A Set of SAS Macros for Producing Customized Reports: Documentation of Summary Statistic Macros

INTRODUCTION.. 2

STEPS IN WRITING REPORT. 2

REPORT MACROS. 3

WREPORT. 3

COLSET. 3

MOVE. 3

NMOVE. 5

PROCEDURE MACROS. 6

BREAKDN.. 7

CHISQP. 9

FREQDIS. 11

GLMP. 13

LOGISTP. 16

MIXEDP. 18

PHREGP. 21

REGP. 24

LIFETAB.. 26

GENMODP. 28

MEDIANP. 30

HTML REPORTS. 32

HTML REPORT EXAMPLE 1. 36

HTML REPORT EXAMPLE 2. 38

HTML REPORT EXAMPLE 3. 40

ADDING SPECIAL CHARACTERS. 42

INTRODUCTION

This document provides documentation on how to use a set of macros that produce customized report tables. These macros are used extensively by statisticians at the Division of Biostatistics at the University of Minnesota to generate statistical reports for DSMB and other purposes. These macros have been presented at several SAS conferences over the years (first presented in 1994: "A Set of SAS Macros for Producing Customized Reports" by Greg Grandits and Ken Svendsen, published in the proceedings of the 19th SAS User's Group International). These papers also provide documentation and give some background into the development. Most recently these macros have been enhanced to generate html tables which can be inserted into Word or posted as web pages.

Briefly, producing customized reports is a high priority in both research and business settings. SAS has several procedures that generate reports, but only the PUT statement within a data step gives complete flexibility to placement of text and data values. The advent of ODS, while helpful, still does not provide the flexibility needed, leaving the data step as the only alternative. However, data step processing and writing PUT statements is cumbersome to program. The described macros, working behind the scenes do all the tedious work. and allow the user to produce customized reports that are easy to program and give complete flexibility to placement of text and data values. Information that can be placed on the report page include text, summary statistics (counts, means, etc) and more complex statistics (regression coefficients, hazard ratios, etc) and significance levels (P-values).

The user first defines columns and column widths across the report page. Text or data values (usually summary statistics) are then moved to these columns and specified lines using macros MOVE or NMOVE. Items can be moved to single columns or to combined columns with the items automatically centered. Lines to where items are moved can be listed or a starting line can be given along with the number of lines to be printed with optional skip patterns. Summary statistics are accessible using one of several available macros that run SAS procedures and output information to datasets. Statistics are placed in array type names (e.g., M1-M20) which can be easily moved to the report page after a SET statement. An example of a call is %nmove (M1-M20,col=2 3,line=12L10,pat=1st/1).

The enhancements to make html table generations allow html formatting such as font type, size, and format, and shading of cells. In addition, multiple statistics can be placed in one column such as mean ± SD giving the final appearance a journal finished appearance.

STEPS IN WRITING REPORT

Report programs are made up of the following statements:

%WREPORT statement that indicates a new report is starting

%COLSET statement that defines columns and column widths across the report page. This macro needs to be called before any move statements and is called once for each report.

%MOVE statements which move text to the report. Features include centering, underlining, and repeating text.

SET statement(s) that read in statistical information from a SAS dataset. These one observation datasets will have been generated from one of the statistical generating macros described later.

%NMOVE statements which place the statistical information to the report page.

Note: Macro WREPORT that allows creation of html tables replaces macro REPORT that produced only text tables. WREPORT can also produce text tables.

REPORT MACROS

WREPORT

%WREPORT is used simply as %WREPORT which indicates the start of a new report.

COLSET

%COLSET is used as follows:

%COLSET (column1 size column2 size … )

Example: %COLSET (25 9 9 2X 9 9)

This statement sets up 5 columns. The first column is 25 positions long and the last 4 columns are 9 positions long. Two spaces are placed between column 3 and column 4 the last 3 columns. This is used to set off text from other text. (Note: the XX only applies to text based tables. Also for HTML tables the size of the columns are only relative to the total size).

MOVE

%MOVE is used as follows:

%MOVE (‘string 1’:’string 2’:…, line=, col=, center=, under=, pat=, mfirst=)

‘string1’:’string2’:’stringN’ character strings enclosed by single or double quotes and separate by colons.

Example: ‘Men’:’Women’:’Total’ (3 text strings)

line = line numbers to which strings are moved. This can be given in 2 ways illustrated by example.

line = 12 21 33 moves strings to lines 12, 21, and 33

line = 12L3 moves strings to lines 12, 13, and 14 (12 is the starting line followed by 3 consecutive lines)

col = defined column numbers to which strings are moved. This can be given in 3 ways illustrated by example.

col = 3 4 8 moves strings to defined columns 3, 4, and 8

col = 2-3 4-5 6-7 moves strings to columns formed by combining columns 2-3, 4-5, and 6-7

col = 2.4 moves strings to columns 2 through 4 (Shortcut for col = 2 3 4)

center = set to “Y” or “1” if centering of text is desired; set to “N” or “2” if left justification is desired. Default is Y.

under = set to “Y” or “1” if string is to underlined; set to “N” or “2” if string is not to be underlined. Default is N.

pat = used to skip lines; must be used with the second method of specifying lines

line = 10L18

pat = 6^TH/1 skips 1 line after every 6^th line

mfirst= determined whether lines or columns move fastest when placing text; set to L for lines and C for columns. Default is L.

Note: HTML parameters are described in the section on HTML reports.

Example: %MOVE (‘Total Cholesterol’:’LDL-C’:’HDL-C’, col=1, line=10L6, pat=3^rd/1);

This statement would produce the following text in lines 10-12 and 13-15. Note that the text strings are repeated until the total lines are exhausted.

Total Cholesterol

LDL-C

HDL-C

Total Cholesterol

LDL-C

HDL-C

NMOVE

%NMOVE is used as follows:

%NMOVE (varlist, line=, col=, pat=, mfirst=, fmt=, scaler =, specfmt=);

The line, col, pat, and mfirst parameters are identical to those in %MOVE.

varlist= list of variables to move to the report page; this can be specified in two formats

var 1 var 2 var 3 a list separated by spaces

var1-var(n) a starting and ending variable with same root name; n total variables to be moved.

fmt = used to format the numeric values.

fmt = 6.2 values are displayed with 2 decimals.

scaler = multiplies values by this number before printing. Useful for converting fractions to percentages or vice versa.

scaler = 0.01 multiplies values by 0.01 before printing

specfmt= used to apply a SAS format other than the default.

specfmt=pvalue5.3 uses the SAS internal format to display the numbers

PROCEDURE MACROS

Here is a summary of each macro and statistics available to place on a report.

