Merge pull request #12 from Novartis/10-jsm2024

index.qmd

@@ -13,6 +13,17 @@ development problems. In order to highlight the versatility of that package,
 we have developed a collection of case studies covering a diverse set of 
 clinical questions that were addressed via Bayesian modelling with the `brms` package. 
 
+::: {.content-visible when-profile="public"}
+::: {.callout-tip}
+## Short course at JSM -- Portland, OR -- August 5th 2024
+
+We will offer a short course based on this material at the 2024 Joint Statistical Meetings in Portland on August 5. The materials are available here:
+
+- [Slides](workshops/jsm2024/slides/bamdd_jsm2024.pdf) & [additional material (see readme)](https://github.com/Novartis/bamdd/tree/main/workshops/jsm2024)
+
+:::
+:::
+
 ## Navigating the site
 
 The material is organized as follows:
@@ -26,6 +37,7 @@ studies. You can find a [listing with descriptions on this page](src/02_case_stu
 as exposing and injecting custom stan code to `brms` models,
 and efficient parallel computation to support sampling in `brms`.
 
+
 ## Updates
 
 This web-site is intended as a live document with updates as
@@ -34,6 +46,7 @@ appropiate. Key changes to the web-site are tracked here:
 
 
 ::: {.content-visible when-profile="public"}
+- 31st July 2024: Linked materials from JSM Portland 2024 short course
 - 5th June 2024: Added @sec-pos on assessing probability of success of
   in Phase-III based on a single-arm trial in Phase-II
 - 29th April 2024: Third edition course from Paul Bürkner with the new

src/01b_basic_workflow.qmd

@@ -91,7 +91,8 @@ set.seed(5886935)
 #| eval: true 
 #| include: false
 # invisible to the reader additional setup steps, which are optional
-{{< include setup.R >}}
+# {{< include setup.R >}}
+source("setup.R")
 ```
 
 As `brms` models are translated to Stan model files, which must be

src/01c_priors.qmd

@@ -9,7 +9,10 @@ bibliography: references.bib
 
 # Model setup & priors {#sec-model-priors}
 
-```{r, include=FALSE}
+```{r, eval=TRUE,echo=TRUE,message=FALSE,warning=FALSE}
+#| code-fold: true
+#| code-summary: "Show R setup"
+
 library(ggplot2)
 library(dplyr)
 library(knitr)
@@ -31,7 +34,8 @@ theme_set(theme_bw(12))
 
 ```{r, include=FALSE, echo=FALSE, eval=TRUE}
 # invisible to the reader additional setup steps, which are optional
-{{< include setup.R >}}
+# {{< include setup.R >}}
+source("setup.R")
 ```
 
 Within the Bayesian regression modeling in Stan framework `brms`

src/02a_meta_analysis.qmd

@@ -46,7 +46,8 @@ set.seed(593467)
 
 ```{r, include=FALSE, echo=FALSE, eval=TRUE}
 # invisible to the reader additional setup steps, which are optional
-{{< include setup.R >}}
+# {{< include setup.R >}}
+source("setup.R")
 ```
 
 ## Background

src/02ab_meta_analysis_trtdiff.qmd

@@ -24,7 +24,8 @@ set.seed(979356)
 
 ```{r, include=FALSE, echo=FALSE, eval=TRUE}
 # invisible to the reader additional setup steps, which are optional
-{{< include setup.R >}}
+# {{< include setup.R >}}
+source("setup.R")
 ```
 
 ## Background

src/02ac_meta_analysis_strata.qmd

@@ -56,7 +56,8 @@ options(width=120, digits=2)
 
 ```{r, include=FALSE, echo=FALSE, eval=TRUE}
 # invisible to the reader additional setup steps, which are optional
-{{< include setup.R >}}
+# {{< include setup.R >}}
+source("setup.R")
 ```
 
 ## Background

src/02b_dose_finding.qmd

@@ -52,7 +52,8 @@ set.seed(8979476)
 
 ```{r, include=FALSE, echo=FALSE, eval=TRUE}
 # invisible to the reader additional setup steps, which are optional
-{{< include setup.R >}}
+# {{< include setup.R >}}
+source("setup.R")
 ```
 
 ```{r get_annualized_rates}

src/02c_dose_escalation.qmd

@@ -72,7 +72,8 @@ set.seed(8794567)
 
 ```{r, include=FALSE, echo=FALSE, eval=TRUE}
 # invisible to the reader additional setup steps, which are optional
-{{< include setup.R >}}
+# {{< include setup.R >}}
+source("setup.R")
 ```
 
