chemo.org.uk https://www.chemo.org.uk/ discussing oncology pharmacy, data manipulation and other general thoughts quarto-1.5.56 Sat, 10 May 2025 23:00:00 GMT White Axes, line labels and transparency in Survival Curves Calum Polwart https://www.chemo.org.uk/posts/TransparentSurvivalCurves.html

Background

I don’t know if I’m actually going to use the results of this yet, but having spent more time than I’d hoped to trying to figure out how to do this, I thought I should document it. I’m producing a survival curve using ggsurvplot, but I want to present them on a dark background. That means I want to make my plot background transparent and make all the black stuff (lines, text etc) white so that it has good contrast on the dark background. I’ve almost never seen examples of this in the wild - either because in design terms someone says its a bad thing, /or/ its hard to do in R so people don’t try?

I’ve read enough data science blogs to know that having to move your eyes from a line to a legend to work out which one is which is also considered bad, and so I’m also going to label my lines.

Lets set up a basic plot

My plots have risk tables on them which makes life all the more ‘exciting’. We will start with the official example plot - survival in the lung cancer data-set. I’ve made some tweaks to the standard off the shelf plot to get us started.

Code 
require(survival)
require(survminer)
fit<- survfit(Surv(time, status) ~ sex, data = lung)

# Basic survival curves
ggsurvplot(
    fit, 
    data = lung, 
1    title = "The official example Survival Curve",
2    palette = c("#E7B800", "#2E9FDF"),
3    pval = TRUE,
4    risk.table = TRUE,
5    tables.height = 0.2
    )
1
Add a title
2
Change the colours for the two survival lines
3
Add a p-value annotation to the plot
4
Add the risk table below the plot
5
Set the risk table to be 20% of the plot height.

Now to customise that plot theme

ggsurvplot works using ggplot and so you might expect that you can take its result and throw a few theme() statements at it and magically make things look different. Of course if it was that easy, I’d not be writing a blog about it! ggsurvplot knows this is an issue so it lets you define the theme /within/ the call to the plot. The reason this strange approach is used, is that ggsurvplot is actually creating two plots - the risk table is a plot as well.

So for simplicity, lets say I wanted to change the axis to be red and a bit thicker I can use this:

Code 
ggsurvplot(
    fit, 
    data = lung,
    title = "The official example Survival Curve",
    palette = c("#E7B800", "#2E9FDF"),
    pval = TRUE,
    risk.table = TRUE,
    tables.height = 0.2,
1    ggtheme = theme(
            axis.line = element_line(colour = "red", linewidth=2),
                        axis.ticks = element_line(colour = "red", linewidth = 1.5),
                        axis.ticks.length=unit(.25, "cm")
                     ) 
    )
1
Add a theme statement to change the axis.line to red, and the tick colour, width and length.

This is all fairly well documented and straightforward so far. And we can add a transparent background using the theme_transparent() in the ggtheme statement.

Code 
ggsurvplot(
    fit, 
    data = lung,
    title = "The official example Survival Curve",
    palette = c("#E7B800", "#2E9FDF"),
    pval = TRUE,
    risk.table = TRUE,
    tables.height = 0.2,
1    ggtheme = theme_transparent() +
            theme( 
                axis.line = element_line(colour = "red", linewidth=2), 
                            axis.ticks = element_line(colour = "red", linewidth = 1.5), 
                            axis.ticks.length=unit(.25, "cm") 
                        ) 
    )
1
Add a theme_transparent statement much like you would with ggplot2 normally, but within the ggtheme statement.

If you’ve not already got dark mode enabled on the site, now might be a good time to enable it! You can use the slider at the top right of the screen. However, already there are a few things that even then aren’t behaving as you might expect! The title has vanished. The tick-marks (on the main plot) are in black not red and my decision to thicken the axis has suddenly made the risk table axis really obvious. You can add a tables.theme statement as well which would let you handle its axis differently. I’m also going to switch off the text labels with - tables.y.text = FALSE.

Code 
ggsurvplot(
    fit, 
    data = lung,
    title = "The official example Survival Curve",
    palette = c("#E7B800", "#2E9FDF"),
    pval = TRUE,
    risk.table = TRUE,
    tables.height = 0.2,
1    tables.y.text = FALSE,
    ggtheme = theme_transparent() + 
2            theme( text = element_text(colour = "yellow"),
                axis.line = element_line(colour = "red", linewidth=2), 
3                            axis.ticks.x = element_line(colour = "red", linewidth = 1.5),
                axis.ticks.y = element_line(colour = "red", linewidth = 1.5),
                            axis.ticks.length=unit(.25, "cm") 
                        ), 
4    tables.theme = theme_cleantable()
    )
1
Removing the text labels just for tidiness
2
Make the text a different colour. When I’m experimenting with graphs like this I often use a mix of horrible colours to check what I’m changing, then flip to my final colour I want later.
3
Specify the axis ticks for x and y separately as they must be defined separately elsewhere in the ggsurvplot theme :-(
4
Adding a different theme to the risk table

You will notice we still have black text on the p value and the risk numbers and that we are still missing the title and the title on the risk table. The title is fairly easy to fix. The black text for the p-value is however not so simple. These bits of text are actually geom_text and the colour is specified when they are placed rather than using a theme. ggsurvplot has provided a method for the numbers at risk so we can use that.

Code 
ggsurvplot(
    fit, 
    data = lung,
    title = "The official example Survival Curve",
    palette = c("#E7B800", "#2E9FDF"),
    pval = TRUE,
    risk.table = TRUE,
    tables.height = 0.2,
    tables.y.text = FALSE, 
1    tables.col = "magenta",
    ggtheme = theme_transparent() + 
            theme( text = element_text(colour = "yellow"), 
2                plot.title = element_text(colour = "green", size = 30),
3                                plot.tag=element_text(colour="orange"),
                                plot.subtitle = element_text(colour = "blue", size = 25),
                axis.line = element_line(colour = "red", linewidth=2), 
                            axis.ticks.x = element_line(colour = "red", linewidth = 1.5), 
                axis.ticks.y = element_line(colour = "red", linewidth = 1.5), 
                            axis.ticks.length=unit(.25, "cm") 
                        ), 
    tables.theme = theme_cleantable() 
    )
1
Use tables.col to specify the text colour, or alternatively you could use tables.col = “strata” to get the line colours as the text colour.
2
Specifying a plot title, and in this example a size will restore the title.
3
If you are using a subtitle, or plot tags you will need to specify these too.

If you’d like the Number at Risk to be smaller and on the left we can add it as a theme to tables.theme()

Code 
ggsurvplot(
    fit, 
    data = lung,
    title = "The official example Survival Curve",
    palette = c("#E7B800", "#2E9FDF"),
    pval = TRUE,
    risk.table = TRUE,
    tables.height = 0.2,
    tables.y.text = FALSE, 
    tables.col = "magenta",
    ggtheme = theme_transparent() + 
            theme( text = element_text(colour = "yellow"), 
                plot.title = element_text(colour = "green", size = 30), 
                                plot.tag=element_text(colour="orange"), 
                                plot.subtitle = element_text(colour = "blue", size = 25), 
                axis.line = element_line(colour = "red", linewidth=2), 
                            axis.ticks.x = element_line(colour = "red", linewidth = 1.5), 
                axis.ticks.y = element_line(colour = "red", linewidth = 1.5), 
                            axis.ticks.length=unit(.25, "cm") 
                        ), 
    tables.theme = theme_cleantable() +
            theme( 
1                plot.title = element_text(size = 20, hjust = 0 )
            )
    )
1
Reduce the size and horizontally justify (left align) the title

Hacking the p-value colour

We’ve now done as much as we can achieve within ggsurvplot directly. So lets plot this all with white instead of all these hideous colours. We will save this as an object so that we can hack the p value colour.

Code 
par(bg = "#000000", fg = "#FFFFFF")
ggsurvplot(
    fit, 
    data = lung,
    title = "The official example Survival Curve",
    palette = c("#E7B800", "#2E9FDF"),
    pval = TRUE,
    risk.table = TRUE,
    tables.height = 0.2,
    tables.y.text = FALSE, 
1    tables.col = "white",
    ggtheme = theme_transparent() + 
            theme( text = element_text(colour = "white"),
                plot.title = element_text(colour = "white", size = 30),
                                plot.tag=element_text(colour="white"),
                                plot.subtitle = element_text(colour = "white", size = 25),
                axis.line = element_line(colour = "white", linewidth=2),
                            axis.ticks.x = element_line(colour = "white", linewidth = 1.5),
                axis.ticks.y = element_line(colour = "white", linewidth = 1.5),
                            axis.ticks.length=unit(.25, "cm") 
                        ), 
    tables.theme = theme_cleantable() +
            theme( 
                plot.title = element_text(size = 20, hjust = 0 )
            )
2    ) -> myPlot
3print(myPlot)
1
Change all the colours to white
2
Save the plot to an object called myPlot
3
Use print to view the plot

If we now examine the structure of myPlot with summary(myPlot) we get:

Code 
summary(myPlot)
               Length Class      Mode
plot           11     gg         list
table          11     gg         list
data.survplot  10     data.frame list
data.survtable 10     data.frame list

Which tells us there is 4 parts to this object. Two data tables containing the data, and two gg class objects (plots in essence) once called ‘plot’ which has the survival curve, the other ‘table’ which has the risk table. Examining the plot will give us more information:

Code 
summary(myPlot$plot)
data: time, n.risk, n.event, n.censor, surv, std.err, upper, lower,
  strata, sex [208x10]
mapping:  x = ~time, y = ~surv
scales:   y, ymin, ymax, yend, yintercept, ymin_final, ymax_final, lower, middle, upper, y0, colour, fill, x, xmin, xmax, xend, xintercept, xmin_final, xmax_final, xlower, xmiddle, xupper, x0 
faceting: <ggproto object: Class FacetNull, Facet, gg>
    compute_layout: function
    draw_back: function
    draw_front: function
    draw_labels: function
    draw_panels: function
    finish_data: function
    init_scales: function
    map_data: function
    params: list
    setup_data: function
    setup_params: function
    shrink: TRUE
    train_scales: function
    vars: function
    super:  <ggproto object: Class FacetNull, Facet, gg>
-----------------------------------
mapping: colour = ~strata 
geom_step: direction = hv, na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity 

mapping: x = ~x, y = ~y 
geom_blank: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity 

mapping: colour = ~strata 
geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity 

mapping: x = ~x, y = ~y 
geom_text: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity

