Read the results

Start with the README in the returned delivery directory. It links the actual reports and states what completed. Then open the requested endpoint workbook.

Choose the workbook

File	Use
01_ORR_pathway_first_report.xlsx	Response pathway and gene results
02_Survival_pathway_first_report.xlsx	Survival pathway and gene results
03_Integrated_review_report.xlsx	Cross-endpoint pathway/gene review
01_ORR_exposure_first_report.xlsx	Response results for named features
02_Survival_exposure_first_report.xlsx	Survival results for named features
03_Exposure_integrated_review_report.xlsx	Cross-endpoint exposure review

Only requested, available endpoints should supply results. An absent endpoint or empty placeholder is not a reason to add an unrequested analysis. If Excel writers are unavailable, inspect the reported CSV fallback.

Read these sheets first

1. 00_READ_FIRST and 02_Parameter_Settings: scope, thresholds and settings.
2. Endpoint statistics and method details: actual numbers, method and scale.
3. Diagnostics, sensitivity and skipped/failed rows: limitations and exclusions.
4. Column_Dictionary and the method/claim guides: interpretation of each field.
5. Integrated review: compare endpoints after reading their separate results.

Pathway-first gene reports use Gene_Stats, Gene_Method_Details, Gene_Explain and Gene_Full_Audit. Exposure-first uses the corresponding Exposure_* sheets. Keep the full audit tables when sharing a compact display.

Read the effect scale

Scale	Null	Meaning
OR / RR	1	Relative odds / risk for the stated contrast
Log-OR	0	Log odds ratio; exponentiate estimate and CI together for OR
HR	1	Relative event hazard under the stated Cox summary
Log-HR	0	Log hazard ratio; exponentiate estimate and CI together for HR
RiskDiff	0	Absolute difference in outcome probability
MeanDiff	0	Difference in the original outcome units
RMST_Diff	0	Difference in restricted mean survival up to the specified horizon
SurvDiff	0	Difference in survival probability at the specified horizon

Read the contrast too: per unit, per SD, two doses and intervention policies answer different questions. HR is not a probability ratio or an absolute gain in survival. Screening ORs and formal risk differences are different results.

Report a result

Give the cohort, endpoint coding, contrast, covariates, method, scale, estimate, CI and actual P/FDR family. Add the support level and the main diagnostic finding. Association, screening, statistical mediation and functional mechanism evidence are distinct claims. Method agreement or a small P value does not prove causality.

FDR is adjusted over the requested family, not just rows selected for display. Retain failed or unestimable rows in the scope accounting. Continue with diagnostics before choosing headline results.