13 Capstone: End-to-End Mini Workflow

This chapter ties everything together: read data → derive ADSL → produce TLFs → render a report.

13.1 Parameters

# You could parametrize paths via YAML; here we keep inline defaults.
dm_path <- "data/sdtm/dm.sas7bdat"
ex_path <- "data/sdtm/ex.sas7bdat"

13.2 1) Read (or Synthesize) SDTM

library(haven); library(dplyr); library(lubridate)


Attaching package: 'dplyr'

The following objects are masked from 'package:stats':

    filter, lag

The following objects are masked from 'package:base':

    intersect, setdiff, setequal, union


Attaching package: 'lubridate'

The following objects are masked from 'package:base':

    date, intersect, setdiff, union

if (file.exists(dm_path)) {
  dm <- read_sas(dm_path)
} else {
  dm <- tibble::tibble(
    STUDYID = "XYZ123",
    USUBJID = sprintf("XYZ-%03d", 1:60),
    ARM = sample(c("Placebo","Active"), 60, replace=TRUE),
    AGE = round(rnorm(60, 60, 8)),
    SEX = sample(c("M","F"), 60, replace=TRUE),
    RANDDT = as.Date("2025-01-15") + sample(0:40, 60, replace=TRUE)
  )
}
if (file.exists(ex_path)) {
  ex <- read_sas(ex_path)
} else {
  ex <- tibble::tibble(
    USUBJID = dm$USUBJID,
    EXSTDTC = dm$RANDDT + sample(0:3, nrow(dm), replace=TRUE)
  )
}

13.3 2) Derive ADSL (Minimal Demo)

adsl <- dm |>
  left_join(ex, by="USUBJID") |>
  mutate(
    TRT01P = ARM,
    TRT01PN = as.integer(factor(ARM, levels=c("Placebo","Active"))),
    TRT01A = TRT01P,
    TRT01AN = TRT01PN,
    SAFFL = "Y",          # demo only; define rules in real life
    FASFL = "Y"
  ) |>
  dplyr::select(STUDYID.x, USUBJID, TRT01P, TRT01PN, TRT01A, TRT01AN, AGE, SEX, EXSTDTC, SAFFL, FASFL)

13.4 3) TLFs

library(gt); library(ggplot2); library(survival)

tbl1 <- adsl |>
  group_by(TRT01P) |>
  summarise(N=n(),
            mean_age = mean(AGE), sd_age = sd(AGE),
            pct_female = mean(SEX=="F")*100)
tbl1_gt <- gt(tbl1) |> tab_header(title="Table 1. Baseline by Treatment")
tbl1_gt

Description of Planned Arm	N	mean_age	sd_age	pct_female
Table 1. Baseline by Treatment
Placebo	226	75.04867	8.503715	60.61947
Screen Failure	52	75.09615	9.699928	69.23077
Xanomeline High Dose	184	74.01087	7.939656	48.36957
Xanomeline Low Dose	181	75.29834	8.277778	60.77348

set.seed(123)
adsl$time <- rexp(nrow(adsl), rate=ifelse(adsl$TRT01P=="Active", 0.08, 0.1))
adsl$status <- rbinom(nrow(adsl), 1, 0.7)
fit <- survfit(Surv(time, status) ~ TRT01P, data=adsl)
# reuse plotting function from prior chapter
ggsurv <- function(fit) {
  ss <- summary(fit)
  dd <- data.frame(time=ss$time, surv=ss$surv, strata=rep(names(fit$strata), fit$strata))
  ggplot(dd, aes(x=time, y=surv, linetype=strata)) + geom_step() + theme_minimal() +
    labs(title="KM Curve (Toy)", x="Time", y="Survival", linetype="Treatment")
}
#ggsurv(fit)

13.5 4) Save Outputs

# Example: Save Table 1 as PNG
#gtsave(tbl1_gt, "tlf-table1.png")

Challenge: Convert this chapter into a parameterized report (e.g., treatment subset or different cohort) and render multiple outputs.

--- title: "Capstone: End-to-End Mini Workflow" --- This chapter ties everything together: **read data → derive ADSL → produce TLFs → render a report.** ## Parameters ```{r} # You could parametrize paths via YAML; here we keep inline defaults. dm_path <- "data/sdtm/dm.sas7bdat" ex_path <- "data/sdtm/ex.sas7bdat" ``` ## 1) Read (or Synthesize) SDTM ```{r} library(haven); library(dplyr); library(lubridate) if (file.exists(dm_path)) { dm <- read_sas(dm_path) } else { dm <- tibble::tibble( STUDYID = "XYZ123", USUBJID = sprintf("XYZ-%03d", 1:60), ARM = sample(c("Placebo","Active"), 60, replace=TRUE), AGE = round(rnorm(60, 60, 8)), SEX = sample(c("M","F"), 60, replace=TRUE), RANDDT = as.Date("2025-01-15") + sample(0:40, 60, replace=TRUE) ) } if (file.exists(ex_path)) { ex <- read_sas(ex_path) } else { ex <- tibble::tibble( USUBJID = dm$USUBJID, EXSTDTC = dm$RANDDT + sample(0:3, nrow(dm), replace=TRUE) ) } ``` ## 2) Derive ADSL (Minimal Demo) ```{r} adsl <- dm |> left_join(ex, by="USUBJID") |> mutate( TRT01P = ARM, TRT01PN = as.integer(factor(ARM, levels=c("Placebo","Active"))), TRT01A = TRT01P, TRT01AN = TRT01PN, SAFFL = "Y", # demo only; define rules in real life FASFL = "Y" ) |> dplyr::select(STUDYID.x, USUBJID, TRT01P, TRT01PN, TRT01A, TRT01AN, AGE, SEX, EXSTDTC, SAFFL, FASFL) ``` ## 3) TLFs ```{r} library(gt); library(ggplot2); library(survival) tbl1 <- adsl |> group_by(TRT01P) |> summarise(N=n(), mean_age = mean(AGE), sd_age = sd(AGE), pct_female = mean(SEX=="F")*100) tbl1_gt <- gt(tbl1) |> tab_header(title="Table 1. Baseline by Treatment") tbl1_gt ``` ```{r} set.seed(123) adsl$time <- rexp(nrow(adsl), rate=ifelse(adsl$TRT01P=="Active", 0.08, 0.1)) adsl$status <- rbinom(nrow(adsl), 1, 0.7) fit <- survfit(Surv(time, status) ~ TRT01P, data=adsl) # reuse plotting function from prior chapter ggsurv <- function(fit) { ss <- summary(fit) dd <- data.frame(time=ss$time, surv=ss$surv, strata=rep(names(fit$strata), fit$strata)) ggplot(dd, aes(x=time, y=surv, linetype=strata)) + geom_step() + theme_minimal() + labs(title="KM Curve (Toy)", x="Time", y="Survival", linetype="Treatment") } #ggsurv(fit) ``` ## 4) Save Outputs ```{r} # Example: Save Table 1 as PNG #gtsave(tbl1_gt, "tlf-table1.png") ``` **Challenge:** Convert this chapter into a **parameterized report** (e.g., treatment subset or different cohort) and render multiple outputs.