How to make nice publishable adverse event tables using tidyverse

Wed, 12 Feb 2020 00:00:00 +0000

This blog post is just an answer to a colleague to provide R code for the generation of Adverse Event tables. And it is also nice to have the code available when I need it in the future. Probably I will pull my hair at the horrible code, but this gives room to enhance it later.

Functions

First I define all functions to be used. I reuse some of the ideas in the post where I show how to make publishable tables using purrr here.

ae_n_pct <-  function(data, var, group, level = 1) {
  var <- ensym(var)
  group <- ensym(group)

  data %>% 
    group_by(subjectid, !!group, !!var) %>% 
    summarise(n = sum(!!var)) %>% 
    group_by(!!group, !!var) %>% 
    summarise(n_ae = sum(n),
            n_pat = n()) %>% 
    group_by(!!group) %>% 
    mutate(N_pat = sum(n_pat),
           pct = round(n_pat/N_pat*100,digits = 1),
           txt = paste0(n_pat, " (", pct, "%)")) %>%
    filter(!!var %in% !!level) %>% 
    ungroup %>% 
    select(!!group, txt) %>% 
    deframe
}

ae_N_n_pct <-  function(data, var, group, level = 1) {
  var <- ensym(var)
  group <- ensym(group)

  data %>% 
    group_by(subjectid, !!group) %>% 
    summarise(n = sum(!!var)) %>% 
    mutate(!!var := if_else(n==0, 0, 1)) %>% 
    group_by(!!group, !!var) %>% 
    summarise(n_ae = sum(n),
            n_pat = n()) %>% 
    group_by(!!group) %>% 
    mutate(N_pat = sum(n_pat),
           pct = round(n_pat/N_pat*100,digits = 1),
           txt = paste0("[", n_ae,"] ", n_pat, " (", pct, "%)")) %>%
    mutate(txt = if_else(n_ae == 0, "[0] 0 (0%)", txt)) %>% 
    filter(!!var %in% !!level) %>% 
    ungroup %>% 
    select(!!group, txt) %>% 
    deframe
}

stats_exec <- function(f, data, var, group, ...){
    exec(f, data, var, group, !!!(...))
}

Then I do a bit of data wrangling. The mock-up data can be downloaded in .rds format here

adae <- readRDS("adae.rds") %>% 
  mutate(anyae = if_else(is.na(pt), 0, 1),
         sae = if_else(is.na(pt), 0, sae)
  ) %>%  
  group_by(subjectid) %>% 
  mutate(n_ae = sum(anyae),
         one_ae = n_ae == 1,
         two_ae = n_ae == 2,
         three_plus_ae = n_ae > 2,
         anysae = max(sae)) %>% 
  ungroup

Summary of Adverse Events

The summary of Adverse Events is a nice table just summing up the adverse events in the trial. Note the “[N] n (%)”-format which is the number of events, number of patients with events and percentage of patients with event.

arms <- c("Active", "Control")
total_n <- n_distinct(adae$subjectid)

header_ae <- adae %>%
  group_by(trt, subjectid) %>%
  summarise(n=n()) %>%
  group_by(trt) %>%
  summarise(n = n()) %>%
  ungroup() %>%
  mutate(armtxt = arms) %>%
  mutate(txt = paste0(armtxt, " (N=", n, ")")) %>%
  select(txt) %>%
  deframe

## `summarise()` has grouped output by 'trt'. You can override using the `.groups`
## argument.

ae_summary_table <- tribble(
    ~text,  ~var, ~f,
  "Number of AEs", "anyae", "ae_N_n_pct",
  "Number of patients with any AEs?", "anyae", "ae_n_pct",
  "Number of patients with one AE", "one_ae", "ae_n_pct",
  "Number of patients with two AE", "two_ae", "ae_n_pct",
  "Number of patients with three or more AEs", "three_plus_ae", "ae_n_pct",
  "Number of SAEs", "sae", "ae_N_n_pct",
  "Number of patients with any SAEs?", "anysae","ae_n_pct"
)

ae_summary_table %>%
  mutate(data = list(adae),
         group = "trt",
         param = list(level = 1)) %>%
  mutate(res = pmap(list(f, data, var, group, param), stats_exec)) %>%
  mutate(id = map(res,names)) %>%
  unnest(c(res, id)) %>%
  mutate(id = paste0("txt", id)) %>%
  pivot_wider(values_from = res, names_from = id) %>%
  select(text, starts_with("txt")) %>%
  kable(col.names = c("Parameter", header_ae),
        caption = "Summary of Adverse Events",
        booktabs = TRUE)

