A COMPUTER PROGRAM PACKAGE

FOR CANCER SURVIVAL STUDIES

Timo Hakulinen

Robert Gibberd

Kamal Abeywickrama

Bengt Söderman

June 1997

To download the complete package as a selfexpandable file click here

Version 2 of the package (SURV2) is also available.

In order to run the package in a DOS box under MS windows, the following device drivers (included in the package) must be installed:

1. DOSXMSF.EXE must be installed in the same directory as the executables or in the path. DOSXMSF.EXE is a DOS extender that is required in order to run 32-bit programs under MS-DOS.

2. DOSXNT.386 must be installed (by calling it in SYSTEM.INI). For example, add the following entry in the SYSTEM.INI file under the [386Enh] section:

device=C:\windows\dosxnt.386

(assuming DOSXNT.386 is in c:\windows). DOSXNT.386 is a Windows device driver that is necessary if you want to run the FORTRAN application in an MS-DOS session under Windows. Some reports indicate that this must be the last line in the [386Enh] section of SYSTEM.INI.

 


CONTENTS


0. Introduction

The purpose of this program package is to calculate relative survival rates and to perform certain statistical tests on these calculated statistics. For the theoretical background and terminology of these methods, refer to Ederer et al. (1961), Hakama and Hakulinen (1977), Hakulinen (1977), Hakulinen (1982), Hakulinen and Abeywickrama (1985), and Hakulinen et. al. (1987). The references are given in Section 7. The creation of data files to enable a regression analysis extension based on the GLIM statistical package (Hakulinen and Tenkanen 1987) is also included together with the GLIM macros.

This package was designed for mainframe computers at the Finnish Cancer Registry at the request of the Organization of European Cancer Institutes, and subsequently modified for micro-computers at the University of Newcastle, Australia.

Available versions and their characteristics are described in Appendix C.


1. General description

The computer package has three parts (Appendix A). The first is the TABULATN program, the second the PRINTNG program and the third (which is optional) is the GROUPTST program. The GROUPTST program may be used in conjunction with the first two programs or separately. In the following section, an overview of the procedure for using the first two programs by themselves is given; the details for each program are then given in the following three sections (sections 2, 3 and 4). The use of the regression model is described in section 5.

Instructions to install the programs on your computer are given in Appendix C. Examples of output are given in Appendix B.

To use the relative survival analysis programs it is necessary to create the following data files on your computer.

The creation of these files is not done by the package, but by the user using his own programs and/or editors. The formats of these files are given in subsequent sections.

Once the files have been created, the first program to run is called TABULATN which processes the patient data and population mortality files in accordance with the specifications of the parameter file, to create a file containing the data required to produce the life-tables. The formatting and writing of the life-table files is done by the program PRINTNG which also requires a parameter file to specify the required life-table columns required. An example showing the dialogue between the computer and and the user when running these two programs is given below (input typed by the user on the keyboard is in boldface). The default file names are given in square brackets and can be obtained by pressing <return>.

C:\SURV16>tabulatn
Tabulation program Version 1.6
Finnish Cancer Registry October 1996
The date is: 15.11.1996

Type the tabulation parameter file-name [tabu.par]mel.par<return>
Type the patient data file-name [cancer.dat]mel.dat<return>
Type the population mortality file-name [popmort.par]<return>
Type the output filename [tabulatn.lis]meltab.lis<return>

Stop - Program terminated

C:\SURV16>printng
Printing program Version 1.6
Finnish Cancer Registry October 1996
The date is: 15.11.1996

Type the printl parameter file name [printl.par]<return>
Type the output file-name [printng.lis]mel.lis<return>
Type the print2 parameter file.name [print2.par]<return>
Type the glim parameter file-name [glim.par]<return>

Stop - Program terminated

C:\SURV16>


2. Input and output files needed for the TABULATN program

The first program to run is TABULATN and this program requires the following input files (you must set your default directory to the one, which holds the programs, if the input files are in another directory, you must specify the path). The default file names are used here, although the user can define his own names as shown in the previous example.

The output files are called:

The file TABULATN.LIS can be written to any directory if you specify the path. The latter two files (TEMP.DAT and TRANS.DAT) are used to transmit data to the printing program and will be written to the same directory where the programs are

2.1 Patient datafile (cancer.dat)

This file contains the patient data and is a 'rectangular' file with each line (record) containing patient data (fields) as follows:

         field 1   field 2   field 3

record 1     903         0        41
record 2     903         0        36
record 3     906         0        27

The first few lines of the sample data set are shown below where the file description is:

        name    starting width codes
                  column

field 1 site           1     3 see below
field 2 sex            4     1 0 = male, 1= female
field 3 age            5     3 years
field 4 month of diag. 8     2 MM
field S year of diag. 10     2 YY
field 6 treatment     12     2 not given here
field 7 stage         14     1 0: unknown;
                               1: localized;
                               2 - 9: non-localized.
field 8 status at end 20     1 0: alive;
        of follow-up           1: dead;
                               2: lost to follow-up
field 9 survival time 21     2 in completed years/

site codes: 900:1ip,901: lid, 902: ear,903: face, 904: other location in head and neck,905: trunk, 906: upper limbs, 907: lower limbs,908: multiple location,909: undefined

9030 41 7534111 7
9030 36 1534111 4
9060 2712534131 0
9031 46 8534111 1
9071 6910534121 3
9051 31 1534111 1
9051 29 9534121 3
9051 3610534111 1
9041 39 2534111 1
9071 44 6534111 7
9010 70 8534111 8
9050 36 4534111 1

The program does not require that the fields have any special width or position as the parameter file (tabu.par) provides the file specifications.

2.2 Population mortality file (popmort.dat)

This file describes the mortality of the general population and allows the relative survival ratios to be calculated. It contains the probability of death (or survival) and the life expectancy for single year age-groups for both sexes, for the relevant calendar years. These data are usually obtained from the official publications of mortality for the country or state from which the patients are registered. The file has two parts: annual expected probabilities of death for the general population and life expectancy for the general population. The file consists of blocks of records, each block has 10 records. The blocks are:

The probabilities should be multiplied by 10*5 and the expected lifetimes should be given in years. The data are for one-year age groups in the range 0 - 99 years. The first record in each block is for the ages 0-9, the second for the ages 10-19, and so on. The tenth record is for the ages 90-99. Ages higher than 99 years can be accommodated by using the AGE parameter in the LIFE namelist. A record for annual expected probabilities of death has the format 10(I7) (i.e. a field width of 7 columns rightjustified). A record of the expected lengths of life, for the general population, has the format of 10(F7.2) (i.e. a field width of 7 characters with 2 decimal points, rightjustified). If the expected lengths of life are not available, calculations of mean length of life are not possible (XAGE = O in LIFE namelist). If there is no mortality file at all (INEX = 0 in LIFE namelist) only observed survival rates can be calculated.

If the values are needed for a more recent calender year interval than the last one in the popmort.dat file, values in the most recent time period are used by the program. Similarly, if results for ages greater than 99 years are needed, values for the age of 99 years are used.

If the year of the beginning of follow-up (e.g. diagnosis) is after the last year available in the POPMORT.DAT file, the patient's record is rejected. Hence, it is suggested that, for example, the death probability and life expectation values for 1976-80 be duplicated for 1981-85 as well, if there is a patient diagnosed in 1982 when the last year available for general mortality is 1980.

