Équipe ONCOSTAT
Modules de cours de R

Analyses statistiques & tables standards

Dan Chaltiel

Avec le package 📦 {grstat}

Plan du module

1. Analyse statistique
   • Modèles logistiques (glm) et modèles de Cox (coxph)
   • Formatage en table
   • Tests d’hypothèses
2. Sorties standards avec 📦 {grstat}
   • Tables des Adverse Events
   • Graphiques des Adverse Events
   • Création de projet standard (DEV)

Analyse statistique

Introduction

• Pour les analyses statistiques, on utilise généralement des tests stats et des modèles de régression

• Dans ce cours, on verra :

  • Les tests pour variables numériques et catégorielles
  • Les modèles linéaires, logistiques, et de Cox

Tests : variables numériques

Pour les variables numériques, on utilise soit le t-test, soit le test de Wilcoxon.

• Data
• 2-sample t-test
• 1-sample t-test
• 2-sample Wilcoxon
• Attention!

db = mtcars2 %>% 
  select(mpg, am)
db
#> # A tibble: 32 × 2
#>    mpg        am        
#>    <labelled> <labelled>
#>  1 21.0       manual    
#>  2 21.0       manual    
#>  3 22.8       manual    
#>  4 21.4       auto      
#>  5 18.7       auto      
#>  6 18.1       auto      
#>  7 14.3       auto      
#>  8 24.4       auto      
#>  9 22.8       auto      
#> 10 19.2       auto      
#> # ℹ 22 more rows

t = t.test(mpg ~ am, data=db)
t
#> 
#>  Welch Two Sample t-test
#> 
#> data:  mpg by am
#> t = -3.7671, df = 18.332, p-value = 0.001374
#> alternative hypothesis: true difference in means between group auto and group manual is not equal to 0
#> 95 percent confidence interval:
#>  -11.280194  -3.209684
#> sample estimates:
#>   mean in group auto mean in group manual 
#>             17.14737             24.39231

t$p.value
#> [1] 0.001373638

t = t.test(mpg~1, mu=17, data=db)
t
#> 
#>  One Sample t-test
#> 
#> data:  mpg
#> t = 2.9008, df = 31, p-value = 0.006788
#> alternative hypothesis: true mean is not equal to 17
#> 95 percent confidence interval:
#>  17.91768 22.26357
#> sample estimates:
#> mean of x 
#>  20.09062

t$p.value
#> [1] 0.00678831

w = wilcox.test(mpg ~ am, data=db)
w
#> 
#>  Wilcoxon rank sum test with continuity correction
#> 
#> data:  mpg by am
#> W = 42, p-value = 0.001871
#> alternative hypothesis: true location shift is not equal to 0

w$p.value
#> [1] 0.001871391

L’interface traditionnelle est à risque d’erreur!

t.test(mtcars$mpg, mtcars$am)
#> 
#>  Welch Two Sample t-test
#> 
#> data:  mtcars$mpg and mtcars$am
#> t = 18.413, df = 31.425, p-value < 2.2e-16
#> alternative hypothesis: true difference in means is not equal to 0
#> 95 percent confidence interval:
#>  17.50519 21.86356
#> sample estimates:
#> mean of x mean of y 
#>  20.09062   0.40625

Ici, am est binaire 0/1 et est traité comme une valeur numérique (moyenne 0.4).

Tests : variables catégorielles

Pour les variables catégorielles, on utilise soit le test du Chi², soit le test exact de Fisher.

• Data
• 2-sample Chi-squared test
• 2-sample Fisher exact test

db = mtcars2 %>% 
  select(am, vs)
db
#> # A tibble: 32 × 2
#>    am         vs        
#>    <labelled> <labelled>
#>  1 manual     vshaped   
#>  2 manual     vshaped   
#>  3 manual     straight  
#>  4 auto       straight  
#>  5 auto       vshaped   
#>  6 auto       straight  
#>  7 auto       vshaped   
#>  8 auto       straight  
#>  9 auto       straight  
#> 10 auto       straight  
#> # ℹ 22 more rows

tbl = table(db$am, db$vs)
tbl
#>         
#>          straight vshaped
#>   auto          7      12
#>   manual        7       6

cs = chisq.test(tbl)
cs
#> 
#>  Pearson's Chi-squared test with Yates' continuity correction
#> 
#> data:  tbl
#> X-squared = 0.34754, df = 1, p-value = 0.5555

cs$p.value
#> [1] 0.5555115

tbl = table(db$am, db$vs)
cs = fisher.test(tbl)
cs
#> 
#>  Fisher's Exact Test for Count Data
#> 
#> data:  tbl
#> p-value = 0.4727
#> alternative hypothesis: true odds ratio is not equal to 1
#> 95 percent confidence interval:
#>  0.09441436 2.61414549
#> sample estimates:
#> odds ratio 
#>   0.511233

cs$p.value
#> [1] 0.4726974

Modélisation

• Un modèle statistique décrit la relation entre une ou plusieurs variables explicatives (\(X_1, X_2, \dots, X_p\)) et une variable réponse (\(Y\)).

• On veut prédire \(Y\) ou expliquer son lien avec \(X\). \[ Y = \beta_0 + \beta_1 X_1 + \beta_2 X_2 + \dots + \beta_p X_p + \varepsilon \]

• Selon le type de réponse, on utilisera un modèle de régression différent, avec pour chacun des hypothèses spécifiques à valider.

Modélisation en R

• En R, la modélisation repose sur des fonctions comme lm(), glm() et coxph().

• La syntaxe standard est la suivante :

  formula = y ~ x1 + x2
  fit = regression_function(formula, data=my_data)

• L’objet fit contient toutes les informations de notre modèle.

• On peut visualiser le résultat avec la summary(fit), ou mieux avec broom::tidy(fit) ou gtsummary::tbl_regression().

Données d’exemple

• Pour commencer, on va utiliser les données simulées du dataset ae de grstat_example() contenant : le grade des adverse events aegr, le fait qu’il soit SAE sae, et la causalité aerel.

• On va d’abord modéliser aegr (modèle linéaire), puis sae (modèle logistique).

• Aperçu des données :

  db = grstat::grstat_example(seed=100)
  db$ae %>% select(subjid, aegr, sae, aerel) %>% print(n=2)
  #> # A tibble: 606 × 4
  #>   subjid  aegr sae   aerel       
  #>    <int> <dbl> <fct> <chr>       
  #> 1      1     4 No    Radiotherapy
  #> 2      1     2 No    Radiotherapy
  #> # ℹ 604 more rows

Modèle mixte

Il y a plusieurs lignes par patient donc il aurait fallu rajouter un effet aléatoire, mais ça dépasse le cadre de ce cours.

Modèle linéaire (Y continu)

fit = lm(aegr ~ sae + aerel, data=db$ae)

• fit
• summary()
• confint()
• tidy()
• gtsummary

fit
#> 
#> Call:
#> lm(formula = aegr ~ sae + aerel, data = db$ae)
#> 
#> Coefficients:
#>                 (Intercept)                        saeNo  aerelExperimental treatment  
#>                     2.66664                     -0.95142                      0.02929  
#>                  aerelOther            aerelRadiotherapy      aerelStandard treatment  
#>                    -0.06962                     -0.02662                     -0.07004

summary(fit)
#> 
#> Call:
#> lm(formula = aegr ~ sae + aerel, data = db$ae)
#> 
#> Residuals:
#>     Min      1Q  Median      3Q     Max 
#> -1.6959 -0.6886 -0.6452  0.3544  3.3548 
#> 
#> Coefficients:
#>                             Estimate Std. Error t value Pr(>|t|)    
#> (Intercept)                  2.66664    0.14944  17.844  < 2e-16 ***
#> saeNo                       -0.95142    0.13460  -7.069 4.37e-12 ***
#> aerelExperimental treatment  0.02929    0.12824   0.228    0.819    
#> aerelOther                  -0.06962    0.12828  -0.543    0.588    
#> aerelRadiotherapy           -0.02662    0.12491  -0.213    0.831    
#> aerelStandard treatment     -0.07004    0.12886  -0.544    0.587    
#> ---
#> Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#> 
#> Residual standard error: 0.985 on 600 degrees of freedom
#> Multiple R-squared:  0.07785,    Adjusted R-squared:  0.07016 
#> F-statistic: 10.13 on 5 and 600 DF,  p-value: 2.5e-09