Macro Name	SAS Procedure	Description
Breakdn	SUMMARY	Stores descriptive statistics for a list of analyses variables by levels of 1 or more class variables
Chisqp	FREQ	CHISQP uses PROC FREQ to compute Cochran-Mantel-Haenszel Chi-square tests and p-values for the association between a list of dependent categorical variables and an independent categorical variable, adjusted (optionally) for a set of other categorical covariates.
Freqdis	FREQ	Stores counts, percents, cumulative counts and cumulative percents for a class variable for each level of a second class variable
Glmp	GLM	Stores statistics from GLM for 1 or more independent variables for a list of dependent variables Statistics include p-values from factors in the model, least square means, contrasts and p-values from the contrasts.
Logistp	LOGIST	Stores statistics from LOGIST for 1or more independent variables for a list of dependent variables Statistics include estimates, SE estimates, odds ratios and confidence intervals
Mixedp	MIXED	Stores statistics from MIXED for 1or more independent variables for a list of dependent variables Statistics include p-values from factors in the model, contrasts and p-values from the contrasts.
Phregp	PHREG	Stores statistics from PHREG for 1or more independent variables for a list of dependent variables Statistics include estimates, SE estimates, hazard ratios and confidence intervals.
Regp	REG	Stores regression coefficients, standard errors, t-statistics, and p-values for 1 or more independent variables for a list of dependent variables
Lifetab		Macro LIFETAB reads the dataset output in the OUTSURV option of PROC LIFETEST and outputs the survival values for specified times values for each strata.
Genmodp	GENMOD	Stores statistics from PROC GENMOD.

BREAKDN

Overview

BREAKDN uses PROC SUMMARY to compute descriptive statistics for a list of analyses variables for each combination of a list of class variables and outputs them to a new one observation data set. The statistics are contained in array like names as described below.

Statistics Available Names on Output Dataset

N n1-n?

MEAN m1-m?

MEDIAN med1-med?

STD s1-s?

STDERR se1-se?

MIN min1-min?

MAX max1-max?

SUM sum1-sum?

25% Percentile LQ1-LQ?

75% Percentile UQ1-UQ?

where ? is the product of the number of analyses variables and the number of combinations defined by the class variables.

Example Call

%breakdn(data=rand, class=gender 2T group 6, var=age sbp dbp,

sfirst=class, where=where clinic = ‘A’, out=c1) ;

Note: Class factors must be sequential and start with the number 1.

Macro with default and optional parameters listed

%breakdn(data=_LAST_, class=, var=, sfirst=VAR, where=, print=NO, out=OUT,

nzero=);

Required: class and var

Optional: all others

Description of parameters

data SAS dataset read

class List of class variables. Statistics on the variables listed in VAR are calculated for each combination of class levels and optionally for subtotals and grand totals. Each class variable is listed by the number of levels (needs to begin with one and continue through the number specified). A "T" appended to the number indicates that totals are desired. Class variables must start with the number 1 and be sequential.

Example: class=gender 2T group 6

There are two class variables, gender and group with 2 and 6 levels, respectively. Statistics for each level of gender (and total) by each level of group will be computed and stored in a dataset (18=3*6 combinations in the order gender=1, group=1,2,3,4,5,6 followed by gender=2, group=1,2,3,4,5,6 followed by combined genders , group=1,2,3,4,5,6. Note that the class variable to the right moves the "fastest".

var= List of variables that statistics are to be computed

Example: var=age dbp sbp

sfirst= specifies the order in which the array elements of the statistics are stored. The default is sfirst=var which means for a fixed combination of the class variables, statistics for each of the variables listed in VAR are stored consecutively. By setting sfirst=class, for a fixed variable listed in VAR, statistics for all combinations of the class variables will be stored consecutively. The following examples illustrate the effect of SFIRST on the order the elements are stored.

Examples:

%breakdn(data=rand,class=gender 2T,var=age dbp sbp,out=t1,sfirst=var);

n's for age, dbp, sbp for gender=1 are stored in n1-n3.

n's for age, dbp, sbp for gender=2 are stored in n4-n6.

n's for age, dbp, sbp for gender=total are stored in n7-n9.

%breakdn(data=rand,class=gender 2T,var=age dbp sbp,out=t2,sfirst=class);

n's for age for gender=1,2 and total are stored in n1-n3.

n's for dbp for gender=1,2 and total are stored in n4-n6.

n's for sbp for gender=1,2 and total are stored in n7-n9.

The statistics are stored similarly for the other statistics.

where= optional parameter which allows the user to add a WHERE clause to the call.

Example: where=where ondrug = 1

out= SAS dataset to which the statistics are written.

print= If set equal to yes, the output from PROC SUMMARY is displayed. Used to verify results.

nzero= If set equal to y then missing n’s get the value 0 rather than missing (displayed as a period (.).

CHISQP

Overview

CHISQP uses PROC FREQ to compute Cochran-Mantel-Haenszel Chi-square tests and p-values for the association between a list of dependent categorical variables and an independent variable, adjusted (optionally) for a set of other categorical covariates. These statistics are output to a one observation data set. The statistics are contained in array like names as described below

Statistics Available Names on Output Dataset

P-value p1-p?

Chi-square statistic chisq1-chisq?

where ? is the number of analyses (number of dependent variables).

Example Call:

%chisqp(data=rand, dlist= fever nausea headache, group=trt, clist=site, pdata=p);

This would generate the following PROC FREQ

PROC FREQ DATA=rand;

TABLES site*trt*(fever nausea headache)/cmh noprint;

Note: Class factors must be sequential and start with the number 1.

Macro with default parameters listed:

%chisqp(data=_LAST_, dlist=, group=, clist=, test=G, where=, pdata= ) ;

Required: dlist, group, pdata

Optional: all others

Description of parameters

data= SAS dataset to read

dlist= List of categorical dependent variables.

group= Categorical variable to be tested for relationship with each dependent variable.

clist= List of categorical variables to be used as covariates.

test= Type of Chi-square test. The default is test=G, which does the General Association test. Set test=C for Correlation 1df or test=R for Row Means Differ.

where= Optional parameter which allows the user to add a WHERE clause to the call.

