Further survival analyses

Set up

Let us first load the packages required.

library(CDMConnector)
library(CohortSurvival)
library(dplyr)
library(cmprsk)
library(survival)

We will create a cdm reference to use our example MGUS2 survival dataset.

cdm <- CohortSurvival::mockMGUS2cdm()

The CohortSurvival package focuses on Kaplan-Meier survival estimates and cumulative incidence in a competing-risk setting. It does not fit more complex models, such as Cox proportional hazards or Fine and Gray models, directly. However, the format the data has to be in to be inputted to well-known modelling functions from packages like survival or cmprsk can be retrieved from OMOP data with addCohortSurvival(). Let us see how to do it in both single-event and competing-risk survival settings.

Further analysis with single event survival

To get the time and status information we need for the coxph function in the package survival, for instance, we only need to call addCohortSurvival(). The stratification variables need to be columns previously added to the cohort by the user. status is 1 for people with the outcome event and 0 for people censored before the event. time is the number of days from cohort entry to the event or censoring.

input_survival_single <- cdm$mgus_diagnosis |>
       addCohortSurvival(
       cdm = cdm,
       outcomeCohortTable = "death_cohort",
       outcomeCohortId = 1
       ) 

input_survival_single |> 
  glimpse()
#> Rows: ??
#> Columns: 13
#> $ cohort_definition_id <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1…
#> $ subject_id           <int> 1, 2, 3, 4, 5, 6, 7, 8, 10, 11, 12, 13, 14, 15, 1…
#> $ cohort_start_date    <date> 1981-01-01, 1968-01-01, 1980-01-01, 1977-01-01, …
#> $ cohort_end_date      <date> 1981-01-01, 1968-01-01, 1980-01-01, 1977-01-01, …
#> $ age                  <dbl> 88, 78, 94, 68, 90, 90, 89, 87, 79, 86, 89, 87, 8…
#> $ sex                  <fct> F, F, M, M, F, M, F, F, F, M, F, M, F, M, F, F, M…
#> $ hgb                  <dbl> 13.1, 11.5, 10.5, 15.2, 10.7, 12.9, 10.5, 12.3, 9…
#> $ creat                <dbl> 1.30, 1.20, 1.50, 1.20, 0.80, 1.00, 0.90, 1.20, 1…
#> $ mspike               <dbl> 0.5, 2.0, 2.6, 1.2, 1.0, 0.5, 1.3, 1.6, 2.3, 2.3,…
#> $ age_group            <chr> ">=70", ">=70", ">=70", "<70", ">=70", ">=70", ">…
#> $ days_to_exit         <int> 30, 25, 46, 92, 8, 4, 151, 2, 136, 2, 108, 10, 14…
#> $ status               <dbl> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1…
#> $ time                 <dbl> 30, 25, 46, 92, 8, 4, 151, 2, 136, 2, 108, 10, 14…

We can decide to change some of the default parameters in this function. Information on all these can be found in ?addCohortSurvival. For instance, we can choose to exclude people with an outcome only 180 days before index date, instead of anytime, and follow them up for only one year. We can also decide to use cohort_end_date as the outcome date variable and censor them at a particular date, for instance, the 1st of January of 1994. We see how that gives us different results:

cdm$mgus_diagnosis |>
       addCohortSurvival(
       cdm = cdm,
       outcomeCohortTable = "death_cohort",
       outcomeWashout = 180,
       followUpDays = 365
       ) |>
  filter(cohort_start_date > "1993-01-01") |>
  glimpse()
#> Rows: ??
#> Columns: 13
#> $ cohort_definition_id <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1…
#> $ subject_id           <int> 213, 671, 941, 1245, 1248, 1340, 1374, 1376, 1380…
#> $ cohort_start_date    <date> 1994-01-01, 1994-01-01, 1994-01-01, 1994-01-01, …
#> $ cohort_end_date      <date> 1994-01-01, 1994-01-01, 1994-01-01, 1994-01-01, …
#> $ age                  <dbl> 93, 81, 85, 82, 87, 67, 68, 67, 69, 66, 79, 86, 8…
#> $ sex                  <fct> F, M, M, M, M, F, F, F, M, M, M, F, F, F, F, F, M…
#> $ hgb                  <dbl> 12.8, 13.0, 13.6, 11.4, 12.7, 12.2, 9.2, 13.7, 15…
#> $ creat                <dbl> 1.1, 1.4, 1.1, 1.5, 1.5, 1.4, 1.8, 1.1, 0.8, 2.0,…
#> $ mspike               <dbl> 0.8, 1.3, 1.5, 1.4, 0.5, 1.2, 0.5, 1.5, 0.0, 0.0,…
#> $ age_group            <chr> ">=70", ">=70", ">=70", ">=70", ">=70", "<70", "<…
#> $ days_to_exit         <dbl> 19, 43, 12, 1, 10, 46, 40, 41, 22, 31, 6, 57, 52,…
#> $ status               <dbl> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0…
#> $ time                 <dbl> 19, 43, 12, 1, 10, 46, 40, 41, 22, 31, 6, 57, 52,…
cdm$mgus_diagnosis |>
       addCohortSurvival(
       cdm = cdm,
       outcomeCohortTable = "death_cohort",
       outcomeDateVariable = "cohort_end_date",
       censorOnDate = as.Date("1994-01-01")
       ) |>
    filter(cohort_start_date > "1993-01-01") |>
  glimpse()
#> Rows: ??
#> Columns: 13
#> $ cohort_definition_id <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1…
#> $ subject_id           <int> 213, 671, 941, 1245, 1248, 1340, 1374, 1376, 1380…
#> $ cohort_start_date    <date> 1994-01-01, 1994-01-01, 1994-01-01, 1994-01-01, …
#> $ cohort_end_date      <date> 1994-01-01, 1994-01-01, 1994-01-01, 1994-01-01, …
#> $ age                  <dbl> 93, 81, 85, 82, 87, 67, 68, 67, 69, 66, 79, 86, 8…
#> $ sex                  <fct> F, M, M, M, M, F, F, F, M, M, M, F, F, F, F, F, M…
#> $ hgb                  <dbl> 12.8, 13.0, 13.6, 11.4, 12.7, 12.2, 9.2, 13.7, 15…
#> $ creat                <dbl> 1.1, 1.4, 1.1, 1.5, 1.5, 1.4, 1.8, 1.1, 0.8, 2.0,…
#> $ mspike               <dbl> 0.8, 1.3, 1.5, 1.4, 0.5, 1.2, 0.5, 1.5, 0.0, 0.0,…
#> $ age_group            <chr> ">=70", ">=70", ">=70", ">=70", ">=70", "<70", "<…
#> $ days_to_exit         <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0…
#> $ status               <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0…
#> $ time                 <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0…