confint(fit)
#>                                  2.5 %     97.5 %
#> (Intercept)                  2.3731462  2.9601320
#> saeNo                       -1.2157688 -0.6870811
#> aerelExperimental treatment -0.2225645  0.2811367
#> aerelOther                  -0.3215431  0.1823044
#> aerelRadiotherapy           -0.2719339  0.2186922
#> aerelStandard treatment     -0.3231122  0.1830387

broom::tidy(fit, conf.int=TRUE)
#> # A tibble: 6 × 7
#>   term                        estimate std.error statistic  p.value conf.low conf.high
#>   <chr>                          <dbl>     <dbl>     <dbl>    <dbl>    <dbl>     <dbl>
#> 1 (Intercept)                   2.67       0.149    17.8   1.89e-57    2.37      2.96 
#> 2 saeNo                        -0.951      0.135    -7.07  4.37e-12   -1.22     -0.687
#> 3 aerelExperimental treatment   0.0293     0.128     0.228 8.19e- 1   -0.223     0.281
#> 4 aerelOther                   -0.0696     0.128    -0.543 5.88e- 1   -0.322     0.182
#> 5 aerelRadiotherapy            -0.0266     0.125    -0.213 8.31e- 1   -0.272     0.219
#> 6 aerelStandard treatment      -0.0700     0.129    -0.544 5.87e- 1   -0.323     0.183

gtsummary::tbl_regression(fit, conf.int=TRUE)

Characteristic	Beta	95% CI 1	p-value
Serious AE
Yes	—	—
No	-0.95	-1.2, -0.69	<0.001
aerel
Cancer	—	—
Experimental treatment	0.03	-0.22, 0.28	0.8
Other	-0.07	-0.32, 0.18	0.6
Radiotherapy	-0.03	-0.27, 0.22	0.8
Standard treatment	-0.07	-0.32, 0.18	0.6
1 CI = Confidence Interval

Modèle linéaire : hypothèses

• Linéarité
• Normalité des résidus
• Homoscédasticité

La relation entre \(Y\) et les covariables \(X\) doit être linéaire.

Sur la courbe des résidus en fonction des valeurs modélisées, on s’attend à une courbe horizontale.

plot(fit, which = 1)

Une courbure indique que la relation entre les prédicteurs et la réponse n’est pas linéaire → un terme polynomial ou une transformation pourrait être nécessaire. Une forme en entonnoir suggère une hétéroscédasticité (voir which = 3).

Sur le QQplot, les points doivent suivre la droite diagonale.

plot(fit, which = 2)

Une déviation en queue de distribution (points qui s’écartent de la diagonale aux extrémités) indique des résidus non normaux → potentiellement des valeurs aberrantes ou une distribution asymétrique. Une forte courbure peut suggérer une distribution en cloche trop aplatie ou trop pointue.

La variance des résidus doit être constante, on s’attend à une courbe horizontale.

plot(fit, which = 3)

lmtest::bptest(fit)
#> 
#>  studentized Breusch-Pagan test
#> 
#> data:  fit
#> BP = 12.98, df = 5, p-value = 0.02356

Une forme en entonnoir (dispersion des points qui augmente avec les valeurs prédites) indique une hétéroscédasticité → un modèle robuste ou une transformation des variables pourrait être nécessaire. Une variation systématique des points (ex. un motif en vague) suggère que l’erreur dépend des prédicteurs.

Modèle logistique (Y binaire)

fit2 = glm(sae ~ aegr + aerel, data=db$ae, family=binomial(link="logit"))
#family: binomial (logit, probit...), gaussian (identity, log...), poisson, ...

• fit
• summary()
• confint()
• tidy()
• gtsummary

fit2
#> 
#> Call:  glm(formula = sae ~ aegr + aerel, family = binomial(link = "logit"), 
#>     data = db$ae)
#> 
#> Coefficients:
#>                 (Intercept)                         aegr  aerelExperimental treatment  
#>                      3.6240                      -0.7305                       0.7356  
#>                  aerelOther            aerelRadiotherapy      aerelStandard treatment  
#>                     -0.3606                       0.2361                       0.2495  
#> 
#> Degrees of Freedom: 605 Total (i.e. Null);  600 Residual
#> Null Deviance:       391.4 
#> Residual Deviance: 346.8     AIC: 358.8

summary(fit2)
#> 
#> Call:
#> glm(formula = sae ~ aegr + aerel, family = binomial(link = "logit"), 
#>     data = db$ae)
#> 
#> Coefficients:
#>                             Estimate Std. Error z value Pr(>|z|)    
#> (Intercept)                   3.6240     0.4314   8.401  < 2e-16 ***
#> aegr                         -0.7305     0.1171  -6.240 4.38e-10 ***
#> aerelExperimental treatment   0.7356     0.5025   1.464    0.143    
#> aerelOther                   -0.3606     0.4109  -0.877    0.380    
#> aerelRadiotherapy             0.2361     0.4394   0.537    0.591    
#> aerelStandard treatment       0.2495     0.4602   0.542    0.588    
#> ---
#> Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#> 
#> (Dispersion parameter for binomial family taken to be 1)
#> 
#>     Null deviance: 391.36  on 605  degrees of freedom
#> Residual deviance: 346.78  on 600  degrees of freedom
#> AIC: 358.78
#> 
#> Number of Fisher Scoring iterations: 5

confint(fit2)
#>                                  2.5 %     97.5 %
#> (Intercept)                  2.8229887  4.5202857
#> aegr                        -0.9636990 -0.5029897
#> aerelExperimental treatment -0.2238604  1.7736458
#> aerelOther                  -1.1819895  0.4403776
#> aerelRadiotherapy           -0.6279850  1.1101385
#> aerelStandard treatment     -0.6482995  1.1742253

broom::tidy(fit2, conf.int=TRUE)
#> # A tibble: 6 × 7
#>   term                        estimate std.error statistic  p.value conf.low conf.high
#>   <chr>                          <dbl>     <dbl>     <dbl>    <dbl>    <dbl>     <dbl>
#> 1 (Intercept)                    3.62      0.431     8.40  4.41e-17    2.82      4.52 
#> 2 aegr                          -0.730     0.117    -6.24  4.38e-10   -0.964    -0.503
#> 3 aerelExperimental treatment    0.736     0.502     1.46  1.43e- 1   -0.224     1.77 
#> 4 aerelOther                    -0.361     0.411    -0.877 3.80e- 1   -1.18      0.440
#> 5 aerelRadiotherapy              0.236     0.439     0.537 5.91e- 1   -0.628     1.11 
#> 6 aerelStandard treatment        0.250     0.460     0.542 5.88e- 1   -0.648     1.17

gtsummary::tbl_regression(fit2, conf.int=TRUE)

Characteristic	log(OR) 1	95% CI 1	p-value
AE grade	-0.73	-0.96, -0.50	<0.001
aerel
Cancer	—	—
Experimental treatment	0.74	-0.22, 1.8	0.14
Other	-0.36	-1.2, 0.44	0.4
Radiotherapy	0.24	-0.63, 1.1	0.6
Standard treatment	0.25	-0.65, 1.2	0.6
1 OR = Odds Ratio, CI = Confidence Interval

Modèle logistique : hypothèses

• Linéarité
• Normalité des résidus
• Homoscédasticité

La relation entre \(Y\) et les covariables \(X\) doit être linéaire.

Sur la courbe des résidus en fonction des valeurs modélisées, on s’attend à une courbe horizontale.

plot(fit2, which = 1)

Une courbure indique que la relation entre les prédicteurs et la réponse n’est pas linéaire → un terme polynomial ou une transformation pourrait être nécessaire. Une forme en entonnoir suggère une hétéroscédasticité (voir which = 3).

La normalité des résidus n’est PAS une hypothèse du modèle logistique.

plot(fit2, which = 2)

La variance des résidus doit être constante, on s’attend à une courbe horizontale.

plot(fit2, which = 3)

Une forme en entonnoir (dispersion des points qui augmente avec les valeurs prédites) indique une hétéroscédasticité → un modèle robuste ou une transformation des variables pourrait être nécessaire. Une variation systématique des points (ex. un motif en vague) suggère que l’erreur dépend des prédicteurs.

