143 patients [18]. All statistical analysis was performed using STATA V.14.
Results
Data from 155 patients (1550 patient-years) were analyzed. All the patients were Caucasic and lived in Center or South of Italy. During follow-up, 21 patients had a CV event (1.35 events per 100 patient/years): 8 cases of myocardial infarction or unstable angina pectoris, 3 cases of stable angina pectoris, 2 cases of TIA, 4 cases of PAD, 4 cases of HF. No fatal events were reported. The primary outcome was adjusted to fit each CV risk algorithm, leaving 21, 15, 11, 17, and 15 CV events for Framingham, QRISK2, RRS, SCORE, and CUORE, respectively. As the RRS is not applicable to patients with diabetes or younger than 45 years, these patients (n = 58) were excluded, and only 97 patients were included in the analysis of RRS. Patients characteristics are summarized in Table 2.
c-statistic scores of 0.7679 (95% CI 0.64768 to 0.88812), 0.864 (95% CI 0.79675 to 0.93278), 0.7575 (95% CI 0.65784 to 0.85708), 0.8660 (95% CI 0.79428 to 0.93772), and 0.7183 (95% CI 0.57795 to 0.85862) for SCORE, CUORE, FRS, QRISK2, and RRS respectively were found (Fig 1A–1E).
Overall, the multiplicative factors do not seems to improve the performances of none of the algorithms: c-statistic scores of 0.7679 (95% CI 0.64768 to 0.88812), 0.8648 (95% CI 0.79675 to 0.93278), 0.7584 (95% CI 0.65889 to 0.85782), 0.8664 (95% CI 0.79452 to 0.93834), and 0.7183 (95% CI 0.57795 to 0.85862) for SCORE*1.5 (p = ns vs SCORE), CUORE*1.5 (p = ns vs CUORE), FRS*1.5 (p = ns vs FRS), QRISK2-RA (p = ns vs QRISK2), and RRS*1.5 (p = ns vs RRS) respectively were found (Fig 2A–2E).
Areas under the curve (AUC)-values (95% CI) are 0.7679 (95% CI 0.64768 to 0.88812), 0.864 (95% CI 0.79675 to 0.93278), 0.7575 (95% CI 0.65784 to 0.85708), 0.8660 (95% CI 0.79428 to 0.93772), and 0.7183 (95% CI 0.57795 to 0.85862) for SCORE (A), CUORE (B), FRS (C), QRSIK2 (D), and RRS (E), respectively.
Areas under the curve (AUC)-values (95% CI) are 0.7679 (95% CI 0.64768 to 0.88812), 0.8648 (95% CI 0.79675 to 0.93278), 0.7584 (95% CI 0.65889 to 0.85782), 0.8664 (95% CI 0.79452 to 0.93834), and 0.7183 (95% CI 0.57795 to 0.85862) for SCORE*1.5 (A), CUORE*1.5 (B), FRS*1.5 (C), QRISK2-RA (D), and RRS*1.5 (E), respectively.
Overall, a discrepancy between predicted risk and observed CV events was found (Fig 3A–3E).
Observed (dark grey bars) vs predicted (light grey bars) CV events (%) in deciles, sextiles, and septiles for SCORE (A), SCORE*1.5 (B), CUORE (C), CUORE*1.5 (D), FRS (E), FRS*1.5 (F), QRISK2 (G), QRISK2-RA (H), RRS (I), and RRS*1.5 (L).
CUORE score underestimated CV risk in middle and top deciles and HL test demonstrated poor model fit (p = 0.045); the multiplication by the factor of 1.5 did not improve the performance (HL test p = 0.045). QRISK2 and QRISK2-RA underestimated CV risk in middle deciles and overestimated CV risk in top deciles, and HL test did not demonstrate poor model fit (p = 0.162). CV risk predicted by FRS and FRS*1.5 was lower than observed CV events in every decile, but HL test did not yield a poor model fit (p = 0.26 for FRS and p = 0.259 for FRS*1.5). SCORE and SCORE*1.5 tended to underestimate CV risk in every sextile considered and HL test provided poor model fit (p<0.5). HL test yield poor model fit also for RRS and RRS*1.5 as well (p<0.05).