## `summarise()` has grouped output by 'subjectid'. You can override using the
## `.groups` argument.
## `summarise()` has grouped output by 'trt'. You can override using the `.groups`
## argument.
## `summarise()` has grouped output by 'subjectid', 'trt'. You can override using
## the `.groups` argument.
## `summarise()` has grouped output by 'trt'. You can override using the `.groups`
## argument.
## `summarise()` has grouped output by 'subjectid', 'trt'. You can override using
## the `.groups` argument.
## `summarise()` has grouped output by 'trt'. You can override using the `.groups`
## argument.
## `summarise()` has grouped output by 'subjectid', 'trt'. You can override using
## the `.groups` argument.
## `summarise()` has grouped output by 'trt'. You can override using the `.groups`
## argument.
## `summarise()` has grouped output by 'subjectid', 'trt'. You can override using
## the `.groups` argument.
## `summarise()` has grouped output by 'trt'. You can override using the `.groups`
## argument.
## `summarise()` has grouped output by 'subjectid'. You can override using the
## `.groups` argument.
## `summarise()` has grouped output by 'trt'. You can override using the `.groups`
## argument.
## `summarise()` has grouped output by 'subjectid', 'trt'. You can override using
## the `.groups` argument.
## `summarise()` has grouped output by 'trt'. You can override using the `.groups`
## argument.

Table 1: Summary of Adverse Events
Parameter	Active (N=81)	Control (N=80)
Number of AEs	[149] 74 (91.4%)	[165] 79 (98.8%)
Number of patients with any AEs?	74 (86%)	79 (92.9%)
Number of patients with one AE	14 (17.3%)	12 (15%)
Number of patients with two AE	13 (16%)	16 (20%)
Number of patients with three or more AEs	49 (60.5%)	52 (65%)
Number of SAEs	[4] 4 (4.9%)	[2] 2 (2.5%)
Number of patients with any SAEs?	6 (7.4%)	6 (7.5%)

Adverse Events by System Organ Class and Preferred term

This table is almost a listing, but it gives a nice overview of all Adverse Events in the trial. First we need to make function which does most of the work.

ae_table_fns <- function(data, filtervar){

  filtervar = ensym(filtervar)

data %>%
  group_by(trt) %>%
  mutate(N_pat = n_distinct(subjectid)) %>%
  filter(!!filtervar == 1)  %>%
  group_by(subjectid, trt, N_pat, soc, pt) %>%
  summarise(n_ae = n()) %>%
  filter(!is.na(pt)) %>%
  group_by(trt, N_pat, soc, pt) %>%
  summarise(n_pat = n(),
            n_ae = sum(n_ae)) %>%
  mutate(pct = round(n_pat/N_pat*100,digits = 1),
         txt = paste0("[", n_ae,"] ", n_pat, " (", pct, "%)"),
         arm = paste0("arm", trt)) %>%
  ungroup %>% select(arm, soc, pt, txt) %>%
  pivot_wider(values_from = txt, names_from = arm) %>%
  mutate_at(vars(starts_with("arm")), ~if_else(is.na(.), "", .)) %>%
  arrange(soc, pt) %>%  group_by(soc2 = soc) %>%
  mutate(soc = if_else(row_number() != 1, "", soc)) %>% ungroup() %>% select(-soc2)
}

adae %>%
  bind_rows(adae, .id="added") %>%
  mutate(pt = if_else(added == 2, "#Total", pt)) %>%
  mutate(all = 1) %>%
  ae_table_fns("all") %>%
  knitr::kable(col.names = c("System Organ Class", "Preferred Term", header_ae),
               caption = " Adverse Events by System Organ Class and Preferred term*",
         booktabs = TRUE,
         longtable = TRUE)

## `summarise()` has grouped output by 'subjectid', 'trt', 'N_pat', 'soc'. You can
## override using the `.groups` argument.
## `summarise()` has grouped output by 'trt', 'N_pat', 'soc'. You can override
## using the `.groups` argument.

