--- title: "Describing `crmPack` Objects" output: rmarkdown::html_vignette editor: visual number-sections: true embed-resources: TRUE reference-location: margin citation-location: margin bibliography: vignettes.bib vignette: > %\VignetteIndexEntry{Describing `crmPack` Objects} %\VignetteEngine{knitr::rmarkdown} %\VignetteEncoding{UTF-8} --- ## Introduction Objects created by `crmPack` are almost always S4 objects. Like all S4 objects, by default they do not render in a particularly user-friendly way. ``` r cs <- CohortSizeDLT(intervals = 0:2, cohort_size = c(1, 3, 5)) cs ``` ``` #> An object of class "CohortSizeDLT" #> Slot "intervals": #> [1] 0 1 2 #> #> Slot "cohort_size": #> [1] 1 3 5 ``` Fortunately, a little known feature of `knitr` can put this right at little or no cost to the end user: in the simplest case, demonstrated below, all that needs to be done is to reference the object in a markdown or Quarto chunk. ``` r cs ```
Defined by the number of toxicities so far observed
No of toxicities
Lower Upper Cohort size
0 1 1
1 2 3
2 Inf 5
> The `knit_print`methods provided by `crmPack` are not intended to be fully customisable or > comprehensive. We do believe, however, that they cover the vast majority of use-cases and are > easily extended using the techniques described later in this vignette. > Formatting of these objects currently only works for HMTL output. If another format - such as PDF > or Microsoft Word - is required, our suggested workaround is to create the HTML output and then > print or save the document to the required format. ## How is this done? When running code at the console, the result of an R function or statement that is not assigned to an object is `print`ed. (Unless, of course, it is returned `invisible`ly.) The same process appears to happen when the chunks of a markdown or Quarto document are evaluated. But that is not quite the case. Instead, the result is passed to an S3 function called `knit_print` [@Xie2024]. It is the results of running `knit_print` on the returned expression that appear in the rendered document. As a simple demonstration of the concept, consider: ``` r knit_print.DustySpringfield <- function(x, ...) { "I just don't know what to do with myself" } lyric <- 10 lyric #> [1] 10 class(lyric) <- "DustySpringfield" lyric #> I just don't know what to do with myself ``` The actions of `knit_print` are entirely arbitrary, but this mechanism provides developers with an easy way to provide nicely-rendered versions of any objects that are rendered by `knitr`. We have provided such methods for (almost) all `crmPack` classes. ## Using `knit_print` in `crmPack` By default, all that needs to be done is to reference the object to be printed in a markdown or quarto chunk. This is equivalent to `knit_print(object)`. However, the `knit_print` methods for most `crmPack` classes have optional arguments that can be used to customise the way in which the object is rendered. To change the default value of any parameter to `knit_print` the function must be called explicitly: `knit_print(cs, tox_label = "DLAE")`. ### Common customisations The most commonly needed customisations are to alter the way in which participants and toxicities are described. These are handled by the `label` and `tox_label` arguments to `knit_print`. These arguments can be provided either as a scalar or a vector of length 2. If a vector, the first element is taken to describe a single instance and the second any other number of instances. If a scalar, it is converted to a vector, whose first element is the scalar value provided and the second the scalar with `"s"` appended[^1]. [^1]: Except for `tox_label = "toxicity"`, which becomes `tox_label = c("toxicity", "toxicities")`. So, for example: ``` r CohortSizeConst(3) ``` A constant size of 3 participants. ``` r knit_print(CohortSizeConst(3), label = "subject") ``` A constant size