Some lines from the Finnish popmort.dat file are given below:

03571 00361 00217 00158 00118 00107 00087 00095 00080 00069 male
00060 00061 00080 00075 00086 00087 00103 00118 00165 00164
27164 30274 26369 31006 34874 31343 35211 47619 45000 14286
02825 00284 00140 00110 00082 00075 00059 00051 00049 00042 female
00039 00043 00053 00052 00053 00056 00063 00073 00094 00091
63.37 64.71 63.94 63.08 62.17 61.25 60.31 59.36 58.42 57.47 male
56.51 55.54 54.57 53.62 52.66 51.70 50.75 49.80 48.86 47.94
02.81 02.68 02.62 02.38 02.23 02.15 01.90 01.67 01.73 01.73
69.84 70.86 70.06 69.16 68.24 67.29 66.34 65.38 64.41 63.45 female
62.47 61.50 60.52 59.55 58.58 57.62 56.65 55.68 54.72 53.77

2.3 Parameterfile for TABULATN program (tabu.par)

This file is used to specify all the parameters needed by the program to produce the life-tables. The components of the file are:

A section of the tabu.par file included in package is shown:

SKIN MELANOMA 1953-1982
 &LIFE DGN=5,4,0, ISTOP=15, VNUMB=9, SURV=9, CLOSE=82,12,31,
  STAT=8, JSTA=0, LSTA=2, AGE=3,0,99,
  SEX=2,0,1, YEAR=51,85,5, MEAN=3, LIS=20, LIST=20, TEST=0 &END
(I3,Il,I3,3I2,Il,Il,I2)
ALL MALES
 &TABLE NV=2, IVl=0,0 &END
ALL FEMALES
 &TABLE IVl=l,l &END
MALES, AGES 0-39, NON-LOCALIZED
 &TABLE NV=2,3,7, IVl=0,0, IV2=0,39, IV3=2,9 &END
MALES, AGES 40-54, NON-LOCALIZED
 &TABLE IV2=40,54 &END etc.

2.4 LIFE namelist (information on the inputfiles)

The parameter values are given in the form (starting from column 2):

&LIFE paraml = F.FF,param2 = NN, K*MM, param3 = nn,
char = 'AAAA', 'BBBB' &END

where F.FF is a real number, NN, MM, K are integers and AAAA and BBBB are character strings. The parameter list can be continued to the next line after a comma. The & character starting a namelist must be in the second column. The first column of each continuation line must also be blank.

The list of parameters, their format, default values and contents are given in the following table.

PARAMETER FORMAT DEFAULT CONTENT
                   VALUE
NFMT           I       1 the number of lines needed to give the input
                         format for the patient data file;
                         maximum allowed = 3.
LIST           I      10 the number of patient records to be listed from
                         the patient file for verification.
VNUMB          I       0 the number of input fields per patient to be
                         read (maximum allowed = 95).
DGN            I   0,0,0 the field numbers giving the date of diagnosis
                         or entry into the study.
                         1st component: year;
                         2nd component: month; and
                         3rd component: day.
BOF            I   0,0,0 the common starting date: (for follow-up studies
                         with common beginning of follow-up date only).
                         Format is YYYY or YY, MM, DD.
EOF            I   0,0,0 the field numbers giving the patient closing
                         dates (e.g. death date): field for year, month, day.
CLOSE          I   0,0,0 the common closing date: (if there is one,)
                         give year, month and day (YYYY or YY, MM, DD).
NOEST          I       0 0 = substitution of erroneous or missing month
                         or day codes by 7 and 15.22, respectively
                         (every occasion is listed);
                         1 = case is rejected and an error notice is given.
RANGE          F    50.0 if the beginning of follow-up is not within
                         the 'RANGE' years from the closing date,
                         an error notice is given.
ISTOP          I      l5 the number of life table intervals (e.g. usually
                         years) to be calculated for life-table output.
TUNIT          I     1,1 1st component: survival calculation unit
                         (1 = year, 2 = month);
                         2nd component: length of life table interval in
                         these units. (N.B. Possible values are only:
                         1 if unit is years,
                         1, 2, 3, 4 or 6 if unit is months.
                         Other values are rejected).
SURV           I       0 the field number giving the survival time
                         ( 0 = there is no such variable).
                         N.B. This must be given in months or years
                         as in the first component of TUNIT.
STAT           I       0 the field number giving the closing status (alive,
                         dead or other).
JSTA           I 10*99999 codes in the STAT field for those alive
                         at the closing date. Up to 10 values allowed.
LSTA           I 10*999999 the codes in STAT field for those lost
                         to follow-up before the closing date (CLOSE).
                         Up to 10 values allowed.
INEX           I      11 default value of 11 if the population mortality
                         file is to be used, else 0 means population
                         mortality file not needed as only the observed
                         survival rates are desired.
NOTQ           I       0 0 = expected annual probabilities of death are
                         in the population mortality file;
                         1 = expected annual probabilities of
                         survival are given instead of those of death.
AGE            I  0,0,99 1st component: the number of the field containing
                         the age at the beginning of follow-up
                         (100 = there is no such field and the ages are to
                         be calculated by the program);
                         2nd component: smallest value used for the age in
                         years for the population mortality file; 
                         3rd component: largest value used for the age in
                         years for the population mortality file.
                         N.B. The range of ages at the beginning
                         of the follow-up should be within the age range of
                         the population mortality file, otherwise cases
                         will be deleted.
BIRTH          I   0,0,0 the field numbers giving the date of birth:
                         year, month, day (YY or YYYY, MM, DD );
                         needed if the AGE parameter = 100.
SEX            I   0,0,1 the field number containing the sex of patient
                         ( 101 = all patients are males,
                         102 = all patients are females);
                         2nd component: the code for male;
                         3rd component: the code for female.
YEAR           I   0,0,5 1st component: the first calender year for
                         which the expected annual probability of death
                         and expectation of life are given in the
                         population mortality file;
                         2nd component: the last calender year for which
                         the probability and expectation of life are given
                         in the population mortality file;
                         3rd component: the number of calender years
                         covered by each set of probabilities of death and
                         life expectancy (e.g. for 1951-55,1956-60,
                         .. 1976-80 the components would take the
                         values 51, 80, 5 )
OTHER          I   0,0,0 population mortality data for more than
                         one general population may be used.
                         The first component defines the field
                         containing the population the patient is from,
                         and the 2nd and 3rd components give minimum
                         and maximum values used for defining the population.
LIS            I      10 number of patient records for which the expected
                         annual and cumulative survival probabilities
                         and expected length of life are listed.
TNO            I       1 starting number (appears in the title) of the
                         first life table to be produced.
TNOD           I       1 increment in the life-table number from one
                         life-table to the next. 
MEAN           I       0 the field number whose mean and standard error
                         of the mean are to be calculated for each table
                         (usually age at diagnosis).
                         0 = calculation not needed.
XAGE           I       1 calculation of the mean length of life (0 = not
                         needed, 1 = needed). Can be calculated only if
                         the life table interval length is one year.
R              F       0 fraction (r) representing constant lower than
                         expected asymptotic survival due to persistent
                         excess risk of death from cancer
                         (see Hakama and Hakulinen (1977).
LERR           I    1000 maximum number of error notices allowed before
                         aborting program.
MERR           I    1000 maximum number of patient rejections allowed
                         before aborting program.
LISTH          I       0 0 = the parameters of the present run and life
                         tables are printed;
                         1 = the parameters in a previously run life-table
                         file (trans.dat) are listed and no tables produced.
TEST           I       0 extra diagnostic printing desired (0 = No, 1
                         = Yes. Used in testing the program).

2.5 TABLE namelist (information on the life-tables to be produced)

The parameter values are given in the form (starting from column 2):

&TABLE paraml = F.FF, param2 = NN, K*MM, param3 =NN &END

where F.FF is a real number and NN, MM, K are integers. The parameter list can be continued to the next line after a comma. The first column of each continuation line must be blank. An example is:

 &TABLE NV = 2,6,8,1, MV = 1,2,1,1, IVl = 2,2 IV2 =1,2,8,9,
  IV3 = 5,7, IV4 = 1,4 &END

The list of parameters, their format, default values and contents are given in the following table.