Table 2: Adverse Events by System Organ Class and Preferred term*
System Organ Class	Preferred Term	Active (N=81)	Control (N=80)
Blood and lymphatic system disorders	#Total	[1] 1 (1.2%)	[2] 2 (2.5%)
	Increased tendency to bruise		[1] 1 (1.2%)
	Neutropenia	[1] 1 (1.2%)
	Thrombocytopenia		[1] 1 (1.2%)
Cardiac disorders	#Total	[2] 2 (2.5%)	[2] 2 (2.5%)
	Palpitations	[2] 2 (2.5%)	[2] 2 (2.5%)
Eye disorders	#Total	[2] 2 (2.5%)	[5] 5 (6.2%)
	Dry eye		[2] 2 (2.5%)
	Eye irritation	[2] 2 (2.5%)	[1] 1 (1.2%)
	Vision blurred		[2] 2 (2.5%)
Gastrointestinal disorders	#Total	[42] 32 (39.5%)	[25] 19 (23.8%)
	Abdominal discomfort	[2] 2 (2.5%)	[1] 1 (1.2%)
	Abdominal pain	[1] 1 (1.2%)
	Abdominal pain upper	[1] 1 (1.2%)	[1] 1 (1.2%)
	Angular cheilitis	[1] 1 (1.2%)
	Constipation		[1] 1 (1.2%)
	Diarrhoea	[3] 3 (3.7%)
	Diverticulum intestinal	[1] 1 (1.2%)
	Dyspepsia		[1] 1 (1.2%)
	Flatulence		[1] 1 (1.2%)
	Gastritis		[2] 2 (2.5%)
	Gastrooesophageal reflux disease		[2] 2 (2.5%)
	Glossodynia		[1] 1 (1.2%)
	Lip ulceration		[1] 1 (1.2%)
	Mouth ulceration	[1] 1 (1.2%)	[2] 2 (2.5%)
	Nausea	[25] 22 (27.2%)	[11] 10 (12.5%)
	Oral mucosal blistering	[1] 1 (1.2%)	[1] 1 (1.2%)
	Paraesthesia oral	[1] 1 (1.2%)
	Tooth loss	[1] 1 (1.2%)
	Vomiting	[4] 4 (4.9%)
General disorders and administration site conditions	#Total	[12] 12 (14.8%)	[12] 11 (13.8%)
	Asthenia		[1] 1 (1.2%)
	Fatigue	[4] 4 (4.9%)	[5] 5 (6.2%)
	Impaired healing	[1] 1 (1.2%)
	Influenza like illness		[1] 1 (1.2%)
	Infusion site swelling	[1] 1 (1.2%)
	Injection site bruising		[1] 1 (1.2%)
	Injection site reaction		[1] 1 (1.2%)
	Malaise		[1] 1 (1.2%)
	Nodule	[1] 1 (1.2%)
	Pain	[1] 1 (1.2%)
	Pyrexia	[4] 4 (4.9%)	[2] 2 (2.5%)
Immune system disorders	#Total	[1] 1 (1.2%)	[2] 2 (2.5%)
	Hypersensitivity	[1] 1 (1.2%)	[2] 2 (2.5%)
Infections and infestations	#Total	[19] 18 (22.2%)	[38] 31 (38.8%)
	Borrelia infection		[1] 1 (1.2%)
	Bronchitis	[1] 1 (1.2%)	[1] 1 (1.2%)
	Conjunctivitis		[1] 1 (1.2%)
	Conjunctivitis bacterial		[1] 1 (1.2%)
	Diverticulitis		[1] 1 (1.2%)
	Epididymitis		[1] 1 (1.2%)
	Furuncle	[1] 1 (1.2%)
	Gastroenteritis	[2] 2 (2.5%)
	Gastroenteritis viral		[1] 1 (1.2%)
	Gingival abscess		[1] 1 (1.2%)
	Herpes virus infection		[1] 1 (1.2%)
	Infected skin ulcer	[1] 1 (1.2%)
	Influenza	[1] 1 (1.2%)	[2] 2 (2.5%)
	Localised infection	[1] 1 (1.2%)	[1] 1 (1.2%)
	Nail bed infection		[1] 1 (1.2%)
	Nasopharyngitis	[5] 5 (6.2%)	[11] 10 (12.5%)
	Oral herpes		[1] 1 (1.2%)
	Otitis media	[1] 1 (1.2%)
	Respiratory tract infection		[1] 1 (1.2%)
	Rhinitis		[1] 1 (1.2%)
	Sinusitis	[1] 1 (1.2%)
	Tinea versicolour		[1] 1 (1.2%)
	Upper respiratory tract infection	[4] 4 (4.9%)	[8] 8 (10%)
	Urinary tract infection	[1] 1 (1.2%)	[1] 1 (1.2%)
	Urinary tract infection bacterial		[1] 1 (1.2%)
Injury, poisoning and procedural complications	#Total	[3] 3 (3.7%)	[7] 7 (8.8%)
	Arthropod bite		[2] 2 (2.5%)
	Arthropod sting		[1] 1 (1.2%)
	Contusion		[1] 1 (1.2%)
	Incorrect dose administered	[1] 1 (1.2%)
	Joint dislocation	[1] 1 (1.2%)
	Limb injury		[1] 1 (1.2%)
	Road traffic accident		[1] 1 (1.2%)
	Skin wound	[1] 1 (1.2%)	[1] 1 (1.2%)