Example: where=where ondrug eq 1

pdata= SAS dataset to which the statistics are written.

FREQDIS

Overview

FREQDIS uses PROC SUMMARY to obtain counts, percents, cumulative counts and cumulative percents for a class variable for each level of another class variable (if given) and outputs them to a one observation data set. This macro is useful for displaying the distribution of a variable for each level of another variable.

Statistics Available Names on Output Dataset

Counts n1-n?

Cumulative counts cn1-cn?

Percents p1-p?

Cumulative percents cp1-cp?

where ? is the product of the number of levels of the class variables.

Example Call

%freqdis(data=rand, class=gender 2T agecat 5T,out=table1) ;

Call with default and optional parameters listed

%freqdis(data=_LAST_, class=, where=, out=&class ) ;

Required: class

Optional: data where out

Note: Class factors must be sequential and start with the number 1.

Description of parameters

data SAS dataset to read

class List of up to two class variables for which counts, percents, cumulative counts and cumulative percents are to be computed. Each class variable is followed by the number of levels (need to begin with one and continue through the number specified). A T appended to the number of levels signifies that totals are desired. When two class variables are specified, the counts, percents, cumulative counts and cumulative percents for the second variable are calculated for each level of the first.

Example: class=gender 2T agecat 5T

n1-n6 contain the counts for each of the 5 age categories and the total over all ages for gender=1.

n7-n12 contain the counts for each of the 5 age categories and the total over all ages for gender=2.

n13-n18 contain the counts for each of the 5 age categories and the total over all ages for both genders combined.

p1-p18 contain the percents as above.

cn1-cn18 contain the cumulative counts as above.

cp1-cp18 contain the cumulative percents as above.

where= Optional parameter which allows the user to add a WHERE clause to the call.

Example: where=where ondrug eq 1

out= SAS dataset to which the statistics are written.

GLMP

Overview

GLMP runs PROC GLM(s) and stores statistics generated from the analyses into one or more SAS datasets. Statistics include p-values from factors in the model, least square means (adjusted means) and contrasts and p-values from the contrasts.

Statistics Available Names on Output Dataset(s)

For independent variables

P-value from ANOVA p1-p?

F-statistic from ANOVA f1-f?

If lsmean statement used

Adjusted (LSM) Means lsm1-lsm?

LSM Standard Errors lsmse1-lsmse?

If estimate statement used

Estimate e1-e?

Standard error of estimate se1-se?

T-value for estimate t1-t?

P-value for estimate p1-p?

If contrast statement used

F-value from contrast f1-f?

DF from contrast df1-df?

P-value from contrast p1-p?

See below for value of ? for each of the types of statistics and how they are stored.

Example calls:

%let listd = chold hdld ldld trigd;

%let listb = cholbl hdlbl ldlbl trigbl;

%glmp(data=rand, class=group, dlist=&listd, ilist=group clinic, blist=&listb,

factor=group, pdata=pvals, lsmean=yes, lsdata=lsmeans);

%glmp(data=rand, class=trt, dlist=&listd, factor=trt, pdata=pvals,

estimate = '1 v 3' trt 1 0 -1/'2 v 3' trt 0 1 -1, condata=con,

contrast=%str('2 df test' trt 1 0 -1, trt 0 1 -1));

Macro with default and optional parameters listed

%glmp(data=_LAST_, class=, dlist=, ilist=, blist=, factor=,

where=, pdata=&factor, lsmean=NO, lsclass=&class,

lsdata=&lsclass, estimate=, estdata=,contrast=,condata=);

Description of parameters

data= SAS dataset to read

class= class variable(s) for GLM. List all variables that are to be treated as categorical variables

dlist= dependent variable list. An analysis is run for each variable.

ilist= independent variable list. This list will be used for each dependent variable given in dlist.

blist= optional list of variables where the i'th variable in the list is added as a covariate to the i'th analysis. It is used mainly to adjust for the baseline value of the dependent variable.

Note: if blist is not given then only one GLM is run with multiple dependent variables in the model statement. If blist is given then a separate GLM is run for each dependent variable.

factor= independent variable(s) for which p-values from the ANOVA table are obtained

pdata= SAS dataset(s) in which statistics are stored for the variables listed in factor. There is one dataset per factor with one observation in each dataset. The variables on the dataset are:

p1-p? The p-values for the i'th analyses (? Total analyses)

f1-f? The F-statistics for the i'th analyses.

lsmean= set to YES if least square means are desired.

lsclass= the class variable for which LS means are desired

lsdata= SAS dataset containing the LS means. This dataset contains one observation with the following variables:

lsm1-lsm? = the LS means

lsmse1-lsmse? = the standard errors of the LS means

? = the number of levels of the class variable times the number of dependent variables. For example, if the class variable has 5 levels and there are 2 dependent variables then ? = 10; LS1-LS5 contain the LS means for the first dependent variable and LS6-LS10 contain the LS means for the second dependent variable.

Note: If you have 2 or more class variables and request LS means for one of them then PROC GLM does not give the "correct" results (the adjusted means it computes assume a balanced design). In this case define indicator variables for the levels of the additional class variables and add them as independent variables in the model. The mean of the adjusted means will then be equal to the overall unadjusted mean!

estimate= If given, an estimate statement(s) is generated to perform 1 df contrasts. Multiple contrasts are separated by a slash (/).

Example: estimate = '1 v 3' trt 1 0 -1/ '2 v 3' trt 0 1 -1

This would perform two contrasts, the first contrast comparing the first and third levels of trt; the second comparing the second and third levels of trt.

estdata= The dataset the estimates are stored. The variables are:

e1-e? contrast (estimate) values

se1-se? standard errors of the estimates

t1-t? t-statistics

p1-p? p-values

? is the total estimate statements times the number of dependent variables. They are stored in the order generated. If there are 2 dependent variables and 2 contrasts then

e1 = estimate for analyses 1, contrast 1

e2 = estimate for analyses 1, contrast 2

e3 = estimate for analyses 2, contrast 1

e4 = estimate for analyses 2, contrast 2

contrast= This generates the contrast statement in GLM to perform >1 df contrasts. Constasts are separted by a slash (/).