N.B. Values of previous TABLE namelists are used as default values for the next TABLE namelist.

PARAMETER FORMAT DEFAULT CONTENT
                   VALUE
NV             I     5*0 the number of the fields defining the criteria
                         of patient selection for the life table (at most 5
                         variables are allowed; if none are specified
                         then all patients are selected).
                         The components are studied by the program in order
                         from 1 to 5. If, for example, NV(3) = 0, then
                         components NV(4) and NV(5) do not have any effect
                         on the output.
MV             I     5*1 for each field specified by NV above, the number
                         of continuous intervals of values from which
                         selection can be made (maximum number of intervals
                         per field is 4).
IV1 - IV5      I     8*0 the upper and lower bounds of the selection
                         intervals for a field in the order: lower, upper,
                         lower, upper, etc. (each of the possible 5
                         arrays IV1 - IV5 has 8 elements i.e. 4 pairs
                         allowed at most per field).
R              F value in LIFE namelist
                         the asymptotic r used
                         in excess risk of death which enables
                         life-table specific values for R.
                         (see R in LIFE namelist).
TNO            I         value specified by TNO and TNOD in LIFE namelist.
                         number of life table


3. Using PRINTNG program

This program creates a file containing the life-tables and plots of the survival functions. It requires two data files that have been created by TABULATN and up to 3 parameter files as listed below:

The output files are:

3.1 Parameter file for PRINTNG program (printl.par)

This file is used to specify all the parameters needed by the program to produce the life-tables. The components of the file are:

The printl.par file included in the package is shown:

 &PRINT COLS=1,2,17,18,19,3,4,6,7, PLOT=3,6,9, TOTEST=l,
  GLIM=3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18 &END

3.2 PRINT namelist (columns for printing and plotting)

This namelist has the same format as those in sections 2.4 and 2.5. The list of parameters, their format, default values and contents are given in the following table.

PARAMETER FORMAT DEFAULT CONTENT
                   VALUE
TABLES         I    9999 the numbers of the life-tables to be printed.
                         (9999 = all tables to be printed. Example:
                         TABLES = 1, - 5, 8, 9, -12 where '9, - 12'
                         denotes 9, 10, 11, 12).
MAXCOL         I      10 number of optional life-tables columns to
                         be printed (maximum = 10, see section 3.3)
COLS           I 1,...,10 numbers of columns to be printed.
CFRMS          A         this contains the row format of the columns, from
                         column 1 to 26.
NFRMS          A         headline format of the columns, from column 1
                         to 26, as with CFRMS but with A formats.
PLOT           I     6*0 number of columns to be plotted (at most 6
                         columns can be plotted; plotting is possible
                         if ISTOP < 30).
CHARS          A ' 0 ', ' * ', ' X ', ' + ', ' - ', ' % '
                         symbols to be used in the plotting (in order of use).
TIT1           I       0 1 = new general title used,
                         0 = no change.
TIT2           I       0 1 = new life-table title,
                         0 = no change.
TOTEST         I       0 1 = tables need to be tested by the GROUPTST
                         program; the necessary parameters are supplied
                         by the NTEST namelist.
                         0 = tests not needed.
GLIM           I    80*0 the numbers of the life-tables to be involved
                         in the analysis with GLIM, first component = 0
                         if no GLIM analysis is needed.
DATE           I/A       processing date needed in printout in the form
                         DD/MM/YY.

3.3 Life-table columns for printing and plotting

The first 5 'compulsory' columns are always printed. The others are optional and only those specified by COLS are printed (see the default values of COLS in section 3.2). Similarly, the quantities plotted are only those specified by PLOT. The methods of calculation are given in Hakulinen (1982). For an example of the output see Appendix B.

NUMBER SYMBOL PRINTED     CONTENT
none   I      always      life-table time interval (xi-xx+1).
none   L      always      number of patients alive at beginning of
                          time interval li.
none   D      always      number of deaths among patients during time
                          interval di .
none   W      always      number of withdrawals alive among patients
                          during time interval wi
none   L'     always      L - W/2 = (li - wi / 2) 'effective' number
                          of patients in time interval.
1      P      recommended interval-specific observed survival rate
                          ( pi ) of patients.
2      2*SEP  optional    twice the standard error of P.
3      CP     recommended cumulative observed survival rate
                          ( ip0 ) of patients.
4      2*SECP optional    twice standard error of CP.
5      CP*    recommended cumulative expected survival rate (ip0*)
                          (Hakulinen 1982), with Chiang's (1968)
                          approximation.
6      CR     recommended cumulative relative survival rate 
                          corresponding to CP
7      2*SECR recommended twice standard error of CR.
8      SP*    optional    cumulative expected survival rate (Hakulinen
                          1982), with simplified approximation: ip0*,
                          in which pi* = 1-di*/(li* - wi*/2 ).
9      SR     optional    cumulative relative survival rate
                          corresponding to SP*.
10     2*SESR optional    twice standard error of SR.
11     E2P*   optional    cumulative expected survival rate (ip0*)
                          with Ederer II method.
12     E2R    optional    cumulative relative survival rate
                          corresponding to E2P*.
13     2*SE2R optional    twice standard error of E2R.
14     ElP*   optional    cumulative expected survival rate (ip0*)
                          with Ederer I method.
15     ElR    optional    cumulative relative survival rate
                          corresponding to ElP*.
16     2*SElR optional    twice standard error of ElR.
17     P*     recommended interval-specific expected survival
                          rate,( Pi* ).
18     R      recommended interval-specific relative survival rate
                          corresponding to P*.
19     2*SER  recommended twice standard error of R.
20     L*     optional    expected number of persons in the comparable
                          general population, constructed from the given one,
                          alive at the beginning of interval ( li* ).
21     D*     optional    expected number of deaths for the above
                          general population during interval ( di* ).
22     W*     optional    expected number of withdrawals alive for
                          that general population during interval ( wi* ).
23     DEL*   optional    expected 'delta' deaths as given in
                          Hakulinen (1982).
24     NEXTL* not needed  li + 1* (see L* above).
25     INCP*  optional    expected annual probability
                          corresponding to CP*.
26     INSP*  optional    expected annual probability
                          corresponding to SP*.

3.4 Text below the life-table

'Effective sample size at termination', is the number of patients whose complete follow-up would produce a standard error equal to that observed for the last value of CP. The mean and the standard error of the mean is given if a variable is specified by MEAN in the LIFE namelist. The expectation of life for a general population, matched by age and sex with the characteristics of the patients, the expectation of life for the patients and the proportion of expected life lost, are printed according to the instructions given by XAGE and R in LIFE namelist (R is also in TABLE namelist).


4. GROUPTST program for statistical tests

This program performs the maximum likelihood tests described in Hakulinen et. al. (1987), a logrank extension (Hakulinen et. al. 1987) and, for tests on two groups only, the Charles Brown (1984) test. GROUPTST can be run in conjunction with the TABULATN and PRINTNG program (section 4.1) or by itself (section 4.2). For an example of the output see Appendix B.

Let k be the index referring to the groups being compared and i the index referring to the follow-up interval. For these tests, the tables are condensed by pooling the follow-up interval with the previous one if any of the lki or dki are equal to zero. If any of the dki are zero in the first interval, the first interval is pooled with the second.

If tests with the proportional hypothesis, H1, are desired, the life-tables are all cut-off before the relative survival rate estimates under the equality hypothesis, H0, exceed unity. If the life-tables thus shortened happen to become less than two intervals in length, the cut-off is abandoned and the tests with the proportional hypothesis are not performed.