This table with the added time and status information should be enough to call any advanced function, like the aforementioned Cox Proportional Hazards model:

survival::coxph(survival::Surv(time, status) ~ age + sex, data = input_survival_single)
#> Call:
#> survival::coxph(formula = survival::Surv(time, status) ~ age + 
#>     sex, data = input_survival_single)
#> 
#>          coef exp(coef) se(coef)      z        p
#> age  0.061622  1.063561 0.003402 18.114  < 2e-16
#> sexM 0.358258  1.430835 0.065693  5.454 4.94e-08
#> 
#> Likelihood ratio test=391.2  on 2 df, p=< 2.2e-16
#> n= 1384, number of events= 963
survival::survdiff(survival::Surv(time, status) ~ sex, data = input_survival_single)
#> Call:
#> survival::survdiff(formula = survival::Surv(time, status) ~ sex, 
#>     data = input_survival_single)
#> 
#>         N Observed Expected (O-E)^2/E (O-E)^2/V
#> sex=F 631      423      471      4.88      9.67
#> sex=M 753      540      492      4.67      9.67
#> 
#>  Chisq= 9.7  on 1 degrees of freedom, p= 0.002

Further analysis with competing risk survival

For competing-risk settings, we need to use the same function that adds time and status information, but twice. We first add time and status information for the outcome, then for the competing outcome. Then we combine those variables to identify which outcome, if any, happened first for each individual so that we can feed the result to subsequent models. In the coding below, status = 0 means censored, status = 1 means the event of interest, and status = 2 means the competing outcome.


# Add all status and time information for both outcomes
  input_survival_cr <- cdm$mgus_diagnosis |>
    addCohortSurvival(cdm, "progression") |>
    dplyr::rename(
      "outcome_time" = "time",
      "outcome_status" = "status"
    ) |>
     addCohortSurvival(cdm, "death_cohort") |>
    dplyr::rename(
      "competing_outcome_time" = "time",
      "competing_outcome_status" = "status"
    )
  
  # Collect and combine the two event processes
  input_survival_cr <- input_survival_cr |>
    dplyr::collect() |>
    dplyr::mutate(
      time = pmin(outcome_time, competing_outcome_time),
      status = factor(
        dplyr::if_else(competing_outcome_time <= outcome_time, 2 * competing_outcome_status, outcome_status))
    ) |>
    dplyr::select(-c("outcome_time", "outcome_status", "competing_outcome_time", "competing_outcome_status"))

We can use the package cmprsk to fit a Fine and Gray model to the competing risk data. We first change our sex covariate to numeric, and then we can run the analysis:

input_survival_cr <- input_survival_cr |>
  dplyr::mutate(sex = dplyr::if_else(sex == "M", 0, 1))

covs <- data.frame(input_survival_cr$age, input_survival_cr$sex)
names(covs) <- c("age", "sex")

summary(cmprsk::crr(ftime = input_survival_cr$time,
            fstatus = input_survival_cr$status,
            cov1 = covs,
            failcode = 1,
            cencode = 0))
#> Competing Risks Regression
#> 
#> Call:
#> cmprsk::crr(ftime = input_survival_cr$time, fstatus = input_survival_cr$status, 
#>     cov1 = covs, failcode = 1, cencode = 0)
#> 
#>        coef exp(coef) se(coef)     z p-value
#> age -0.0192     0.981  0.00585 -3.28   0.001
#> sex  0.2871     1.333  0.19309  1.49   0.140
#> 
#>     exp(coef) exp(-coef)  2.5% 97.5%
#> age     0.981       1.02 0.970 0.992
#> sex     1.333       0.75 0.913 1.945
#> 
#> Num. cases = 1384
#> Pseudo Log-likelihood = -726 
#> Pseudo likelihood ratio test = 8.32  on 2 df,

Disconnect from the cdm database connection

We finish by disconnecting from the cdm database connection.

cdmDisconnect(cdm)