of 3 subjects. Dose units are defined by the `units` parameter. By default, no units are printed. ``` r d <- Data(doseGrid = c(0.1, 0.3, 0.9, 2.5, 5, 10, 15)) d ``` No participants are yet evaluable. The dose grid is 0.1, 0.3, 0.9, 2.5, 5, 10 and 15. ``` r knit_print(d, units = "mg/dL") ``` No participants are yet evaluable. The dose grid is 0.1 mg/dL, 0.3 mg/dL, 0.9 mg/dL, 2.5 mg/dL, 5 mg/dL, 10 mg/dL and 15 mg/dL. The format used to display dose levels (and other information in other classes) can be changed with the `fmt` parameter: ``` r knit_print(d, units = "mg/dL", fmt = "%.2f") ``` No participants are yet evaluable. The dose grid is 0.10 mg/dL, 0.30 mg/dL, 0.90 mg/dL, 2.50 mg/dL, 5.00 mg/dL, 10.00 mg/dL and 15.00 mg/dL. `biomarker_label` and `biomarker_units` allow the representation of a biomarker to be customised. ``` r x <- .DefaultDualEndpointRW() x ``` The relationships between dose and toxicity and between dose and PD biomarker will be modelled simultaneously. A probit log normal model will describe the relationship between dose and toxicity: $$ \Phi^{-1}(Tox | d) = f(X = 1 | \theta, d) = \alpha + \beta \cdot log(d/d^*) $$\n where d* denotes a reference dose. The prior for θ is given by\n$$ \boldsymbol\theta = \begin{bmatrix}\alpha \\ \beta\end{bmatrix}\sim N \left(\begin{bmatrix} 0.00 \\ 1.00\end{bmatrix} , \begin{bmatrix} 1.00 & 0.00 \\ 0.00 & 1.00\end{bmatrix} \right) $$ \n\n The reference dose will be 1.00. The PD biomarker response `w` at dose `d` is modelled as $$ w(d) \sim N(f(d), \sigma_w^2) $$ where f(d) is a first order random walk such that $$ f(d) = \beta_{W_i} - \beta_{W_{i - 1}}\sim N(0, 0.01 \times (d_i - d_{i - 1})) $$ ``` r knit_print(x, biomarker_name = "CRP", biomarker_units = "mg/dL") ``` The relationships between dose and toxicity and between dose and PD biomarker will be modelled simultaneously. A probit log normal model will describe the relationship between dose and toxicity: $$ \Phi^{-1}(Tox | d) = f(X = 1 | \theta, d) = \alpha + \beta \cdot log(d/d^*) $$\n where d* denotes a reference dose. The prior for θ is given by\n$$ \boldsymbol\theta = \begin{bmatrix}\alpha \\ \beta\end{bmatrix}\sim N \left(\begin{bmatrix} 0.00 \\ 1.00\end{bmatrix} , \begin{bmatrix} 1.00 & 0.00 \\ 0.00 & 1.00\end{bmatrix} \right) $$ \n\n The reference dose will be 1.00. The PD biomarker response `w` at dose `d` is modelled as $$ w(d) \sim N(f(d), \sigma_w^2) $$ where f(d) is a first order random walk such that $$ f(d) = \beta_{W_i} - \beta_{W_{i - 1}}\sim N(0, 0.01 \times (d_i - d_{i - 1})) $$ ## Rendering complex classes Some `crmPack` classes have slots whose values are themselves `crmPack` classes. `CohortSizeMax` is a simple example. In these cases, the slot values are each passed to `knit_print` in turn. ``` r .DefaultCohortSizeMax() ``` The maximum of the cohort sizes defined in the following rules:
Defined by the dose to be used in the next cohort
Dose
Lower Upper Cohort size
0 10 1
10 Inf 3
Defined by the number of toxicities so far observed
No of toxicities
Lower Upper Cohort size
0 1 1
1 Inf 3
`knit_print` methods for sub-classes of `RuleDesign` (and related classes) offer slightly more control. Here, an overall header for the rendition of the object is provided by the `title` parameter (whose value defaults to "Design" and the slot values are separated by sub-headers. The styling of the overall header and sub-headers is controlled by the `level` parameter. The default value of `level` is `2L`, and the level of slots is defined recursively to be one more than the level of the parent slot[^2]. Class-specific parameters are passed to slot-specific `knit_print` methods using `...