This shows there are 4 parts to this particular plot (your plots may vary!). Importantly the 4th part is geom_text - this likely contains our p-value information. These sit in something called a layer and we should look at the 4th Layer. We are interested in the aesthetics so we need to look at the aes_params:

Code 
myPlot$plot$layers[[4]]$aes_params
$label
[1] "p = 0.0013"

$size
[1] 5

$hjust
[1] 0

$family
[1] ""

This confirms we are looking that p-value and shows that no colour has been specified. Adding a colour is as easy as:

Code 
myPlot$plot$layers[[4]]$aes_params$colour <- "red"
print(myPlot)

Labelling the line instead of a legend

I also wanted to edit the legend to annotate those alongside my lines rather than at the top of the window as a legend. We should therefore return to our ggsurvplot command to disable the legend:

Code 
ggsurvplot(
    fit, 
    data = lung,
    title = "The official example Survival Curve",
    palette = c("#E7B800", "#2E9FDF"),
    pval = TRUE,
1    legend = "none",
    risk.table = TRUE,
    tables.height = 0.2,
    tables.y.text = FALSE, 
    tables.col = "white",
    ggtheme = theme_transparent() + 
            theme( text = element_text(colour = "white"), 
                plot.title = element_text(colour = "white", size = 30), 
                                plot.tag=element_text(colour="white"), 
                                plot.subtitle = element_text(colour = "white", size = 25),
                axis.line = element_line(colour = "white", linewidth=2), 
                            axis.ticks.x = element_line(colour = "white", linewidth = 1.5),
                axis.ticks.y = element_line(colour = "white", linewidth = 1.5), 
                            axis.ticks.length=unit(.25, "cm") 
                        ), 
    tables.theme = theme_cleantable() +
            theme( 
                plot.title = element_text(size = 20, hjust = 0 ) 
            )
    ) -> myPlot 
2myPlot$plot$layers[[4]]$aes_params$colour <- "white"
print(myPlot)
1
Disable the legend
2
Update the p-value colour

As myPlot$plot contains the survival curve we can use ggannotate to, erm, annotate it! Strata are ordered in the order of the factor for their grouping. So in our example at the very start we created a survival object with:

fit<- survfit(Surv(time, status) ~ sex, data = lung)

and ‘sex’ is where the strata comes from. Unhelpfully they are coded as sex = 1 or 2. I gather that the correct allocation is 1 = Male and 2 = Female. It would probably be better to fix this before we create the survival object. So lets go back and do that:

Code 
1require(forcats)

lung$sex |>
2    as.character() |>
3    fct_recode ( "Male" = "1", "Female" = "2") -> lung$sex

4fit<- survfit(Surv(time, status) ~ sex, data = lung)

5ggsurvplot(
    fit,
    data = lung,
    title = "The official example Survival Curve",
    palette = c("#E7B800", "#2E9FDF"),
    pval = TRUE,
    risk.table = TRUE,
    tables.height = 0.2,
    tables.y.text = FALSE
    )
1
I prefer to use tidyverse forcats package for factors, but you could use base R to achieve the same
2
The data is currently stored as a number and not a factor. forcats can’t process a number it needs a character or factor so lets make it a character
3
Re-code the data and save it back into the lung$sex data column. You could of course save it elsewhere and use that if you didn’t want to affect the source data
4
Refit the data with the coded values
5
Re run the graph, I’ve used a shortened version here with the legend re-enabled to show which is which for colours - yellow is male and blue is female

To annotate the graphs we will need to know:

  • Where we want the labels to be
  • The label colours
  • The label names

The label colours is ‘easy’ we are already defining that with the line palette = c(“#E7B800”, “#2E9FDF”) in the plot and it would be sensible to use labels of the same colour as the line.

The label names is also ‘easy’ as we simply need to take the factor levels. These will be applied to the colours in order. So the first factor gets #E7B800 as a colour.

Positioning by code is harder. You may simply decide to do it using some hard entered numbers of the co-ordinates you want to use. But I thought this little bit of code might work:

Code 
1line_colours =  c("#E7B800", "#2E9FDF")
2label_names = levels(lung$sex)

# Get the lower line name
3lower_line <- surv_median(fit)$strata[surv_median(fit)$median == min(surv_median(fit)$median, na.rm=T)]
upper_line <- surv_median(fit)$strata[surv_median(fit)$median == max(surv_median(fit)$median, na.rm=T)]

# establish 75% of x-axis
sum_fit <- summary(fit)
x75 <- max(sum_fit$time, na.rm=T) * 0.75

# get the lower line level at 75% of the way along the line
lower_line_y <- sum_fit$surv[sum_fit$time >= x75 & sum_fit$strata == lower_line][1]

# get the upper line level at 75% of the way along the line
upper_line_y <- sum_fit$surv[sum_fit$time >= x75 & sum_fit$strata == upper_line][1]

# annotate the lower line with the right hand to corner of the label at 75% & lower_line (with some padding)

4myPlot$plot <- myPlot$plot + annotate(
            "text",
            x= x75,
            y = lower_line_y,
            label= label_names[surv_median(fit)$strata ==lower_line],
            colour=line_colours[surv_median(fit)$strata ==lower_line],
            hjust = 1,
            vjust = 1
            )

# annotate the upper line above the line

myPlot$plot <- myPlot$plot + annotate(
            "text",
            x= x75,
            y = upper_line_y,
            label= label_names[surv_median(fit)$strata ==upper_line],
            colour=line_colours[surv_median(fit)$strata ==upper_line],
            hjust = 0,
            vjust = -2 # this is a bit unpredictable
            )

print(myPlot)
1
define the line colours as a vector
2
define the label names for the graph
3
establish where to plot the annotations
4
plot the annotations

For completeness - we can create this all as a single piece of code, and we will replace the line colouration inf the ggsurvplot function so that it couldn’t be swapped around in error.

Code 
require(survival)
require(survminer)
require(forcats)

1data(lung)

lung$sex |>
    as.character() |> 
    fct_recode ( "Male" = "1", "Female" = "2") -> lung$sex 

fit<- survfit(Surv(time, status) ~ sex, data = lung) 

line_colours =  c("#E7B800", "#2E9FDF")  
label_names = levels(lung$sex) 

ggsurvplot(
    fit, 
    data = lung,
    title = "The official example Survival Curve",
2    palette = line_colours,
    pval = TRUE,
    legend = "none", 
    risk.table = TRUE,
    tables.height = 0.2,
    tables.y.text = FALSE, 
    tables.col = "white", 
    ggtheme = theme_transparent() + 
            theme( text = element_text(colour = "white"), 
                plot.title = element_text(colour = "white", size = 30), 
                                plot.tag=element_text(colour="white"), 
                                plot.subtitle = element_text(colour = "white", size = 25),
                axis.line = element_line(colour = "white", linewidth=2), 
                            axis.ticks.x = element_line(colour = "white", linewidth = 1.5),
3                axis.title.x = element_text(colour = "white", hjust=1),
                axis.title.y = element_text(colour = "white", hjust=1),
                axis.ticks.y = element_line(colour = "white", linewidth = 1.5), 
                            axis.ticks.length=unit(.25, "cm") 
                        ), 
    tables.theme = theme_cleantable() +
            theme( 
                plot.title = element_text(size = 20, hjust = 0 ) 
            )
    ) -> myPlot 

# change the p-value colour
myPlot$plot$layers[[4]]$aes_params$colour <- "white" 

# Get the lower line name
lower_line <- surv_median(fit)$strata[surv_median(fit)$median == min(surv_median(fit)$median, na.rm=T)] 
upper_line <- surv_median(fit)$strata[surv_median(fit)$median == max(surv_median(fit)$median, na.rm=T)] 

# establish 75% of x-axis
sum_fit <- summary(fit) 
x75 <- max(sum_fit$time, na.rm=T) * 0.75 

# get the lower line level at 75% of the way along the line
lower_line_y <- sum_fit$surv[sum_fit$time >= x75 & sum_fit$strata == lower_line][1] 

# get the upper line level at 75% of the way along the line
upper_line_y <- sum_fit$surv[sum_fit$time >= x75 & sum_fit$strata == upper_line][1] 

# annotate the lower line with the right hand to corner of the label at 75% & lower_line (with some padding)

myPlot$plot <- myPlot$plot + annotate( 
            "text", 
            x= x75, 
            y = lower_line_y, 
            label= label_names[surv_median(fit)$strata ==lower_line], 
            colour=line_colours[surv_median(fit)$strata ==lower_line],
            hjust = 1,
            vjust = 1
            )

# annotate the upper line above the line

myPlot$plot <- myPlot$plot + annotate(
            "text",
            x= x75, 
            y = upper_line_y, 
            label= label_names[surv_median(fit)$strata ==upper_line], 
            colour=line_colours[surv_median(fit)$strata ==upper_line],
            hjust = 0,
            vjust = -2 # this is a bit unpredictable 
            )

print(myPlot)
1
reload the lung data-set as we’ve been messing with it
2
insert the line colour vector
3
expose the axis labels which I missed earlier!

Finally, if we want to save the plot as a vector or png, we need to tell ggsave not to apply a background colour!

Code 
ggsave("myPlot.png", bg="transparent")
Saving 8 x 8 in image

We used R version 4.3.0[1] and the following R packages: cowplot v. 1.1.3[2], DT v. 0.33[3], glue v. 1.8.0[4], knitr v. 1.50[5–7], qicharts2 v. 0.7.5[8], rmarkdown v. 2.29[9–11], survival v. 3.8.3[12,13], survminer v. 0.5.0[14], tidyverse v. 2.0.0[15].