Investigations	#Total	[21] 17 (21%)	[15] 13 (16.2%)
	Alanine aminotransferase increased	[13] 11 (13.6%)	[9] 9 (11.2%)
	Biopsy prostate	[1] 1 (1.2%)
	Blood bilirubin increased		[2] 2 (2.5%)
	Blood pressure decreased	[1] 1 (1.2%)
	Blood pressure increased	[1] 1 (1.2%)
	Blood triglycerides increased		[1] 1 (1.2%)
	Chest X-ray abnormal		[1] 1 (1.2%)
	Hepatic enzyme increased	[3] 3 (3.7%)
	Platelet count decreased		[2] 2 (2.5%)
	Transaminases increased	[1] 1 (1.2%)
	Weight increased	[1] 1 (1.2%)
Metabolism and nutrition disorders	#Total		[3] 3 (3.8%)
	Decreased appetite		[1] 1 (1.2%)
	Hyperlipidaemia		[1] 1 (1.2%)
	Hypertriglyceridaemia		[1] 1 (1.2%)
Musculoskeletal and connective tissue disorders	#Total	[7] 7 (8.6%)	[6] 6 (7.5%)
	Arthralgia	[1] 1 (1.2%)
	Back pain	[1] 1 (1.2%)
	Groin pain		[1] 1 (1.2%)
	Intervertebral disc protrusion	[1] 1 (1.2%)
	Musculoskeletal pain	[1] 1 (1.2%)	[1] 1 (1.2%)
	Neck pain		[2] 2 (2.5%)
	Pain in extremity	[2] 2 (2.5%)	[1] 1 (1.2%)
	Rheumatoid arthritis	[1] 1 (1.2%)
	Rotator cuff syndrome		[1] 1 (1.2%)
Neoplasms benign, malignant and unspecified (incl cysts and polyps)	#Total	[1] 1 (1.2%)
	Malignant melanoma	[1] 1 (1.2%)
Nervous system disorders	#Total	[8] 8 (9.9%)	[17] 15 (18.8%)
	Anosmia	[1] 1 (1.2%)
	Dementia	[1] 1 (1.2%)
	Dizziness	[2] 2 (2.5%)	[6] 5 (6.2%)
	Dysgeusia	[1] 1 (1.2%)	[1] 1 (1.2%)
	Headache	[2] 2 (2.5%)	[4] 4 (5%)
	Hypoaesthesia		[2] 2 (2.5%)
	Muscle contractions involuntary	[1] 1 (1.2%)
	Paraesthesia		[2] 2 (2.5%)
	Taste disorder		[1] 1 (1.2%)
	Tension headache		[1] 1 (1.2%)
Psychiatric disorders	#Total	[2] 2 (2.5%)	[2] 2 (2.5%)
	Anxiety	[1] 1 (1.2%)
	Insomnia		[2] 2 (2.5%)
	Terminal insomnia	[1] 1 (1.2%)
Renal and urinary disorders	#Total	[1] 1 (1.2%)	[1] 1 (1.2%)
	Dysuria		[1] 1 (1.2%)
	Renal mass	[1] 1 (1.2%)
Reproductive system and breast disorders	#Total	[1] 1 (1.2%)	[1] 1 (1.2%)
	Hypomenorrhoea	[1] 1 (1.2%)
	Uterine polyp		[1] 1 (1.2%)
Respiratory, thoracic and mediastinal disorders	#Total	[8] 8 (9.9%)	[7] 7 (8.8%)
	Cough	[3] 3 (3.7%)	[2] 2 (2.5%)
	Dysphonia		[1] 1 (1.2%)
	Dyspnoea	[1] 1 (1.2%)	[1] 1 (1.2%)
	Epistaxis	[2] 2 (2.5%)
	Interstitial lung disease		[1] 1 (1.2%)
	Nasal discomfort		[1] 1 (1.2%)
	Oropharyngeal pain	[2] 2 (2.5%)
	Throat tightness		[1] 1 (1.2%)
Skin and subcutaneous tissue disorders	#Total	[14] 12 (14.8%)	[17] 17 (21.2%)
	Acne	[1] 1 (1.2%)
	Alopecia	[4] 4 (4.9%)	[2] 2 (2.5%)
	Blister	[4] 3 (3.7%)	[1] 1 (1.2%)
	Erythema	[1] 1 (1.2%)
	Hyperhidrosis	[1] 1 (1.2%)
	Night sweats		[1] 1 (1.2%)
	Pain of skin		[1] 1 (1.2%)
	Pruritus		[2] 2 (2.5%)
	Purpura		[1] 1 (1.2%)
	Rash	[2] 2 (2.5%)	[6] 6 (7.5%)
	Rash erythematous		[1] 1 (1.2%)
	Urticaria	[1] 1 (1.2%)	[2] 2 (2.5%)
Social circumstances	#Total		[1] 1 (1.2%)
	Stress at work		[1] 1 (1.2%)
Surgical and medical procedures	#Total	[2] 2 (2.5%)	[1] 1 (1.2%)
	Parathyroidectomy	[1] 1 (1.2%)
	Rehabilitation therapy	[1] 1 (1.2%)
	Rheumatoid nodule removal		[1] 1 (1.2%)
Vascular disorders	#Total	[2] 2 (2.5%)	[1] 1 (1.2%)
	Hypertension	[2] 2 (2.5%)	[1] 1 (1.2%)
NA	#Total	[12] 12 (14.8%)	[7] 6 (7.5%)