4.1 Running GROUPTEST in conjunction with the other two programs

In order to inform the PRINTNG program that the GROUPTST program will follow, TOTEST must be set equal to 1 in the PRINT namelist. Two files will be created by the PRINTNG program and these are required as input files:

There is one output file containing the results:

The parameters for GROUPTST program are given in the NTEST namelist. The NTEST namelist will be read by the PRINTNG program and the required data passed onto the GROUPTST program only if TOTEST is set to 1 in the PRINT namelist.

This namelist in the print2.par file has the same format as those in sections 2.4 and 2.5. The list of parameters, their format, default values and contents are given in the following table.

PARAMETER FORMAT DEFAULT CONTENT
                   VALUE
IT1            I    10*0 the numbers of the life-tables forming the
                         first group of tables to be tested statistically.
                         In the printout, the order of the tables is the
                         same as that for the TABLE namelists. One group
                         can have up to 10 tables.
IT2 - IT5      I    10*0 same as for IT1. Up to five groups can
                         be tested. These parameters are for the second
                         to fifth groups.
ITE            I       0 1 = values of the Rs and the Cs are to be printed
                         at each iteration. Any overstepping of permitted
                         bounds and their corrections are also reported.
                         0 = above values are not printed.
OBS            I       0 1 = the tests should be performed on the observed
                         rates as well.
                         0 = tests are not performed on the observed rates.
PROP           I       1 1 = the tests with the proportional hypothesis
                         are to be performed. However, if the length of
                         the pooled and shortened life-tables is less than
                         two intervals, these tests are skipped
                         (see section 4.0).
                         0 = tests using the proportional hypothesis are
                         to be skipped.
MAXIT          I    1000 the maximum number of iterations allowed
                         per test before stopping computations.
                         This is a safeguard against computer
                         time wastage.
MI             I      30 the maximum number of intervals to be taken for
                         the test. In the analysis, the number of
                         intervals actually taken for the
                         test may be smaller.
PREC1          F  0.0001 should be equal to about 1000 times PREC2
                         or 0.001 whichever is less. This is the
                         preliminary precision obtained
                         in the iterations.
PREC2          F 0.0000001 should be equal to the final precision
                         required in the iterations. PREC2 should
                         be less than PREC1.

Most default values need not be changed to run the program. The program will work if only the table numbers required are specified for at least the first one of ITl,...,IT5. If one of these groups is not specified, the program ignores that and subsequent groups.

4.2 Running GROUPTST by itself

The program can be run by itself using the NTEST namelist of section 4.1 and an input data file containing the data

The input parameter file (group.par) should contain the NTEST namelist. An example is given as:

&NTEST ITl=1,2, IT2=3,4,5,6 &END

The data file (group.dat) should contain the data on all required tables. All lines must be in the format:

(3I 10,F10.8,I 10).

1. The first line contains the table number (I10 format).

2. One line for data on each interval. Each line contains: First the number '-2' (I10 format), next the number of deaths, D, (I10 format), then the number, L, (I10 format), then the expected survival rate, P*, (F10.8) format), and finally the number of withdrawals, W, (I10 format). Note: If only observed rates are to be tested, P* = 1.0 for all intervals.

3. The end of table line with: The number, '0'. (I10 format).

An example from the supplied data:

        -2       239      14590.97015077       119
        -2       222      11010.97059095        78
        -2       130       8010.97034484        62
        -2        61       6090.968927'0        58
        -2        51       4900.96874708        51
        -2        29       3880.96888268        34
        -2        33       3250.96553618        32
        -2        17       2600.96896255        23
        -2        11       2200.96804982        21
        -2        10       1880.96844620        15
        -2         5       1630.96691990        15
        -2         8       1430.96530479        13
        -2         2       1220.96865773        10
        -2         3       1100.97027075         9
        -2         4        980.96735400        10
         0
         2
        -2      209       18350.97992235       140
        -2      178       14860.98181677       109
        -2 ........... etc.

The numbers -2 and 0 indicate to the program that it is reading data for an interval and the end of table record, respectively.

N.B. In this file the data for tables should be in strictly increasing order, e.g. tables in the order 2, 5, 6, 12 are allowed, but not those in the order 2, 8, 3, 4 or 3, 4, 6, 6, 8. Deviations from instructions

will cause errors or program failure.


5. Regression analysis with GLIM

The PRINTNG program will write data files in a format that can be read by GLIM for a regression analysis of the hazard function. The independent variables are determined by the variables NV in the TABLE namelist used to create the life-tables. The parameter GLIM must be defined in the PRINT namelist to give the numbers of life-tables needed in the GLIM analyses.

The structure of the GLIM data file is defined using the GLIMPAR namelist in the glim.par file used by the PRINTNG program. The list of parameters, their format, default values and contents are given in the following table.

PARAMETER FORMAT DEFAULT CONTENT
                   VALUE
NV             I    10*0 the numbers of the fields characterizing the
                         independent variables to be included in the GLIM
                         analyses.
IVl-IV10       I 20*(-1) the upper and lower bounds of the class
                         intervals corresponding to each of the variables
                         in NV (IV1 for the first variable, IV2 for the
                         second, etc.)
IC1-IC10       I 1,2,..,10 the codes to be assigned for each
                         class (IC1 for the first variable, etc.).
RC1-RC10       F    10*0 alternative quantitative codes to be assigned
                         for each class (RC1 for the first variable, etc.). 
CWIDTH         I     3*0 the numbers of consecutive follow-up intervals
                         to be included in each output file.

An example of the GLIMPAR namelist is:

 &GLIMPAR NV=2,3,7,
  IV2=0,39,40,54,55,69,70,99,
  IVl=0,0,1,1,
  IV3=0,1,2,9,
  IC2=2,3,1,4,
  IC3=2,1,
  RC2=30,47,62,75,
  CWIDTH=5,5,5 &END

In the above parameter file the patient groups are defined by variables 2 (sex), 3 (age) and 7(stage). The age-groups are 0 - 39, 40 - 54, 55 - 69 and 70 - 99 years. The group 55 - 69 years will be used as a reference category (code = 1) in GLIM (because this group is large). Age group 0 - 39 will receive code 2, 40 - 54 code 3 and 70 - 99 code 4. 'Central' ages 30, 47, 62 and 75 years are given for the age groups for an alternative way of modelling the age variable. Data for the first five follow-up intervals will be included in the first output file (follow l.dat), that for the next five intervals in the second output file (follow2.dat), and that for the 11th to 15th intervals in the third output file (follow3.dat).