Sensitivity and specificity of the 10% and 20% cut-off points for CV risk for FRS, QRISK2, CUORE, and RRS and of the 1% and 5% cut-off points for CV risk for SCORE are reported in Table 3.
Among patients with CV event, 17.6% was at “low risk” and 52.9% at “low-intermediate risk” according to SCORE, 80% was at “low risk” and 93.3% was at “low-intermediate risk” according to CUORE, 76.2% was at “low risk” and 95.2% was at “low-intermediate risk” according to FRS, 20% was at “low risk” and 60% was at “low-intermediate risk” according to QRISK2, 81.8% was at “low risk” and 90.9% was at “low-intermediate risk” according to RRS, 5.9% was at “low risk” and 52.9% was at “low-intermediate risk” according to SCORE*1.5, 66.7% was at “low risk” and 86.7% was at “low-intermediate risk” according to CUORE*1.5, 61.9% was at “low risk” and 85.7% was at “low-intermediate risk” according to FRS*1.5, 46.7% was at “low risk” and 80% was at “low-intermediate risk” according to QRISK2-RA, and 45.4% was at “low risk” and 81.8% was at “low-intermediate risk” according to RRS*1.5.
Discussion
The prevention of CV events in general population still remains a challenge. In patients with inflammatory arthritis, an increased burden of CV comorbidity and mortality is well established [10]. Despite in psoriatic arthritis an increased CV risk has been demonstrated [19], little is known about CV risk algorithm in these patients. Overall, the five algorithms evaluated in this study underestimated CV risk. A relatively good discrimination between patients with or without CV events has been demonstrated, with areas under the ROC curves between 0.7183 for RSS and 0.8660 for QRISK2, and calibration of all five algorithms was poor to moderate. In 2015/2016, EULAR suggested to adapt CV risk algorithms by a factor of 1.5 (or to include the independent risk factor RA in QRISK2) in all the patients with RA and virtually other inflammatory arthritis, such as PsA. Notably, in this study the adaptation suggested by EULAR did not increase the discriminative ability and calibration of none of the five algorithms under consideration. Despite the discriminative ability in patients with PsA is comparable to general population, the five algorithms performed less well in terms of calibration [12–17]. Particularly, a poor model fit with a significant different distribution of observed events compared to predicted ones has been demonstrated for CUORE, SCORE, and RSS, but not for FRS and QRISK2. Most of CV events reported in this study occurred in patients at “low risk” or at “low-intermediate risk”. In general population and in patients with PsA as well, people at “low risk” or “low-intermediate risk” are less likely to receive any preventive treatment. Consequently, according to the results of our study, it seems crucial to redefine cut-off values for low and intermediate CV risk in patients PsA. Data reported in the present study demonstrate that, among all the evaluated models, QRISK2 and QRISK2-RA show the best discriminative ability and calibration in Caucasic patients with PsA from Center and South of Italy. These results are in line with other studies. In 2014 Eder and coworkers highlighted that the FRS can underestimate the extent of subclinical atherosclerosis in patients with psoriasis and PsA [20]. Moreover, Shen and coworkers tested the performance of several CV risk scores, including FRS, QRISK2, SCORE, 10-year atherosclerotic cardiovascular disease risk algorithm (ASCVD) from the American College of Cardiology and the American Heart Association, and the EULAR-recommended modified versions by 1.5 multiplication factor in discriminating subclinical atherosclerosis in PsA patients. Also in this case, results showed that all CV risk scores underestimated the subclinical atherosclerosis risk. Further, EULAR-recommended modification improved the sensitivity of FRS and ASCVD only to a moderate level [21]. In a recent cross-sectional descriptive study, Martinez-Vidal and coworkers analyzed the cardiovascular risk according to SCORE charts in 102 PsA patients and studied the presence of subclinical CV disease by common carotid ultrasound. SCORE underestimated the presence of subclinical CV disease in PsA patients. Indeed, according to SCORE charts, above 71% of the patients showed “intermediate risk”, 25.5% “high risk” and 4% “very high risk”. After the ultrasound evaluation, 26.5% of patients were reclassified at “very high risk” due to the presence of atherosclerotic plaques [22].
