In professorbeautiful/NNTbiomarkerHome: Calculate Design Parameters for Biomarker Validation Studies

r input$biomarkerReportTitle

---

Step 1. r stepsTableInitial[1, "SteppingStone"]:

• r stepsTableInitial[1,"Question"]

r ClinicalScenario.

Step 2. r stepsTableInitial[2, "SteppingStone"] :

• r stepsTableInitial[2,"Question"]

The biomarker is intended to help r input$who, by identifying who should receive r input$Option_Treat and who should receive r input$Option_Wait.

Step 3. r stepsTableInitial[3, "SteppingStone"]:

• r stepsTableInitial[3,"Question"]

Currently, the proportion of patients who should receive prevalence is number needed to treat to help one (NNT) is The biomarker test will be useful can create a clinical consensus supporting using the test for clinical decisions. if the NNT among test-positive patients, NNT_Pos, is less than NNT_Lower = r NNTlower, and if the NNT among test-negative patients, NNT_Neg, is greater than NNT_Upper = r NNTupper.

Therefore we choose targets NNT_Pos = r input$NNTpos and NNT_Neg = r input$NNTneg. This performance should suffice to create a clinical consensus supporting using the test for clinical decisions. These values correspond to positive predictive value = PPV = r 1/input$NNTpos, and negative predictive value = NPV = r 1-1/input$NNTneg.

Step 4. r stepsTableInitial[4, "SteppingStone"]:

• r stepsTableInitial[4,"Question"]

If the biomarker test achieves these predictive values, the benefit to patients will be r input$SpecificBenefit.

Step 5. r stepsTableInitial[5, "SteppingStone"]:

• r stepsTableInitial[5, "Question"]

The retrospective study will recruit r input$NpatientsProspective patients. r input$follow_up. If the test divides the r input$NpatientsProspective patients into roughly r input$percentPositive% positive and r 100 - input$percentPositive% negative, and if the estimates match the hoped-for values NNT_Pos = r input$NNTpos and NNT_Neg = r input$NNTneg, then the confidence intervals would be (r round(digits=3, rValues$PPV_ProspectiveInterval)) for PPV, and (r round(digits=3, rValues$NPV_ProspectiveInterval)) for NPV, or equivalently (r round(digits=3, rValues$NNTpos_ProspectiveInterval)) for NNT_Pos, and (r round(digits=3, rValues$NNTneg_ProspectiveInterval)) for NNT_Neg.

r input$ProspectiveStudyNotes

Step 6. r stepsTableInitial[6, "SteppingStone"]:

• r stepsTableInitial[6,"Question"]

The proportion of patients r input$BestToTreatDescription is assumed to be r round(100*input$prevalence)%. Combining that with the target PPV and NPV, the required sensitivity (SN) and specificity (SP) are r rValues$sensitivityPercent% and r rValues$specificityPercent%, respectively (contra-Bayes Theorem). To get a sense of the accuracy of anticipated estimates in the retrospective (case/control) portion of the study, we consider anticipated results for samples sizes r input$samplesizeCases cases and r input$samplesizeControls controls. For example, if the estimates SN = r round(input$samplesizeCases* rValues$sensitivity)/r input$samplesizeCases = r round(100*input$samplesizeCases* rValues$sensitivity/input$samplesizeCases) % and SP = r round(input$samplesizeControls* rValues$specificity)/r input$samplesizeControls = r round(100*input$samplesizeControls* rValues$specificity/input$samplesizeControls) % are observed, then the corresponding confidence intervals will be (r round(digits=3, rValues$Se_RetrospectiveInterval)) for SN, and (r round(digits=3, rValues$Sp_RetrospectiveInterval)) for SP. For NNT_Pos and for NNT_Neg, the Bayes predictive intervals will be (r round(digits=3, rValues$NNTpos_RetrospectiveInterval)) for NNT_Pos , and (r round(digits=3, rValues$NNTneg_RetrospectiveInterval)) for NNT_Neg . (These predictive intervals derive from assuming independent Jeffreys priors for SN and SP, sampling from joint independent posteriors incorporating the anticipated results, and applying Bayes theorem).

professorbeautiful/NNTbiomarkerHome documentation built on June 7, 2019, 8:28 a.m.