References

1. R Core Team. (2023). R: A language and environment for statistical computing. R Foundation for Statistical Computing. https://www.R-project.org/
2. Wilke, C. O. (2024). cowplot: Streamlined plot theme and plot annotations for “ggplot2”. https://CRAN.R-project.org/package=cowplot
3. Xie, Y., Cheng, J., & Tan, X. (2024). DT: A wrapper of the JavaScript library “DataTables”. https://CRAN.R-project.org/package=DT
4. Hester, J., & Bryan, J. (2024). glue: Interpreted string literals. https://CRAN.R-project.org/package=glue
5. Xie, Y. (2014). knitr: A comprehensive tool for reproducible research in R. In V. Stodden, F. Leisch, & R. D. Peng (Eds.), Implementing reproducible computational research. Chapman; Hall/CRC.
6. Xie, Y. (2015). Dynamic documents with R and knitr (2nd ed.). Chapman; Hall/CRC. https://yihui.org/knitr/
7. Xie, Y. (2025). knitr: A general-purpose package for dynamic report generation in R. https://yihui.org/knitr/
8. Anhoej, J. (2024). qicharts2: Quality improvement charts. https://CRAN.R-project.org/package=qicharts2
9. Xie, Y., Allaire, J. J., & Grolemund, G. (2018). R markdown: The definitive guide. Chapman; Hall/CRC. https://bookdown.org/yihui/rmarkdown
10. Xie, Y., Dervieux, C., & Riederer, E. (2020). R markdown cookbook. Chapman; Hall/CRC. https://bookdown.org/yihui/rmarkdown-cookbook
11. Allaire, J., Xie, Y., Dervieux, C., McPherson, J., Luraschi, J., Ushey, K., Atkins, A., Wickham, H., Cheng, J., Chang, W., & Iannone, R. (2024). rmarkdown: Dynamic documents for r. https://github.com/rstudio/rmarkdown
12. Terry M. Therneau, & Patricia M. Grambsch. (2000). Modeling survival data: Extending the Cox model. Springer.
13. Therneau, T. M. (2024). A package for survival analysis in r. https://CRAN.R-project.org/package=survival
14. Kassambara, A., Kosinski, M., & Biecek, P. (2024). survminer: Drawing survival curves using “ggplot2”. https://CRAN.R-project.org/package=survminer
15. Wickham, H., Averick, M., Bryan, J., Chang, W., McGowan, L. D., François, R., Grolemund, G., Hayes, A., Henry, L., Hester, J., Kuhn, M., Pedersen, T. L., Miller, E., Bache, S. M., Müller, K., Ooms, J., Robinson, D., Seidel, D. P., Spinu, V., … Yutani, H. (2019). Welcome to the tidyverse. Journal of Open Source Software, 4(43), 1686. https://doi.org/10.21105/joss.01686

Reuse

This post is released under CC-BY-4.0 https://creativecommons.org/licenses/by/4.0/. Data shown is sourced from the Lung Dataset in the survminer R package

Citation

BibTeX citation:
@misc{polwart2025,
  author = {Polwart, Calum},
  title = {White {Axes,} Line Labels and Transparency in {Survival}
    {Curves}},
  date = {2025-05-11},
  url = {https://www.chemo.org.uk/posts/TransparentSurvivalCurves.html},
  langid = {en}
}
For attribution, please cite this work as:
1. Polwart, C. (2025, May 11). White Axes, line labels and transparency in Survival Curves. Chemo.org.uk. https://www.chemo.org.uk/posts/TransparentSurvivalCurves.html
]]> Data R https://www.chemo.org.uk/posts/TransparentSurvivalCurves.html Sat, 10 May 2025 23:00:00 GMT Interaction checking tools Calum Polwart https://www.chemo.org.uk/posts/Story-behind-interaction-poster.html I’ve just published a poster and thought I’d share some of the thinking behind it.

Where did the idea come from?

This poster wasn’t my idea. Perhaps it wasn’t really anyone’s idea but two people were pretty key in developing the idea. Chloë Waterson, then working at County Durham and Darlington NHS Foundation Trust and Rajinder Nijjar, working for Guys’ and St Thomas’ and NHS England were the real instigators. We spoke with a few collaborators along the way but these two have become co-investigators on this project because they both asked a question around the same time.

Chloë was first to drop me a note asking if I’d seen that Liverpool Cancer Drug Interaction checker might be being shut down and what did I think about this. That was nearly 18 months ago. And what did I think? Well my initial though was a mixture of “Oh that’s a shame”, “I’m old enough to remember using a book to look up interactions” and “Sounds like its a money issue, someone will sort it”. I did also briefly ponder why that particular tool had become so popular. We briefly spoke about the idea of being able to check what was a good alternative, but did not much more. I did spend a couple of hours studying how the website actually worked and seeing if the data would be available in an archived form if the site was pulled. Anyway, move on a few weeks and Raj sends an email round some colleagues working for NHS England asking what people are using instead. She’d been using UpToDate but was finding problems. Seemingly new interactions surfacing for drugs we’d used together for years and some drugs not being listed (domperidone was her example).

I joined Raj and Chloë together in a Teams call. We tried inviting others who had shown interest on the BOPA Forum but didn’t get any takers. But we had the start of an idea. We wanted to know what was the best alternative. We were still talking as if Liverpool was the gold standard.

We bounced some ideas around about how we could do that. Creating our own test scenarios. Using specificity and sensitivity to measure things. A little searching, followed by more detailed searching revealed we weren’t the first to tackle the question of how good a drug interaction checking tool is. A little more work and we were uncovering papers describing oncology specific tools.

— Content on this website has been produced with reasonable care. However, errors can occur. Opinions change over time. Code that worked at the time of publication may no longer work. Users of all content from this site do so at their own risk.

Reuse

Citation

BibTeX citation:
@misc{polwart2024,
  author = {Polwart, Calum},
  title = {Interaction Checking Tools},
  date = {2024-10-11},
  url = {https://www.chemo.org.uk/posts/Story-behind-interaction-poster.html},
  langid = {en}
}
For attribution, please cite this work as:
1. Polwart, C. (2024, October 11). Interaction checking tools. Chemo.org.uk. https://www.chemo.org.uk/posts/Story-behind-interaction-poster.html
]]> https://www.chemo.org.uk/posts/Story-behind-interaction-poster.html Thu, 10 Oct 2024 23:00:00 GMT Formatting qicharts2 Calum Polwart https://www.chemo.org.uk/posts/Formatting-qicharts2.html

Brief background

I was asked recently about the best visualisation for a quality improvement project. This was a project that had aimed to reduce the frequency of errors detected by a pharmacy team. This was a specific error, and so would be relatively rare. With a common event, you can easily plot something like daily intervention rates and for me I’d usually suggest using the NHS Plot the Dots[1] package as it has a lot of ability to customise and is really rather well written. But for rare events a time between events plot (often called a ‘T’ plot) is much better. This is the forgotten cousin in SPC charting, and often omitted from the most common SPC tools. NHSRplotthedots can’t do ‘t’ plots. But an R package called qicharts can. qichart appears to have been superseded by qicharts2[2] which produces a ggplot2 object which, I thought at least, would mean we could style the graphics more easily using the ggplot grammar of graphics.

Difference between a standard SPC and T plot

This seems limited in description, but the control lines are calculated in a different way. A moving average is used instead of a mean/median. This means you cant just relabel the existing SPC charts other tools make. There is a rather nice NHS PDF available here that outlines the manual process to build the data.

A T plot is plotting the time (usually in days) between events happening. If you introduce a new process to reduce errors, you’d hope to see the time between events increase. It is as simple as that. The rest of the SPC rules - more than 6 values above the control line a value outside of the 3 sigma bands etc are the same as normal. Although I’ve not seen it written down, I think ‘impossible’ values (less than zero) are also not shown in the sigma ranges. And certainly the tool all check that the time between events isn’t zero. If you have two events on the same day, you need to use a fraction of a day to capture the second event. (If that happened lots you might be better using hours as a measure instead of days).

Lets create a very simple example

We might as well use the example data provided on that rather nice NHS PDF document. It looks to be measuring time between falls.

Code 
# Create an object with the dates of the falls
fallsDates <- as.Date(c(
    "2014-03-02", "2014-03-06", "2014-03-07", "2014-03-15", "2014-03-22",
    "2014-04-01", "2014-04-11", "2014-04-14", "2014-04-26", 
    "2014-05-03", "2014-05-13", "2014-05-28",
    "2014-06-04", "2014-06-10", "2014-06-14",  "2014-06-21", "2014-06-30" ))

# Calculate the time between events
events <- c(NA, diff(fallsDates))

The NHS PDF has one feature I don’t like, it uses dates along the x-axis. I don’t think that’s great practice, as dates on an x-axis should usually be on a time scale and these are actually the specific dates of events. There may be times that’s useful - such as picking out events around holiday periods etc. But usually the x-axis for a t-plot should be a sequential line of event numbers.

In its simplest form qiplots2 will give as a t-chart with just this code:

Code 
require(qicharts2)

qic(
    y = events,
    x = 1:length(events),
    chart = "t"
)

There are clearly documented parameters to change titles, axis labels etc. So you can quickly move to a graph like this:

Code 
qic(
    y = events,
    x = 1:length(events),
    chart = "t",
    title = "Time between falls",
    subtitle = "SPC 't' chart",
    xlab = "Fall number",
    ylab = "Time (days)"
)

And you can either freeze data or split the data at a certain point if you want to see if an intervention is helping. For the sake of demonstration we will assume some falls reduction intervention was introduced after the 10th fall.

Code 
qic(
    y = events,
    x = 1:length(events),
    chart = "t",
    title = "Time between falls",
    subtitle = "SPC 't' chart",
    xlab = "Fall number",
    ylab = "Time (days)",
    part = 10
)

Formatting the result

This is a ggplot, but it hasn’t been created in a conventional ggplot way. That is you are not entering code along the lines of:

data |>
    ggplot2(aes(x = .., y = ..)) +
    geom_qichart()

If you were, your formatting would usually be controlled in the geom. So what can we do?