Serious Adverse Events by System Organ Class and Preferred term

Then the sub-table with only the serious adverse events.

adae %>%
  bind_rows(adae, .id="added") %>%
  mutate(pt = if_else(added == 2, "#Total", pt)) %>%
  ae_table_fns("sae") %>%
  knitr::kable( col.names = c("System Organ Class", "Preferred Term", header_ae),
               caption = "Serious Adverse Events by System Organ Class and Preferred term*",
         booktabs = TRUE,
         longtable = TRUE)

## `summarise()` has grouped output by 'subjectid', 'trt', 'N_pat', 'soc'. You can
## override using the `.groups` argument.
## `summarise()` has grouped output by 'trt', 'N_pat', 'soc'. You can override
## using the `.groups` argument.

Table 3: Serious Adverse Events by System Organ Class and Preferred term*
System Organ Class	Preferred Term	Active (N=81)	Control (N=80)
General disorders and administration site conditions	#Total	[1] 1 (1.2%)
	Pyrexia	[1] 1 (1.2%)
Infections and infestations	#Total		[1] 1 (1.2%)
	Epididymitis		[1] 1 (1.2%)
Neoplasms benign, malignant and unspecified (incl cysts and polyps)	#Total	[1] 1 (1.2%)
	Malignant melanoma	[1] 1 (1.2%)
Reproductive system and breast disorders	#Total		[1] 1 (1.2%)
	Uterine polyp		[1] 1 (1.2%)
Surgical and medical procedures	#Total	[2] 2 (2.5%)
	Parathyroidectomy	[1] 1 (1.2%)
	Rehabilitation therapy	[1] 1 (1.2%)

Usually there is also a table of probable/possible study treatment related AE/SAEs, and maybe also a AE/SAE of special interest table. They are made similarly to the SAE table.

Most common Adverse events

Lastly a table of the most common Adverse events. It is easy to change the treshold.

adae %>%
  group_by(trt) %>%
  mutate(N_pat = n_distinct(subjectid)) %>%
  group_by(subjectid, trt, N_pat, soc, pt) %>%
  summarise(n_ae = n()) %>%
  group_by(trt, N_pat, soc, pt) %>%
  summarise(n_pat = n(),
            n_ae = sum(n_ae)) %>%
  mutate(pct = round(n_pat/N_pat*100,digits = 1),
         txt = paste0("[", n_ae,"] ", n_pat, " (", pct, "%)"),
         arm = paste0("arm", trt)) %>%
  group_by(pt) %>%
  mutate(N_pat = sum(n_pat),
         pct_tot = N_pat /total_n) %>%
  filter(pct_tot>0.05) %>%
  ungroup %>%
  select(arm, soc, pt, txt, pct_tot) %>%
  pivot_wider(values_from = txt, names_from = arm) %>%
  mutate_at(vars(starts_with("arm")), ~if_else(is.na(.), "", .)) %>%
  arrange(desc(pct_tot)) %>% select( pt, everything()) %>% select(-pct_tot, -soc)  %>%
  knitr::kable( col.names = c( "Preferred Term", header_ae),
               caption = "Most common Adverse Events (more than 5 percent) by Preferred Term",
         booktabs = TRUE,
         longtable = TRUE)

## `summarise()` has grouped output by 'subjectid', 'trt', 'N_pat', 'soc'. You can
## override using the `.groups` argument.
## `summarise()` has grouped output by 'trt', 'N_pat', 'soc'. You can override
## using the `.groups` argument.