The structure of each file will be: records in free format, one corresponding to each follow-up interval in the order:

An example of the first few lines of the followl.dat file for the above parameter file is:

1 1 2 2 0 0 5 139.0 0.99861 0.00 35.00 0.00 0.00 0.00
2 1 2 2 0 0 1 134.0 0.99860 0.00 35.00 0.00 0.00 0.00
3 1 2 2 0 0 7 133.0 0.99860 0.00 35.00 0.00 0.00 0.00
4 1 2 2 0 0 2 126.0 0.998S9 0.00 35.00 0.00 0.00 0.00
5 1 2 2 0 0 0 124.0 0.99856 0.00 35.00 0.00 0.00 0.00
1 1 3 2 0 0 6 104.0 0.99523 0.00 44.00 0.00 0.00 0.00
2 1 3 2 0 0 7  98.0 0.99487 0.00 44.00 0.00 0.00 0.00
3 1 3 2 0 0 5  91.0 0.99437 0.00 44.00 0.00 0.00 0.00

To use GLIM the data file FOLLOWl.DAT, FOLLOW2.DAT or FOLLOW3.DAT is read in and fitted to a model involving linear predictors of age, sex, follow-up time etc. These predictor variables are defined by the glim.par file as shown in the previous section. The GLIM macro INPUT.GLM contains the necessary GLIM commands to:

The user has to specify the number of records to be read in from the data file FOLLOW*.DAT which can be obtained from the end of the file PRINTlNG.LIS. An example of using the Glim macro follows:

$units 80$
$input 13$
?filename input.glm
$echo
$dat fu sex age stage vl v2 nd Id ps v3 agel v4 vS v6$ $fac fu 5 sex 2 age 4 stage 2 $
$c macro pd prepares data for plotting $c
$mac pd $cal obs=-%log(ns/ps/ld):exp=%exp(%lp):f=age+4*(stage-1)$ $fac f 8$endmac
$c
$c macros pme --plot male expected: pmo plot male observed
$c macros pfe --plot female expected: pfo plot female observed $c
$mac pme $cal %re=%lt(sex,2)$plot exp fu 'cabdjhik' f$endmac $mac pmo
$cal %re=%lt(sex,2)$plot obs fu 'cabdjhik' f$endmac $mac pfe $cal %re=%ge(sex,2)$plot
exp fu 'cabdjhik' f$endmac $mac pfo $cal %re=%ge(sex,2)$plot obs fu 'caWjhik'
f$endmac $c
$c input data file name: followl, or ... follow3 $c
$dinp 12 followl.dat
$c
1 1 2 2 0 0 5 139.0 0.99861 0.00 35.00 0.00 0.00 0.00
2 1 2 2 0 0 1 134.0 0.99860 0.00 35.00 0.00 0.00 0.00
3 1 2 2 0 0 7 133.0 0.99860 0.00 35.00 0.00 0.00 0.00
4 1 2 2 0 0 2 126.0 0.99859 0.00 35.00 0.00 0.00 0.00
etc.
$warning$
$mac ml $cal exlp = -%exp(%lp)
$cal %fv = ~oexp(exlp)*ld*ps $endmac
$mac m2 $cal %dr = 1/(%fv*exlp) $endmac
$mac m3 $cal %va = %fv*(1-%fv/ld) $endmac
$mac m4
$cal %di = 2*(%yv*%1Og(%yv/%fv) + (Id - %yv)*%log((1 - %yv/ld)/(1 - %fv/ld)))
$endmac$
$own ml m2 m3 m4
$cal %lp = 0.8$
$cal ns = ld - nd $yvar ns$
$fit$
$echo$
$return$
$echo$
$ $c you must rewind charmel 13 to read in this macro$
$finish$

In the above example, the $unit 80$ command is used to specify the number of records in the file followl.dat. Then the $input 13$ command is used to read in ~he macro, which is read in after specifying its name viz. input.glm. The macro is now read and then prompts for the input data file. You enter the filename: followl.dat

The data is read and displayed on the screen. The first model deviance is displayed. Additional models can be fitted in the normal way. e.g.

$fit + fu$d e$

To plot the observed or expected hazard functions for males or females use the macros:

$use pd$ ! calculates the hazard functions and the factors

$use pme$ ! plots male expected hazard for sample data

$use pfo$ ! plots female observed hazard for sample data

NOTE: The model used in this GLIM macro assumes that the patients have an excess or increased mortality (hazard) relative to the general population. If, however, the observed excess hazard is negative then this can be either due to:

(a) statistical fluctuations for one or two groups, but the underlying trend reveals a positive excess hazard;

(b) there is no excess hazard for the patients, which may occur after 5 years of follow-up at which stage the patients are 'cured'.

In the case of (a), a parsimonious model should be chosen that reflects the underlying trends and not the statistical fluctuations. If a full model is chosen (i.e. with all independent variables and interactions included) or a model with too many parameters, then the model will fail to converge as it cannot model the negative excess hazard. An error message will be given by GLIM caused by zero derivatives or lack of convergence.

In the case of (b), the model is not appropriate for all follow-up intervals and you should only include those for which there is a positive excess hazard or select a multiplicative model.

N.B. The GLIM software is available from NAG Ltd and authorized distributors.


6. Acknowledgements

The authors would like to thank the staff of the Finnish Cancer Registry, the Australian Institute of Health, Canberra and the Cancer Registries in Australia for their most valuable help and support, and users all over the world for their continuing interest.


7. References

Brown, C.C (1984): The statistical comparison of relative survival rates. Biometrics 39: 941-948.

Chiang, C.L. (1968): Introduction to Stochastic Processes in Biostatistics. Wiley, New York.

Ederer, F., Axtell, L.M. and Cutler, S.J. (1961): The relative survival rate: a statistical methodology. Natl. Cancer Inst. Monogr. 6:101-121.

Hakama, M. and Hakulinen, T. (1977): Estimating the expectation of life in cancer survival studies with incomplete follow-up information. J. Chron. Dis. 30: 585-597.

Hakulinen, T. (1977): On long-term relative survival rates, J. Chron. Dis. 30: 431443.

Hakulinen, T. (1982): Cancer survival corrected for heterogeneity in patient withdrawal. Biometrics 38:933-942.

Hakulinen, T. and Abeywickrama, K.H. (1985): A computer program package for relative survival analysis. Comp. Progr. Biomed. 19: 197-207.

Hakulinen, T. and Tenkanen, L. (1987): Regression analysis of relative survival rates. Appl. Stat. 36: 309-317

Hakulinen, T., Tenkanen, L., Abeywickrama, K. and Päivärinta, L. (1987): Testing equality of relative survival patterns based on aggregated data. Biometrics 43: 313-325.


Appendix A. Flowchart describing the system

To be included later


Appendix B. Example

TABU.PAR

 &LIFE DGN=5,4,0, ISTOP=15, VNUMB=9, SURV=9, CLOSE=82,12,31,
  STAT=8, JSTA=0, LSTA=2, AGE=3,0,99,
  SEX=2,0,1, YEAR=51,85,5, MEAN=3, LIS=20, LIST=20, TEST=0 /
(I3,I1,I3,3I2,I1,I1,I2)
ALL MALES
 &TABLE NV=2, IV1=0,0 /
ALL FEMALES



TABULATN.LIS

1 ------------------------------------------------------------------------------------------------------------------------------
  LIFE TABLE   999999       31.5.1997 
  ------------------------------------------------------------------------------------------------------------------------------


 SKIN MELANOMA 1953-1982                                                         
  &LIFE DGN=5,4,0, ISTOP=15, VNUMB=9, SURV=9, CLOSE=82,12,31,                    
   STAT=8, JSTA=0, LSTA=2, AGE=3,0,99,                                           
   SEX=2,0,1, YEAR=51,85,5, MEAN=3, LIS=20, LIST=20, TEST=0 /                    
 (I3,I1,I3,3I2,I1,I1,I2)                                                         
 ALL MALES                                                                       
  &TABLE NV=2, IV1=0,0 /                                                         
 ALL FEMALES                                                                     
  &TABLE IV1=1,1 /                                                               
1 ------------------------------------------------------------------------------------------------------------------------------
  LIFE TABLE   999999       31.5.1997        SKIN MELANOMA 1953-1982                         
  ------------------------------------------------------------------------------------------------------------------------------


  LIFE TABLE SETUP:

  RUN IDENTIFICATION ..............ANYRUN

  MAX NUMBER OF TABLE INTERVALS ... 15
  LENGTH OF TABLE INTERVAL ........  1  YEAR 
  NUMBER OF INPUT VARIABLES .......  9
  NUMBER OF FORMAT CARDS ..........  1
  WITHDRAWAL STATUS VARIABLE NO:...  8  = STAT
  CODES FOR THOSE ALIVE ...........  0 
  CODES FOR LOST FROM FOLLOW-UP....  2 
  SURVIVAL TIME VARIABLE ..........  9  = SURV
  CLOSING DATE ....................   82-12-31