Well firstly, it is a ggplot object. So you can apply ggplot theme actions. For example:

Code 
require(ggplot2)
Loading required package: ggplot2
Code 
qic(
    y = events,
    x = 1:length(events),
    chart = "t",
    title = "Time between falls",
    subtitle = "SPC 't' chart",
    xlab = "Fall number",
    ylab = "Time (days)",
    part = 10
) +
    theme_bw()

Adding the theme, means if the theme has a legend it will appear, but we can turn that off as usual with a ggplot theme legend position none statement. All this means we can edit the plotting canvas, the size of text, the fonts, etc. But the actual data lines and the grey shaded 3 sigma zones (produced with a geom_ribbon are beyond easy reach).

Code 
qic(
    y = events,
    x = 1:length(events),
    chart = "t",
    title = "Time between falls",
    subtitle = "SPC 't' chart",
    xlab = "Fall number",
    ylab = "Time (days)",
    part = 10
) +
    theme_bw() +
    theme(text = element_text(colour = "blue") , 
             legend.position = "none")

Formatting the lines and shading

I couldn’t see an easy way to change the lines and shading. It might be possible by hacking a grob produced from the plot but frankly that felt like far too much effort. I might as well try and draw the graph from the raw data! This is open source software - which means we can see the code that produces the graph and thats always worth a look in these situations. The key functionfor plotting lies here.

plot.qic <- function(x, title, ylab, xlab, subtitle, caption, part.labels,
                     n.facets,
                     nrow, ncol, scales, show.labels, show.grid, show.95,
                     decimals, flip, dots.only, point.size,
                     x.format, x.angle, x.pad,
                     y.expand, y.percent, y.percent.accuracy, strip.horizontal,
                     # col.line = '#5DA5DA', 
                     # col.signal = '#F15854', 
                     # col.target = '#059748',
                     ...)

Which in its-self is interesting. Because there are three commented out variables in the function which may be a future development.


  col1      <- '#8C8C8C' # rgb(140, 140, 140, maxColorValue = 255) # grey
  col2      <- getOption('qic.linecol', default = '#5DA5DA')       # blue
  # col3      <- getOption('qic.signalcol', default = '#F15854')     # red
  col3      <- getOption('qic.signalcol', default = '#FAA43A')     # amber
  col4      <- getOption('qic.targetcol', default = '#059748')     # green
  col5      <- '#C8C8C8' # rgb(200, 200, 200, maxColorValue = 255) # light grey
  cols      <- c('col1' = col1,
                 'col2' = col2,
                 'col3' = col3,
                 'col4' = col4,
                 'col5' = col5)

So, actually - the plotting function is looking up its colours from a set of options called qic.[variable] and applying some defaults if these are not set. That means we can set those variables and change the colours.

Code 
currentOptions <- options()
options(qic.linecol = "red")

qic(
    y = events,
    x = 1:length(events),
    chart = "t",
    title = "Time between falls",
    subtitle = "SPC 't' chart",
    xlab = "Fall number",
    ylab = "Time (days)",
    part = 10
) +
    theme_bw() +
    theme(text = element_text(colour = "blue") , 
             legend.position = "none")

That leaves just the ribbon to try and edit.

# Add control limits and centre and target lines
  if (getOption('qic.clshade', default = TRUE) &
      sum(!is.na(x$lcl))) {
    p <- p +
      geom_ribbon(aes_(ymin = ~ lcl, ymax = ~ ucl),
                  fill = 'grey87',
                  alpha = 0.4)
  } else {
    p <- p +
      geom_line(aes_(y = ~ lcl), colour = col1, na.rm = T)
    
    p <- p +
      geom_line(aes_(y = ~ ucl), colour = col1, na.rm = T)
  }

So there is a further option, completely undocumented as far as I can see, that switches on / off the shading.

Code 
currentOptions <- options()
options(qic.linecol = "red", qic.clshade = F)

qic(
    y = events,
    x = 1:length(events),
    chart = "t",
    title = "Time between falls",
    subtitle = "SPC 't' chart",
    xlab = "Fall number",
    ylab = "Time (days)",
    part = 10
) +
    theme_bw() +
    theme(text = element_text(colour = "blue") , 
             legend.position = "none")

And there we have it. A chart with line colours edited, text adjusted and the shading removed. Why the package authors didn’t do it with geom_qichart I don’t quite know. But if you need a customisable qichart for a time series, I hope these notes help.

— Content on this website has been produced with reasonable care. However, errors can occur. Opinions change over time. Code that worked at the time of publication may no longer work. Users of all content from this site do so at their own risk.

References

1. Reading, C., Wellesley-Miller, S., Turner, Z., Jemmett, T., Smith, T., Mainey, C., & MacKintosh, J. (2021). NHSRplotthedots: Draw XmR charts for NHSE/i ’making data count’ programme. https://CRAN.R-project.org/package=NHSRplotthedots
2. Anhoej, J. (2024). qicharts2: Quality improvement charts. https://CRAN.R-project.org/package=qicharts2

Reuse

Citation

BibTeX citation:
@misc{polwart2024,
  author = {Polwart, Calum},
  title = {Formatting Qicharts2},
  date = {2024-10-11},
  url = {https://www.chemo.org.uk/posts/Formatting-qicharts2.html},
  langid = {en}
}
For attribution, please cite this work as:
1. Polwart, C. (2024, October 11). Formatting qicharts2. Chemo.org.uk. https://www.chemo.org.uk/posts/Formatting-qicharts2.html
]]> https://www.chemo.org.uk/posts/Formatting-qicharts2.html Thu, 10 Oct 2024 23:00:00 GMT Self-assessment of Compliance with British Oncology Pharmacy Association (BOPA) Standards for Clinical Verification of Systemic Anti-Cancer Therapies (SACT) https://www.chemo.org.uk/posters/BOPAClinicalPharmacy.html This poster was an attempt to measure resources and its relationship to compliance with verification.

The sample size was not enough to make reliable guesses on resource vs capacity. However, there was clearly variation in compliance.

— Content on this website has been produced with reasonable care. However, errors can occur. Opinions change over time. Code that worked at the time of publication may no longer work. Users of all content from this site do so at their own risk.

Reuse

Citation

BibTeX citation:
@misc{polwart2024,
  author = {Polwart, Calum and Gabriel, Sumantha},
  title = {Self-Assessment of {Compliance} with {British} {Oncology}
    {Pharmacy} {Association} {(BOPA)} {Standards} for {Clinical}
    {Verification} of {Systemic} {Anti-Cancer} {Therapies} {(SACT)}},
  date = {2024-10-11},
  url = {https://doi.org/10.6084/m9.figshare.26962102},
  doi = {10.6084/m9.figshare.26962102},
  langid = {en}
}
For attribution, please cite this work as:
1. Polwart, C., & Gabriel, S. (2024, October 11). Self-assessment of Compliance with British Oncology Pharmacy Association (BOPA) Standards for Clinical Verification of Systemic Anti-Cancer Therapies (SACT). https://doi.org/10.6084/m9.figshare.26962102
]]> BOPA Poster Verification https://www.chemo.org.uk/posters/BOPAClinicalPharmacy.html Thu, 10 Oct 2024 23:00:00 GMT Selection of an optimal drug-drug interaction database for use in oncology and haematology settings https://www.chemo.org.uk/posters/InteractionsPoster.html This poster started life following the announcement that the Liverpool Cancer Drug Interaction website was closing down. Discussions ensued about good alternatives. We identified 6 alternative candidates and assessed each against Liverpool, against a previously published paper[1] and an adaption of a method[2] to combine the results of multiple tools.

— Content on this website has been produced with reasonable care. However, errors can occur. Opinions change over time. Code that worked at the time of publication may no longer work. Users of all content from this site do so at their own risk.

References

1. Marcath, L. A., Xi, J., Hoylman, E. K., Kidwell, K. M., Kraft, S. L., & Hertz, D. L. (2018). Comparison of Nine Tools for Screening Drug-Drug Interactions of Oral Oncolytics. Journal of Oncology Practice, 14(6), e368–e374. https://doi.org/10.1200/JOP.18.00086
2. Abarca, J., Colon, L. R., Wang, V. S., Malone, D. C., Murphy, J. E., & Armstrong, E. P. (2006). Evaluation of the performance of drug-drug interaction screening software in community and hospital pharmacies. Journal of managed care pharmacy : JMCP, 12(5), 383–389. https://doi.org/10.18553/jmcp.2006.12.5.383

Reuse

Citation

BibTeX citation:
@misc{polwart2024,
  author = {Polwart, Calum and Waterson, Chloe and Nijjar, Rajinder},
  title = {Selection of an Optimal Drug-Drug Interaction Database for
    Use in Oncology and Haematology Settings},
  date = {2024-10-11},
  url = {https://doi.org/10.6084/m9.figshare.26965504},
  doi = {10.6084/m9.figshare.26965504},
  langid = {en}
}
For attribution, please cite this work as:
1. Polwart, C., Waterson, C., & Nijjar, R. (2024, October 11). Selection of an optimal drug-drug interaction database for use in oncology and haematology settings. https://doi.org/10.6084/m9.figshare.26965504
]]> BOPA Poster Prescribing https://www.chemo.org.uk/posters/InteractionsPoster.html Thu, 10 Oct 2024 23:00:00 GMT Engaging with Social Media Calum Polwart https://www.chemo.org.uk/posts/Engaging_posters.html This has been bugging me for a little while. I was going to do a twitter thread about it. But actually, twitter has been bugging me to. So I thought I might post it to mastodon, but of course no-one will read it there. So I’m going back to the ‘olden days’ and blogging about it. I’m posting it up a few days before #BOPA2024 in the hope that it stirs a few thoughts.

If I happen to have picked your social media post or your poster to comment on, please don’t be offended. They were mostly picked at random from posters in my area of work, that I’ve looked at this weekend, to show what people do that could improve. I could have picked 100’s of other examples. And I’m sure people will reply with examples of mine that are worse.