Analyses de survie

Pour l’exemple, on va simuler des données de survie selon une loi exponentielle, avec 3 covariables : le bras de traitement arm, un biomarqueur binaire bm_bin, et un biomarqueur continu bm_cont.

#' Simulated dataset
#'
#' @param n number of observations
#' @param beta_trt,beta_bm_cont,beta_bm_cont the true coefficients
#' @param eos non-informative censoring (end of study)
#' @param seed RNG seed
surv_example = function(n=300, 
                  beta_trt=log(0.7), beta_bm_bin=log(1.5), beta_bm_cont=log(0.8),
                  eos=30, seed=42){
  set.seed(seed)
  # Generate covariates
  baseline = tibble(
    subjid = seq(n),
    arm = rep(c(0,1), each=n/2),   
    bm_bin = rbinom(n, 1, 0.4),
    bm_cont = rnorm(n),
  )

  rtn = baseline %>%
    mutate(
      # Generate exponential survival time
      log_predlin = beta_trt*arm + beta_bm_bin*bm_bin + beta_bm_cont*bm_cont,
      lambda0 = log(2)/10 / exp(beta_bm_bin*mean(bm_bin) + beta_bm_cont*mean(bm_cont)),
      lambda = lambda0 * exp(log_predlin),
      t_real = rexp(n, rate=lambda),

      # Administrative censorship
      t = pmin(t_real, eos),
      event = as.numeric(t_real<=eos),

      # labels
      arm =  factor(arm,  labels=c("Control", "Treatment")),
      bm_bin = factor(bm_bin, labels=c("Negative", "Positive")),
    )

  rtn %>%
    select(subjid, arm, t, event, bm_bin, bm_cont)
} 

df = surv_example()
df
#> # A tibble: 300 × 6
#>    subjid arm          t event bm_bin    bm_cont
#>     <int> <fct>    <dbl> <dbl> <fct>       <dbl>
#>  1      1 Control 29.4       1 Positive -0.0407 
#>  2      2 Control 23.7       1 Positive -1.55   
#>  3      3 Control 28.6       1 Negative  1.17   
#>  4      4 Control 30         0 Positive -0.274  
#>  5      5 Control  1.38      1 Positive -0.468  
#>  6      6 Control  2.18      1 Negative -1.24   
#>  7      7 Control  0.942     1 Positive -0.00776
#>  8      8 Control 30         0 Negative -0.800  
#>  9      9 Control  5.31      1 Positive -0.533  
#> 10     10 Control  7.79      1 Positive  1.29   
#> # ℹ 290 more rows

Kaplan Meier & Log Rank

• survfit
• Logrank
• KM plot (binaire)
• KM plot (continu)

library(survival)
library(ggsurvfit)

km = survfit(Surv(t, event) ~ arm, data=df)
km
#> Call: survfit(formula = Surv(t, event) ~ arm, data = df)
#> 
#>                 n events median 0.95LCL 0.95UCL
#> arm=Control   150    126   11.0    7.79    15.9
#> arm=Treatment 150    114   15.9   13.99    21.7

tidy_survfit(km, times=c(12,24)) %>% 
  select(strata, everything())
#> # A tibble: 4 × 16
#>   strata         time n.risk n.event n.censor cum.event cum.censor estimate std.error conf.high conf.low
#>   <fct>         <dbl>  <dbl>   <dbl>    <dbl>     <dbl>      <dbl>    <dbl>     <dbl>     <dbl>    <dbl>
#> 1 arm=Control      12     70      80        0        80          0    0.467    0.0873     0.554    0.393
#> 2 arm=Control      24     39      31        0       111          0    0.260    0.138      0.341    0.198
#> 3 arm=Treatment    12     97      53        0        53          0    0.647    0.0604     0.728    0.575
#> 4 arm=Treatment    24     54      43        0        96          0    0.360    0.109      0.446    0.291
#> # ℹ 5 more variables: estimate_type <chr>, estimate_type_label <chr>, monotonicity_type <chr>,
#> #   strata_label <chr>, conf.level <dbl>

survdiff(Surv(t, event) ~ arm, data=df)
#> Call:
#> survdiff(formula = Surv(t, event) ~ arm, data = df)
#> 
#>                 N Observed Expected (O-E)^2/E (O-E)^2/V
#> arm=Control   150      126      106      3.74      6.72
#> arm=Treatment 150      114      134      2.96      6.72
#> 
#>  Chisq= 6.7  on 1 degrees of freedom, p= 0.01

survdiff(Surv(t, event) ~ arm + strata(bm_bin), data=df)
#> Call:
#> survdiff(formula = Surv(t, event) ~ arm + strata(bm_bin), data = df)
#> 
#>                 N Observed Expected (O-E)^2/E (O-E)^2/V
#> arm=Control   150      126      109      2.76      5.15
#> arm=Treatment 150      114      131      2.29      5.15
#> 
#>  Chisq= 5.1  on 1 degrees of freedom, p= 0.02

Documentation: https://www.danieldsjoberg.com

survfit2(Surv(t, event) ~ arm, data=df) %>% 
  ggsurvfit(, linetype_aes = TRUE) +
  add_confidence_interval() +
  add_risktable(
    risktable_stats = c("n.risk", "cum.censor", "cum.event")
  ) +
  theme_ggsurvfit_KMunicate() +
  scale_y_continuous(limits = c(0, 1)) +
  scale_x_continuous(expand = c(0.02, 0)) +
  theme(legend.position="inside", legend.position.inside = c(0.85, 0.85))

model <- coxph(Surv(t, event) ~ bm_cont, data=df, x=TRUE)
contsurvplot::plot_surv_area(time="t", status="event", variable="bm_cont", data=df, model=model)

Modèle de Cox (Y survie)

fit_bare  = coxph(Surv(t, event) ~ arm, data=df)
fit_adj = coxph(Surv(t, event) ~ arm + bm_bin + bm_cont, data=df)
fit_strat = coxph(Surv(t, event) ~ arm + strata(bm_bin) + bm_cont, data=df)

• fit
• summary()
• tidy()
• fit_strat
• gtsummary

fit_adj
#> Call:
#> coxph(formula = Surv(t, event) ~ arm + bm_bin + bm_cont, data = df)
#> 
#>                    coef exp(coef) se(coef)      z      p
#> armTreatment   -0.31975   0.72633  0.12998 -2.460 0.0139
#> bm_binPositive  0.26931   1.30907  0.13201  2.040 0.0413
#> bm_cont        -0.16892   0.84458  0.07046 -2.397 0.0165
#> 
#> Likelihood ratio test=16.87  on 3 df, p=0.0007502
#> n= 300, number of events= 240

summary(fit_adj)
#> Call:
#> coxph(formula = Surv(t, event) ~ arm + bm_bin + bm_cont, data = df)
#> 
#>   n= 300, number of events= 240 
#> 
#>                    coef exp(coef) se(coef)      z Pr(>|z|)  
#> armTreatment   -0.31975   0.72633  0.12998 -2.460   0.0139 *
#> bm_binPositive  0.26931   1.30907  0.13201  2.040   0.0413 *
#> bm_cont        -0.16892   0.84458  0.07046 -2.397   0.0165 *
#> ---
#> Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#> 
#>                exp(coef) exp(-coef) lower .95 upper .95
#> armTreatment      0.7263     1.3768    0.5630    0.9371
#> bm_binPositive    1.3091     0.7639    1.0106    1.6956
#> bm_cont           0.8446     1.1840    0.7356    0.9697
#> 
#> Concordance= 0.585  (se = 0.02 )
#> Likelihood ratio test= 16.87  on 3 df,   p=8e-04
#> Wald test            = 16.82  on 3 df,   p=8e-04
#> Score (logrank) test = 16.92  on 3 df,   p=7e-04

confint(fit_adj)
#>                      2.5 %      97.5 %
#> armTreatment   -0.57450305 -0.06500688
#> bm_binPositive  0.01058439  0.52804332
#> bm_cont        -0.30702431 -0.03081770

broom::tidy(fit_adj, conf.int=TRUE)
#> # A tibble: 3 × 7
#>   term           estimate std.error statistic p.value conf.low conf.high
#>   <chr>             <dbl>     <dbl>     <dbl>   <dbl>    <dbl>     <dbl>
#> 1 armTreatment     -0.320    0.130      -2.46  0.0139  -0.575    -0.0650
#> 2 bm_binPositive    0.269    0.132       2.04  0.0413   0.0106    0.528 
#> 3 bm_cont          -0.169    0.0705     -2.40  0.0165  -0.307    -0.0308

broom::tidy(fit_strat, conf.int=TRUE)
#> # A tibble: 2 × 7
#>   term         estimate std.error statistic p.value conf.low conf.high
#>   <chr>           <dbl>     <dbl>     <dbl>   <dbl>    <dbl>     <dbl>
#> 1 armTreatment   -0.308    0.130      -2.36  0.0181   -0.564   -0.0527
#> 2 bm_cont        -0.175    0.0703     -2.48  0.0130   -0.313   -0.0369

gtsummary::tbl_regression(fit_adj, conf.int=TRUE)

