PEANUTS: Case-Cohort Studies

 

Input commands are in blue.
Statements generated by PEANUTS are in green.

The data consists of individual records on breast cancer incidence and radiation dose in a cohort of women who received repeated chest floroscopy as part of their treatment for tuberculosis.  These data consist of a sample of 20 percent of the cohort members and all breast cancer cases, whether or not they were included in the sample.  This is a subset of the data described by Z. Hrubec et al in Cancer Research, 49, 1989.
 

Describe and read the data.  Since not all the variables are needed, a format statement is used.

NAMES id bcflag byr bmo bday bc icd8 xyr xmo xday iyr imo iday dose agex @
FORMAT '(f4.0,f2.0,t12,f4.0,2f3.0,t23,f1.0,f4.0,t28,3f3.0,t37,3f3.0,t60,f8.3,t78,f5.2)' @  ! Exclude bilateral breast cancers
TRAN IF bc > 1 THEN DELETE ENDIF @
 INPUT casechrt.dat @
Input from casechrt.dat
        395 records read        395 records used                                   0 records rejected   Workspace for 300 variables.
17 are currently defined.  Up to 283 new variables can be created

The MISS command indicates that values of dose less than zero are to be treated as missing.  Dose is rescaled and a number of other transformations are made to compute age at exposure and age at entry and other quantities.
MISS dose < 0 @
! Change to Gray
TRAN dose = dose/100 @
! Compute dates and ages
TRAN xyr = xyr + 1900;
 iyr = iyr + 1900 ;
 idate = julian(iyr,imo,iday)/365.25 ;
 xdate = julian(xyr,xmo,xday)/365.25 ;
 bdate = julian(byr,bmo,bday)/365.25 ;
 xage = xdate - bdate ;
iage = idate - bdate ; @

Indicate the data are from a case-cohort study with a 20 percent sampling fraction. and specify the case indicator (EVENT) and the time variable.  In this analysis the time scale for the background hazard is attained age, but women do not enter the study until their first treatment.  Thus, iage is the entry-time variable.
CSCHT 0.2 @
EVENT bcflag @
ENTRY iage @
TIME xage @

Fit a linear dose response model with no effect modifiers.  This model can be written as:

 LINE 1 dose @   FIT @
 Sorting data on entry time ...
*** WARNING: Exit time exceeds entry time for 3 records
Subjects dropped from analysis
         Iter  Step Deviance
            0    0   793.3531
            1    0   782.7261
           2    0   782.0896
           3    0   782.0786
           4    0   782.0786
Hazard function regression model
Case-cohort design with sampling fraction  0.200
Product additive excess RR model {RR = T0*(1 + T1 + T2 + ...)}
XAGE     is used for survival time.
BCFLAG   is used for outcomes
IAGE     is used for entry (left-truncation) times
                            Parameter Summary Table
 # Name                          Estimate     Std.Err.    Test Stat.  P value
-- -----------------------  ------------  ----------  -----------  -------
Linear term 1
1 DOSE ....................      0.7731      0.3976        1.944   0.052
Records used =      390         Records dropped           =      5
    Deviance  =     782.079      Free parameters           =       1
     Number of risk sets       =      75

Add age at exposure as an effect modifier.  The new model is :

LOGL 1 agex @
FIT @
Sorting data on entry time ...
        Iter  Step Deviance
         0    0   782.0786
         1    0   781.2375
         2    0   780.6490
         3    0   780.5376
         4    0   780.5328
         5    0   780.5325
         6    0   780.5325
Hazard function regression model
Case-cohort design with sampling fraction  0.200
Product additive excess RR model {RR = T0*(1 + T1 + T2 + ...)}
 XAGE     is used for survival time.
 BCFLAG   is used for outcomes
 IAGE     is used for entry (left-truncation) times
                             Parameter Summary Table
  # Name                    Estimate       Std.Err.     Test Stat.  P value
-- -------------------   ------------    ----------   -----------  -------
Linear term 1
 1 DOSE ................      3.328          4.118       0.8083     0.419
Log-linear term 1
 2 AGEX ................    -0.06408     0.05946    -1.078      0.281
Records used =      390         Records dropped           =       5
Deviance  =     780.533         Free parameters            =        2
                                              Number of risk sets       =      75

Return to HiroSoft's Home Page
Back to the directory of examples
Back to the PEANUTS overview