The scenario

You’ve done a piece of work, you’ve analysed your data and somehow shrunk your data to 3-500 words to submit as an abstract to your flashy conference. Its been accepted. (Well done - remember most conferences are rejecting a proportion of submissions). You’ve hopefully now produced a beautiful poster and had it nicely printed. Its on display at the conference. Your job is done, right?

Displaying your poster

Surely, all you have to do with your poster is pin it to the poster board, stand near it for your allotted time and be ready to answer some questions, guide people through your poster. After-all, that’s why people come to conferences isn’t it. There is nothing else left for you to do, if no-one engages with you and your poster either it is the audience’s fault for not having enough insight to realise how ground breaking your scientific discovery is, or your science was rubbish and you should never present again. Right?

It doesn’t all stop when you display the poster. In-fact, your work is only just beginning! Firstly, who is your real audience. The 200 people who might walk past the poster at your allotted time? The other delegates who have gone for coffee instead? Or everyone in your profession in the world? (OK that sounds a bit scary - but there is a very high chance there are people who are interested in what you’ve done who aren’t at your conference.)

If you are into social media, you’ll know it is very common to post he proverbial “selfie” with your poster. So well done - you’ve shared your poster with the world… haven’t you?

Engaging Social Media - is probably its own Science

Public relations, publicity and whatever you collectively label “influencer” culture as is almost certainly its own science. Major PR departments test different engagement approaches (that’s what they call getting you to spend time thinking about their product) using A versus B analysis in much the same way as “proper” science people might test two drugs against each other to work out what is best. I’m not in any way claiming to be an “influencer” or an expert on Social Media. But there are some things I can tell you are “wrong”. With “wrong” being that they don’t help me to engage with your ‘product’. Your product is your research, your poster and your take home message. I’m hoping by addressing some of what I find hard to engage with, others can learn from other people’s mistakes and improve their own poster engagement.

Does Social Media really matter?

I’ve presented posters at conferences, been put on a poster board at the very far corner and not a single person has spoken to me. That’s pretty de-moralising.

I’ve never posted a poster on social media and not been able to at least see people looked at the poster. And almost always, someone replies, re-tweets/re-posts or likes your poster. If they don’t, it might be that you’ve missed some of the things below that I think affect engagement.

If one person not at the conference engages in a full conversation online about my poster, and no-one bothers to stop and ask questions in person - I still think that was worth displaying the poster.

The Selfie

I’m sure “selfies” are great at facilitating engagement. Your ‘followers’ are scrolling their endless stream of cat pictures and suddenly see a familiar face. They stop. They read that you are at a conference. Great. You’ve got the start of some engagement.

Indeed that sort of thing is what happens as you browse the poster walls at a conference. People stop to look at posters of people they know. Two things then come to mind here: -

  1. If your friends are stopping by, you could have sent them a text with your results;

  2. At least let them look at your poster.

In the @Florez_Lab example above, the presenters are unfortunately in front of the poster. If only they’d stepped aside of the poster like this tweet:

Now you may well say “but I cant read the poster” - but you can usually download the image from the social media platform and zoom it (even on a phone). Here is the same image cropped to just the poster:

Zoomed image from @coffeemommy’s tweet above

That image is just about readable.

While aesthetically, this angled picture is probably far more pleasing to the eye than the harshness of a straight on poster picture, it doesn’t make reading the poster easier:

No-one says you can only post one image. Here is a good example of the aesthetically pleasing and the poster for the interested reader

Since the purpose of my blog is to highlight things that could improve. Shiny posters, reflecting light are hard to read. So if you are on the ball - export a copy of the poster as a “.png” file to your phone from your ‘desktop publishing’ software (OK you all use powerpoint - but it can do it too!). Then when you are tweeting why not simply use that picture with your selfie?

The key message

If you’ve ever followed any of Mike Morrison’s better poster videos, you’ll probably know he is a massive fan of a single take away sentence describing the findings. In @coffeemommy’s tweet above it appears at the top of the third column of text. I’m not sure I know what the words mean - but I cant see that she anywhere tweeted the words:

“Patients, care partners, and community members have contextual knowledge which is not only relevant to the design and development of processes, products and services, but also, through the process of co-creation, these individuals tend to become further involved, motivated, and influenced by the very element they are co-creating.”

Search engines can’t search words on poster pictures (at least, not in a robust and reliable way). So if I search for “co-creation” and “patients” the picture won’t show up.

If @coffeemommy had replied to her selfie with the summary text she would have made her post more searchable and her message more engaging. That said, as she says - she seems to have been kept busy at the poster show so maybe she didn’t need more engagement. However, planned in advance, while writing the poster, she might have spotted the key message was actually quite wordy (too long for a standard 280 character tweet).

The angled tweet from @hafaydeefe above does this very nicely - Two sentences tell me what the poster is about.

The other thing worth considering is ALT text. Introduced as an alternative to showing the picture in circumstances where the picture isn’t rendered - for instance a very slow internet connection, or someone using a screen reader who has visual impairment. ALT text will make the picture far more searchable.

The Posters

I set out to blog about how people use Social Media to engage their posters, not to talk poster design. But, I will re-iterate my desire for people to watch Mike Morrison’s original video. And this is important - watch the video. Learn from the video. Using (abusing) his template, without watching the video may be a worse sin in poster publishing than sticking to the age old scientific poster in three columns format.

White space is crucial to readability

Justification of text is dreadful for readability

Here are some decent attempts to deliver some of what Mike talks about:

We will ignore the fact they are comiting the ‘stand in front of my poster’ sin! I haven’t tried to read their key messages, I suspect they are still too wordy.

@JorgeOct_ has got the selfie thing almost right - only just obstructing his poster. But please @JorgeOct_ - go and re-listen to Mike’s thoughts on white space, margins, text justification etc. I’m not sure if Mike actually mentions it - but underline is another graphics design sin that reduced readability.

In case you’ve not seen Mike’slatest work - you might want to check out his latest video:

He is less an advocate of the bill board and more an advocate of changing the rules these days.

The QR Code

While Mike wasn’t the first person to use a QR code on a poster, they do seem more popular since he promoted it. Mike’s suggestion was that the QR code would actually lead to a whole paper about the poster. I don’t think I’ve ever seen it done.

I like a QR code. At the conference it is absolutely the easiest way to access a ‘hard’ copy of the poster. On Social Media… it is frustrating to say the least - just tag a link in the post!

Depending how you create your QR code, or where you link it to (fig.share is my favourite, ResearchGate is also good) - you can see how often people engage with your poster. You’ll find people interacting months and years later sometimes. I find people locate things on Research Gate more than they maybe do on fig.share. But I post the poster to fig.share as it lets me embargo it and get a DOI for it. I then link the DOI on ResearchGate after the conference.

The Hashtag

Most conferences will have a #hashtag. It goes without saying that your social media post needs to tag that hash tag to be found by interested people. And make sure you get the right hashtag. I don’t think there is a year goes by without me tagging #BOPA23 instead of #BOPA2023 etc.

Tagging others

If you tag other people they usually get notified about your post. That’s a bit like stopping them as they walk past your poster. But there is a good chance they may re-post your post. If you only have 3 followers on twitter, this may well help your work get more noticed. If you are keen to build your social media following, then that probably increases the chances of you picking up some new follows too.

Don’t make your post a hashtag and twitter tag spam. If need be post them as replies. Keep the messages readable.

A bit shy?

I’m British. I’m not sure Social Media quite works for the traditional British introverted sciencey person. We don’t always like to shout about what we’ve done. But you did a thing, and got it accepted by your peers at a conference. You should be proud. But, if you aren’t sure - why not use someone else to promote your work. Just say to them “Hey, I’m not sure I get all this ‘post a selfie’ thing with my poster but I know it helps engagement. I don’t suppose you’d post it to look less contrived”. I’ve done that before with people and it works well in my opinion. In fact, once you realise it might be a thing - you see it happen far more than the selfies:

Tell a story

I tried this last year, and I think it worked well. Its my general approach to presenting anyway, to try and explain the story rather than just present some data.

Oh and when I said “a bit shy” - @chloewaterson88 didn’t seem that shy with this quote tweet of the thread…

Because we were telling a story about HOW we did the analysis as well as why and the results we were able to tag a few people outside of the particular conference sphere we were attending.

The platform

So far, I’ve been talking about twitter. Twitter (I refuse to be an X user!) thinks of its self as a town square. A place where people can openly sharing information. Prior to recent developments it was pretty decent for that. It still is. But you may well have objections about its owner, its politics, its values etc. There are other platforms. I do post some stuff to mastondon - although I have seen far far less engagement and a bit like taking a poster to a conference and no-one talking to you - posting and no engagement is a bit disheartening.

If you don’t want to be on twitter - work out where else is worth you sharing stuff. But remember - someone had to do it first on twitter for it to become a thing. So perhaps you need to be the first and tell others about it?

Prepare in advance

You will be busy at the conference. Do you actually have time to be writing out tweets about your poster and adding ALT text and finding people to tag in. You already know your message. You already have a copy of your poster. Why not key it up to post (twitter lets you schedule posts, as do various other free platforms that link to twitter). Then you can concentrate on the replies.

This always catches me off guard - I will go into my phone an hour after it posted and see 15 notifications having forgotten I’d pre-set it to post at a certain time.

That’s all folks…

Anyway, I hope this is helpful. I hope to see LOTS of social media engagement at #BOPA24 in 12 days time. There are 135 posters. There is less than 2 hours when poster presenters are to be present with their poster - so less than 1 minute per poster. So don’t be surprised if you need to do something to get people to notice your poster.

— Content on this website has been produced with reasonable care. However, errors can occur. Opinions change over time. Code that worked at the time of publication may no longer work. Users of all content from this site do so at their own risk.