Characteristic	log(HR) 1	95% CI 1	p-value
arm
Control	—	—
Treatment	-0.32	-0.57, -0.07	0.014
bm_bin
Negative	—	—
Positive	0.27	0.01, 0.53	0.041
bm_cont	-0.17	-0.31, -0.03	0.017
1 HR = Hazard Ratio, CI = Confidence Interval

Modèle de Cox: hypothèses

• HHP table
• HHP plot
• Linearity spline
• Linearity plot

cox.zph(fit_adj)
#>         chisq df     p
#> arm      3.74  1 0.053
#> bm_bin   1.88  1 0.171
#> bm_cont  1.33  1 0.249
#> GLOBAL   6.55  3 0.088

cox.zph(fit_adj) %>% survminer::ggcoxzph()

fit_spline = coxph(Surv(t, event) ~ pspline(bm_cont), data=df)
broom::tidy(fit_spline)
#> # A tibble: 2 × 5
#>   term                     estimate std.error statistic p.value
#>   <chr>                       <dbl>     <dbl>     <dbl>   <dbl>
#> 1 pspline(bm_cont), linear   -0.166    0.0697      5.68  0.0172
#> 2 pspline(bm_cont), nonlin   NA       NA           1.87  0.607

fit_num = coxph(Surv(t, event) ~ bm_cont, data=df)
survminer::ggcoxfunctional(fit_num)

Choix des groupes de référence

Le groupe de référence est le premier niveau de la variable. Si ce n’est pas un facteur, R prendra le premier par ordre alphabétique.

df2 = df %>% 
  mutate(arm = fct_relevel(arm, "Treatment"),
         bm_bin = fct_relevel(bm_bin, "Positive"))
levels(df$arm)
#> [1] "Control"   "Treatment"
levels(df2$arm)
#> [1] "Treatment" "Control"

Les modèles :

m1 = coxph(Surv(t, event) ~ arm + bm_bin, data=df)
m2 = coxph(Surv(t, event) ~ arm + bm_bin, data=df2)
coef(m1)
#>   armTreatment bm_binPositive 
#>     -0.3081165      0.2808774
coef(m2)
#>     armControl bm_binNegative 
#>      0.3081165     -0.2808774

Test de modèles emboîtés

• La plupart des modèles donnent un test de Wald pour chaque élément.

• Pour les variables catégorielles, on peut être intéressé par un test du rapport de vraissemblance qui teste la variable globalement (pas les niveaux indépendemment).

• Il faut alors utiliser la fonction anova() sur deux modèles emboîtés.

fit1 = coxph(Surv(t, event) ~ arm + bm_bin, data=df)
fit2 = coxph(Surv(t, event) ~ arm + bm_bin + bm_cont, data=df)

anova(fit1, fit2)
#> Analysis of Deviance Table
#>  Cox model: response is  Surv(t, event)
#>  Model 1: ~ arm + bm_bin
#>  Model 2: ~ arm + bm_bin + bm_cont
#>    loglik Chisq Df Pr(>|Chi|)  
#> 1 -1220.7                      
#> 2 -1217.8 5.778  1    0.01623 *
#> ---
#> Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Conclusion

• Il y a beaucoup de façons de faire des modèles de régression en R.

• Il y a beaucoup de packages différents.

  • Possibilités infinies
  • On peut parfois s’y perdre un peu

• Attention à la validité des packages, qui reste difficile à démontrer (contrairement à SAS)

Générer du contenu avec 📦 {grstat}

Le package grstat 📦

• Package interne au BBE
• Reprend les macros SAS validées
• Propose des sorties standardisées

Version

Cette présentation a été réalisée avec grstat v0.1.0.9009.

Documentation

Accessible dans RStudio et en ligne:

Fonctions

Fonctions stables et validées :

• Tableaux AE : ae_table_grade() & ae_table_soc()
• Graphiques AE : ae_plot_grade() & butterfly_plot()

Fonctions en cours de développement :

• Swimmerplot & Waterfall-plot RECIST
• Forest-plot
• Sorties standard pour les modèles principaux
• Vignettes/Book de documentation

Liste

La liste des fonctions est accessible dans les références de la doc.

Données d’exemple

On va utiliser une base de données d’exemple.

library(grstat)
db = grstat_example(r=0.4)
attach(db, warn.conflicts=FALSE)

db
#> $enrolres
#> # A tibble: 200 × 4
#>    subjid arm       arm3        crfname 
#>     <int> <fct>     <fct>       <chr>   
#>  1      1 Control   Treatment B enrolres
#>  2      2 Control   Control     enrolres
#>  3      3 Treatment Treatment A enrolres
#>  4      4 Control   Control     enrolres
#>  5      5 Treatment Treatment A enrolres
#>  6      6 Treatment Treatment A enrolres
#>  7      7 Treatment Control     enrolres
#>  8      8 Treatment Treatment A enrolres
#>  9      9 Control   Treatment B enrolres
#> 10     10 Control   Treatment B enrolres
#> # ℹ 190 more rows
#> 
#> $ae
#> # A tibble: 622 × 7
#>    subjid aesoc                                           aeterm                 aegr sae   aerel crfname
#>     <int> <chr>                                           <chr>                 <dbl> <fct> <chr> <chr>  
#>  1      2 Eye disorders                                   Eyelid disorders          1 No    Expe… ae     
#>  2      2 Cardiac disorders                               Heart failures            4 Yes   Stan… ae     
#>  3      2 Surgical and medical procedures                 Therapeutic procedur…     1 No    Other ae     
#>  4      3 Respiratory, thoracic and mediastinal disorders Pulmonary vascular d…     1 No    Stan… ae     
#>  5      3 Surgical and medical procedures                 Diagnostic procedures     3 No    Canc… ae     
#>  6      3 Neoplasms benign, malignant, and unspecified    Malignant neoplasms       1 Yes   Radi… ae     
#>  7      4 Skin and subcutaneous tissue disorders          Skin and subcutaneou…     1 No    Canc… ae     
#>  8      4 Investigations                                  Blood analyses            3 No    Stan… ae     
#>  9      5 Endocrine disorders                             Adrenal gland disord…     2 No    Expe… ae     
#> 10      5 Ear and labyrinth disorders                     Vertigo and balance …     1 No    Canc… ae     
#> # ℹ 612 more rows
#> 
#> $date_extraction
#> [1] "2024/01/01"
#> 
#> $datetime_extraction
#> [1] "2024-01-01 01:00:00 CET"

EDCimport

db n’est pas (encore) une base EDCimport mais fonctionne de la même façon.

