This meta-analysis was registered in PROSPERO (CRD 42020211821), following the PRISMA 2020 checklist.25 The authors searched Pubmed, EMBASE, and Cochrane Library databases, including all studies from the respective inception of these databases to October 14, 2020. The investigator performed the literature retrieval, research selection, data extraction, and quality assessment.

The search strategy included a variety of terms related to psoriasis and organ-based comorbidities. The detailed search strategies of Pubmed, Embase, and Cochrane Library were put in the supplementary material.

According to the PECO strategy, articles that met the following criteria would be included. Population: the authors included studies that contained both psoriasis and the general population; Exposure: psoriasis with regular diagnosis (no age limit); Comparative: the people without psoriasis and from the same research environment as psoriasis; Outcome: the record results were organ-based comorbidities (cardiovascular diseases, kidney diseases, liver diseases, respiratory diseases, spleen diseases, and cerebrovascular system-related diseases). Publications with insufficient and reduplicated data were excluded. The restriction of study type and language were not considered in the initial search.

Full text of articles meeting the eligibility criteria was selected for data extraction. Data was extracted by the first author, year of publication, country, demographic, number of patients, and follow-up time. For example, the authors intended to include studies that were clearly defined as diseases with the kidney, such as chronic kidney disease, etc. In addition, studies that had nothing, but abnormal testing data were excluded. If a study provided data with different types of psoriasis (mild, moderate, and severe), the authors combined them to obtain an estimate of all psoriatic patients. In order to ensure the accuracy of data extraction, any difference found in the data can be addressed by referring to the original article. The Newcastle-Ottawa Quality Assessment Scale was used to assess the quality of included studies, with a maximum score of 9 points, representing the highest quality. The scale assessed the quality of each study in three areas, including (1) Case and control subjects’ recruitment; (2) Comparability between the two groups; (3) Identification of results of interest. The characteristics of the studies and evaluation results were displayed in Table 1.

Statistical analyses were performed by RevMan 5.4 (the Cochrane collaboration) and Stata version 12.0 (StataCorp LP, College Station, TX77845). The results were reflected in pooled RR (pRR) and 95% CI. All statistical tests were bilateral. The p-values of less than 0.05 were considered significant. The Inconsistency test (I²) statistic was used to assess the heterogeneity. The I² value of 0%, 25%, 50%, and 75% indicated no, low, medium, and high heterogeneity, respectively. Sensitivity analyses, subgroup analyses, and meta-regression were performed to identify potential sources of heterogeneity. Besides, the funnel plots and Egger’s weighted regression were used to assess potential publication bias.

The pRRs and corresponding 95% CIs were estimated by a random-effect meta-analysis. A logarithmic transformation was performed for the estimated RRs of maximum adjusted effect size and the CIs. pRRs were subjected to Z-test and sensitivity analysis and used to assess the combined stability effect. Publication bias was measured by constructing funnel plots. A symmetrical inverted funnel indicated that there is no high likelihood of publication bias. Finally, a study was removed to assess its impact on the merger. The similarity of combined RRs before and after removing the study indicated that there was a high certainty in the results. When heterogeneity existed between studies, the authors applied the random-effect model and used the Mantel-Haenszel analysis method.

The search strategy (Fig. 1) produced 7,412 potentially relevant references (3,862 in Pubmed, 2,984 in EMBASE, and 566 in the Cochrane Library). Finally, fifteen retrospective cohort studies remained, containing 216,348 psoriatic patients and 9,896,962 general people. All studies were based on a managed database, relied on diagnostic codes to identify and verify the diagnosis of psoriasis and organ-based comorbidities.

Figure 1.

Flow diagram of study selection.

PRISMA 2020 flow diagram for new systematic reviews which included searches of databases and registers only.

(0.41MB).

The main clinical characteristics and quality evaluation of the included studies were summarized in Table 1. These studies were conducted in seven countries: the United States, the United Kingdom, Germany, Denmark, Italy, Taiwan, and South Korea. The mean age of psoriatic patients ranged from 12.0 to 69.5 years, and females accounted for 32.58% to 59.19%. For the general population, the mean age ranged from 12.0 to 70.6 years, and females accounted from 22.14% to 67.42%. Only two studies did not report the average age,26,27 and one study mainly investigated children with psoriasis.28 So, the span of age characteristics was large after synthesizing data. There was little specific organ-based comorbidities data of race. Therefore, no subgroup analysis was conducted on race.28,29

The risks all included organ-based comorbidities between psoriatic patients and the general population were described in Fig. 2. The authors mainly studied the involvements of the heart, kidneys, lung, brain, and liver. In 15 studies,26–40 they contained 216,348 psoriatic patients and 9,896,962 control subjects. It showed that psoriatic patients had a significantly increased risk of these organ-based comorbidities, and pRR was 1.20 (95% CI 1.11‒1.31, I²=98%, p<0.001). After removing the study involving mainly children,28 pRR became to 1.17 (95% CI 1.08‒1.27, I²=98%, p<0.001).

Figure 2.

Forest plot of the risks of comorbidities in psoriatic patients comparing the general population.

(0.47MB).

