Sebastian Wesseling | PubMed Health

PubMed2003

Comparison of eight computer programs for receiver-operating characteristic analysis.

BACKGROUND: ROC analysis is widely accepted to assess and compare diagnostic validity of laboratory tests. Within the last few years, many new ROC programs have become available but have not been systematically evaluated. The aim of this study was to assess different ROC programs regarding their ease of use, mathematical correctness, final output, and their compatibility with other graphics programs. METHODS: Eight available programs running under Windows (AccuROC, Analyse-It, CMDT, GraphROC, MedCalc, mROC, ROCKIT, and SPSS) were evaluated. ROC analyses of prostate-specific antigen and related values were performed from a dataset of 928 men with prostate cancer and benign prostatic hyperplasia and corresponding subsets. Criteria such as data input, data output, and correctness and completeness of results were used to evaluate the practicability of the programs. RESULTS: Although the programs produced equivalent results (areas under the curves and their characteristics), we observed deficiencies concerning input of data, processing of the output data, and completeness of the results. Analyse-It, AccuROC, and MedCalc exhibited good performance, but each program had different shortcomings. Only GraphROC could compare curves at a certain sensitivity or specificity cutoff. CONCLUSIONS: Adequate ROC analysis and ROC plotting cannot be performed with a single program. Analyse-It, AccuROC, and MedCalc can be recommended with certain limitations. Further improvements of the programs are necessary.

Area Under Curve↗

PubMed2003

Determination of non-alpha1-antichymotrypsin-complexed prostate-specific antigen as an indirect measurement of free prostate-specific antigen: analytical performance and diagnostic accuracy.

BACKGROUND: A new assay measures prostate-specific antigen (PSA) not complexed to alpha(1)-antichymotrypsin (nACT-PSA) after removing PSA complexed to ACT by use of anti-ACT antibodies. We evaluated nACT-PSA and its ratio to total PSA (tPSA) as alternatives to free PSA (fPSA) and its ratio to tPSA in differentiating prostate cancer (PCa) and benign prostatic hyperplasia (BPH) in patients with tPSA of 2-20 micro g/L. METHODS: PSA in serum of 183 untreated patients with PCa and 132 patients with BPH was measured retrospectively on the chemiluminescence immunoassay analyzer LIAISON(R) (Byk-Sangtec Diagnostica) with the LIAISON tPSA and LIAISON fPSA assays. The nACT-PSA fraction was determined with a prototype assay measuring the residual PSA after precipitation of ACT-PSA with an ACT-precipitating reagent. RESULTS: nACT-PSA was higher than fPSA in samples with fPSA concentrations <1 microg/L but lower in samples with >1 microg/L fPSA. The median ratios of fPSA/tPSA and of nACT-PSA/tPSA were significantly different between patients with BPH and PCa (19.4% vs 12.2% and 17.4% vs 13.0%, respectively). Within the tPSA ranges tested (2-20, 2-10, and 4-10 microg/L), areas under the ROC curves for the fPSA/tPSA ratios were significantly larger than those for nACT-PSA/tPSA. In the tPSA ranges <10 microg/L, the areas under the ROC curves for fPSA/tPSA were significantly larger than those for tPSA, whereas the areas for nACT-PSA/tPSA were not. At decision limits for 95% sensitivity and specificity, both ratios significantly increased specificity and sensitivity, respectively, compared with tPSA, but the fPSA/tPSA ratio showed higher values. CONCLUSIONS: nACT-PSA and its ratio to tPSA provide lower diagnostic sensitivity and specificity than fPSA/tPSA. The fPSA/tPSA ratio represents the state-of-the-art method for differentiating between PCa and BPH.

Aged↗