 Such a model is straightforward to implement in `brms`. Below, we use a nonlinear formula specification, in order to allow the prior for the intercept to be specified on the log scale. 

src/02cb_tte_dose_escalation.qmd

@@ -144,7 +144,8 @@ gt_format <- function(x) {
 
 ```{r, include=FALSE, echo=FALSE, eval=TRUE}
 # invisible to the reader additional setup steps, which are optional
-{{< include setup.R >}}
+# {{< include setup.R >}}
+source("setup.R")
 ```
 
 ## Example trial

src/02e_multiple_imputation.qmd

@@ -76,7 +76,8 @@ options(brms.normalize=FALSE)
 
 ```{r, include=FALSE, echo=FALSE, eval=TRUE}
 # invisible to the reader additional setup steps, which are optional
-{{< include setup.R >}}
+# {{< include setup.R >}}
+source("setup.R")
 ```
 
 Let us consider some simulated data for a trial in chronic obstructive

src/02g_longitudinal.qmd

@@ -45,7 +45,8 @@ control_args <- list(adapt_delta = 0.95)
 
 ```{r, include=FALSE, echo=FALSE, eval=TRUE}
 # invisible to the reader additional setup steps, which are optional
-{{< include setup.R >}}
+# {{< include setup.R >}}
+source("setup.R")
 ```
 
 ```{r get_data, echo = TRUE}

src/02h_mmrm.qmd

@@ -51,7 +51,8 @@ options(width=120)
 
 ```{r, include=FALSE, echo=FALSE, eval=TRUE}
 # invisible to the reader additional setup steps, which are optional
-{{< include setup.R >}}
+# {{< include setup.R >}}
+source("setup.R")
 ```
 
 ```{r include=FALSE}

src/02i_time_to_event.qmd

@@ -58,7 +58,8 @@ options(width=120)
 
 ```{r, include=FALSE, echo=FALSE, eval=TRUE}
 # invisible to the reader additional setup steps, which are optional
-{{< include setup.R >}}
+# {{< include setup.R >}}
+source("setup.R")
 ```
 
 

src/02j_network_meta_analysis.qmd

@@ -38,7 +38,8 @@ control_args <- list(adapt_delta=0.95)
 
 ```{r, include=FALSE, echo=FALSE, eval=TRUE}
 # invisible to the reader additional setup steps, which are optional
-{{< include setup.R >}}
+# {{< include setup.R >}}
+source("setup.R")
 ```
 
 

src/02l_single_arm_pos.qmd

@@ -47,7 +47,8 @@ theme_set(theme_bw(base_size=18))
 
 ```{r, include=FALSE, echo=FALSE, eval=TRUE}
 # invisible to the reader additional setup steps, which are optional
-{{< include setup.R >}}
+# {{< include setup.R >}}
+source("setup.R")
 ```
 
 

workshops/jsm2024/README.md

@@ -0,0 +1,28 @@
+# Course materials 
+
+These are the materials accompanying the short course "Applied Modelling in Drug Development using brms" at the Joint Statistical Meetings in Portland, OR on August 5, 2024. 
+
+Instructors:
+- David Ohlssen
+- Andrew Bean
+- Bjoern Holzhauer
+
+Additional contributors:
+- Paul Buerkner
+- Sebastian Weber
+- Lukas Widmer
+- Cong Zhang
+
+## Folder structure
+
+- `schedule.png` has the course schedule
+- `slides/bamdd_jsm2024.pdf` is the main slide deck for the course
+- `slides/R/` is a directory containing the R code from the case studies presented in the course
+- `exercises/` contains two sets of hands-on exercises in the form of Quarto files 
+- `exercises/solutions` has solutions from the organizers
+
+## Exercise instructions
+
+Course participants can find the exercises in the shared Posit Cloud space for the course (link provided during the course). See slides 38-44 in the main slides for instructions.
+
+