AE par grade : tableaux

• Total
• Par bras
• SAE
• Population
• Postprod

ae_table_grade(df_ae=ae, df_enrol=enrolres, arm=NULL) %>%
  as_flextable(header_show_n=TRUE)

	Treatment arm
	All patients (N=200)
Patient maximum AE grade
No declared AE	3 (2%)
Grade 1	33 (16%)
Grade 2	60 (30%)
Grade 3	51 (26%)
Grade 4	40 (20%)
Grade 5	13 (6%)
Patient had at least one AE of grade
No declared AE	3 (2%)
Grade ≥ 1	197 (98%)
Grade ≥ 2	164 (82%)
Grade ≥ 3	104 (52%)
Grade ≥ 4	53 (26%)
Grade = 5	13 (6%)
Patient had at least one AE of grade
No declared AE	3 (2%)
Grade 1	160 (80%)
Grade 2	121 (60%)
Grade 3	64 (32%)
Grade 4	46 (23%)
Grade 5	13 (6%)

ae_table_grade(df_ae=ae, df_enrol=enrolres, arm="arm") %>%
  as_flextable(header_show_n=TRUE)

	Treatment arm		Total
	Control (N=80)	Treatment (N=120)
Patient maximum AE grade
No declared AE	3 (4%)	0 (0%)	3 (2%)
Grade 1	22 (28%)	11 (9%)	33 (16%)
Grade 2	24 (30%)	36 (30%)	60 (30%)
Grade 3	19 (24%)	32 (27%)	51 (26%)
Grade 4	11 (14%)	29 (24%)	40 (20%)
Grade 5	1 (1%)	12 (10%)	13 (6%)
Patient had at least one AE of grade
No declared AE	3 (4%)	0 (0%)	3 (2%)
Grade ≥ 1	77 (96%)	120 (100%)	197 (98%)
Grade ≥ 2	55 (69%)	109 (91%)	164 (82%)
Grade ≥ 3	31 (39%)	73 (61%)	104 (52%)
Grade ≥ 4	12 (15%)	41 (34%)	53 (26%)
Grade = 5	1 (1%)	12 (10%)	13 (6%)
Patient had at least one AE of grade
No declared AE	3 (4%)	0 (0%)	3 (2%)
Grade 1	66 (82%)	94 (78%)	160 (80%)
Grade 2	44 (55%)	77 (64%)	121 (60%)
Grade 3	21 (26%)	43 (36%)	64 (32%)
Grade 4	12 (15%)	34 (28%)	46 (23%)
Grade 5	1 (1%)	12 (10%)	13 (6%)

ae %>%
  filter(sae=="Yes") %>%
  ae_table_grade(df_enrol=enrolres, arm="arm", ae_label="SAE") %>%
  as_flextable(header_show_n=TRUE)

	Treatment arm		Total
	Control (N=80)	Treatment (N=120)
Patient maximum SAE grade
No declared SAE	60 (75%)	90 (75%)	150 (75%)
Grade 1	7 (9%)	3 (2%)	10 (5%)
Grade 2	4 (5%)	4 (3%)	8 (4%)
Grade 3	2 (2%)	8 (7%)	10 (5%)
Grade 4	7 (9%)	10 (8%)	17 (8%)
Grade 5	0 (0%)	5 (4%)	5 (2%)
Patient had at least one SAE of grade
No declared SAE	60 (75%)	90 (75%)	150 (75%)
Grade ≥ 1	20 (25%)	30 (25%)	50 (25%)
Grade ≥ 2	13 (16%)	27 (22%)	40 (20%)
Grade ≥ 3	9 (11%)	23 (19%)	32 (16%)
Grade ≥ 4	7 (9%)	15 (12%)	22 (11%)
Grade = 5	0 (0%)	5 (4%)	5 (2%)
Patient had at least one SAE of grade
No declared SAE	60 (75%)	90 (75%)	150 (75%)
Grade 1	7 (9%)	4 (3%)	11 (6%)
Grade 2	6 (8%)	5 (4%)	11 (6%)
Grade 3	4 (5%)	10 (8%)	14 (7%)
Grade 4	7 (9%)	10 (8%)	17 (8%)
Grade 5	0 (0%)	5 (4%)	5 (2%)

Pour décrire une sous-population, il faut filtrer df_enrol:

enrolres2 = enrolres %>%
  filter(arm=="Control")
ae %>%
  ae_table_grade(df_enrol=enrolres2, arm="arm") %>%
  as_flextable(header_show_n=TRUE)

	Treatment arm		Total
	Control (N=80)	Treatment (N=0)
Patient maximum AE grade
No declared AE	3 (4%)	0 (NA)	3 (4%)
Grade 1	22 (28%)	0 (NA)	22 (28%)
Grade 2	24 (30%)	0 (NA)	24 (30%)
Grade 3	19 (24%)	0 (NA)	19 (24%)
Grade 4	11 (14%)	0 (NA)	11 (14%)
Grade 5	1 (1%)	0 (NA)	1 (1%)
Patient had at least one AE of grade
No declared AE	3 (4%)	0 (NA)	3 (4%)
Grade ≥ 1	77 (96%)	0 (NA)	77 (96%)
Grade ≥ 2	55 (69%)	0 (NA)	55 (69%)
Grade ≥ 3	31 (39%)	0 (NA)	31 (39%)
Grade ≥ 4	12 (15%)	0 (NA)	12 (15%)
Grade = 5	1 (1%)	0 (NA)	1 (1%)
Patient had at least one AE of grade
No declared AE	3 (4%)	0 (NA)	3 (4%)
Grade 1	66 (82%)	0 (NA)	66 (82%)
Grade 2	44 (55%)	0 (NA)	44 (55%)
Grade 3	21 (26%)	0 (NA)	21 (26%)
Grade 4	12 (15%)	0 (NA)	12 (15%)
Grade 5	1 (1%)	0 (NA)	1 (1%)

On peut modifier le résultat avant de passer en flextable:

ae %>%
  ae_table_grade(df_enrol=enrolres, arm="arm") %>%
  filter(!variable %in% c("Grade 1", "Grade 2")) %>%
  as_flextable(header_show_n=TRUE)

	Treatment arm		Total
	Control (N=80)	Treatment (N=120)
Patient maximum AE grade
No declared AE	3 (4%)	0 (0%)	3 (2%)
Grade 3	19 (24%)	32 (27%)	51 (26%)
Grade 4	11 (14%)	29 (24%)	40 (20%)
Grade 5	1 (1%)	12 (10%)	13 (6%)
Patient had at least one AE of grade
No declared AE	3 (4%)	0 (0%)	3 (2%)
Grade ≥ 1	77 (96%)	120 (100%)	197 (98%)
Grade ≥ 2	55 (69%)	109 (91%)	164 (82%)
Grade ≥ 3	31 (39%)	73 (61%)	104 (52%)
Grade ≥ 4	12 (15%)	41 (34%)	53 (26%)
Grade = 5	1 (1%)	12 (10%)	13 (6%)
Patient had at least one AE of grade
No declared AE	3 (4%)	0 (0%)	3 (2%)
Grade 3	21 (26%)	43 (36%)	64 (32%)
Grade 4	12 (15%)	34 (28%)	46 (23%)
Grade 5	1 (1%)	12 (10%)	13 (6%)

AE par grade : graphiques

• Total
• Par bras
• Absolues

ae_plot_grade(df_ae=ae, df_enrol=enrolres, arm=NULL) #340

ae_plot_grade(df_ae=ae, df_enrol=enrolres, arm="ARM", 
              variant=c("max", "sup")) #453

⚠ Attention aux valeurs absolues pour des essais non équilibrés !

ae_plot_grade(df_ae=ae, df_enrol=enrolres, arm="ARM", 
              type="absolute") #567

AE par SOC : tableaux

• Total
• Par bras
• Population
• Avec Term

ae_table_soc(df_ae=ae, df_enrol=enrolres, arm=NULL) %>%
  as_flextable()