`. [^2]: Because markdown header styles are defined only for six levels, the greatest value for `level`, including values generated by nested calls, is `6`. ``` r knit_print(.DefaultDesign()) ``` ## Design ### Dose toxicity model A logistic log normal model will describe the relationship between dose and toxicity: $$ p(Tox | d) = f(X = 1 | \theta, d) = \frac{e^{\alpha + \beta \cdot log(d/d_{ref})}}{1 + e^{\alpha + \beta \cdot log(d/d_{ref})}} $$\n where d~ref~ denotes a reference dose. The prior for θ is given by\n$$ \boldsymbol\theta = \begin{bmatrix}\alpha \\ log(\beta)\end{bmatrix}\sim N \left(\begin{bmatrix}-0.85 \\ 1.00\end{bmatrix} , \begin{bmatrix} 1.00 & -0.50 \\ -0.50 & 1.00\end{bmatrix} \right) $$ \n\n The reference dose will be 56.00. ### Stopping rule If either of the following rules are `TRUE`: - If both of the following rules are `TRUE`: - ≥ 3 cohorts dosed: If 3 or more cohorts have been treated. - P(0.2 ≤ prob(DLE | NBD) ≤ 0.35) ≥ 0.5: If the probability of toxicity at the next best dose is in the range [0.20, 0.35] is at least 0.50. - ≥ 20 patients dosed: If 20 or more participants have been treated. ### Escalation rule
Defined by highest dose administered so far
Dose
Min Max Increment
0 20 1.00
20 Inf 0.33
### Use of placebo Placebo will not be administered in the trial. ### Dose recommendation The dose recommended for the next cohort will be chosen in the following way. First, doses that are ineligible according to the increments rule will be discarded. Next, any dose for which the mean posterior probability of toxicity being in the overdose range - (0.35, 1] - is 0.25 or more will also be discarded. Finally, the dose amongst those remaining which has the highest chance that the mean posterior probability of toxicity is in the target toxicity range of 0.2 to 0.35 (inclusive) will be selected. ### Cohort size The maximum of the cohort sizes defined in the following rules:
Defined by the dose to be used in the next cohort
Dose
Lower Upper Cohort size
0 30 1
30 Inf 3
Defined by the number of toxicities so far observed
No of toxicities
Lower Upper Cohort size
0 1 1
1 Inf 3
### Observed data No participants are yet evaluable. The dose grid is 1, 3, 5, 10, 15, 20, 25, 40, 50, 80 and 100. ### Starting dose The starting dose is 3. Slot headers can be customised using the `sections` parameter. `sections` should be a named vector. Names should be valid slot names for the object being rendered and values the requested slot headers. ``` r knit_print( .DefaultDesign(), level = 4, sections = c( "nextBest" = "Selection of the dose for the following cohort", "startingDose" = "Initial dose" ) ) ``` $$Output not shown.$$ > It is not possible to omit slots from the rendition of a `crmPack` object. If you need to do this, > you can either render the required slots individually, or override the definition of `knit_print` > for the super class as demonstrated below. ## Restoring console-like behaviour To restore the default behaviour for `crmPack` objects, simply wrap the object in a call to `normal_print()`. ``` r normal_print(cs) #> An object of class "CohortSizeDLT" #> Slot "intervals": #> [1] 0 1 2 #> #> Slot "cohort_size": #> [1] 1 3 5 ``` ## Accessing the output of `knit_print` One of the parameters of `knitr::knit_print` is `asis`, with a default value of `TRUE`. `asis` has the same effect as setting the chunk option `output` to `asis`. This is achieved by returning an object of class `knit-asis`. Setting `asis` to `FALSE` will display the raw HTML code generated by `knit_print` to be displayed. Alternatively, it may allow easier manipulation of the return value. ``` r csOutput1 <- knit_print(CohortSizeDLT(intervals = 0:2, cohort_size = c(1, 3, 5))) class(csOutput1) #> [1] "knit_asis" csOutput1 ```
Defined by the number of toxicities so far observed
No of toxicities
Lower Upper Cohort size
0 1 1
1 2 3
2 Inf 5
``` r csOutput2 <- knit_print(CohortSizeDLT(intervals = 0:2, cohort_size = c(1, 3, 5)), asis = FALSE) class(csOutput2) #> [1] "character" csOutput2 #> [1] "\n\n \n\n\n\n\n \n \n \n \n \n \n\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n\n
