The correlation between the clinical-laboratory findings and the histopathological changes of renal glomeruli in forty diabetics of which renal biopsies were performed on 14 cases and autopsies on the remaining 26 cases was studied by the multivariate analysis such as principal component analysis, multivariate regression analysis and canonical correlation analysis.
I have proposed the grading system for the statistical analysis of pathological changes of the renal glomeruli. The abnormal morphological changes such as diffuse lesion, nodular lesion, and arteriolosclerotic changes were picked up in each photographic picture and points were given according to their severity.
Fourteen variables of clinical-laboratory findings such as age of patient, duration of diabetes mellitus, urinary protein, PSP test, edema, BUN, serum total protein, serum α
2-globulin fraction, serum A/G ratio, serum cholesterol, fasting blood sugar, retinal findings, systolic blood pressure and diastolic blood pressure were adopted.
Calculation for multivariate analysis was carried out with the computer NEAC 2200 model 500.Result:
1) Principal component analysis revealed that nine clinical variables (urinary protein, FSP test, edema, BUN, serum total protein, serum α
2-globulin fraction, serum A/G ratio, retinal findings and systolic blood pressure) had close relation to pathological changes of diabetic glomerulosclerosis.
2) The clinical-laboratory score of each patient in order to estimate the pathological changes of diffuse lesion, nodular lesion, and arteriolosclerotic changes respectively was calculated by multiple regression analysis. Multiple correlation ccefficient was.841 (p<.001) for a diffuse lesion, .904 (p<.001) for a nodular lesion, and. 752 (p<.05) for a arteriolosclerotic change respectively.
3) The canonical analysis for predicting pathological changes of renal glomeruli composed the following linear combination formula:
Clinical score=-(.046×age)-(.023×duration) + (.316×urinary protein)-(.717×PSP)-(.099×edema)-(.185×BUN)-(.238×serum total protein) + (.206×serum A/G ratio)-(.161×serum a2-globulin fraction) + (.054×FBS) + (.010×serum cholesterol) + (.323×retinal finiding) + (.114×systolic blood pressure)-(.131×diastolic blood pressure) Pathological score= (.343×nodular lesion) + (.791×diffuse lsion)-(.086×arteriolosclerotic change)
Pathological changes were found to be separated into the following three groups with the agreement of 92.5%:(1) slight to moderate diffuse lesion only, (2) diffuse lesion and slight nodular lesion, and (3) marked diffuse and nodular lesion.
These results show that multivariate analysis is useful in predicting the pathological changes of diabetic glomerulosclerosis and the prognosis of diabetics with nephropathy.
View full abstract