	All patients (N=200)
CTCAE SOC	G1	G2	G3	G4	G5	NA	Tot
Congenital, familial and genetic disorders	24 (12%)	14 (7%)	7 (4%)	4 (2%)	2 (1%)		51 (26%)
Social circumstances	17 (8%)	15 (8%)	7 (4%)	6 (3%)			45 (22%)
Surgical and medical procedures	24 (12%)	5 (2%)	9 (4%)	3 (2%)	2 (1%)		43 (22%)
Neoplasms benign, malignant, and unspecified	16 (8%)	12 (6%)	2 (1%)	7 (4%)	2 (1%)		39 (20%)
Eye disorders	19 (10%)	6 (3%)	8 (4%)	2 (1%)			35 (18%)
Pregnancy, puerperium and perinatal conditions	18 (9%)	8 (4%)	6 (3%)	1 (0%)			33 (16%)
Cardiac disorders	11 (6%)	7 (4%)	4 (2%)	6 (3%)	2 (1%)		30 (15%)
Injury, poisoning and procedural complications	12 (6%)	8 (4%)	8 (4%)	1 (0%)	1 (0%)		30 (15%)
Immune system disorders	12 (6%)	8 (4%)	4 (2%)	3 (2%)			27 (14%)
Endocrine disorders	10 (5%)	7 (4%)	4 (2%)	3 (2%)			24 (12%)
Hepatobiliary disorders	13 (6%)	9 (4%)	1 (0%)	1 (0%)			24 (12%)
Psychiatric disorders	15 (8%)	3 (2%)	3 (2%)	2 (1%)	1 (0%)		24 (12%)
Vascular disorders	10 (5%)	8 (4%)	2 (1%)	1 (0%)	1 (0%)		22 (11%)
Musculoskeletal and connective tissue disorders	8 (4%)	6 (3%)	2 (1%)	5 (2%)			21 (10%)
Respiratory, thoracic and mediastinal disorders	9 (4%)	5 (2%)	2 (1%)	1 (0%)	1 (0%)		18 (9%)
Ear and labyrinth disorders	9 (4%)	6 (3%)	1 (0%)				16 (8%)
Metabolism and nutrition disorders	7 (4%)	3 (2%)	4 (2%)				14 (7%)
General disorders and administration site conditions	8 (4%)	3 (2%)		1 (0%)			12 (6%)
Nervous system disorders	6 (3%)	5 (2%)		1 (0%)			12 (6%)
Skin and subcutaneous tissue disorders	4 (2%)	5 (2%)	1 (0%)	2 (1%)			12 (6%)
Infections and infestations	5 (2%)	4 (2%)		1 (0%)			10 (5%)
Blood and lymphatic system disorders	6 (3%)	1 (0%)	1 (0%)		1 (0%)		9 (4%)
Gastrointestinal disorders	2 (1%)	2 (1%)	4 (2%)	1 (0%)			9 (4%)
Reproductive system and breast disorders	6 (3%)	3 (2%)					9 (4%)
Investigations	4 (2%)	2 (1%)	1 (0%)	1 (0%)			8 (4%)
Renal and urinary disorders	1 (0%)	3 (2%)	1 (0%)	1 (0%)			6 (3%)
No Declared AE						3 (2%)	3 (2%)

ae_table_soc(df_ae=ae, df_enrol=enrolres, arm="arm") %>%
  as_flextable()

	Control (N=80)							Treatment (N=120)
CTCAE SOC	G1	G2	G3	G4	G5	NA	Tot	G1	G2	G3	G4	G5	NA	Tot
Congenital, familial and genetic disorders	12 (15%)	5 (6%)	1 (1%)	2 (2%)			20 (25%)	12 (10%)	9 (8%)	6 (5%)	2 (2%)	2 (2%)		31 (26%)
Social circumstances	8 (10%)	4 (5%)	2 (2%)				14 (18%)	9 (8%)	11 (9%)	5 (4%)	6 (5%)			31 (26%)
Surgical and medical procedures	16 (20%)	1 (1%)	2 (2%)				19 (24%)	8 (7%)	4 (3%)	7 (6%)	3 (2%)	2 (2%)		24 (20%)
Neoplasms benign, malignant, and unspecified	7 (9%)	5 (6%)		3 (4%)	1 (1%)		16 (20%)	9 (8%)	7 (6%)	2 (2%)	4 (3%)	1 (1%)		23 (19%)
Eye disorders	7 (9%)	1 (1%)	3 (4%)	1 (1%)			12 (15%)	12 (10%)	5 (4%)	5 (4%)	1 (1%)			23 (19%)
Pregnancy, puerperium and perinatal conditions	8 (10%)	2 (2%)	1 (1%)				11 (14%)	10 (8%)	6 (5%)	5 (4%)	1 (1%)			22 (18%)
Cardiac disorders	6 (8%)	3 (4%)	2 (2%)	1 (1%)			12 (15%)	5 (4%)	4 (3%)	2 (2%)	5 (4%)	2 (2%)		18 (15%)
Injury, poisoning and procedural complications	5 (6%)	6 (8%)	4 (5%)				15 (19%)	7 (6%)	2 (2%)	4 (3%)	1 (1%)	1 (1%)		15 (12%)
Immune system disorders	7 (9%)	2 (2%)	2 (2%)				11 (14%)	5 (4%)	6 (5%)	2 (2%)	3 (2%)			16 (13%)
Endocrine disorders	4 (5%)	4 (5%)	2 (2%)	1 (1%)			11 (14%)	6 (5%)	3 (2%)	2 (2%)	2 (2%)			13 (11%)
Hepatobiliary disorders	1 (1%)	5 (6%)		1 (1%)			7 (9%)	12 (10%)	4 (3%)	1 (1%)				17 (14%)
Psychiatric disorders	6 (8%)	2 (2%)	1 (1%)				9 (11%)	9 (8%)	1 (1%)	2 (2%)	2 (2%)	1 (1%)		15 (12%)
Vascular disorders	7 (9%)	2 (2%)	1 (1%)				10 (12%)	3 (2%)	6 (5%)	1 (1%)	1 (1%)	1 (1%)		12 (10%)
Musculoskeletal and connective tissue disorders	3 (4%)		1 (1%)	1 (1%)			5 (6%)	5 (4%)	6 (5%)	1 (1%)	4 (3%)			16 (13%)
Respiratory, thoracic and mediastinal disorders	1 (1%)	3 (4%)	2 (2%)				6 (8%)	8 (7%)	2 (2%)		1 (1%)	1 (1%)		12 (10%)
Ear and labyrinth disorders	4 (5%)	3 (4%)					7 (9%)	5 (4%)	3 (2%)	1 (1%)				9 (8%)
Metabolism and nutrition disorders	4 (5%)	1 (1%)	1 (1%)				6 (8%)	3 (2%)	2 (2%)	3 (2%)				8 (7%)
General disorders and administration site conditions	4 (5%)	1 (1%)		1 (1%)			6 (8%)	4 (3%)	2 (2%)					6 (5%)
Nervous system disorders	2 (2%)			1 (1%)			3 (4%)	4 (3%)	5 (4%)					9 (8%)
Skin and subcutaneous tissue disorders	3 (4%)	2 (2%)					5 (6%)	1 (1%)	3 (2%)	1 (1%)	2 (2%)			7 (6%)
Infections and infestations	4 (5%)	2 (2%)					6 (8%)	1 (1%)	2 (2%)		1 (1%)			4 (3%)
Blood and lymphatic system disorders	3 (4%)	1 (1%)					4 (5%)	3 (2%)		1 (1%)		1 (1%)		5 (4%)
Gastrointestinal disorders								2 (2%)	2 (2%)	4 (3%)	1 (1%)			9 (8%)
Reproductive system and breast disorders	1 (1%)	1 (1%)					2 (2%)	5 (4%)	2 (2%)					7 (6%)
Investigations	1 (1%)	1 (1%)	1 (1%)				3 (4%)	3 (2%)	1 (1%)		1 (1%)			5 (4%)
Renal and urinary disorders		3 (4%)					3 (4%)	1 (1%)		1 (1%)	1 (1%)			3 (2%)
No Declared AE						3 (4%)	3 (4%)

Idem, pour décrire une sous-population, il faut filtrer df_enrol:

enrolres2 = enrolres %>% head(50)
ae %>%
  ae_table_soc(df_enrol=enrolres2, arm="arm") %>%
  as_flextable()

