OBJECTIVE

To assess the possibilities of polynomial quantile regression analysis in prediction of time till psoriatic arthritis onset depending on age of manifestation of moderate-to-severe psoriasis.

MATERIAL AND METHODS

We analyzed the data of psoriasis patient registry of the Russian Society of Dermatovenereologists and Cosmetologists. Distribution normality was tested using the Shapiro—Wilk test. Data heteroscedasticity was assessed using scatter diagrams. In accordance with quantile definition, we determined 95%, 75%, 50% and 25% quantiles of time to development of psoriatic arthritis. The bootstrap method was used to obtain 95% confidence intervals (CI) for quantile values. Lines of 95%, 75%, 50% and 25% quantiles were obtained by polynomial quantile regression. To test the assumption about admissibility and expediency of polynomial quantile regression, we compared 95% CI and appropriate quantile regression lines.

RESULTS

There were 760 patients. We found no normal distribution in all ranges of age of psoriasis onset. The absence of homoscedasticity and normal distribution of data ruled out parametric statistical analysis and linear regression analysis. Comparing the positions of 95% CI of the calculated subrange values of 95%, 75%, 50% and 25% quantiles for time till psoriatic arthritis onset and quantile regression lines for the same quantiles, we found that parabolic quantile regression data are in good agreement with calculation of subrange quantiles.

CONCLUSION

Polynomial quantile regression is advisable to predict time till psoriatic arthritis onset depending on age of onset of moderate-to-severe psoriasis.