  DATE OF DIAGNOSIS VARIABLES:
          YEAR   VARIABLE   5 =DGN(1)
          MONTH  VARIABLE   4 =DGN(2)

  END-OF-FOLLOW-UP VARIABLES (BLANK=NOT GIVEN):

  THE MEAN AND THE S.E. OF THE MEAN, OF THE VARIABLE NUMBERED    3  IS COMPUTED

  INPUT FORMAT:

 (I3,I1,I3,3I2,I1,I1,I2)                                 

  INDEX VARIABLES FOR EXPECTED PROBABILITIES AND EXPECTATIONS OF LIFE:

        1. INDEX:  3 RANGE   0-  99             = AGE (AT DIAGNOSIS)
        2. INDEX:  2 RANGE   0-   1             = SEX (MALE, FEMALE)
        3. INDEX:  5 RANGE  51-  85  BY 5 YEARS =YEAR (CALENDAR, OF DATA GIVEN)

     1400 EXPECTED DEATH PROBABILITIESS ARE READ FROM FILE 11

     1400 EXPECTED LIFE TIMES ARE READ FROM FILE 11
1 ------------------------------------------------------------------------------------------------------------------------------
  LIFE TABLE   999999       31.5.1997        SKIN MELANOMA 1953-1982                         
  ------------------------------------------------------------------------------------------------------------------------------



 ......... DATA LIST OF    20 CASES ..........

 1:ST ROW(S). TABLE AND POTENTIAL INTERVALS,STATUS AND DATA:
 2:ND ROW(S). ANNUAL AND CUMULATIVE EXPECTED PROBABILITIES:
  ------------------------------------------------------------------------------------------------------------------------------


      8    30     1   903     0    41     7    53    41     1     1     7
  .9947 .9945 .9936 .9935 .9934 .9926 .9918 .9910 .9900 .9887 .9871 .9868 .9853 .9836 .9821 .9947 .9893 .9830 .9766 .9702
  .9630 .9552 .9465 .9371 .9264 .9145 .9024 .8892 .8746 .8589
  expected life time:  28.31
      5    30     1   903     0    36     1    53    41     1     1     4
  .9962 .9960 .9960 .9957 .9958 .9954 .9947 .9942 .9932 .9924 .9925 .9918 .9906 .9891 .9890 .9962 .9921 .9882 .9839 .9798
  .9753 .9701 .9645 .9580 .9507 .9436 .9358 .9270 .9169 .9068
  expected life time:  32.66
      1    30     1   906     0    27    12    53    41     3     1     0
  .9975 .9973 .9973 .9974 .9974 .9972 .9972 .9969 .9969 .9968 .9962 .9963 .9956 .9956 .9952 .9975 .9948 .9922 .9896 .9871
  .9843 .9815 .9785 .9756 .9724 .9688 .9653 .9610 .9568 .9523
  expected life time:  40.70
      2    30     1   903     1    46     8    53    41     1     1     1
  .9959 .9959 .9956 .9955 .9951 .9949 .9944 .9936 .9935 .9929 .9926 .9914 .9908 .9904 .9893 .9959 .9918 .9874 .9829 .9782
  .9732 .9677 .9615 .9553 .9485 .9415 .9334 .9248 .9160 .9062
  expected life time:  28.73
      4    30     1   907     1    69    10    53    41     2     1     3
  .9624 .9581 .9518 .9501 .9452 .9405 .9304 .9236 .9145 .9064 .8942 .8836 .8705 .8693 .8566 .9624 .9220 .8776 .8339 .7882
  .7413 .6898 .6370 .5826 .5281 .4722 .4172 .3632 .3157 .2705
  expected life time:  10.60
      2    30     1   905     1    31     1    53    41     1     1     1
  .9984 .9983 .9982 .9985 .9984 .9983 .9981 .9981 .9983 .9982 .9978 .9977 .9975 .9974 .9972 .9984 .9967 .9949 .9934 .9918
  .9900 .9882 .9863 .9846 .9828 .9807 .9784 .9759 .9733 .9706
  expected life time:  42.50
      4    30     1   905     1    29     9    53    41     2     1     3
  .9984 .9984 .9984 .9988 .9986 .9985 .9984 .9983 .9985 .9985 .9983 .9982 .9978 .9980 .9978 .9984 .9968 .9952 .9940 .9926
  .9911 .9895 .9878 .9863 .9848 .9831 .9813 .9792 .9772 .9751
  expected life time:  44.36
      2    30     1   905     1    36    10    53    41     1     1     1
  .9979 .9977 .9977 .9979 .9978 .9977 .9974 .9972 .9971 .9972 .9969 .9965 .9963 .9960 .9958 .9979 .9957 .9934 .9912 .9890
  .9868 .9842 .9815 .9786 .9759 .9728 .9693 .9658 .9619 .9578
  expected life time:  37.86
      2    30     1   904     1    39     2    53    41     1     1     1
  .9977 .9977 .9971 .9974 .9972 .9970 .9967 .9966 .9965 .9963 .9957 .9952 .9948 .9948 .9942 .9977 .9954 .9925 .9899 .9871
  .9841 .9809 .9776 .9742 .9705 .9663 .9617 .9567 .9518 .9463
  expected life time:  35.10
      8    30     1   907     1    44     6    53    41     1     1     7
  .9968 .9965 .9959 .9962 .9956 .9955 .9951 .9949 .9946 .9945 .9935 .9929 .9926 .9922 .9912 .9968 .9934 .9893 .9855 .9812
  .9768 .9720 .9671 .9619 .9566 .9504 .9436 .9367 .9294 .9212
  expected life time:  30.54
      9    30     1   901     0    70     8    53    41     1     1     8
  .9376 .9331 .9278 .9237 .9167 .9110 .9034 .8958 .8859 .8772 .8644 .8524 .8376 .8229 .8154 .9376 .8749 .8117 .7498 .6874
  .6262 .5657 .5067 .4489 .3938 .3404 .2902 .2431 .2000 .1631
  expected life time:   8.69
      2    30     1   905     0    36     4    53    41     1     1     1
  .9962 .9960 .9960 .9957 .9958 .9954 .9947 .9942 .9932 .9924 .9925 .9918 .9906 .9891 .9890 .9962 .9921 .9882 .9839 .9798
  .9753 .9701 .9645 .9580 .9507 .9436 .9358 .9270 .9169 .9068
  expected life time:  32.66
      2    30     1   905     0    53     6    53    41     1     1     1
  .9846 .9833 .9807 .9807 .9797 .9776 .9751 .9720 .9709 .9671 .9642 .9629 .9597 .9561 .9526 .9846 .9682 .9495 .9312 .9123
  .8918 .8697 .8453 .8207 .7937 .7653 .7368 .7072 .6761 .6441
  expected life time:  18.72
      1    30     1   905     1    48     3    53    51     3     1     0
  .9956 .9952 .9946 .9949 .9944 .9936 .9930 .9924 .9926 .9914 .9908 .9893 .9883 .9874 .9866 .9956 .9908 .9855 .9805 .9750
  .9687 .9620 .9546 .9476 .9394 .9308 .9208 .9101 .8986 .8866
  expected life time:  26.96
      1    30     1   907     0    66     9    53    41     2     1     0
  .9528 .9509 .9473 .9442 .9388 .9339 .9286 .9237 .9178 .9092 .9004 .8939 .8859 .8702 .8619 .9528 .9060 .8583 .8104 .7608
  .7105 .6598 .6095 .5594 .5086 .4579 .4094 .3627 .3156 .2720
  expected life time:  10.64
      1    30     1   905     1    18     6    53    41     2     1     0
  .9991 .9991 .9990 .9994 .9992 .9992 .9992 .9991 .9993 .9992 .9992 .9992 .9991 .9991 .9991 .9991 .9982 .9971 .9965 .9958
  .9950 .9942 .9934 .9927 .9919 .9912 .9904 .9896 .9887 .9878
  expected life time:  54.72
     10    30     1   907     0    49     5    53    41     1     1     9
  .9891 .9886 .9872 .9868 .9849 .9842 .9833 .9807 .9787 .9768 .9737 .9727 .9709 .9678 .9639 .9891 .9778 .9653 .9525 .9381
  .9233 .9078 .8903 .8714 .8512 .8289 .8062 .7827 .7575 .7302
  expected life time:  21.73
      4    30     1   905     1    42     8    53    41     1     1     3
  .9972 .9970 .9968 .9967 .9966 .9962 .9956 .9955 .9952 .9948 .9946 .9945 .9935 .9933 .9933 .9972 .9943 .9911 .9879 .9846
  .9808 .9765 .9721 .9675 .9625 .9573 .9520 .9458 .9395 .9332
  expected life time:  32.36
      2    30     1   907     0    75     3    53    41     1     1     1
  .9083 .8949 .8931 .8867 .8757 .8641 .8562 .8420 .8262 .8200 .8006 .7839 .7669 .7574 .7307 .9083 .8128 .7259 .6437 .5636
  .4870 .4170 .3511 .2901 .2379 .1905 .1493 .1145 .0867 .0634
  expected life time:   6.55
      2    30     1   903     0    83     5    53    11     2     1     1
  .8277 .8062 .7890 .7897 .7773 .7495 .7652 .7028 .7025 .6923 .6584 .6620 .6770 .7143 .7439 .8277 .6674 .5265 .4158 .3232
  .2423 .1854 .1303 .0915 .0634 .0417 .0276 .0187 .0134 .0099
  expected life time:   4.03
 *** ERROR, CASE    2979 REJECTED FOR INVALID INDEX. DATA:
   903     1   102     4    82    41     0     0     0
 *** ERROR, CASE    2980 REJECTED FOR INVALID INDEX. DATA:
   903     0   100    11    80    41     0     0     2
 *** ERROR, CASE    3296 REJECTED FOR FORMAT
1 .....  INPUT STATISTICS  ........