Defined by the number of toxicities so far observed
No of toxicities
Lower Upper Cohort size
0 1 1
1 2 3
2 Inf 5
\n\n" ``` But with the chunk option `output` set to `asis`... ``` r cat(csOutput2) #> #> #> #> #> #> #> #> #> #> #> #> #> #> #> #> #> #> #> #> #> #> #> #> #> #> #> #> #> #> #>
Defined by the number of toxicities so far observed
No of toxicities
Lower Upper Cohort size
0 1 1
1 2 3
2 Inf 5
``` ## Providing your own `knit_print` method {#sec-override} If the methods provided in `crmPack` don't do what you want, it's easy to roll your own, using standard S3 techniques. The formal arguments to `knitr::knit_print` are `x` and `...`. Additional arguments can be added after `...`. As an example, consider `knit_print.NextBestNCRM`, which currently returns a paragraph of text: ``` r .DefaultNextBestNCRM() ``` The dose recommended for the next cohort will be chosen in the following way. First, doses that are ineligible according to the increments rule will be discarded. Next, any dose for which the mean posterior probability of toxicity being in the overdose range - (0.35, 1] - is 0.25 or more will also be discarded. Finally, the dose amongst those remaining which has the highest chance that the mean posterior probability of toxicity is in the target toxicity range of 0.2 to 0.35 (inclusive) will be selected. You might feel this is better presented as a bulleted list. You can achieve this as follows[^3]: [^3]: For simplicity, the `tox_label` and `asis` parameters, which are defined in the current implementation of the function, are omitted in this custom implementation. They should be preserved in any "real world" customisation. ``` r knit_print.NextBestNCRM <- function(x, ...) { knitr::asis_output( paste0( "The dose recommended for the next cohort will be chosen in the following ", "way.\n\n- First, doses that are ineligible according to the increments rule ", "will be discarded.\n- Next, any dose for which the mean posterior probability of ", " toxicity being in the overdose range - (", x@overdose[1], ", ", x@overdose[2], "] - is ", x@max_overdose_prob, " or more will also be discarded.\n- Finally, the dose amongst those remaining ", "which has the highest chance that the mean posterior probability of toxicity ", "is in the target toxicity range of ", x@target[1], " to ", x@target[2], " (inclusive) will be selected.\n\n" ) ) } registerS3method("knit_print", "NextBestNCRM", knit_print.NextBestNCRM) .DefaultNextBestNCRM() ``` The dose recommended for the next cohort will be chosen in the following way. - First, doses that are ineligible according to the increments rule will be discarded. - Next, any dose for which the mean posterior probability of toxicity being in the overdose range - (0.35, 1] - is 0.25 or more will also be discarded. - Finally, the dose amongst those remaining which has the highest chance that the mean posterior probability of toxicity is in the target toxicity range of 0.2 to 0.35 (inclusive) will be selected. ## Class coverage `crmPack` defines 125 classes. Custom `knit_print` methods exist for 92 of them. Of the remaining 33 classes, 22 are virtual classes that will never be directly instantiated by end users. That leaves 11 classes for which `knit_print` methods may be useful. These classes are listed below. |Class | |:----------------------| |DualSimulationsSummary | |NextBestEWOC | |Simulations | |StoppingExternal | |DualSimulations | |GeneralSimulations | |Samples | |DASimulations | |IncrementsMaxToxProb | |EffFlexi | |McmcOptions | The majority of these classes relate to simulation of the operating characteristics of CRM trials. Reporting of this information is likely to need customisation that is beyond the scope of a simple function[^4]. [^4]: The `crmPack` team is considering the creation of markdown or Quarto templates that may assist in this area, but consider this to be a long-term ambition.