workshops/jsm2024/exercises/ex1_historical_control.qmd

@@ -0,0 +1,154 @@
+---
+title: 'brms exercises: historical controls'
+author: 'your-name-here'
+date: today
+output: 
+  html_document:
+    toc: true
+    embed-resources: true
+---
+
+```{r setup, include=FALSE}
+knitr::opts_chunk$set(echo = TRUE)
+```
+
+## Background
+
+These exercises accompany the case study [Use of Historical Control
+Data](TODO:link) as part of the training course "Workshop
+on applied modeling in drug development using brms."
+
+## Preliminary code
+
+First load the required packages and set some options as described in the case study.
+
+```{r, warning = FALSE, message = FALSE}
+here::i_am("exercises/ex1_historical_control.qmd")
+library(here)
+
+library(ggplot2)
+library(dplyr)
+library(knitr)
+library(brms)
+library(posterior)
+library(bayesplot)
+library(RBesT)
+theme_set(theme_bw())
+
+options(
+  # how many processor cores would you like to use?
+  mc.cores = 4, 
+  # how would you like to access Stan?
+  brms.backend = "cmdstanr",
+  # cache model binaries
+  cmdstanr_write_stan_file_dir=here::here("_brms-cache"),
+  # no need to normalize likelihoods
+  brms.normalize = FALSE,
+  # when you are storing your model to file, 
+  # how shall it be updated?
+  brms.file_refit = "on_change"  
+  # alternatives: "never", "always"
+  # use "never" for production
+)
+
+set.seed(254335)
+```
+
+Then create the dataset used in the example:
+
+```{r}
+AS_region <- bind_cols(RBesT::AS, region=sample(c("asia", "europe", "north_america"), 8, TRUE))
+kable(AS_region)
+```
+
+And fit the model as shown in the case study:
+
+```{r}
+model <- bf(r | trials(n) ~ 1 + (1 | study), family=binomial)
+
+model_prior <- prior(normal(0, 2), class="Intercept") +
+    prior(normal(0, 1), class="sd", coef="Intercept", group="study")
+
+map_mc_brms  <- brm(model, AS_region, prior = model_prior,
+                    seed = 4767,
+                    silent = 2, refresh = 0)
+
+```
+
+## Exercise 1
+
+Create a posterior predictive check based on the predictive
+distribution for the response rate.  
+Steps:
+* Use `posterior_predict` to create samples from the predictive
+  distribution of outcomes per trial.
+* Use `sweep(predictive, 2, AS_region$n, "/")` to convert these
+  samples from the predictive distribution of the outcome counts to
+  samples from the predictive distribution for responder rates.
+* Use `ppc_intervals` from `bayesplot` to create a plot showing your
+  results.
+
+```{r}
+# Your solution here -----------------------------------------------------------
+
+
+```
+
+
+## Exercise 2
+
+Redo the analysis with region, but treat region as a fixed
+   effect. Evaluate the informativeness of the obtained MAP priors.
+   The model formula for `brms` should look like `region_model_fixed
+   <- bf(r | trials(n) ~ 1 + region + (1 | study), family=binomial)`.
+Steps:
+* Consider the prior for the region fixed effect first. The reference
+  region is included in the intercept. The reference region is
+  implicitly defined by the first level of the variable region when
+  defined as `factor`.
+  - Define `asia` to be the reference region in the example. Also
+    include a level `other` in the set of levels.
+  - Assume that an odds-ratio of $2$ between regions can be seen as
+    very large such that a prior of $\mbox{N}(0, (\log(2)/1.96)^2)$
+    for the region main effect is adequate.
+* Obtain the MAP prior for each region by using the
+  `AS_region_all` data frame defined below and apply
+  `posterior_linpred` as shown above.
+* Convert the MCMC samples from the MAP prior distribution into
+  mixture distributions with the same code as above.
+* Calculate the ESS for each prior distribution with the `ess`
+  function from `RBesT`.
+
+```{r}
+
+AS_region_all <- data.frame(region=c("asia", "europe", "north_america", "other")) %>%
+    mutate(study=paste("new_study", region, sep="_"), r=0, n=6)
+
+# Your solution here -----------------------------------------------------------
+
+
+```
+
+
+## Exercise 3
+
+Run the analysis for the normal endpoint in the `crohn` data set of
+`RBesT`. Refer to the `RBesT` vignette for a [normal endpoint](https://cran.r-project.org/web/packages/RBesT/vignettes/introduction.html)
+   on more details and context.  
+Steps:
+* Use as `family=gaussian` and use the `se` response modifier in place
+  of `trials` to specify a known standard error.
+* Use the same priors as proposed in the vignette.
+* Compare the obtained MAP prior (in MCMC sample form) from `RBesT`
+  and `brms`.
+
+```{r}
+crohn <- RBesT::crohn
+
+# Your solution here -----------------------------------------------------------
+
+
+```
+
+
+