	Control (N=18)							Treatment (N=32)
CTCAE SOC	G1	G2	G3	G4	G5	NA	Tot	G1	G2	G3	G4	G5	NA	Tot
Congenital, familial and genetic disorders	3 (17%)	2 (11%)	1 (6%)				6 (33%)	6 (19%)	1 (3%)	1 (3%)				8 (25%)
Injury, poisoning and procedural complications		5 (28%)	4 (22%)				9 (50%)	2 (6%)	1 (3%)	2 (6%)				5 (16%)
Social circumstances	3 (17%)	2 (11%)	1 (6%)				6 (33%)	4 (12%)	2 (6%)	1 (3%)	1 (3%)			8 (25%)
Neoplasms benign, malignant, and unspecified	2 (11%)						2 (11%)	1 (3%)	4 (12%)	2 (6%)	1 (3%)			8 (25%)
Cardiac disorders	1 (6%)			1 (6%)			2 (11%)	2 (6%)	2 (6%)	1 (3%)	1 (3%)			6 (19%)
Surgical and medical procedures	1 (6%)		2 (11%)				3 (17%)	2 (6%)	2 (6%)	1 (3%)				5 (16%)
Hepatobiliary disorders		1 (6%)		1 (6%)			2 (11%)	4 (12%)	1 (3%)					5 (16%)
Musculoskeletal and connective tissue disorders	1 (6%)						1 (6%)	1 (3%)	3 (9%)		2 (6%)			6 (19%)
Psychiatric disorders		1 (6%)					1 (6%)	3 (9%)		1 (3%)	1 (3%)	1 (3%)		6 (19%)
Pregnancy, puerperium and perinatal conditions		1 (6%)					1 (6%)	3 (9%)	2 (6%)					5 (16%)
Ear and labyrinth disorders	1 (6%)						1 (6%)	2 (6%)	2 (6%)					4 (12%)
Endocrine disorders	1 (6%)	1 (6%)	1 (6%)				3 (17%)			2 (6%)				2 (6%)
Eye disorders	2 (11%)						2 (11%)	2 (6%)	1 (3%)					3 (9%)
Skin and subcutaneous tissue disorders	1 (6%)	1 (6%)					2 (11%)		2 (6%)	1 (3%)				3 (9%)
Immune system disorders	1 (6%)	1 (6%)					2 (11%)		2 (6%)					2 (6%)
Respiratory, thoracic and mediastinal disorders								2 (6%)	1 (3%)		1 (3%)			4 (12%)
Vascular disorders	2 (11%)						2 (11%)		2 (6%)					2 (6%)
Gastrointestinal disorders									2 (6%)		1 (3%)			3 (9%)
General disorders and administration site conditions	1 (6%)						1 (6%)	2 (6%)						2 (6%)
Investigations	1 (6%)	1 (6%)	1 (6%)				3 (17%)
Metabolism and nutrition disorders								2 (6%)		1 (3%)				3 (9%)
Renal and urinary disorders										1 (3%)	1 (3%)			2 (6%)
Reproductive system and breast disorders								2 (6%)						2 (6%)
Nervous system disorders	1 (6%)						1 (6%)
No Declared AE						1 (6%)	1 (6%)

ae_table_soc(df_ae=ae, df_enrol=enrolres, arm="ARM", 
             term="aeterm") %>%
  as_flextable()