Example: Contrast = %str('2 df test' trt 1 0 -1, trt 1 -1 0)

The %str is required.because a comma is used in the parameter.

condata= The dataset the contrasts are stored. The variables are:

f1-f? F-statistics

p1-p? P-values

df1-df? Degrees of freedom of the contrast

? is the total number of contrast statements times the number of dependent variables. They are stored in the same order as in the estimate statements (see above).

where= Optional parameter which allows the user to add a WHERE clause to the call.

Example: where=where ondrug eq 1

LOGISTP

Overview

LOGISTP uses proc LOGIST to perform logistic regression and saves results for factors of interest into SAS datasets. Results for each factor is placed in a separate dataset.

Statistics Available Names on Output Dataset(s)

For class variables

Wald Chi-Square statistic chisq1-chisq?

P-value from Chi-square test p1-p?

Likelihood Ratio Chi-Square lrchi1-lrchi? (for global test)

For non-class variables

Estimate (beta coefficient) e1-e?

Standard error of estimate se1-se?

Z-statistic (beta/SE) z1-z?

P-value for estimate p1-p?

Odds ratio o1-o?

Upper bound of 95% CI u1-u?

Lower bound of 95% CI l1-l?

For all variables the overall likelihood ratio chi-square statistic (lrchi1-lrchi?) and associated p-value (plr1-plr?) are placed on the dataset. Also on all datasets is the number of events (NE1-NE?), number of non-events (NC1-NC?) and total observations in analysis (NT1-NT?)

? is the number of analyses (dependent variables)

Example Call

%logistp(data=rand, dlist=chd cvd allcause, ilist=age dbp chol prevmi,

factor=prevmi ,out=mi, order=d, units=10 10 30 1) ;

Macro with default and optional parameters listed

%logistp(data=_LAST_, dlist=, ilist=,class=, factor=, units=, where=, out=&factor, order=D) ;

Description of parameters

data= SAS dataset to read

class= class variable(s) for LOGIST. List all variables that are to be treated as categorical variables

dlist= dependent variable list. An analysis is run for each variable. Each variable should have 2 values, a value for the event and a value for a non event (usually 1 for event and 0 for non event)

ilist= independent variable list. This list will be used for each dependent variable given in dlist.

factor= independent variable(s) in model for which statistics are to be obtained.

units= Used to obtain odds ratios for more than a unit difference. Before the odds ratios are calculated, the coefficients are multiplied by the value of this parameter. The default is 1. This is given as a list over the factors. If there are 3 factors then there should be 3 units given in the list.

where= optional parameter which allows the user to add a WHERE clause

Example: where=where ondrug eq 1

out= SAS dataset(s) in which statistics are stored. There is one dataset for each factor.

order= Set to D to estimate the probability that the dependent variables is equal to 1 (rather than the default which models the probability equal to 0)

MIXEDP

Overview

MIXEDP runs proc MIXED and stores estimates, standard errors, t-statistics and p-values for factors of interest into SAS datasets. Each factor is placed in a separate dataset. Results from estimate and contrast statements can also be put in SAS datasets.

Statistics Available Names on Output Dataset(s)

For class variables

F-statistic f1-f?

P-value from F-statistic p1-p?

For non-class variables

Estimate (beta coefficient) e1-e?

Standard error of estimate se1-se?

T-statistic (beta/SE) t1-t?

P-value for estimate p1-p?

If estimate statement used

Estimate e1-e?

Standard error of estimate se1-se?

T-value for estimate t1-t?

P-value for estimate p1-p?

If contrast statement used

F-value from contrast f1-f?

DF from contrast df1-df?

P-value from contrast p1-p?

See below for value of ? for each of the types of statistics and how they are stored.

Example Call:

%mixedp (data=rand, class=ptid active stratum, dlist = sbp dbp,

ilist=onmeds stratum active stratum*active,

blist=sbpbl dbpbl,

random=int/type=un sub=ptid,

factor=stratum*active,

out=interaction,

estimate= 'Stratum' stratum 1 -1/'Active' Active 1 -1,

estdata=con);

Macro with default and optional parameters listed

%mixedp(data=_LAST_, class=, dlist=, ilist=, blist=, factor=,random=, repeated=, out=&factor,

estimate=, estdata=, contrast=,condata=, lsmeans=, where=, option=noclprint);

Description of parameters

data= SAS dataset read

dlist= Dependent variable list. A separate MIXED is run for each variable in dlist.

ilist= Independent variable list. This same list will be used for each dependent variable in the MIXED analysis.

blist= Optional list of variables where the i'th variable in the list is added as a covariate to the i'th analysis. It is used mainly to adjust for the baseline value of the dependent variable.

option= Any legal option on the PROC MIXED statement

class= List of class variables including grouping unit

random= The entire RANDOM statement used in MIXED

Example: random = intercept/sub=ptid g;

repeated= The entire REPEATED statement used in MIXED

Example: repeated = visit/sub=ptid;

lsmeans= include factors desired. No output dataset is created.

factor= Independent variable(s) for which statistics from PROC MIXED are to be obtained.

where= Optional parameter which allows the user to add a WHERE clause

out= SAS dataset(s) in which statistics are stored for the variables listed in factor. There is one dataset per factor with one observation in each dataset.

estimate= if given, an estimate statement(s) is generated to perform 1 df contrasts. Multiple contrasts are separated by a slash (/).

Example: estimate = '1 v 3' trt 1 0 -1/ '2 v 3' trt 0 1 -1

This would perform two contrasts, the first contrast comparing the first and third levels of trt; the second comparing the second and third levels of trt.

estdata= The dataset the estimates are stored. The variable are:

e1-e? contrast values

se1-se? standard errors of the estimates

t1-t? t-statistics

p1-p? p-values

? is the total estimate statements times the number of dependent variables. They are stored in the order generated. If there are 2 dependent variables and 2 contrasts then

e1 = estimate for analyses 1, contrast 1

e2 = estimate for analyses 1, contrast 2

e3 = estimate for analyses 2, contrast 1

e4 = estimate for analyses 2, contrast 2

contrast= This generates the contrast statement to perform >1 df contrasts. Contrasts are separates by a slash (/).

Example: Contrast = %str('2 df test' trt 1 0 -1, trt 1 -1 0)

The %str is required.

condata= The dataset the contrasts are stored. The variables are:

f1-f? F-statistics

p1-p? P-values

df1-df? Degrees of freedom of the contrast

? is the total number of contrast statements times the number of dependent variables. They are stored in the same order as in the estimate statements (see above).

where= optional parameter which allows the user to add a WHERE clause to the call.

