MRMC design, VOPT design VOPT, MRMC

 

Welcome ! Deutsche Version Statistical services Design and analysis SAS® programming Statistical consulting Statistical training ROC curve ROC software tools Topics: Clin. Laboratory Method validation Passing-Bablok-regression Topics: Software Download Web-Shop ACOMED statistics contact Presentation Publications References About

 

ROC curve, ROC curves, ROC analysis

You are at: Diagnostics - statistical methods

ACOMED statistics - service provider for diagnostic studies

ACOMED statistics is specialized in design and analysis of diagnostic studies for evaluation of performance of diagnostic tests. Calculation of sensitivity and specificity using a 2x2 cross table seems to be so easy, however, diagnostic studies are more complex.

These links might be of interest for you:

Excel-Tool: Sensitivity, Specificity, PPV, NPV for 2x2 tables, with confidence intervals

Special report: Measure of diagnostic accuracies in screening trials

Statistical aspects of diagnostic studies

  • Defining diagnostic pathway, including reality (Intended use, clinical application, clinical situation)
  • Definition of diagnostic accuracs criterion / reference method (gold standard)
  • Design of the study - which phase (I, II, III, IV) should be chosen?
  • Aspects in case of screening trials
  • Special designs of diagnostic studies - e.g. multi-reader-multi-case (MRMC), verification of only positively tested (VOPT)
  • Inclusion and exclusion criteria: "case control" (known disease) vs. "cohort" (suspected to have disease)
  • Sample size estimation (internal tools, software Medcalc®, PASS)
  • Formulation of joined null hypotheses for binary measures
  • Criteria for cut-off determination
  • Biases due to selection, non-perfect reference method, missing diagnostic accuracy criterion, sub-groub-analyses, inclusion criteria
  • Design of data entry, data management, data validation, data transfer
  • Handling of missing values, outliers, erroneous values
  • Documentation, Design of CRF
  • Analysis of ROC curve, cut-off-determination

Statistical analyses (Binary estimates, ROC curve, partial AUC, cut-off etc.)

  • Estimation of binary measures (sensitivity, specificity, predictive values, diagnostic likelihood ratios)
  • Estimation of ROC curve, estimation of AUC, partial ROC, partial AUC
  • Comparison of sensitivity / specificity (relative measures rTPF, rFPF, McNemar-Tests)
  • Comparison of ROC-curves
  • Cut-off estimation
  • Application of generalized linear models (GLM) for evaluation of influence of covariates on diagnostic accuracy measures (regression framework)

Publication according STARD statement

  • E.g. Usage of only validated methods
  • E.g. how participants were recruited?
  • E.g. Reporting of results with confidence intervals
  • E.g. Handling of missing values, outliers etc.
  • E.g. Discuss clinical applicability of findings

All these analyses are conducted according to an internal GCP compliant SOP-system.

Statistical analyses are performed using following software: SAS¬ (if requested, also in Medcalc®, SPSS® 15.0/18.0, NCSS and R.)

Using SAS®, ACOMED statistics can provide analyses performed under regulated conditions (complete traceability, formal exclusion of any manipulation of data and output).

Further information about ROC analysis and related tools you can obtain from following web-sites about ROC curves (in German) and Excel-tools for ROC analysis.

 
ROC curve, ROC curves, DAC method, ROC analysis
Copyright: ACOMED   
diagnostic studies
Diagnostic trials clinical diagnostic studies diagnostic accuracy