		Control (N=80)							Treatment (N=120)
CTCAE SOC	CTCAE v4.0 Term	G1	G2	G3	G4	G5	NA	Tot	G1	G2	G3	G4	G5	NA	Tot
Congenital, familial and genetic disorders	Chromosomal abnormalities	6 (8%)	1 (1%)		1 (1%)			8 (10%)	5 (4%)	3 (2%)	2 (2%)				10 (8%)
	Congenital nervous system disorders	2 (2%)	2 (2%)					4 (5%)	4 (3%)	1 (1%)					5 (4%)
	Familial hematologic disorders	4 (5%)			1 (1%)			5 (6%)	2 (2%)	1 (1%)	3 (2%)	2 (2%)			8 (7%)
	Hereditary connective tissue disorders	2 (2%)	2 (2%)	1 (1%)				5 (6%)	5 (4%)	4 (3%)	1 (1%)		2 (2%)		12 (10%)
Social circumstances	Cultural issues	3 (4%)	1 (1%)					4 (5%)	2 (2%)	1 (1%)	2 (2%)	1 (1%)			6 (5%)
	Economic conditions affecting care	1 (1%)	1 (1%)	1 (1%)				3 (4%)	1 (1%)	3 (2%)		2 (2%)			6 (5%)
	Family support issues	1 (1%)						1 (1%)	4 (3%)	4 (3%)	3 (2%)	2 (2%)			13 (11%)
	Social and environmental issues	3 (4%)	2 (2%)	1 (1%)				6 (8%)	4 (3%)	3 (2%)		1 (1%)			8 (7%)
Surgical and medical procedures	Device implantation procedures	10 (12%)						10 (12%)	3 (2%)		1 (1%)	2 (2%)	1 (1%)		7 (6%)
	Diagnostic procedures	4 (5%)		1 (1%)				5 (6%)	5 (4%)	1 (1%)	4 (3%)		1 (1%)		11 (9%)
	Surgical complications	1 (1%)	1 (1%)					2 (2%)	1 (1%)	3 (2%)	2 (2%)	1 (1%)			7 (6%)
	Therapeutic procedures	1 (1%)		1 (1%)				2 (2%)	1 (1%)						1 (1%)
Neoplasms benign, malignant, and unspecified	Benign neoplasms	2 (2%)	1 (1%)		1 (1%)			4 (5%)	2 (2%)	4 (3%)	1 (1%)	2 (2%)			9 (8%)
	Malignant neoplasms	3 (4%)	2 (2%)					5 (6%)	3 (2%)	2 (2%)	1 (1%)	2 (2%)			8 (7%)
	Neoplasms unspecified	2 (2%)			1 (1%)	1 (1%)		4 (5%)	3 (2%)						3 (2%)
	Tumor progression		2 (2%)		1 (1%)			3 (4%)	2 (2%)	1 (1%)			1 (1%)		4 (3%)
Eye disorders	Corneal disorders	3 (4%)			1 (1%)			4 (5%)	1 (1%)		2 (2%)	1 (1%)			4 (3%)
	Eyelid disorders	2 (2%)						2 (2%)	4 (3%)		2 (2%)				6 (5%)
	Retinal disorders	1 (1%)		1 (1%)				2 (2%)	4 (3%)	2 (2%)	1 (1%)				7 (6%)
	Vision disorders	2 (2%)	1 (1%)	2 (2%)				5 (6%)	5 (4%)	3 (2%)					8 (7%)
Pregnancy, puerperium and perinatal conditions	Breastfeeding issues	4 (5%)	1 (1%)					5 (6%)			2 (2%)				2 (2%)
	Fetal complications	2 (2%)						2 (2%)	1 (1%)	2 (2%)	2 (2%)				5 (4%)
	Labor and delivery complications	1 (1%)	1 (1%)	1 (1%)				3 (4%)	6 (5%)	3 (2%)					9 (8%)
	Pregnancy complications	2 (2%)						2 (2%)	3 (2%)	1 (1%)	1 (1%)	1 (1%)			6 (5%)
Injury, poisoning and procedural complications	Poisonings	1 (1%)	3 (4%)	1 (1%)				5 (6%)	2 (2%)			1 (1%)	1 (1%)		4 (3%)
	Procedural complications	1 (1%)	2 (2%)	1 (1%)				4 (5%)	2 (2%)		1 (1%)				3 (2%)
	Radiation-related toxicities	1 (1%)	1 (1%)					2 (2%)	1 (1%)	2 (2%)	2 (2%)				5 (4%)
	Traumatic injuries	2 (2%)	1 (1%)	2 (2%)				5 (6%)	4 (3%)		1 (1%)				5 (4%)
Cardiac disorders	Cardiac arrhythmias	2 (2%)	1 (1%)					3 (4%)	2 (2%)			4 (3%)			6 (5%)
	Cardiac valve disorders	1 (1%)		1 (1%)				2 (2%)	1 (1%)	1 (1%)	1 (1%)	1 (1%)	2 (2%)		6 (5%)
	Coronary artery disorders	1 (1%)	1 (1%)					2 (2%)	1 (1%)	4 (3%)					5 (4%)
	Heart failures	3 (4%)	1 (1%)	1 (1%)	1 (1%)			6 (8%)	1 (1%)		1 (1%)				2 (2%)
Immune system disorders	Autoimmune disorders	4 (5%)						4 (5%)	1 (1%)	1 (1%)		2 (2%)			4 (3%)
	Hypersensitivity conditions	2 (2%)						2 (2%)	2 (2%)	1 (1%)	1 (1%)	1 (1%)			5 (4%)
	Immunodeficiency	2 (2%)	1 (1%)					3 (4%)			1 (1%)				1 (1%)
	Inflammatory responses		1 (1%)	2 (2%)				3 (4%)	3 (2%)	5 (4%)					8 (7%)
Endocrine disorders	Adrenal gland disorders	1 (1%)	3 (4%)					4 (5%)	1 (1%)	2 (2%)					3 (2%)
	Parathyroid gland disorders	1 (1%)		1 (1%)				2 (2%)			2 (2%)	1 (1%)			3 (2%)
	Pituitary gland disorders	3 (4%)	1 (1%)	1 (1%)				5 (6%)	3 (2%)	1 (1%)					4 (3%)
	Thyroid gland disorders	1 (1%)			1 (1%)			2 (2%)	2 (2%)	1 (1%)		1 (1%)			4 (3%)
Psychiatric disorders	Anxiety disorders	1 (1%)						1 (1%)	3 (2%)		1 (1%)				4 (3%)
	Mood disorders	2 (2%)						2 (2%)				1 (1%)			1 (1%)
	Sleep disorders	3 (4%)	1 (1%)					4 (5%)	5 (4%)						5 (4%)
	Substance-related disorders		1 (1%)	1 (1%)				2 (2%)	2 (2%)	1 (1%)	1 (1%)	1 (1%)	1 (1%)		6 (5%)
Hepatobiliary disorders	Bile duct disorders	1 (1%)	2 (2%)					3 (4%)	3 (2%)	1 (1%)					4 (3%)
	Gallbladder disorders		1 (1%)					1 (1%)	3 (2%)	1 (1%)					4 (3%)
	Hepatic failure		1 (1%)		1 (1%)			2 (2%)	3 (2%)	1 (1%)	1 (1%)				5 (4%)
	Liver disorders		1 (1%)					1 (1%)	3 (2%)	1 (1%)					4 (3%)
Vascular disorders	Hypertension-related conditions	2 (2%)	1 (1%)					3 (4%)	1 (1%)	2 (2%)	1 (1%)				4 (3%)
	Hypotension-related conditions	2 (2%)						2 (2%)	2 (2%)	1 (1%)			1 (1%)		4 (3%)
	Vascular hemorrhagic disorders	1 (1%)		1 (1%)				2 (2%)		3 (2%)					3 (2%)
	Venous thromboembolic events	2 (2%)	1 (1%)					3 (4%)		1 (1%)		1 (1%)			2 (2%)
Musculoskeletal and connective tissue disorders	Arthritis and joint disorders	2 (2%)						2 (2%)	3 (2%)	2 (2%)	1 (1%)				6 (5%)
	Bone disorders	1 (1%)			1 (1%)			2 (2%)	1 (1%)	1 (1%)		1 (1%)			3 (2%)
	Connective tissue disorders			1 (1%)				1 (1%)	1 (1%)	2 (2%)		2 (2%)			5 (4%)
	Muscle disorders									2 (2%)		1 (1%)			3 (2%)
Ear and labyrinth disorders	Hearing disorders								1 (1%)	1 (1%)	1 (1%)				3 (2%)
	Labyrinth disorders	1 (1%)						1 (1%)
	Tinnitus	4 (5%)	1 (1%)					5 (6%)	2 (2%)						2 (2%)
	Vertigo and balance disorders		2 (2%)					2 (2%)	3 (2%)	2 (2%)					5 (4%)
Respiratory, thoracic and mediastinal disorders	Lung function disorders		1 (1%)					1 (1%)	4 (3%)	1 (1%)					5 (4%)
	Pleural disorders	1 (1%)	1 (1%)	1 (1%)				3 (4%)				1 (1%)			1 (1%)
	Pulmonary vascular disorders		1 (1%)					1 (1%)	3 (2%)	1 (1%)			1 (1%)		5 (4%)
	Respiratory infections			1 (1%)				1 (1%)	1 (1%)						1 (1%)
Metabolism and nutrition disorders	Fluid and electrolyte disorders	1 (1%)						1 (1%)	1 (1%)	1 (1%)	1 (1%)				3 (2%)
	Lipid metabolism disorders		1 (1%)	1 (1%)				2 (2%)
	Nutritional disorders	3 (4%)						3 (4%)	1 (1%)						1 (1%)
	Vitamin deficiencies								1 (1%)	1 (1%)	2 (2%)				4 (3%)
Nervous system disorders	Headache disorders	1 (1%)						1 (1%)	2 (2%)	2 (2%)					4 (3%)
	Neurological disorders of the central nervous system								2 (2%)	2 (2%)					4 (3%)
	Peripheral neuropathies				1 (1%)			1 (1%)
	Seizure disorders	2 (2%)						2 (2%)		1 (1%)					1 (1%)
General disorders and administration site conditions	Device issues	2 (2%)			1 (1%)			3 (4%)	2 (2%)						2 (2%)
	General physical health deterioration	1 (1%)						1 (1%)	2 (2%)						2 (2%)
	Injection site reactions	1 (1%)						1 (1%)		1 (1%)					1 (1%)
	Pain and discomfort		1 (1%)					1 (1%)		1 (1%)					1 (1%)
Skin and subcutaneous tissue disorders	Dermatitis	1 (1%)						1 (1%)			1 (1%)	1 (1%)			2 (2%)
	Skin and subcutaneous tissue injuries	1 (1%)						1 (1%)		2 (2%)					2 (2%)
	Skin infections	1 (1%)	1 (1%)					2 (2%)	1 (1%)	1 (1%)		1 (1%)			3 (2%)
	Skin pigmentation disorders		1 (1%)					1 (1%)
Infections and infestations	Bacterial infectious disorders	1 (1%)						1 (1%)
	Fungal infectious disorders		1 (1%)					1 (1%)
	Parasitic infectious disorders	1 (1%)						1 (1%)	1 (1%)			1 (1%)			2 (2%)
	Viral infectious disorders	2 (2%)	1 (1%)					3 (4%)		2 (2%)					2 (2%)
Blood and lymphatic system disorders	Bone marrow disorders	1 (1%)						1 (1%)
	Hematologic neoplasms	2 (2%)	1 (1%)					3 (4%)	2 (2%)				1 (1%)		3 (2%)
	Red blood cell disorders								1 (1%)		1 (1%)				2 (2%)
Gastrointestinal disorders	Esophageal disorders											1 (1%)			1 (1%)
	Gastric disorders								2 (2%)	1 (1%)	2 (2%)				5 (4%)
	Intestinal disorders									1 (1%)	2 (2%)				3 (2%)
Reproductive system and breast disorders	Breast disorders								2 (2%)						2 (2%)
	Female reproductive disorders	1 (1%)						1 (1%)	1 (1%)						1 (1%)
	Male reproductive disorders		1 (1%)					1 (1%)	2 (2%)	1 (1%)					3 (2%)
	Menstrual disorders									1 (1%)					1 (1%)
Investigations	Blood analyses			1 (1%)				1 (1%)		1 (1%)					1 (1%)
	Cardiovascular assessments	1 (1%)						1 (1%)	1 (1%)						1 (1%)
	Imaging studies		1 (1%)					1 (1%)	1 (1%)						1 (1%)
	Liver function analyses								1 (1%)			1 (1%)			2 (2%)
Renal and urinary disorders	Bladder disorders		1 (1%)					1 (1%)			1 (1%)				1 (1%)
	Kidney disorders		1 (1%)					1 (1%)	1 (1%)			1 (1%)			2 (2%)
	Urinary tract disorders		1 (1%)					1 (1%)
No Declared AE							3 (4%)	3 (4%)

AE par SOC : graphiques

• Butterfly
• Serious

butterfly_plot(df_ae=ae, df_enrol=enrolres)

ae %>% 
  mutate(serious = sae=="Yes") %>% #colonne logical/booleanZ
  butterfly_plot(df_enrol=enrolres, severe="serious") +
  labs(caption="Darker areas represent Serious Adverse Events")

DEV: gr_new_project()

• Fonction évolutive pour créer un nouveau projet.

• Crée un projet RStudio dans le chemin spécifié

• Crée un squelette standard avec des fichiers pré-remplis

• La fonction va évoluer avec le temps

• Exemple

  gr_new_project("HRNBL2/IDMC_27")

├── main.R
├── NEWS.md
├── R
│   ├── init.R
│   ├── read.R
│   ├── check.R
│   ├── description.R
│   ├── graph.R
│   └── report.R
├── README.md
└── my_proj.Rproj

To be continued

grstat est un package encore jeune et plein de fonctionnalités sont en cours de développement :

• Tables de sorties de modèles
• waterfall_plot() RECIST
• swimmer_plot() RECIST
• Template officer standard
• Création de listes de randomisation
• forest_plot()
• Check automatique de la fiche RECIST standard (TrialMaster)

Vous pouvez demander des fonctionnalités sur https://github.com/Oncostat/grstat/issues.

1