         CASES READ ...............    3296
         REJECTED FOR INVALID INDEX       2 (EXPECTED RATES OR LIFE TIMES)
         REJECTED FOR INVALID DATES       0 (YEAR OUTSIDE THE RANGE)
         REJECTED FOR FORMAT ......       1
         REJ. FOR TRANSFORMATIONS..       0

         LEFT IN TABLES ...........    3293
         WITH ESTIMATED DGN DATE ..       0 (MONTH OR DAY)
         WITH ESTIMATED EOF DATE ..       0 (MONTH OR DAY)
         WITH ESTIMATED BIRTH DATE        0 (MONTH OR DAY)

         TABLES PROCESSED .........      18
                INTO FILE NR ......      21



PRINT1.PAR

 &PRINT COLS=1,2,17,18,19,3,4,6,7, PLOT=3,6,9, TOTEST=1, 
  GLIM=3,-18  /


PRINTNG.LIS



 Printing program:                 Version 1.6
 Finnish Cancer Registry          October 1996
 The date is: 31.5.1997 

  &PRINT COLS=1,2,17,18,19,3,4,6,7, PLOT=3,6,9, TOTEST=1,                        
   GLIM=3,-18  /                                                                 
  &GLIMPAR NV=2,3,7,                                                             
   IV2=0,39,40,54,55,69,70,99,                                                   
   IV1=0,0,1,1,                                                                  
   IV3=0,1,2,9,                                                                  
   IC2=2,3,1,4,                                                                  
   IC3=2,1,                                                                      
   RC2=30,47,62,75,                                                              
   CWIDTH=5,5 /                                                                  
  &NTEST  IT1=1,2, IT2=3,4,5,6 /                                                 
1------------------------------------------------------------------------------------------------------------------------------
 LIFE TABLE      0       31.5.1997 

 THE RUN SEARCHED IS: ALLRUN

 ALL TABLES ARE PROCESSED

 WITH THE   9 COLUMNS:      P  2*SEP     P*      R  2*SER     CP 2*SECP     CR 2*SECR

 THE PLOTTED COLUMNS ARE:     CP     CR     SR

  COLUMN NAME FORMAT:
 (7X,'I       L     D     W        L'''    ,A9   ,A9   ,A9   ,A9   ,A9   ,A9   ,A9   ,A9   ,A9   )     

  COLUMN ROW  FORMAT:
 (1H , I3,1H-,I3, I8,    2I6,    F10.1     ,F9.5 ,F9.5 ,F9.5 ,F9.5 ,F9.5 ,F9.5 ,F9.5 ,F9.5 ,F9.5 )     

 RUN '      ANYRUN' FOUND. IT WAS BORN IN 31.5.1997
1------------------------------------------------------------------------------------------------------------------------------
 LIFE TABLE      1       31.5.1997        SKIN MELANOMA 1953-1982                         
    TABLE TITLE:                       ALL MALES                                       
 ------------------------------------------------------------------------------------------------------------------------------

  TABLE SELECTION:        NV VARIABLE     ALLOWED VALUES

                           2                        0-      0
 ------------------------------------------------------------------------------------------------------------------------------

       I       L     D     W        L'        P    2*SEP       P*        R    2*SER       CP   2*SECP       CR   2*SECR

   0-  1    1459   239   119    1399.5   .82922   .02012   .97015   .85474   .02074   .82922   .02012   .85468   .02074
   1-  2    1101   222    78    1062.0   .79096   .02496   .97059   .81493   .02571   .65588   .02610   .69709   .02774
   2-  3     801   130    62     770.0   .83117   .02700   .97034   .85657   .02782   .54515   .02801   .59807   .03072
   3-  4     609    61    58     580.0   .89483   .02548   .96893   .92352   .02629   .48782   .02865   .55288   .03247
   4-  5     490    51    51     464.5   .89020   .02901   .96875   .91892   .02995   .43426   .02917   .50861   .03416

   5-  6     388    29    34     371.0   .92183   .02787   .96888   .95144   .02877   .40031   .02949   .48481   .03571
   6-  7     325    33    32     309.0   .89320   .03514   .96554   .92509   .03639   .35756   .02986   .44816   .03743
   7-  8     260    17    23     248.5   .93159   .03203   .96896   .96143   .03305   .33310   .03008   .43225   .03904
   8-  9     220    11    21     209.5   .94749   .03082   .96805   .97877   .03184   .31561   .03030   .42422   .04072
   9- 10     188    10    15     180.5   .94460   .03405   .96845   .97538   .03516   .29812   .03057   .41529   .04258

  10- 11     163     5    15     155.5   .96785   .02829   .96692  1.00096   .02926   .28854   .03076   .41679   .04444
  11- 12     143     8    13     136.5   .94139   .04021   .96530   .97523   .04165   .27163   .03120   .40701   .04675
  12- 13     122     2    10     117.0   .98291   .02397   .96866  1.01471   .02474   .26698   .03135   .41531   .04877
  13- 14     110     3     9     105.5   .97156   .03236   .97027  1.00133   .03336   .25939   .03166   .41911   .05115
  14- 15      98     4    10      93.0   .95699   .04208   .96735   .98929   .04350   .24824   .03220   .41676   .05407


