Comprehensive Survival Power Analysis (Expert)

Comprehensive power analysis and sample size calculation for survival studies and clinical trials. Unifies all survival power methods: standard log-rank tests, Cox regression, competing risks, RMST-based analyses, non-inferiority trials, and specialized methods. Supports both simple and complex study designs with accrual periods, variable follow-up, and dropout considerations. Essential for study planning, grant applications, and regulatory submissions.

Usage

comprehensiveSurvivalPower(
  data,
  method_category = "standard",
  calculation_type = "sample_size",
  statistical_method = "log_rank_schoenfeld",
  study_design = "simple",
  alpha = 0.05,
  power = 0.8,
  sample_size = 200,
  allocation_ratio = 1,
  hazard_ratio = 0.75,
  median_survival_control = 12,
  median_survival_treatment = 16,
  accrual_period = 24,
  follow_up_period = 36,
  dropout_rate = 0.05,
  event_rate = 0.6,
  competing_event_rate = 0.2,
  cumulative_incidence_control = 0.4,
  cumulative_incidence_treatment = 0.3,
  rmst_timepoint = 24,
  non_inferiority_margin = 1.25,
  superiority_test = FALSE,
  snp_maf = 0.2,
  genetic_model = "additive",
  number_of_looks = 1,
  spending_function = "lan_demets_obf",
  show_detailed_output = TRUE,
  show_sensitivity_analysis = FALSE,
  show_power_curves = FALSE,
  export_study_design = FALSE,
  genetic_effect_size = 1.5,
  cure_rate_control = 0,
  cure_rate_treatment = 0,
  survival_distribution = "exponential",
  accrual_distribution = "uniform",
  interim_futility_boundary = 0.2,
  binding_futility = FALSE,
  number_of_covariates = 2,
  covariate_correlation = 0,
  interaction_effect_size = 1,
  covariate_distribution = "normal",
  adjust_for_confounders = TRUE,
  show_method_comparison = FALSE
)

Arguments

data

The data as a data frame (optional for power calculations).

method_category

Choose the category of survival analysis method. Standard methods include log-rank tests and Cox regression. Competing risks methods analyze multiple event types. Advanced methods include RMST, non-inferiority, and specialized approaches.

calculation_type

Select what parameter to calculate. Most commonly used are sample size (given desired power) and power (given available sample size).

statistical_method

Choose the specific statistical method. Available options depend on the selected method category. Schoenfeld method is most common for log-rank tests.

study_design

Simple design assumes fixed follow-up time. Complex design includes accrual period and variable follow-up. Sequential designs allow interim analyses.

alpha

Type I error rate. Standard values: 0.05 (two-sided) or 0.025 (one-sided). For non-inferiority trials, consider 0.025.

power

Statistical power (probability of detecting true effect). Standard values: 0.80 (80 percent) or 0.90 (90 percent). Higher power requires larger sample sizes.

sample_size

Total number of subjects in the study. Used when calculating power or detectable effect size from a fixed sample size.

allocation_ratio

Ratio of control to treatment group sizes. 1.0 = equal allocation (1:1), 2.0 = twice as many controls (2:1). Equal allocation is most efficient.

hazard_ratio

Expected hazard ratio. Values < 1 indicate treatment benefit. Common values: 0.75 (25 percent reduction), 0.67 (33 percent reduction), 0.50 (50 percent reduction).

median_survival_control

Expected median survival time in control group. Units should match your study timeline (months, years, etc.).

median_survival_treatment

Expected median survival time in treatment group. Should be longer than control group for beneficial treatments.

accrual_period

Time period for patient recruitment (months/years). Longer accrual periods allow for smaller sample sizes but extend study duration.

follow_up_period

Additional follow-up time after accrual completion. Total study duration = accrual period + follow-up period.

dropout_rate

Annual rate of loss to follow-up (0.05 = 5 percent per year). Higher dropout rates require larger sample sizes to maintain power.

event_rate

Proportion of subjects expected to experience the event during study period. Used for simple design calculations.

competing_event_rate

Proportion experiencing competing events (for competing risks analysis). Should be less than primary event rate.

cumulative_incidence_control

Expected cumulative incidence of primary event in control group (for competing risks analysis).

cumulative_incidence_treatment

Expected cumulative incidence of primary event in treatment group (for competing risks analysis).

rmst_timepoint

Time point for restricted mean survival time analysis. Should be clinically meaningful and within study follow-up period.

non_inferiority_margin

Non-inferiority margin for hazard ratio. 1.25 means treatment is non-inferior if HR < 1.25 (25 percent increase in hazard is acceptable).

superiority_test

After establishing non-inferiority, test for superiority. Requires hierarchical testing approach to control Type I error.

snp_maf

Minor allele frequency for SNP-based survival analysis. Common range: 0.05-0.45.

genetic_model

Genetic inheritance model for SNP analysis. Additive model is most common.

number_of_looks

Number of planned interim analyses for sequential designs. More looks require larger sample sizes to maintain overall Type I error.

spending_function

Method for spending Type I error across interim analyses. O'Brien-Fleming type preserves most alpha for final analysis.

show_detailed_output

Display detailed calculation steps and intermediate results.

show_sensitivity_analysis

Perform sensitivity analysis varying key parameters around their specified values (+/- 10 percent, 20 percent).

show_power_curves

Generate plots showing power vs sample size, effect size, or other parameters.

export_study_design

Generate a comprehensive study design document suitable for protocols and grant applications.

genetic_effect_size

Genetic effect size for SNP-based survival analysis. Represents the hazard ratio per copy of the risk allele.

cure_rate_control

Proportion of subjects in control group who will never experience the event (cure fraction). 0 = no cure, standard survival model.

cure_rate_treatment

Proportion of subjects in treatment group who will never experience the event (cure fraction). Usually higher than control group in effective treatments.

survival_distribution

Assumed underlying survival distribution. Exponential assumes constant hazard; Weibull allows increasing/decreasing hazard; Piecewise allows complex patterns.

accrual_distribution

Pattern of patient enrollment over time. Uniform = constant rate; Increasing/Decreasing = linear change; Exponential = curved pattern.

interim_futility_boundary

Futility boundary for interim analyses in sequential designs. Probability threshold below which study may be stopped for futility.

binding_futility

Whether futility boundaries are binding (must stop if crossed) or non-binding (advisory only).

number_of_covariates

Number of covariates to include in the Cox regression model. More covariates generally require larger sample sizes.

covariate_correlation

Correlation between covariates in the model. Higher correlation reduces effective sample size and increases required sample size.

interaction_effect_size

Effect size for interaction between main exposure and other covariates. Used when testing interaction effects.

covariate_distribution

Distribution type for covariates in the model. Affects power calculations for epidemiological studies.

adjust_for_confounders

Whether to account for confounding variables in power calculations. Generally recommended for epidemiological studies.

show_method_comparison

Compare results across multiple power calculation methods to assess robustness of sample size estimates.

Value

A results object containing:

results$instructions					a html
results$powerResults					a table
results$methodDetails					a table
results$studyDesign					a html
results$assumptions					a table
results$sensitivityTable					a table
results$powerCurvePlot					an image
results$effectSizePlot					an image
results$methodComparison					a table
results$competingRisksDetails					a table
results$sequentialDesign					a table
results$clinicalInterpretation					a html
results$protocolSummary					a html
results$regulatoryNotes					a html
results$technicalDetails					a html

Tables can be converted to data frames with asDF or as.data.frame. For example:

results$powerResults$asDF

as.data.frame(results$powerResults)