Several strengths and weakness of this study should be considered. Only 62.6% of patients were eligible for RRS calculation. A comparison between patients with PsA and age-matched and gender-matched healthy controls has not been made. Furthermore, these results were obtained from a population of Caucasic patients with PsA from Center and South of Italy and thus may not be generalizable to patients with other ethnicity or from other parts of the world. Moreover, HL test for QRISK2 and FRS yielded a p value >0.05, suggesting a good model fit, but it cannot be excluded a low power of the test for these algorithms probably due to small sample size.
In conclusion, this study extensively studied five CV risk algorithms in an Italian population of patients with PsA. Overall, these algorithms and their cut-off for low and intermediate risk tend to provide less accurate prediction of CV risk in patients with PsA compared to general population. Underestimation of CV risk in PsA patients could lead to insufficient treatment of CV risk factors. Prospective and larger studies are required in order to improve CV risk prediction in PsA patients, redefining cut-off values of already available algorithms, adding more biomarkers and disease-related CV risk factors in prediction models, and providing PsA-specific CV risk algorithms.
References
- 1. Ritchlin CT, Colbert RA, Gladman DD. Psoriatic Arthritis. N Engl J Med. 2017;376(10):957–70. pmid:28273019.
- 2. Yim KM, Armstrong AW. Updates on cardiovascular comorbidities associated with psoriatic diseases: epidemiology and mechanisms. Rheumatology international. 2017;37(1):97–105. Epub 2016/05/26. pmid:27221457.
- 3. Costa L, Caso F, D'Elia L, Atteno M, Peluso R, Del Puente A, et al. Psoriatic arthritis is associated with increased arterial stiffness in the absence of known cardiovascular risk factors: a case control study. Clinical rheumatology. 2012;31(4):711–5. Epub 2011/11/25. pmid:22113824.
- 4. Sobchak C, Eder L. Cardiometabolic Disorders in Psoriatic Disease. Current rheumatology reports. 2017;19(10):63. Epub 2017/08/28. pmid:28844116.
- 5. Ramonda R, Lo Nigro A, Modesti V, Nalotto L, Musacchio E, Iaccarino L, et al. Atherosclerosis in psoriatic arthritis. Autoimmunity reviews. 2011;10(12):773–8. Epub 2011/06/21. pmid:21684355.
- 6. Pletcher MJ, Moran AE. Cardiovascular Risk Assessment. The Medical clinics of North America. 2017;101(4):673–88. Epub 2017/06/05. pmid:28577619.
- 7. Arts EE, Popa C, Den Broeder AA, Semb AG, Toms T, Kitas GD, et al. Performance of four current risk algorithms in predicting cardiovascular events in patients with early rheumatoid arthritis. Annals of the rheumatic diseases. 2015;74(4):668–74. Epub 2014/01/07. pmid:24389293.
- 8. Crowson CS, Matteson EL, Roger VL, Therneau TM, Gabriel SE. Usefulness of risk scores to estimate the risk of cardiovascular disease in patients with rheumatoid arthritis. The American journal of cardiology. 2012;110(3):420–4. Epub 2012/04/24. pmid:22521305; PubMed Central PMCID: PMCPMC3398213.
- 9. Ernste FC, Sanchez-Menendez M, Wilton KM, Crowson CS, Matteson EL, Maradit Kremers H. Cardiovascular risk profile at the onset of psoriatic arthritis: a population-based cohort study. Arthritis Care Res (Hoboken). 2015;67(7):1015–21. Epub 2015/01/13. pmid:25581120; PubMed Central PMCID: PMCPMC4482766.