  EFFECTIVE SAMPLE SIZE AT TERMINATION...   720


  MEAN OF VARIABLE  3             ...53.23   (2*STANDARD ERROR OF MEAN=  .852)
  EXPECTATION OF LIFE FOR NORMAL POPULATION
 MATCHED WITH
  THE CHARACTERISTICS OF THE PATIENTS  ... 21.12

  EXPECTATION OF LIFE FOR PATIENTS  ...11.81   (2*SE= 1.009)

  PROPORTION OF EXPECTED LIFE LOST  ...    44.09  % 
1------------------------------------------------------------------------------------------------------------------------------
 LIFE TABLE      1       31.5.1997        SKIN MELANOMA 1953-1982                         
    TABLE TITLE:                       ALL MALES                               
 ------------------------------------------------------------------------------------------------------------------------------


                     SURVIVAL CURVES  *  SYMBOLS USED:       CP<0>       CR<*>       SR

                              PERIOD  %  0===10===20===30===40===50===60===70===80===90==100
                                   1     I                                        0 S      I
                                   2     I                                0 S              I
                                   3     I                          0  S                   I
                                   4     I                       0   S                     I
                                   5     I                     0  S                        I
                                   6     I                   0   S                         I
                                   7     I                 0   S                           I
                                   8     I                0    S                           I
                                   9     I               0    S                            I
                                  10     I              0     S                            I
                                  11     I             0      S                            I
                                  12     I             0     S                             I
                                  13     I            0       S                            I
                                  14     I            0       S                            I
                                  15     I           0        S                            I
                              PERIOD  %  0===10===20===30===40===50===60===70===80===90==100
1------------------------------------------------------------------------------------------------------------------------------
 LIFE TABLE      2       31.5.1997        SKIN MELANOMA 1953-1982                         
    TABLE TITLE:                       ALL FEMALES                                     
 ------------------------------------------------------------------------------------------------------------------------------

  TABLE SELECTION:        NV VARIABLE     ALLOWED VALUES

                           2                        1-      1
 ------------------------------------------------------------------------------------------------------------------------------

       I       L     D     W        L'        P    2*SEP       P*        R    2*SER       CP   2*SECP       CR   2*SECR

   0-  1    1835   209   140    1765.0   .88159   .01538   .97992   .89965   .01570   .88159   .01538   .89957   .01569
   1-  2    1486   178   109    1431.5   .87565   .01744   .98182   .89187   .01777   .77197   .02044   .80416   .02129
   2-  3    1199   111    91    1153.5   .90377   .01737   .98175   .92057   .01769   .69768   .02283   .74229   .02429
   3-  4     997    67    68     963.0   .93043   .01640   .98185   .94763   .01670   .64914   .02412   .70558   .02622
   4-  5     862    50    77     823.5   .93928   .01664   .98211   .95640   .01695   .60973   .02510   .67725   .02788

   5-  6     735    45    58     706.0   .93626   .01839   .98182   .95360   .01873   .57086   .02604   .64826   .02957
   6-  7     632    25    51     606.5   .95878   .01614   .98198   .97637   .01644   .54733   .02661   .63584   .03092
   7-  8     556    16    57     527.5   .96967   .01493   .98100   .98845   .01522   .53073   .02707   .63091   .03218
   8-  9     483    16    46     460.0   .96522   .01709   .98010   .98482   .01743   .51227   .02766   .62326   .03365
   9- 10     421    22    30     406.0   .94581   .02247   .97969   .96542   .02294   .48451   .02858   .60342   .03559

  10- 11     369    11    37     350.5   .96862   .01863   .97944   .98895   .01902   .46931   .02912   .59838   .03712
  11- 12     321     8    35     303.5   .97364   .01839   .98111   .99239   .01875   .45694   .02963   .59665   .03869
  12- 13     278     6    26     265.0   .97736   .01828   .98043   .99687   .01864   .44659   .03014   .59744   .04032
  13- 14     246     8    19     236.5   .96617   .02351   .98125   .98464   .02396   .43148   .03096   .59145   .04243
  14- 15     219     7    23     207.5   .96627   .02507   .98130   .98468   .02555   .41693   .03181   .58574   .04469


  EFFECTIVE SAMPLE SIZE AT TERMINATION...   961


  MEAN OF VARIABLE  3             ...53.83   (2*STANDARD ERROR OF MEAN=  .786)
  EXPECTATION OF LIFE FOR NORMAL POPULATION
 MATCHED WITH
  THE CHARACTERISTICS OF THE PATIENTS  ... 25.53

  EXPECTATION OF LIFE FOR PATIENTS  ...17.86   (2*SE=  .971)

  PROPORTION OF EXPECTED LIFE LOST  ...    30.07  % 
1------------------------------------------------------------------------------------------------------------------------------
 LIFE TABLE      2       31.5.1997        SKIN MELANOMA 1953-1982                         
    TABLE TITLE:                       ALL FEMALES                             
 ------------------------------------------------------------------------------------------------------------------------------


                     SURVIVAL CURVES  *  SYMBOLS USED:       CP<0>       CR<*>       SR

                              PERIOD  %  0===10===20===30===40===50===60===70===80===90==100
                                   1     I                                           0S    I
                                   2     I                                      0S         I
                                   3     I                                  0 S            I
                                   4     I                               0  S              I
                                   5     I                             0   S               I
                                   6     I                            0  S                 I
                                   7     I                          0    S                 I
                                   8     I                          0    S                 I
                                   9     I                         0    S                  I
                                  10     I                       0     S                   I
                                  11     I                      0      S                   I
                                  12     I                      0      S                   I
                                  13     I                     0       S                   I
                                  14     I                     0       S                   I
                                  15     I                    0       S                    I
                              PERIOD  %  0===10===20===30===40===50===60===70===80===90==100


Appendix C. Tecnical Notes

C.0 Hardware

This package was designed for mainframe (Burroughs B5000/B6000/B7000) computers, and subsequently modified to run on IBM AT/PC and compatible micro-computers. It is now maintained on a VAX 4000. Known versions exist on IBM mainframes and Unix systems like SUN.

The output files with the LIS extension (*.LIS) are designed for a page width of 132 characters. To print the files, a 15 inch printer is required, or a 10 inch printer that can be set to condensed mode. The files are Fortran files that contain only ASCII characters except for the formfeed character which allows each table to begin at the top of a page. The files can be read into a word processor and incorporated into a report.

C.1 Installation instructions

C.l.1 Instructions for mainframes

The package is available on 600', 9 track and 1600/6250 bpi magnetic tape. The VAX version is written as a BACKUP saveset and contains the following files:

Non-VAX versions are stored on the tape (no label) as a file containing:

in either ASCII or EBCDIC.

a) Installation of VAX version

Log in under a suitable username, mount the tape as foreign and give a BACKUP command

where 'tapedrive' must be replaced by an actual device name. Usually the programs will run as supplied, but occasionally it will be necessary to recompile and link the program. Use the procedure

SURVIVAL.COM to run the supplied examples.

b) Non-VAX versions

Follow your local operating procedures to read the tape, separate the various files, compile and link the programs, define the files and run the programs. Regardless of the version compare your results with those supplied on the distribution tape.

C.1.2 Instructions for microcomputers

Your computer must be of the dx/dx2 type or else be equipped with a separate numeric co-processor. The programs of this package will not run from the DOS-prompt in Windows. You must stop Windows and return to DOS in order to use the package. Also check that the file CONFIG.SYS contains a line like

FILES = 11 (or more than 11)

The package is normally distributed on a 3.5" 1.44 MB diskette (other sizes/densities are available upon request) that contains the following files.

Create a subdirectory called SURV16 on your hard disk (usually C:)

Copy all files from the diskette to the subdirectory SURV

Run all programs using the sample data and compare the results with those included with the documentation.