Example: where=where ondrug eq 1

PHREGP

Overview

PHREGP uses proc PHREG to perform proportional hazards regression and saves results for factors of interest into SAS datasets. Each factor is placed in a separate dataset

Statistics Available Names on Output Dataset(s)

For class variables

Wald Chi-Square statistic chisq1-chisq?

P-value from F-statistic p1-p?

For non-class variables

Estimate (beta coefficient) e1-e?

Standard error of estimate se1-se?

Z-statistic (beta/SE) z1-z?

P-value for estimate p1-p?

Hazard ratio r1-r?

Upper bound of HR 95% CI u1-u?

Lower bound of HR 95% C l1-l?

Likelihood ratio statistic lrchi1-lrchi?

P-value for LR statistic plr1-plr?

If estimate statement used

Estimate e1-e?

Standard error of estimate se1-se?

Z-value for estimate z1-z?

P-value for estimate p1-p?

Hazard Ratio r1-r?

HR lower 95% CI l1-l?

HR upper 95% CI u1-u?

If contrast statement used

Chi-square value from contrast chisq1-chisq?

P-value from contrast p1-p?

Both class and non-class datasets contain the number of events (NE1-NE?), number of non-events (NC1-NC?) and total observations in analysis (NT1-NT?)

Example Call:

%phregp(data=rand, dlist=chd cvd allcause, ilist=age dbp chol prevmi,

tlist=t t t, strata=clinic, factor=prevmi ,out=mi) ;

Macro with default and optional parameters listed:

%phregp(data=_LAST_, dlist=, ilist=, tlist=, factor=, skipout=

strata=, where=, units=1, ties=breslow, out=&factor) ;

Description of parameters

data= SAS dataset to read

dlist= dependent variable list. An analysis is run for variable. Each variable should be coded 1 for event and 0 if censored.

class= list of class variables. The GLM parameterization is used.

ilist= independent variable list. This list will be used for each dependent variable given in dlist.

tlist= time variable list corresponding to time of event (or censoring) for each variable in dlist

strata= strata variable(s)

factor= independent variable(s) for which statistics from are to be obtained.

units= Used to obtain RR (HR) for more than a unit difference. Before the risk ratios are calculated, the coefficients are multiplied by the value of this parameter. The default is 1. This is given as a list over the factors. If there are 3 factors then there should be 3 units given in the list.

ties= method for handling ties. Default is Breslow

skipout= no datasets are created. Macro used only to generate PHREG code.

where= Optional parameter which allows the user to add a WHERE clause on any other optional coding

Example: where=where ondrug eq 1

out= SAS dataset(s) in which statistics are stored for the variables listed in factor. There is one dataset per factor with one observation in each dataset.

estimate= If given, an estimate statement(s) is generated to perform 1 df contrasts. Multiple contrasts are separated by a slash (/).

Example: estimate = '1 v 3' trt 1 0 -1/ '2 v 3' trt 0 1 -1

This would perform two contrasts, the first contrast comparing the first and third levels of trt; the second comparing the second and third levels of trt.

estdata= The dataset the estimates are stored.

? is the total estimate statements times the number of dependent variables. They are stored in the order generated. If there are 2 dependent variables and 2 contrasts then

e1 = estimate for analyses 1, contrast 1

e2 = estimate for analyses 1, contrast 2

e3 = estimate for analyses 2, contrast 1

e4 = estimate for analyses 2, contrast 2

contrast= This generates the contrast statement to perform >1 df contrasts. Contrasts are separates by a slash (/).

Example: Contrast = %str('2 df test' trt 1 0 -1, trt 1 -1 0)

The %str is required.

condata= The dataset the contrasts are stored. The variables are:

chisq1-chisq? Chi-square statistics

p1-p? P-values

? is the total number of contrast statements times the number of dependent variables. They are stored in the same order as in the estimate statements (see above).

REGP

Overview

REGP runs PROC REG and stores regression coefficients, standard errors, t-statistics, and p-values of covariates from the regression into one or more SAS datasets.

Statistics Available Names on Output Dataset(s)

Regression coefficient (beta) b1-b?

Standard error of beta seb1-seb?

T-statistic (beta/SE) t1-t?

P-value for T-statistic p1-p?

Lower 95% CI for beta l1-l?

Upper 95% CI for beta u1-u?

Where ? is the number of dependent variables.

Example Call:

%let listd = chold hdld ldld trigd ;

%let listb = cholbl hdlbl ldlbl trigbl ;

%let listi = group2 clinica clinicb clinicc stratum2 ;

%regp(data=rand, dlist=&listd, ilist=&listi, blist=&listb,factor=group2, cohort=N, out=results);

Macro with default and optional parameters listed:

%regp(data=_LAST_, dlist=, ilist=, blist=, factor=, cohort=Y, where=, out=&factor) ;

Description of parameters:

data= SAS dataset to read

dlist= dependent variable list.

ilist= independent variable list. This list will be used for each dependent variable given in dlist.

blist= optional list of variables where the i'th variable in the list

is added as a covariate to the i'th analysis. It is used mainly to adjust for the baseline value of the dependent variable.

factor= independent variable(s) for which statistics are obtained.

cohort= if set to Y (default), a single PROC REG is run which excludes cases with any dependent variable having missing data from all regressions. If cohort=N, PROC REG is run separately for each dependent variable

where= optional parameter which allows the user to add a WHERE clause

Example: where=where ondrug eq 1

out= dataset(s) to which statistics are stored. There is one dataset for each variable in factor.

LIFETAB

Overview

LIFETAB reads the dataset created in the OUTSURV option of PROC LIFETEST and outputs the survival values for specified time values for each strata. This macro is useful for displaying life tables estimates from selected points on the graph. Unlike the other statistical macros, the user needs to run the LIFETEST procedure before calling the macro.

Statistics Available Names on Output Dataset(s)

Survival estimates s1-s?

1- survival estimates f1-f?

Upper bound of 95% CI for S u1-u?

Lower bound of 95% CI for S l1-l?

Times t1-t?

Strata value g1-g?

Where ? is the number of time points specified x the number of strata levels. If there were 5 time points specified and 2 levels for strata then ? would be equal to 10 (5 x 2).

