Preparation and exploration of external control datasets

2021-10-01

Source: vignettes/01-data-prep.Rmd

Setup

The purpose of this analysis is to explore the data, specifically the (i) amount of missingness in potential covariates, (ii) small sample sizes among categories of categorical variables, and (iii) the distribution of potential covariates across the treated and control groups prior to any adjustment. The following R packages are used.

Data

The underlying dataset used for the analysis is built using the ecdata package. It consists of patients with advanced non-small cell lung cancer (NSCLC) and is documented here. The pins_nsclc() pins the datasets locally using pins::pin() so that the dataset only needs to be built once. We preprocess the data into a format more suitable for propensity score modeling with the preprocess() function and filter to remove the comparator arm from the RCT in order to emulate a single arm trial.

ecdata::pin_nsclc(update = FALSE) # Pins dataset locally
nsclc <- pin_get("nsclc") %>%
  preprocess()
ec_rct1 <- nsclc %>% # Hypothetical single arm trial with external control
  filter(!(source_type == "RCT" & arm_type == "Comparator"))

All pairwise comparisons that will be used in the analyses are displayed in the table below.

analysis <- new_analysis(nsclc)  

analysis %>%
 html_table() %>%
 scroll_box(width = "100%")
Source
Arm
N
Number ID External Trial External Trial External control Internal control Experimental arm
1 nct02008227_fi_1 Flatiron NCT02008227 Docetaxel Atezolizumab 526 425 425
2 nct01903993_fi_1 Flatiron NCT01903993 Docetaxel Atezolizumab 476 143 144
3 nct02366143_fi_1 Flatiron NCT02366143 Bevacizumab + Carboplatin Atezolizumab + Bevacizumab + Carboplatin 602 336 356
4 nct01351415_fi_1 Flatiron NCT01351415 SOC Bevacizumab + SOC 381 240 245
5 nct01519804_fi_1 Flatiron NCT01519804 Placebo + Platinum + Paclitaxel MetMAb + Platinum + Paclitaxel 1945 54 55
6 nct01496742_fi_1 Flatiron NCT01496742 Placebo + Bevacizumab + Platinum MetMAb + Bevacizumab + Platinum + Paclitaxel 956 70 69
7 nct01496742_fi_2 Flatiron NCT01496742 Placebo + Platinum + Pemetrexed MetMAb + Platinum + Pemetrexed 3244 61 59
8 nct01366131_fi_1 Flatiron NCT01366131 Placebo + Bevacizumab + Carboplatin + Paclitaxel MEGF0444A + Bevacizumab + Carboplatin + Paclitaxel 963 52 52
9 nct01493843_fi_1 Flatiron NCT01493843 Placebo (340 mg) + Carboplatin + Paclitaxel Pictilisib (340 mg) + Carboplatin + Paclitaxel 1274 125 126
10 nct01493843_fi_2 Flatiron NCT01493843 Placebo (340 mg) + Carboplatin + Paclitaxel + Bevacizumab Pictilisib (340 mg) + Carboplatin + Paclitaxel + Bevacizumab 953 79 79
11 nct01493843_fi_3 Flatiron NCT01493843 Placebo (260 mg) + Carboplatin + Paclitaxel + Bevacizumab Pictilisib (260 mg) + Carboplatin + Paclitaxel + Bevacizumab 953 30 62
12 nct02367781_fi_1 Flatiron NCT02367781 Nab-Paclitaxel + Carboplatin Atezolizumab + Nab-Paclitaxel + Carboplatin 87 228 451
13 nct02367794_fi_1 Flatiron NCT02367794 Nab-Paclitaxel + Carboplatin Atezolizumab + Nab-Paclitaxel + Carboplatin 497 340 343
14 nct02657434_fi_1 Flatiron NCT02657434 Carboplatin or Cisplatin + Pemetrexed Atezolizumab + Carboplatin or Cisplatin + Pemetrexed 1536 286 292

Missing data

We count the number of missing values of each potential covariate for each pairwise analysis. Both the number of missing and the proportion missing (of the total number of observations with each arm) are plotted.

n_missing_df <- count_missing(ec_rct1, vars = get_ps_vars())

Proportion missing

autoplot(n_missing_df, yvar = "prop_missing")

Number missing

autoplot(n_missing_df, yvar = "n_missing")

Distribution of covariates

The distributions of the covariates in the unadjusted sample are plotted to check the degree of imbalance prior to adjustment and check for outliers or coding errors. There are a couple of points worth noting:

  • NCT01366131 did not collect smoking information.