Reuse

Citation

BibTeX citation:
@misc{polwart2024,
  author = {Polwart, Calum},
  title = {Engaging with {Social} {Media}},
  date = {2024-09-29},
  url = {https://www.chemo.org.uk/posts/Engaging_posters.html},
  langid = {en}
}
For attribution, please cite this work as:
1. Polwart, C. (2024, September 29). Engaging with Social Media. Chemo.org.uk. https://www.chemo.org.uk/posts/Engaging_posters.html
]]> https://www.chemo.org.uk/posts/Engaging_posters.html Sat, 28 Sep 2024 23:00:00 GMT More High Dose Dexamethasone? Calum Polwart https://www.chemo.org.uk/posts/MoreHigh-Dose-Dexamethasone.html Just before the world stopped for the COVID-19 pandemic, I wrote a blog post on TheDataLab’s website about High Dose Dexamethasone. I was worried that 40mg Dexamethasone seemed to be being dispensed in Primary Care. This is a huge dose and would normally only have use in Haemato-oncology as part of a handful of specialist cancer treatments. In fact, even in secondary care, I would be worried that the risk of dispensing error was too high to justify keeping this in stock.

Distracted by the pandemic I kept an eye on prescribing rates and noticed they were descending. Only last year I discovered Dexamethasone 40mg Tablets had been discontinued. We still don’t know if these dispensing events we saw were real or were data anomalies. I suggested to Chloë Waterson that she might want to take a look at the data when she was looking to learn R; whereby she told me it was no more! She had a look at some even scarier stuff instead![1]

So here is a very brief look at what happened, complete with code for those who like a simple R project

Get the Data from Open Prescribing

Code 
require(tidyverse, warn.conflicts = F)
require(httr, warn.conflicts = F)
require(jsonlite, warn.conflicts = F)
DexId <- "0603020G0AACJCJ" # This is the product code for 40mg Dex on Open Rx

url <- glue::glue("https://openprescribing.net/api/1.0/spending/?code={DexId}&format=json")
response <- GET(url)

# translate the API response into a datatable
data <- fromJSON(rawToChar(response$content))

# translate dates as dates not text
data |>
    mutate(date = as.Date(date)) -> data

# show the data
DT::datatable(data)

Graph the Data

Code 
require(ggplot2)
data |>
    ggplot(aes(y=quantity, x=date)) +
    geom_col() +  # creates a basic column graph
    theme_bw() +  # cleaner theme
    ggtitle("Number of 40mg Dexamethasone tablets dispensed each month") +
    expand_limits(x=as.Date("2024-07-01")) + # stretch the x axis to show all the missing right hand data points
    annotate(   # label when the original blog happened
    'text',
    x = as.Date("2019-08-15"),
    y = 75,
    label = 'Blog\nPublished',
    fontface = 'bold', 
    size = 3.5,
    angle = 0
  ) +
  annotate(
    'curve',
    x = as.Date("2019-11-15"),
    y = 50,
    yend = 28,
    xend = as.Date("2019-11-15"),
    linewidth = 1,
    curvature = 0,
    arrow = arrow(length = unit(0.2, 'cm'))
  ) -> g1

print(g1)

We can repeat the process for the number of prescriptions dispensed instead of the number of tablets dispensed. And we can combine those using cowplot.

Code 
data |>
    ggplot(aes(y=items, x=date)) +
    geom_col() +  # creates a basic column graph
    theme_bw() +  # cleaner theme
    ggtitle("Number of 40mg Dexamethasone prescriptions dispensed each month") +
    expand_limits(x=as.Date("2024-07-01"), y=20) + # stretch the x axis to show all the missing right hand data points
    annotate(   # label when the original blog happened
    'text',
    x = as.Date("2019-08-15"),
    y = 15,
    label = 'Blog\nPublished',
    fontface = 'bold', 
    size = 3.5,
    angle = 0
  ) +
  annotate(
    'curve',
    x = as.Date("2019-11-15"),
    y = 11,
    yend = 6.5,
    xend = as.Date("2019-11-15"),
    linewidth = 1,
    curvature = 0,
    arrow = arrow(length = unit(0.2, 'cm'))
  ) -> g2

print(g2)

Code 
require(cowplot)

cowplot::plot_grid(g1, g2, ncol = 1)

So it definitely looks like I can stop worrying about 40mg Dexamethasone Tablets. But that, instantly made me wonder what happened to the other products I’d been a bit worried by - the 10mg capsule. Well it seems when the 40mg tablet was discontinued a 10mg and 20mg tablet that were launched at the same time!1

Other Tablet Strengths

A browse through the dm+d directory finds these possible suspects:

“Higher” Strength Dexamethasone Tablets
Description BNF Code
Dexamethasone 10mg capsules 0603020G0AABYBY
Dexamethasone 10mg soluble tablets sugar free 0603020G0AACLCL
Dexamethasone 20mg soluble tablets sugar free 0603020G0AACMCM

These products do still appear to be available, based on the dm+d directory at least. If we re-run the data download for each of these codes we can create a combined graph. If there was a genuine clinical use for the 40mg tablets, we might expect to see an increase in 20mg dispensing when the 40mg becomes unavailable.

Get the data

Code 
# Add a strength column to the original data
data |>
    mutate(strength = "40mg") -> data

ids = c("10mg" ="0603020G0AABYBY",
        "10mg" = "0603020G0AACLCL", 
        "20mg" = "0603020G0AACMCM")

# use a loop to repeat the download for each id
for ( DexId in ids ) {

    url <- glue::glue("https://openprescribing.net/api/1.0/spending/?code={DexId}&format=json")
    response <- GET(url)
    newdata <- fromJSON(rawToChar(response$content))
    newdata |>
        mutate(date = as.Date(date)) |>
        mutate(strength = names(ids[ids == DexId]))-> newdata
    
    data <- rbind(data, newdata)
}


url <- glue::glue("https://openprescribing.net/api/1.0/spending/?code={DexId}&format=json")
response <- GET(url)

# translate the API response into a datatable
newdata <- fromJSON(rawToChar(response$content))

# translate dates as dates not text
newdata |>
    mutate(date = as.Date(date)) -> newdata

# show the data
DT::datatable(data)

It looks like we do have some higher strength dispensing happening in 2024!

Graph the new data

That’s certainly not a reassuring picture! While the majority of the use is 10mg, even 10mg is a relatively high dose for prescribing in primary care. The equivalent prednisolone dose for 10mg of dexamethasone is over 65mg.

This graph is more curious. When 40mg tablets were being dispensed (up to the end of 2022), there was a median of 2 prescriptions per month for 22 tablets per month. Between July-19 and June-21, 18 grams of dexamethasone was dispensed. If the 10mg and 20mg tablets directly replaced these, then a similar number of grams should be being dispensed in a similar time span. In fact, since the introduction of 20mg tablets, a median of 7 prescriptions per month and 83 tablets per month have been dispensed.

In the time period July-22 to June-24, 70.47 grams of dexamethasone were dispensed, representing a 4 fold increase in the use of higher strength steroid tablets.

Conclusion

The discontinuation of 40mg Dexamethasone Tablets resulted in a brief period without dispensing of any high strength dexamethasone prescriptions. While high doses may still have been dispensed using lower strength tablets, the potential to identify these as possibly erroneous is increased by the dispensing of large tablet counts to achieve high doses. However, following the introduction of 10 and 20mg tablets there now seems to be a 4 fold increase in the dispensing of high doses of steroids using higher strength tablets. I am not reassured!

— Content on this website has been produced with reasonable care. However, errors can occur. Opinions change over time. Code that worked at the time of publication may no longer work. Users of all content from this site do so at their own risk.

References

1. Waterson, C., & Polwart, C. (2023). Prescribing of Systemic Anti-Cancer Therapies (SACT) Through Primary Care in England. Journal of Oncology Pharmacy Practice, 30(1_suppl). https://doi.org/10.1177/10781552241228011

Footnotes

  1. There were also 4mg and 8mg tablets launched back when the 40mg tablet was launched. While they still have risks of dispensing errors, the consequential risks is lower and 4 and 8mg are doses that would be commonly prescribed in primary care.↩︎

Reuse

This post is released under CC-BY-4.0 https://creativecommons.org/licenses/by/4.0/. Data shown is sourced from Open Prescribing, who obtain it from NHS BSA under an Open Government License.

Citation

BibTeX citation:
@misc{polwart2024,
  author = {Polwart, Calum},
  title = {More {High} {Dose} {Dexamethasone?}},
  date = {2024-08-25},
  url = {https://www.chemo.org.uk/posts/MoreHigh-Dose-Dexamethasone.html},
  langid = {en}
}
For attribution, please cite this work as:
1. Polwart, C. (2024, August 25). More High Dose Dexamethasone? Chemo.org.uk. https://www.chemo.org.uk/posts/MoreHigh-Dose-Dexamethasone.html
]]> OpenPrescribing Data Dexamethasone https://www.chemo.org.uk/posts/MoreHigh-Dose-Dexamethasone.html Sat, 24 Aug 2024 23:00:00 GMT Strange citations? Calum Polwart https://www.chemo.org.uk/posts/StrangeCitations.html I’m not particularly well published despite what some people seem to think. As a result I have Google Scholar set up to tell me if someone cites anything I’ve published. Yes, that might be a little “sad”, but I think its sort of helpful to know if anyone finds anything your write useful!

Back in 2022, I published a paper with two of my colleagues[1]. It describes the real world outcomes of some anticancer drugs. Its an OK paper. Its been cited a couple of times by people discussing real world data. But its also been cited twice by the same authors seemingly incorrectly. The first citation somehow slipped past me. However, the second by[2] caught my attention.

The paper is titled “Biological Mediators and Partial Regulatory Mechanisms on Neuropathic Pain Associated With Chemotherapeutic Agents” published in a Journal called Physiological Research. That may not seem too strange to reference our paper; after-all our paper is about chemotherapy. However, our paper doesn’t discuss neuropathic pain at all. So I hunted out a copy of the paper and tried to read it and understand what it was discussing and why they felt our paper helped their story. I confess, I managed about three lines before finding our paper in the references and locating where we were cited.