Example Call:

%lifetab(data=surv, t=tmort, times = 1 2 3 4 5 6, out=surv, strata=trt);

This call outputs survival estimates at times 1,2,3,4,5,6 for each level of the variable trt.

Macro with default and optional parameters listed:

%lifetab(data=, t=t, times = , inter=, ninter, out= &data, print=Y, zero=)

Description of parameters

data= SAS dataset previously output from the OUTSURV option of proc lifetest

t= name of variable containing event time

times= list of times to output survival curve values for each strata

ninter= number of intervals (used with option inter) of time to output

inter= output survival values at NINTER intervals of this length (used with ninter)

Use TIMES or INTER with NINTER to indicate the times you want the survival estimates to be output. The units should be consistent with the units used for T.

Example: TIMES = .5 1 2 3 4 5 (output at 6 months and then yearly)

Example: NINTER = 5, INTER = 365 (output at 365,365*2, 365*3, 365*4, 365*5)

zero= if set to Y then the survival value at t=0 is output

out= SAS dataset to which statistics are stored. The default is the same as the input dataset.

print= Y prints out values of survival curve at all uncensored points

GENMODP

Overview

GENMODP runs proc genmod and stores statistics from the model into one or more datasets. Default parameters are set for Poisson regression.

Statistics Available Names on Output Dataset(s)

For class variables

Chi-square statistic k1-k?

P-value from chi-square statistic p1-p?

For non-class variables

Estimate (beta coefficient) e1-e?

Standard error of estimate se1-se?

Z-statistic (beta/SE) z1-z?

P-value for estimate p1-p?

Relative Rates r1-r?

Lower 95% CI for RR L1-L?

Upper 95% CI for RR U1-U?

If estimate statement used

Estimate e1-e?

Standard error of estimate se1-se?

Z-value for estimate z1-z?

P-value for estimate p1-p?

Lower 95% CI for RR L1-L?

Upper 95% CI for RR U1-U?

See below for value of ? for each of the types of statistics and how they are stored.

Example Call:

%genmodp(data=temp,

class= ptid yearcat,

dlist= nhosp naids,

ilist= yearcat cd4 ,

offset = lpr,

repeated = subject=pin/type=cs corrw,

factor= yearcat ,

out=out1 ,

estimate = '2005 v 2006' yearcat 1 -1 0|'2006 v 2007' yearcat 0 1 -1,

estdata=est);

Macro with default and optional parameters listed:

%macro genmodp(data=_last_,

class=,

dlist=, ilist=, blist=,

link=log, dist=poisson, offset=,

repeated=,

factor=,out=&factor,

estimate=, estdata=,

units=,

where=,

options=);

Description of parameters:

data= SAS dataset read

dlist= dependent variable list. A separate GENMOD is run for each variable in dlist.

ilist= independent variable list. This same list will be used for each dependent variable in the analysis.

blist= optional list of variables where the i'th variable in the list is added as a covariate to the i'th analysis. It is used mainly to adjust for the baseline value of the dependent variable.

link= link function (default is log)

offset= offset variable need (used for Poisson regression)

option= any legal option on the GENMOD statement

class= list of class variables including grouping unit

repeated= The entire REPEATED statement used in GENMOD. If left blank then no repeated statement is generated.

Example: repeated = subject=pin/type=cs corrw

factor= independent variable(s) for which statistics from GENMOD are to be obtained.

out= dataset(s) in which statistics are stored for the variables listed in factor. There is one dataset per factor with one observation in each dataset.

estimate= If given, an estimate statement(s) is generated to perform 1 df contrasts. Multiple contrasts are separated by a slash (/).

Example: estimate = '1 v 3' trt 1 0 -1/ '2 v 3' trt 0 1 -1

This would perform two contrasts, the first contrast comparing the first and third levels of trt; the second comparing the second and third levels of trt.

estdata= The dataset the estimates are stored. The variable are:

e1-e? contrast values

se1-se? standard errors of the estimates

t1-t? t-statistics

p1-p? p-values

? is the total estimate statements times the number of dependent variables. They are stored in the order generated. If there are 2 dependent variables and 2 contrasts then

e1 = estimate for analyses 1, contrast 1

e2 = estimate for analyses 1, contrast 2

e3 = estimate for analyses 2, contrast 1

e4 = estimate for analyses 2, contrast 2

where= optional parameter which allows the user to add a WHERE clause to the call.

Example: where=where ondrug eq 1

MEDIANP

Overview

MEDIANP runs proc npar1way and stores statistics from the analysis into a dataset.

Statistics Available Names on Output Dataset(s)

Chi-square statistic chisq1-chisq?

P-value from chi-square statistic p1-p?

Degrees of Freedom df1-df?

Example Call:

%medianp (data=weight, class = sex, var=cholbl, test=w),

Print=y);

Macro with default and optional parameters listed:

%medianp (data=_last_, out=out, class=, var=, test=W,

Print=noprint, where = );

Description of parameters:

data= SAS dataset read

var= dependent variable list. A separate NPAR1WAY is run for each variable in var.

class= class variable to be compared for each variable in var.

test= type of test to perform: M (Median); S(Savage); V (VW); W (Wilcoxon).

print= default is NOPRINT of results; set to Y to print results

where= optional parameter which allows the user to add a WHERE clause to the call.

Example: where=where ondrug eq 1

Note: When for TEST=W the chi-square test output is the Kruskal-Wallis which is equivalent to the Wilcoxon test.

HTML REPORTS

The macros can also be used create an HTML table. This HTML file can then be viewed using a web browser or imported into Word. Most of the formatting options that are available in html such as font type, font size, font weight, and indenting and underlining text have been incorporated as well as more general html features such as background and text color. In addition multiple pieces of data can be combined and placed it in a single column with various formatting, such as mean ± SD, N (%), or HR (95% CI) giving the report table a journal finished look. Default options have been set up so that when the html table is inserted into Word the columns spread out proportionately to fill the page margins, making it unnecessary to deal with margin issues.

The most simple HTML table can be made by making the following changes to the usual report:

Add the following bolded parameter to the report call that will contain the html file.

%wreport (repfile = ‘report.txt’, htmlfile=’report.html’); and

Add the following 2 lines at the end of your report

stop;

%makehtml;

The stop statement is required; without it the html file will be incorrect. The macro call to makehtml generates the html file.

There are several enhancements that can be made by using options on the %move and %nmove statements.

