Correlation Between Skin and Affected Organs in 52 Sclerodermic Patients Followed in a Diseases Management Team: Development of a Risk Prediction Model of Organ-Specific Complications

Abstract

Objective: To identify the existence of a correlation among the various organs affected, focusing primarily on immuno-dermatological aspects, and to create a risk prediction model of organ-specific complications.

Material and methods: Fifty-two patients with stable scleroderma, followed between 2015 and 2019, were investigated through an extensive multidisciplinary evaluation in the last year.

Results: Patients with lung involvement presented a worse degree of skin fibrosis than patients without it (p <0.001). No relationship was observed for the heart, kidney, and esophagus. Patients with pulmonary involvement had a lower pressure of the low esophagus sphincter and a higher Warrick score than patients without it (p <0.05). Age was significantly higher in patients with kidney involvement. Diffuse scleroderma patients had a worse pulmonary impairment than limited scleroderma patients (p <0.05). The manometric "sclerodermic" pattern was observed to be the most frequent (55.6%, p <0.05) in dcSSc patients while the sclerodermic and normal pattern were equally represented (41.2 and 32.4% respectively, p <0.05) in lcSSc patients. When compared to the negative serological groups, anti-Scl-70 positive patients presented a worse lung involvement while anti-centromere patients presented a better lung outcome (p <0.05). PM-Scl 100/75 positive patients presented mostly a pulmonary fibrotic pattern (p <0.05) and, also, heart complications were more likely associated with anti PM-Scl 100/75 positivity (p <0.05). The risk prediction model for organ-specific complications had an accuracy of 84.4% (95%CI 78, 89) in complication-site prediction, AUC of 0.871, 86% of sensitivity, and 83% of specificity, Cohen's Kappa (k) of 0.68.

Conclusions: Out of all the organs studied, the skin is the one that correlates with the lung. Patients with a diffuse form of disease presented more frequently the anti Scl-70 antibody and had a worse lung and esophageal involvement (scleroderma pattern) than the negative group. Conversely, patients with limited disease presented all positive for the anti-centromere antibody with a better lung involvement than the negative group, without any difference among the esophageal manometric pattern. Anti PM-Scl 100/75 antibody patients were associated with pulmonary fibrosis and presented cardiac involvement. The model created has demonstrated excellent values of sensitivity, specificity, and accuracy, but further studies are needed for validation.

Keywords: mRSS (Rodnan Score); organ complications; predictive model; scleroderma; systemic sclerosis.

Figure 1

Box plot of the distribution of instrumental parameters by the occurrence of heart, esophagus, kidney and pulmonary involvement. *p < 0.05.

Figure 2

Venn diagram showing the distribution and overlap of organ involvement in the 52 scleroderma patients studied. In dark red color is represented the most common association. The color fades progressively as the association decreases. The majority of patients (27.08%) presented both esophageal and pulmonary involvement while only 1 patient presented all organ involvement. No patient showed only cardiac involvement or kidney–heart or kidney–esophagus involvement.

Figure 3

evaluation of the model’s predictive capacity: ROC curve of all the four organs investigated (left side) and overall (right side): for each organ investigated the AUC of esophagus, heart, lung, and kidney were 0.720, 0.838, 0.809 and 0.84 respectively. Combining the four site-specific probability, the final model presents an AUC of 0.871 with an accuracy of 84.4% (95%CI 78, 89), 86% of sensitivity and 83% of specificity. Cohen’s Kappa (k) of 0.68. Legend: AUC = 0.5: no discrimination; AUC: 0.7 to 0.8: acceptable accuracy; AUC: 0.8 to 0.9 excellent accuracy; AUC >0.9 outstanding.

Figure 4

Overview of the prediction model goodness of fit by organ involvement. For each organ investigated are represented the correct (blue circle) and wrong predictions (red x). The data refer to those reported in Table 4 . For the heart, the model correctly predicted the absence of involvement in all 41 patients and the presence of involvement in seven patients, failing only two cases. For the esophagus, the model predicted the absence of complication in 16 patients and was wrong in seven cases, while it predicted the absence of involvement in 32 cases but was wrong four times. For the lung, the model predicted the absence of involvement in 14 cases and was wrong five times while it predicted the presence of involvement in 34 patients and was wrong five times. Finally, for the kidney, the model predicted the absence of involvement in 35 patients and was wrong three times while it predicted the presence of involvement in 13 patients and was wrong four times.