It turned out the only place there was a citation was in this table:

We are reference 48. Our paper doesn’t mention cabazitaxel or docetaxel at all, nor does it address Ovarian, breast lung or prostate cancer. It does discuss paclitaxel, in a very specific formulation, but like I say, we didn’t mention anything to do with neuropathy in our paper.

I tend to think the best of people, and I wondered if something had very innocently gone wrong in the referencing process. Perhaps a wrong digit in a DOI, or clicking the wrong paper in a reference manager.

While pondering quite what might have happened, I discovered the other paper[3]. Same four authors. This paper is also about chemotherapy induced neuropathy. Published a year earlier. Although when you start to forensically analyse the submissions it was submitted on 5th July last year and accepted for publication 21st July. The[2] paper was submitted on 15th June 2023 and accepted on the 13th September 2023. For those who maybe aren’t familiar with publishing - when you submit you are specifically asked if the paper is being considered elsewhere.

Of course I rushed off and read[3] (well the first 3 lines!) and searched for our citation. Low and behold, the reference was on a table. We were even reference 48 again!

By now, I was starting to be suspicious that this was not an accidental referencing error. Two almost identical articles, with identical tables, submitted in parallel. I decided to write to the editors of both Journals. The editor of Physiological Research replied relatively quickly. He was unaware of the duplicate publication and obviously defended that his reviewers couldn’t check every citation (I hadn’t expected they could!). He promised to contact the authors for explanation.

Interestingly, the authors responded relatively quickly to the editor. It was a single paragraph response, but it was a response. It seems that we were cited to prove that paclitaxel was actually used as chemotherapy.

Reference 48 mentioned that paclitaxel and other categories of agents have been generally used for anti-cancer therapy and thus for this purpose it was cited in the table 1 of our review paper.

As for dual publication, well it seems they don’t think that applies to review articles.

Both papers are review article, but not original research paper.

And that’s where Physiological Research have decided to leave matters. The “Cancer Therapy & Oncology International Journal” have not responded at all.

I posted by thoughts to twitter and one of the guru’s there who spots papermills chipped in to discuss things. I wasn’t 100% sure I could label this as a papermill for picking a dubious reference and plagiarising their own text. However, he looked at the first sentence of both papers:

Cancer is a leading cause of mortality worldwide[@finnerup2021].

Now if you were going to write that sentence, you’d probably reference it to something like the World Health Organisation. Instead Liu et al referenced Finnerup et al - a paper titled “Neuropathic Pain: From Mechanisms to Treatment”. In case it is a surprise for readers, that paper doesn’t even mention cancer mortality!

— Content on this website has been produced with reasonable care. However, errors can occur. Opinions change over time. Code that worked at the time of publication may no longer work. Users of all content from this site do so at their own risk.

References

1. Wadd, N., Peedell, C., & Polwart, C. (2022). Real-World Assessment of Cancer Drugs Using Local Data Uploaded to the Systemic Anti-Cancer Therapy Dataset in England. Clinical Oncology, 34(8), 497–507. https://doi.org/10.1016/j.clon.2022.04.012
2. Liu, Z., Liu, S., Zhao, Y., & Wang, Q. (2024). Biological Mediators and Partial Regulatory Mechanisms on Neuropathic Pain Associated With Chemotherapeutic Agents. Physiological Research, 3/2024, 333–341. https://doi.org/10.33549/physiolres.935162
3. Wang, Q. (2023). Chemotherapy-induced neuropathic pain and underlying mechanisms. Cancer Therapy & Oncology International Journal, 24(4). https://doi.org/10.19080/ctoij.2023.24.556142

Reuse

Citation

BibTeX citation:
@misc{polwart2024,
  author = {Polwart, Calum},
  title = {Strange Citations?},
  date = {2024-08-22},
  url = {https://www.chemo.org.uk/posts/StrangeCitations.html},
  langid = {en}
}
For attribution, please cite this work as:
1. Polwart, C. (2024, August 22). Strange citations? Chemo.org.uk. https://www.chemo.org.uk/posts/StrangeCitations.html
]]> https://www.chemo.org.uk/posts/StrangeCitations.html Wed, 21 Aug 2024 23:00:00 GMT Prescribing of Systemic Anti-Cancer Therapies (SACT) through Primary Care in England https://www.chemo.org.uk/posters/Primary Care Prescribing.html This project started out as a simple exercise in playing with some data and showing a colleague how she might make use of R. I’d looked previously at some data on Open Prescribing regarding high dose dexamethasone which I felt had no place in primary care (or possibly anywhere!) I’ve written about that before on TheDataLab’s website. I initially suggested to Chloë that we might do an update of that data using Open Prescribing’s API. In fact, it turns out Dexamethasone 40mg Tablets have been discontinued. However, in that blog we discussed there were other “worrying” items being recorded as dispensed. Chloë felt it was worth a look and we went for a bit of dive into the data.

I’ve explained a bit more about the back story to the poster on this twitter thread:

Perhaps easiest read on TypeFully’s combined version.

Chloë’s poster was selected as a winning poster for BOPA 2023. Probably not too surprising when you consider she discovered about a million pounds of prescribing of cancer treatments that we didn’t expect.

PDF version of poster is available on Fig Share.

BOPA Poster

Chloë’s work also produced a list of cancer treatments that we felt were questionable for prescribing in primary care. The list is available on Open CodeList’s website if any other researchers wanted to re-use it. All Chloë’s code is available on GitHub too.

— Content on this website has been produced with reasonable care. However, errors can occur. Opinions change over time. Code that worked at the time of publication may no longer work. Users of all content from this site do so at their own risk.

Reuse

Citation

BibTeX citation:
@article{waterson2023,
  author = {Waterson, Chloe and Polwart, Calum},
  publisher = {Sage Journals},
  title = {Prescribing of {Systemic} {Anti-Cancer} {Therapies} {(SACT)}
    Through {Primary} {Care} in {England}},
  journal = {Journal of Oncology Pharmacy Practice},
  volume = {30},
  number = {1\_suppl},
  date = {2023-10-06},
  url = {https://doi.org/10.1177/10781552241228011},
  doi = {10.1177/10781552241228011},
  issn = {1078-1552},
  langid = {en}
}
For attribution, please cite this work as:
1. Waterson, C., & Polwart, C. (2023). Prescribing of Systemic Anti-Cancer Therapies (SACT) through Primary Care in England. Journal of Oncology Pharmacy Practice, 30(1_suppl). https://doi.org/10.1177/10781552241228011
]]> BOPA Poster Prescribing https://www.chemo.org.uk/posters/Primary Care Prescribing.html Thu, 05 Oct 2023 23:00:00 GMT NHSnames Calum Polwart https://www.chemo.org.uk/posts/NHSnameR.html

Introduction

When I did some work on NHS Data in R, I discovered the challenge of NHS Organisation names. Firstly, they are long and clunky:

The Somewhere and Somewhere Else NHS Foundation Trust

is hardly a snappy name. And so people abbreviate it. Often within a locality people fairly consistently abbreviate, but there will be no easy way to unpick that. Using the daft example above you could abbreviate it:

  • TSSE

  • TSSE-FT

  • TSSE-NHS FT

  • SSE

  • S&SE

  • S+SE

  • SSFT

etc

In addition trusts can be known by their 3 letter ODS code, and it is sometimes desirable to translated the ODS code to the full name or a shortened name.

R Package

I have therefore created a few functions to try and make the process simpler and perhaps more consistent. And I’ve rapped these into a simple package.[1]

install.packages("remotes")

remotes::install_gitlab("polc1410/nhsnames")

— Content on this website has been produced with reasonable care. However, errors can occur. Opinions change over time. Code that worked at the time of publication may no longer work. Users of all content from this site do so at their own risk.

References

1. Polwart, C. (2024). NHSnames: Format NHS Organisation Names for Easier Reading. https://gitlab.com/polc1410/nhsnames

Reuse

]]> R Data https://www.chemo.org.uk/posts/NHSnameR.html Sat, 26 Aug 2023 23:00:00 GMT SACTanalyseR Calum Polwart https://www.chemo.org.uk/posts/SACTanalyseR.html

Introduction

When I wrote this paper[1] using a collection of CSV files that had been previously submitted to the National Cancer Registration and Analysis Service I needed t write a few functions to handle the files. Things ranging from merging the CSV files to factorizing data and updating dates of death.

The dataset is in two formats - version 2 and version 3, and is fully described in published literature[2].

R Package

I proposed to the NHS R Community that we could develop this set of functions into an R package, which is now released in an alpha version at: https://gitlab.com/polc1410/sactanalyser

— Content on this website has been produced with reasonable care. However, errors can occur. Opinions change over time. Code that worked at the time of publication may no longer work. Users of all content from this site do so at their own risk.

References

1. Wadd, N., Peedell, C., & Polwart, C. (2022). Real-World Assessment of Cancer Drugs Using Local Data Uploaded to the Systemic Anti-Cancer Therapy Dataset in England. Clinical Oncology, 34(8), 497–507. https://doi.org/10.1016/j.clon.2022.04.012
2. Bright, C. J., Lawton, S., Benson, S., Bomb, M., Dodwell, D., Henson, K. E., McPhail, S., Miller, L., Rashbass, J., Turnbull, A., & Smittenaar, R. (2019). Data Resource Profile: The Systemic Anti-Cancer Therapy (SACT) dataset. International Journal of Epidemiology, 49(1), 15–15l. https://doi.org/10.1093/ije/dyz137

Reuse

]]> R Data https://www.chemo.org.uk/posts/SACTanalyseR.html Tue, 01 Nov 2022 00:00:00 GMT The Utility of an Evidence-based FAQ Document for COVID-19 Vaccination & Patients receiving Systemic Anti-cancer Therapies (SACT) https://www.chemo.org.uk/posters/COVID vaccines.html This poster was presented by Nisha Shaunak at the British Oncology Pharmacy Association Annual Symposium. The poster was developed to review the effectiveness of some collaborative guidelines that were developed during the COVID-19 pandemic. The pragmatic guidelines aimed to help clinicians make sound decisions about the appropriateness of administration of COVID-19 vaccines during the early phases of the pandemic.