Formatting options for %move and %nmove are given in the following table:

Macro Parameter	Description	Default	Other values
Indent	Used to indent text (%move only)	0=none;	1=some indent; 2=more
Fontfamily	any usual font name recognized by html	Times New Roman	Arial; Courier New
Fontsize	size of font including unit of measurement	10pt	12pt (no space between number and pt)
Fontweight	used to bold text;	Normal	Bold; bolder
Fontstyle	used to italicize text	Normal	Italics
Lineheight	Size of cell in proportion to text size	125%	200% (no space between number and %)
Textdecoration	Used for underlining	None	underline
Background	Used for shading or changing backgroung color	White (#ffffff)	Use #d9d9d9 or #b3b3b3 for shading

Note: Default values are controlled through macro makehtml

%macro makehtml

This macro creates the html file. It is not usually needed to change the default values.

Macro Parameter	Description	Default	Other values
border	Size of width in pixels of the frame around the table	2	Any number
frame	Specifies which sides of the frame surrounding the table will be visible	Box – all 4 sides	Above; below; hsides; vsides; lhs; rhs (see html manual)
Rules	Controls the rules (lines) separating cells	Groups	None; all
cellspacing	Amount of space between cells in pixels	0	Any value
Cellpadding	Amount of white space within cell in pixels	1	Any value
Bgcolor	Background color	#ffffff (white)	Any valid color
textcolor	Text color	000000	Any valid color
width	Width of table as proportion of browser width	100% (uses all of browser width)	Other %s or actual widths

In addition, you can change the default values for fontfamily, fontsize, fontweight, fontstyle, and lineheight by setting those parameters in %makehtml.

New parameters for %nmove

combine = y to combine data in same cell (see examples)

fchar = code which tells how to format combined data.

= 1 uses ± to combines 2 numbers

= 1C uses a comma between 2 numbers

= 2 puts second number in parenthesis

= 2P as in 2 but adds percent sign at end of second number

= 3 second and third number are placed in parenthesis and a dash placed between second and third number, Used for HR with CIs.

= 3B second and third number are placed in square brackets and a dash placed between second and third number.

= 4 second and third number are placed in parenthesis and a comma is placed between second and third number.

= 4B second and third number are placed in square brackets and a comma is placed between second and third number.

Example using fchar option:

= 1 45.2 ± 6.7

= 1C 1.18, 1.96

= 2 231 (17.1)

= 2P 231 (17.1%)

= 3 1.02 (0.88 - 1.18)

= 3B 502 [250-788]

= 4 1.02 (0.88, 1.18)

=4B 502 [250,788]

Notes: To use the combine feature the variables to be combined must be specified in array type notation. For example the following call puts the three variables r1, L1, and u1 together in one cell, the variables r2, L2, and u2 in a second cell, etc. The combine option must be set to “y”, otherwise the 21 total variables are treated as individual variables. Also, separate formats can be specified for the pair or triplet of numbers as can separate scaler options.

%nmove(r1-r7 L1-L7 u1-u7, col=4, line=14L7, combine=y, fmt=5.2 5.2 5.2, fchar=3);

%nmove(m1-m7 se1-se7, col=4, line=14L7, combine=y, fmt=5.1 5.1 , fchar=1);

%nmove(sum1-sum7 m1-m7, col=4, line=14L7, combine=y, fmt=5.0 5.1 , fchar=2P,

scaler=1 100);

Some notes on using wreport for making html files

· The lines numbers given in move and nmove statements are used only in a relative sense, .i.e., the lowest line number is the first row of the table, the next line number is the second row of the table, etc. There are no “skipped” rows. The user can use this to their advantage. For example, a footnote text can be moved to a high line number so the footnote text is always at the end of the report.

· The column numbers in %colset are used in a relative sense. Thus, a value of 20 will take twice the space as value of 10. The table will automatically fill the margin when inserted into Word.

· The report displays different grid lines depending on whether or not consecutive line numbers are filled. For example, moving 3 character strings to consecutive lines (e.g. line=22L3) will yield a slightly different looking report then if the strings were moved to lines 22, 24 and 26 (e.g. line=22 24 26). Moving to consecutive lines will give a block looking section which may be nicer to view then having grid lines in between each row. However, this distinction is not honored by Word; but is honored by most browsers.

· Underlines (u=1) on %move statements are ignored. You can include them to make the text report look better.

· You may use html tags within the text strings to do formatting not controlled by the macro formatting parameters. Use the <br> tag in the text to start a new line within the cell. Other useful HTML tags are  to superscript the text between tags and  to subscript the text between tags. If you only wish to bold or italicize a portion of the text in a %move statement, <b></b> and <i></i> may be used. If you really want to emphasize a word, <b><i><u> important word</u></i></b> will give you: important word .

· If the combine option is used more than once in a report then move the largest width characters first (example move the HR and 95%CI before the n and %. Otherwise some text may get truncated.

HTML REPORT EXAMPLE 1

Each table is followed by the code to generate the table.

Number and Percent of Selected Cardiovascular Events by Treatment Group And Hazard Ratio (Experimental/Standard) from Cox Regression Analyses
	Number of Patients With Event		Cox Regression Analyses Experimental/Standard
Endpoint	Experimental	Standard	HR (95% CI)	P-value
Primary CVD	364 (4.5%)	365 (4.4%)	1.02 (0.88, 1.18)	0.771
MI (Fatal/NF)	133 (1.6%)	166 (2.0%)	0.82 (0.65, 1.03)	0.089
Stroke (Fatal/NF)	133 (1.6%)	118 (1.4%)	1.15 (0.90, 1.48)	0.265
CVD Death	152 (1.9%)	143 (1.7%)	1.09 (0.87, 1.37)	0.471
Any CVD Hospitalization ^1	793 (9.7%)	775 (9.3%)	1.05 (0.95, 1.16)	0.307
Revascularization	163 (2.0%)	166 (2.0%)	1.01 (0.82, 1.26)	0.913
TIA	89 (1.1%)	105 (1.3%)	0.87 (0.66, 1.15)	0.330
Angina	202 (2.5%)	190 (2.3%)	1.09 (0.89, 1.33)	0.389
Heart Failure	126 (1.5%)	100 (1.2%)	1.30 (1.00, 1.69)	0.051
Hypertension	22 (0.3%)	18 (0.2%)	1.26 (0.67, 2.34)	0.474
Renal Failure	27 (0.3%)	34 (0.4%)	0.81 (0.49, 1.35)	0.426
^1 Includes events below plus primary events

libname convince '/data/convince/';

data temp;

set convince.convince ;

if abgroup = 'A' then group = 1;

if abgroup = 'B' then group = 0;

if abgroup = 'A' then cgroup = 1 ;

if abgroup = 'B' then cgroup = 2 ;

if abgroup = 'A' then ggroup = 0 ;

if abgroup = 'B' then ggroup = 1 ;