In 14 studies,26–34,36–40 summing 210,856 psoriatic patients and 9,891,470 controls. The pRR for cardiovascular events was 1.20 (95% CI 1.11‒1.30, I²=97%, p<0.001), significantly increased comparing the general population. In 9 studies27,29–31,34–36,38,40 involved 135,345 psoriatic patients and 9,699,285 control subjects, pRR for renal diseases was 1.08 (95% CI 0.87‒1.33, I²=83%, p<0.001). In 8 studies26,27,31–33,35,37,40 with 103,300 psoriatic patients and 4,729,963 control subjects, pRR of cerebrovascular disease events was 1.56 (95% CI 1.20‒2.04, I²=94%, p<0.001). In 9 studies27,30–32,35,36,38,40 with 132,791 psoriatic patients and 9,689,831 control subjects, pRR for the occurrence of pulmonary disease events was 1.05 (95% CI 0.89‒1.23, I²=92%, p<0.001). In 5 studies27,28,35,38,40 involved 46,469 psoriatic patients and 131,154 control subjects, pRR for the occurrence of liver diseases was 1.75 (95% CI 1.33‒2.29, I²=88%, p<0.001), and it was remarkably higher than general population (Fig. 3).

Figure 3.

Forest plot of specific comorbidity risk in psoriatic patients compared with the general population (cardiovascular, renal, cerebrovascular, lung, liver diseases).

(1.27MB).

Psoriatic patients had been reported to suffer from two or more comorbidities simultaneously.41–46 Hence, the authors extracted data of two organs from included studies and calculated pRRs (Fig. 4). pRR for renal diseases plus cardiovascular events was 1.09 (95% CI 1.01‒1.18, I²=85%, p<0.001). pRR for renal diseases plus cerebrovascular events was 1.28 (95% CI 0.99‒1.66, I²=87%, p<0.001). pRR for renal diseases plus liver diseases were 1.46 (95% CI 1.10‒1.94, I²=91%, p<0.001). pRR for cardiovascular events plus liver diseases were 1.41 (95% CI 1.11‒1.80, I²=95%, p<0.001).

Figure 4.

Forest plot of coexisting comorbidity risk in psoriatic patients compared with general population.

(0.85MB).

Three studies were subdividing psoriatic patients into mild and moderate/severe subgroups.31,32,36 They included 97,285 patients with mild psoriasis and 15,016 patients with moderate/severe psoriasis in total. pRR of organ-based comorbidity risk was 0.61 (95% CI 0.46‒0.81, I²=94%, p<0.001) for mild psoriasis, which was lower than that of moderate/severe psoriatic patients (Supplementary Fig. S1).

Sensitivity is an important indicator of literature quality and heterogeneity. One study was excluded to estimate the effect of pRRs at a time. There was no obvious change and a significant decrease in heterogeneity. Sensitivity analyses were performed to assess which study provided greater heterogeneity in the data by the Galbraith plot. All studies showed stable and reliable results (Supplementary Fig. S2). After careful evaluation, it was concluded that every study was not sensitive to pRRs and could not be arbitrarily excluded. As the score results were assessed by Newcastle-Ottawa Quality Assessment Scale for all included studies in Table 1. All studies scored at least 6 points, which was in line with the requirements for continuing study. Meanwhile, Stata's Metaninf command was used to assess the impact of a single study. After excluding one study at a time, the point of estimate did not fall out of 95% CI (Supplementary Fig. S3).

The funnel plot was utilized to analyze the publication bias of 15 studies. The outcome of the graph showed a certain degree of publication bias. The authors further used the Egger method to check and quantify the funnel plot. The p-value was checked to determine whether there was publication bias. The T-value in bias (p=0.854) meant that there was no publication bias (Supplementary Fig. S4).

Regarding the quality of all studies, the NOS score ranged from 6 to 8. The authors determined that 6 studies had a low risk of bias, 7 studies had a medium risk of bias, and 2 studies had a high risk of bias (Table 1).

The authors conducted the subgroup analyses, including regions (Western vs. Asian countries), study period (covering 2000 vs. after 2000), and funding (with vs. without). The research region of the three studies was Asia,26,27,34 the rest was the West. The outcomes showed that Asian countries with pRR of 1.09 (95% CI 0.95‒1.24, I²=98%, p<0.001) and 1.24 (95% CI 1.12‒1.37, I²=98%, p<0.001) for Western countries. The I² was 55.5% and p-value was 0.13 between the subgroups (Supplementary Fig. S5).

Two studies had not clearly pointed out the study duration,38,40 and the rest studies were divided into two groups for further discussion. Four researches’ periods were over the year of 2000,31,33,34,36 and pRR was 1.04 (95% CI 0.89‒1.21, I²=97%, p<0.001). pRR of the rest researches (after 2000) was 1.31 (95% CI 1.17‒1.47, I²=99%, p<0.001). The I² was 82.7% and the p-value was 0.02 between the subgroups (Supplementary Fig. S6).

There was some research that had no funding,27,32 while the rest of the studies were funded. The analysis showed that pRR of unfunded research was 1.33 (95% CI 1.10‒1.62, I²=99%, p<0.001), and pRR of funded ones was 1.18 (95% CI 1.09‒1.29, I²=98%, p<0.001), The I² was 18.7% and p-value was 0.27 between the subgroups (Supplementary Fig. S7).

These results of subgroup analyses, including regions (Western vs. Asian countries), study period (covering 2000 vs. after 2000), and funding (with vs. without), had not provided any specific source of heterogeneity.

According to the characteristics of included studies, new concomitant variables were generated after dummy variables were assigned according to regions (Western vs. Asian countries), study period (covering 2000 vs. after 2000), and funding (with vs. without). The logarithmic LogRR of each study effect indicator RR was the dependent variable. The meta regressions were all carried out (p>0.05), but there was no source of heterogeneity.