  • It is not possible to determine whether a patient is a “Never” smoker in NCT01493843 because there is only a “Former/never” category. Smoking status can therefore not be used for propensity score analyses for this trial.

  • Patients in a race category with fewer than 10 observations were coded as “Other” race.

Age

plot_density(ec_rct1, var = "age")

Days since Dx

plot_density(ec_rct1, var = "days_since_dx", xlim = c(0, 1000))

Race

plot_bar(ec_rct1, var = "race_grouped")

Sex

plot_bar(ec_rct1, var = "sex")

Smoker

plot_bar(ec_rct1, var = "smoker_grouped")

Histology

plot_bar(ec_rct1, var = "histology")

Cancer stage

plot_bar(ec_rct1, var = "stage_grouped")

Categorical variables with small sample sizes

The plots above suggest that some categories of the categorical variables might still have small sample sizes even after preprocessing. Categories of categorical variables with fewer than 10 observations are shown in the table below.

count_by(ec_rct1, vars = get_ps_catvars(), 
         by = c("analysis_num", "source_type"),
         max_n = 9) %>%
  html_table()
analysis_num source_type var value n
4 External race_grouped Asian 7
5 RCT race_grouped Other 2
5 RCT smoker_grouped Never 2
5 RCT stage_grouped Early 9
7 RCT stage_grouped Early 9
8 RCT race_grouped Other 5
10 RCT race_grouped Other 8
11 RCT race_grouped Other 7
12 External race_grouped Black 9
12 External race_grouped Other 5
12 External smoker_grouped Never 9

Save output

saveRDS(analysis, file = "analysis-data-prep.rds")

Session information

## R version 4.0.0 (2020-04-24)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 18.04.5 LTS
## 
## Matrix products: default
## BLAS:   /usr/lib/x86_64-linux-gnu/openblas/libblas.so.3
## LAPACK: /usr/lib/x86_64-linux-gnu/libopenblasp-r0.2.20.so
## 
## locale:
##  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
##  [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
##  [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
##  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
##  [9] LC_ADDRESS=C               LC_TELEPHONE=C            
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       
## 
## attached base packages:
## [1] stats     graphics  grDevices utils     datasets  methods   base     
## 
## other attached packages:
## [1] pins_0.4.5         kableExtra_1.3.4   ggplot2_3.3.5      ecmeta.nsclc_0.1.0
## [5] ecdata_0.2.1       dplyr_1.0.7       
## 
## loaded via a namespace (and not attached):
##  [1] svglite_2.0.0     lattice_0.20-45   tidyr_1.1.4       assertthat_0.2.1 
##  [5] rprojroot_2.0.2   digest_0.6.28     utf8_1.2.2        R6_2.5.1         
##  [9] backports_1.2.1   evaluate_0.14     highr_0.9         httr_1.4.2       
## [13] pillar_1.6.3      rlang_0.4.11      rstudioapi_0.13   data.table_1.14.2
## [17] jquerylib_0.1.4   Matrix_1.3-4      rmarkdown_2.11    pkgdown_1.6.1    
## [21] labeling_0.4.2    textshaping_0.3.5 desc_1.4.0        splines_4.0.0    
## [25] webshot_0.5.2     stringr_1.4.0     munsell_0.5.0     compiler_4.0.0   
## [29] xfun_0.26         pkgconfig_2.0.3   systemfonts_1.0.2 htmltools_0.5.2  
## [33] tidyselect_1.1.1  tibble_3.1.5      fansi_0.5.0       viridisLite_0.4.0
## [37] crayon_1.4.1      withr_2.4.2       rappdirs_0.3.3    grid_4.0.0       
## [41] xtable_1.8-4      jsonlite_1.7.2    gtable_0.3.0      lifecycle_1.0.1  
## [45] DBI_1.1.1         magrittr_2.0.1    scales_1.1.1      stringi_1.7.4    
## [49] cachem_1.0.6      farver_2.1.0      fs_1.5.0          xml2_1.3.2       
## [53] bslib_0.3.0       ellipsis_0.3.2    filelock_1.0.2    ragg_1.1.3       
## [57] generics_0.1.0    vctrs_0.3.8       tools_4.0.0       glue_1.4.2       
## [61] purrr_0.3.4       fastmap_1.1.0     survival_3.2-13   yaml_2.2.1       
## [65] colorspace_2.0-2  rvest_1.0.1       memoise_2.0.0     knitr_1.36       
## [69] sass_0.4.0