- 10. Agca R, Heslinga SC, Rollefstad S, Heslinga M, McInnes IB, Peters MJ, et al. EULAR recommendations for cardiovascular disease risk management in patients with rheumatoid arthritis and other forms of inflammatory joint disorders: 2015/2016 update. Annals of the rheumatic diseases. 2017;76(1):17–28. Epub 2016/10/05. pmid:27697765.
- 11. Taylor W, Gladman D, Helliwell P, Marchesoni A, Mease P, Mielants H. Classification criteria for psoriatic arthritis: development of new criteria from a large international study. Arthritis and rheumatism. 2006;54(8):2665–73. Epub 2006/07/28. pmid:16871531.
- 12. D'Agostino RB Sr., Vasan RS, Pencina MJ, Wolf PA, Cobain M, Massaro JM, et al. General cardiovascular risk profile for use in primary care: the Framingham Heart Study. Circulation. 2008;117(6):743–53. Epub 2008/01/24. pmid:18212285.
- 13. Hippisley-Cox J, Coupland C, Vinogradova Y, Robson J, Minhas R, Sheikh A, et al. Predicting cardiovascular risk in England and Wales: prospective derivation and validation of QRISK2. BMJ (Clinical research ed). 2008;336(7659):1475–82. Epub 2008/06/25. pmid:18573856; PubMed Central PMCID: PMCPMC2440904.
- 14. Doukaki S, Caputo V, Bongiorno MR. Psoriasis and Cardiovascular Risk: Assessment by CUORE Project Risk Score in Italian Patients. Dermatology research and practice. 2013;2013:389031. Epub 2013/10/02. pmid:24082881; PubMed Central PMCID: PMCPMC3780628.
- 15. Ridker PM, Buring JE, Rifai N, Cook NR. Development and validation of improved algorithms for the assessment of global cardiovascular risk in women: the Reynolds Risk Score. Jama. 2007;297(6):611–9. Epub 2007/02/15. pmid:17299196.
- 16. Ridker PM, Paynter NP, Rifai N, Gaziano JM, Cook NR. C-reactive protein and parental history improve global cardiovascular risk prediction: the Reynolds Risk Score for men. Circulation. 2008;118(22):2243–51, 4p following 51. Epub 2008/11/11. pmid:18997194; PubMed Central PMCID: PMCPMC2752381.
- 17. Perk J, De Backer G, Gohlke H, Graham I, Reiner Z, Verschuren M, et al. European Guidelines on cardiovascular disease prevention in clinical practice (version 2012). The Fifth Joint Task Force of the European Society of Cardiology and Other Societies on Cardiovascular Disease Prevention in Clinical Practice (constituted by representatives of nine societies and by invited experts). European heart journal. 2012;33(13):1635–701. Epub 2012/05/05. pmid:22555213.
- 18. Eder L, Wu Y, Chandran V, Cook R, Gladman DD. Incidence and predictors for cardiovascular events in patients with psoriatic arthritis. Annals of the rheumatic diseases. 2016;75(9):1680–6. Epub 2015/10/24. pmid:26493817.
- 19. Puig L. Cardiometabolic Comorbidities in Psoriasis and Psoriatic Arthritis. Int J Mol Sci. 2017;19(1). Epub 2018/01/04. pmid:29295598.
- 20. Eder L, Chandran V, Gladman DD. The Framingham Risk Score underestimates the extent of subclinical atherosclerosis in patients with psoriatic disease. Annals of the rheumatic diseases. 2014;73(11):1990–6. Epub 2013/07/28. pmid:23887287.
- 21. Shen J, Lam SH, Shang Q, Wong CK, Li EK, Wong P, et al. Underestimation of Risk of Carotid Subclinical Atherosclerosis by Cardiovascular Risk Scores in Patients with Psoriatic Arthritis. The Journal of rheumatology. 2017. Epub 2017/11/17. pmid:29142027.
- 22. Martinez-Vidal MP, Fernandez-Carballido C. Is the SCORE chart underestimating the real cardiovascular (CV) risk of patients with psoriatic arthritis? Prevalence of subclinical CV disease detected by carotid ultrasound. Joint Bone Spine. 2017. Epub 2017/07/19. pmid:28716557.