A COMPUTER PROGRAM PACKAGE
FOR CANCER SURVIVAL STUDIES
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.
0. Introduction
1. General description
2. Input and output files needed for the TABULATN program
3. Using PRINTNG program
4. GROUPTST program for statistical tests
5. Regression analysis with GLIM
6. Acknowledgements
7. References
Appendices
A. Flowchart decribing the system
B. Example
C. Technical Notes
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.
1. a file with the patient data, containing fields such as time of entry into the study, age, sex and time of death or censoring.
2. a file containing the population mortality data (annual probability of death or survival and life expectancy of population from which patients are drawn) by age, sex and calendar year.
3. files that contain the details of the analysis to be done, including; titles, file formats, life-table outputs and other parameters.
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:
males, prob. of death, 1st interval (e.g. 1951-1955)
females, prob. of death, 1st interval (e.g. 1951-1955)
males, prob. of death, 2nd interval (e.g. 1956-1960)
females, prob. of death, 2nd interval (e.g. 1951-1960)
.
.
.
males, prob. of death, 6th interval (e.g. 1976-1980)
females, prob. of death, 6th interval (e.g. 1976-1980)
males, life expectancy, 1st interval (e.g. 1951-1955)
females, life expectancy, 1st interval (e.g. 1951-1955)
.
.
.
males, life expectancy, 6th interval (e.g. 1976-1980)
females, life expectancy, 6th interval (e.g. 1976-1980)
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:
1. General title for life-table, 72 characters;
2. LIFE namelist (see section 2.4);
3. Format of patient file (cancer.dat) using FORTRAN syntax,
3*56 characters, use only format specifiers of the forms Tn, nX, Iw or
nIw;
4. Sub-title for the first life-table, 72 characters;
5. TABLE namelist for first life-table (see section 2.5);
6. Sub-title for the second life-table, 72 characters;
7. TABLE namelist for second life-table;
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:
PRINT1.PAR Printing parameters,
PRINT2.PAR Grouptst parameters,
GLIM.PAR GLIM parameters,
TRANS.DAT From Tabulatn,
TEMP.DAT From Tabulatn.
The output files are:
PRINTING.LIS Contains the life-table output and plots,
GROUP.PAR Required for grouptst (optional),
GROUP.DAT Required for grouptst (optional),
FOLLOWl.DAT Required for GLIM (optional),
FOLLOW2.DAT Required for GLIM (optional),
FOLLOW3.DAT Required for GLIM (optional).
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:
1. PRINT namelist (see section 3.2);
2. New general heading (if, and only if, TIT1 > 0,
length 72 characters);
3. New life-table headings (if, and only if, TIT2 >
0, length 72 characters);
4. New PRINT namelist;
5. as 2 above;
6. as 3 above; 7. etc.
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:
GROUP.DAT data from the tables to be tested;
GROUP.PAR parameters from the PRINTNG program or otherwise,
NTEST namelist.
There is one output file containing the results:
GROUPTST.LIS listing of 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:
- sequence number of interval in column 1(1 for the first interval to be included in the file, 2 for the second, etc.);
- code for each variable specifying the patient group in the order given in NV (columns 2 to 6);
- D (number of deaths), L' (effective number exposed) and P* (interval-specific expected survival rate) in columns 7, 8 and 9;
- alternative quantitative codes for the interval specified by parameters RC1 - RC10 in the order given in NV (columns 10 - 14).
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:
- read in the file;
- define the variables, the linear model and the plotting macros;
- run the null model for the supplied data.
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:
· program sources (FORTRAN IV) *.FOR
· compiled and linked programs *.EXE
· example data *.DAT
· example results *.LIS
Non-VAX versions are stored on the tape (no label) as a file containing:
· program sources (FORTRAN IV)
· example data
· example results
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
$BACKUP/LOG tapedrive:SURV.BCK/SELECT=*.* *.*
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.
TABULATN.EXE
PRINTNG.EXE
INPUT3.GLM (for users of GLIM 3.77)
INPUT4.GLM (for users of GLIM 4)
GROUPTST.EXE
MEL.DAT a sample of Finnish melanoma data
POPMORT.DAT Finnish population mortality data
TABU.PAR tabulation parameters
PRINTl.PAR printing parameters
PRINT2.PAR grouptest parameters
GLIM.PAR GLIM parameters
SURV.BAT Batch file to run the examples
Sample output files
Create a subdirectory called SURV16 on your hard disk (usually C:)
MD SURV16
Copy all files from the diskette to the subdirectory SURV
COPY A:*.* C:\SURV16\*.*
Run all programs using the sample data and compare the results with those included with the documentation.