The guidelines are available on the UK SACT Board Website.

COVID Vaccine Poster

The abstracts for this meeting are published in theJournal of Oncology Pharmacy Practice. The poster is available on Research Gate.

There is also a post on this site describing how the flow charts for this guidance were produced in an open, accessible and version controllable format.

— Content on this website has been produced with reasonable care. However, errors can occur. Opinions change over time. Code that worked at the time of publication may no longer work. Users of all content from this site do so at their own risk.

Reuse

Citation

BibTeX citation:
@article{shaunak2021,
  author = {Shaunak, Nisha and Nijjar, Raj and Polwart, Calum and
    Enting, Deborah},
  publisher = {Sage Journals},
  title = {The {Utility} of an {Evidence-based} {FAQ} {Document} for
    {COVID-19} {Vaccination} \& {Patients} Receiving {Systemic}
    {Anti-cancer} {Therapies} {(SACT)}},
  journal = {Journal of Oncology Pharmacy Practice},
  volume = {28},
  number = {2\_suppl},
  date = {2021-10-09},
  url = {https://doi.org/10.1177/10781552221078082},
  doi = {10.1177/10781552221078082},
  issn = {1078-1552},
  langid = {en}
}
For attribution, please cite this work as:
1. Shaunak, N., Nijjar, R., Polwart, C., & Enting, D. (2021). The Utility of an Evidence-based FAQ Document for COVID-19 Vaccination & Patients receiving Systemic Anti-cancer Therapies (SACT). Journal of Oncology Pharmacy Practice, 28(2_suppl). https://doi.org/10.1177/10781552221078082
]]> BOPA Poster Vaccines COVID-19 https://www.chemo.org.uk/posters/COVID vaccines.html Fri, 08 Oct 2021 23:00:00 GMT A National Service Evaluation of Paclitaxel Pre-medication Regimes for the Prevention of Hypersensitivity during a period of Ranitidine Shortage - initial data analysis https://www.chemo.org.uk/posters/Paclitaxel Reactions.html This poster was presented by Emma Foreman at the British Oncology Pharmacy Association Annual Symposium. The poster summarised some work evaluating the impact of omitting ranitidine from paclitaxel infusions. Ranitidine was withdrawn from the market due to concerns about contamination. However, ranitidine had been widely considered an essential component of the premedication for paclitaxel to reduce the risk of allergic reactions.

Foreman, Chambers and myself collated data from a number of centres to evaluate the impact of using an alternative H2 antagonist or no antagonist. This work has subsequently been published in the British Journal of Clinical Pharmacology as described on this page of this site.

Paclitaxel allergy Poster

The abstracts for this meeting are published in theJournal of Oncology Pharmacy Practice. The poster is available on Research Gate.

— Content on this website has been produced with reasonable care. However, errors can occur. Opinions change over time. Code that worked at the time of publication may no longer work. Users of all content from this site do so at their own risk.

Reuse

Citation

BibTeX citation:
@article{foreman2021,
  author = {Foreman, Emma and Chambers, Pinkie and Polwart, Calum},
  publisher = {Sage Journals},
  title = {A {National} {Service} {Evaluation} of {Paclitaxel}
    {Pre-medication} {Regimes} for the {Prevention} of
    {Hypersensitivity} During a Period of {Ranitidine} {Shortage} -
    Initial Data Analysis},
  journal = {Journal of Oncology Pharmacy Practice},
  volume = {28},
  number = {2\_suppl},
  date = {2021-10-09},
  url = {https://doi.org/10.1177/10781552221078082},
  doi = {10.1177/10781552221078082},
  issn = {1078-1552},
  langid = {en}
}
For attribution, please cite this work as:
1. Foreman, E., Chambers, P., & Polwart, C. (2021). A National Service Evaluation of Paclitaxel Pre-medication Regimes for the Prevention of Hypersensitivity during a period of Ranitidine Shortage - initial data analysis. Journal of Oncology Pharmacy Practice, 28(2_suppl). https://doi.org/10.1177/10781552221078082
]]> BOPA Poster Adverse Reactions Paclitaxel H2-Antagonists https://www.chemo.org.uk/posters/Paclitaxel Reactions.html Fri, 08 Oct 2021 23:00:00 GMT Rates of hospital emergency admission following chemotherapy for patients undergoing review by a Non-Medical Prescriber: A useful metric for service evaluation? https://www.chemo.org.uk/posters/emergency_addmisions.html Poster presented at British Oncology Pharmacy Association Annual Symposium 2020 This poster looked to identify the rate of hospital admission following non-medical prescriber review of patients undergoing systemic anti-cancer therapy, as a measure for service evaluation of Non-Medical Prescribing Clinics.

PDF version of poster is available on Research Gate.

— Content on this website has been produced with reasonable care. However, errors can occur. Opinions change over time. Code that worked at the time of publication may no longer work. Users of all content from this site do so at their own risk.

Reuse

Citation

BibTeX citation:
@article{polwart2020,
  author = {Polwart, Calum and Wadd, Nicholas},
  publisher = {Sage Journals},
  title = {Rates of Hospital Emergency Admission Following Chemotherapy
    for Patients Undergoing Review by a {Non-Medical} {Prescriber:} {A}
    Useful Metric for Service Evaluation?},
  journal = {Journal of Oncology Pharmacy Practice},
  volume = {27},
  number = {2\_suppl},
  date = {2020-10-09},
  url = {https://doi.org/10.1177/1078155220985357},
  doi = {10.1177/1078155220985357},
  issn = {1078-1552},
  langid = {en}
}
For attribution, please cite this work as:
1. Polwart, C., & Wadd, N. (2020). Rates of hospital emergency admission following chemotherapy for patients undergoing review by a Non-Medical Prescriber: A useful metric for service evaluation? Journal of Oncology Pharmacy Practice, 27(2_suppl). https://doi.org/10.1177/1078155220985357
]]> BOPA Poster Prescribing https://www.chemo.org.uk/posters/emergency_addmisions.html Thu, 08 Oct 2020 23:00:00 GMT Geographical variation in the use of fluorouracil in the treatment of Early Breast Cancer within England https://www.chemo.org.uk/posters/fluorouracil variations.html This poster was presented by Jess Pealing at the British Oncology Pharmacy Association Annual Symposium. The poster was developed to review the variation in the use of fluorouracil in the management of Early Breast Cancer within England. The ESMO guidelines[1] no longer recommend the routine use of fluoropyramidines, however, we were aware that many centres continued to use fluorouracil and we sought to identify the extent of variation.

The abstracts for this meeting are published in the Journal of Oncology Pharmacy Practice. The poster is available on Research Gate.

A number of people asked following this poster how we developed the map, and we have published a simplified version of the code on Research Gate.

— Content on this website has been produced with reasonable care. However, errors can occur. Opinions change over time. Code that worked at the time of publication may no longer work. Users of all content from this site do so at their own risk.

References

1. Cardoso, F., Kyriakides, S., Ohno, S., Penault-Llorca, F., Poortmans, P., Rubio, I. T., Zackrisson, S., & Senkus, E. (2019). Early breast cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Annals of Oncology, 30(8), 1194–1220. https://doi.org/10.1093/annonc/mdz173

Reuse

Citation

BibTeX citation:
@article{pealing2020,
  author = {Pealing, Jess and Rodriguez, Marga and Polwart, Calum},
  publisher = {Sage Journals},
  title = {Geographical Variation in the Use of Fluorouracil in the
    Treatment of {Early} {Breast} {Cancer} Within {England}},
  journal = {Journal of Oncology Pharmacy Practice},
  volume = {27},
  number = {2\_suppl},
  date = {2020-10-09},
  url = {https://doi.org/10.1177/1078155220985357},
  doi = {10.1177/1078155220985357},
  issn = {1078-1552},
  langid = {en}
}
For attribution, please cite this work as:
1. Pealing, J., Rodriguez, M., & Polwart, C. (2020). Geographical variation in the use of fluorouracil in the treatment of Early Breast Cancer within England. Journal of Oncology Pharmacy Practice, 27(2_suppl). https://doi.org/10.1177/1078155220985357
]]> BOPA Poster Fluorouracil Breast Cancer https://www.chemo.org.uk/posters/fluorouracil variations.html Thu, 08 Oct 2020 23:00:00 GMT High Dose Dexamethasone Brian MacKenna Calum Polwart https://www.chemo.org.uk/posts/High-Dose-Dexamethasone.html Here is a link to a blog post, I wrote for the guys over at TheDataLab about High Dose Dexamethasone. Data Lab tells me this was one of their most read posts. I continue to keep an eye on the dexamethasone data, and in the future will post a “live” update of that data on here.

— Content on this website has been produced with reasonable care. However, errors can occur. Opinions change over time. Code that worked at the time of publication may no longer work. Users of all content from this site do so at their own risk.

Reuse

]]> OpenPrescribing Data Dexamethasone https://www.chemo.org.uk/posts/High-Dose-Dexamethasone.html Thu, 28 Nov 2019 00:00:00 GMT Merging XLS Files https://www.chemo.org.uk/posts/MergingExcelFiles.html This is a very brief placeholder about merging Excel files.

I’ve recently been working on a project which had multiple identically structured, single sheet Excel files. These needed merging to a single file. I’ve done this with CSV files before from the command line with a batch file. Excel took a little more effort but the resulting actions are in the accompanying zip file.

The files are first coverts by a visual basic script to CSV file (requires Excel installed on the PC); all Excel files in the folder where the file is run are converted.

Then the CSV files are converted by the batch script.

Credit for code: This is mostly code reworked slightly from Stack Overflow posts.

— Content on this website has been produced with reasonable care. However, errors can occur. Opinions change over time. Code that worked at the time of publication may no longer work. Users of all content from this site do so at their own risk.

Reuse

]]> https://www.chemo.org.uk/posts/MergingExcelFiles.html Sun, 11 May 2025 23:42:40 GMT