%let evlist = primary mi str cvddth allcvd corevas tia angina chf acchyp renalf ;

%let tmlist = tprimary tmi tstr tcvddth tallcvd tcorevas ttia

tangina tchf tacchyp trenalf ;

keep patno group cgroup &evlist &tmlist soc cstrata country agerand;

%breakdn(data=temp, class= cgroup 2, var=&evlist, out=out1);

%phregp(data=temp, dlist = &evlist, strata=soc cstrata,

tlist= &tmlist, ilist=group, factor=group,out = out2);

%wreport(repfile='example.txt', htmlfile='example.html');

%colset(25 14 14 21 7);

%move('Number and Percent of Selected Cardiovascular Events by Treatment Group<br>'

'And Hazard Ratio (Experimental/Standard) from Cox Regression Analyses',

col=1-0, line=3, fontweight=bold, fontsize=12pt);

%move('Number of Patients<br>With Event', col=2-3, line=9, fontweight=bold);

%move('Cox Regression Analyses<br>Experimental/Standard', col=4-0, line=9,

fontweight=bold);

%move('Experimental':'Standard':'HR (95% CI)':'P-value', col=2.0,line=11,

fontweight=bold);

%move('Endpoint', col=1, center=2, line=11, u=1, fontstyle=italic, fontweight=bold);

%move('Primary CVD ':'Any CVD Hospitalization <sup>1', col=1, center=2, line=14 19,

fontstyle=italic, fontweight=bold);

%move('MI (Fatal/NF)':

'Stroke (Fatal/NF)':

'CVD Death':

'Revascularization':

'TIA':

'Angina':

'Heart Failure':

'Hypertension':

'Renal Failure',

col=1, center=2, line=15L3 20L6, indent=1);

%move('<br>¹ Includes events below plus primary events', col=1-0, line=30, center=2);

set out2;

%nmove(r1-r11 L1-L11 u1-u11, col=4, line=14L4 19L7, combine=y, fmt=5.2 5.2 5.2, fchar=4);

%nmove(p1-p11 , col=5, fmt=6.3);

set out1;

%nmove(sum1-sum22 m1-m22, col=2 3, scaler=1 100, combine=y, fmt=5.0 5.1, fchar=2P);

stop;

run;

%makehtml;

run;

HTML REPORT EXAMPLE 2

Mean Levels of Measures Contributing to The Rasmussen Score and Other Determinations by Gender for Rasmussen Population
Variables Contributing to Score	Men	Women	Total	P (M/W)
Large Artery Elasticity	18.5 ± 5.8	14.5 ± 5.1	16.9 ± 5.8	<.001
Small Artery Elasticity	7.4 ± 3.2	5.2 ± 2.8	6.5 ± 3.3	<.001
Resting Systolic BP (mmHg)	133.2 ± 16.8	130.0 ± 20.8	131.9 ± 18.6	0.009
Resting Diastolic BP (mmHg)	85.3 ± 9.8	80.8 ± 9.7	83.5 ± 10.0	<.001
Exercise SBP (mmHg)	156.3 ± 21.6	153.4 ± 25.1	155.1 ± 23.1	0.062
Exercise DBP (mmHg)	75.1 ± 11.9	70.7 ± 11.1	73.4 ± 11.8	<.001
Change in SBP (mmHg)	23.1 ± 15.4	23.6 ± 15.7	23.3 ± 15.5	0.662
Change in DBP (mmHg)	-10.2 ± 9.0	-10.3 ± 8.7	-10.2 ± 8.9	0.955
LVMI (g/m2)	94.7 ± 25.6	79.0 ± 22.0	88.3 ± 25.4	<.001
Microalbumin (mg/mmoL)	0.8 ± 1.5	1.3 ± 3.2	1.0 ± 2.3	0.005
BNP (pg/dL)	18.4 ± 22.9	30.0 ± 28.7	23.1 ± 26.0	<.001
Other Risk Factors Measured
Total Cholesterol (mg/dl)	199.2 ± 35.0	210.9 ± 38.5	203.9 ± 36.9	<.001
LDL Cholesterol (mg/dl)	126.6 ± 31.2	127.2 ± 34.4	126.9 ± 32.5	0.778
HDL Cholesterol (mg/dl)	45.0 ± 11.4	59.2 ± 17.6	50.7 ± 15.8	<.001
Triglycerides (mg/dl)	146.0 ± 115.6	125.0 ± 116.2	137.6 ± 116.2	0.007
Glucose (mg/dl)	95.5 ± 16.2	91.3 ± 17.0	93.8 ± 16.6	<.001
CRP (mg/L)	0.3 ± 0.4	0.4 ± 0.5	0.3 ± 0.4	<.001
Homocysteine (umol/L)	8.9 ± 2.4	7.7 ± 2.2	8.4 ± 2.4	<.001
Age (y)	51.0 ± 11.4	52.6 ± 11.1	51.7 ± 11.3	0.033
10-Year CHD Risk (%)	7.1 ± 5.6	1.9 ± 2.0	5.0 ± 5.2	<.001
Body Mass Index (kg/m2)	27.8 ± 4.4	26.9 ± 5.5	27.4 ± 4.9	0.008
Number of Patients	565	379	944

%let dlist = c1 c2 rest_sbp rest_dbp ex_sbp ex_dbp _sbp _dbp

lvmindx malb_ bnp total_chol ldl hdl

trig gluc hscrp homocys age chd10;

%breakdn(data=temp, class = gender 2T, var=&dlist bmicalc, out=out1);

%breakdn(data=temp, class = gender 2T, var=bmicalc, out=out2);

%regp(data=temp, dlist = &dlist bmicalc, ilist=gender, factor=gender,

cohort=N, out=out3);

%wreport (htmlfile = 'rasmussen.html' );

%colset(35 25 25 25 10 );

%move('Mean Levels of Measures Contributing to The Rasmussen Score<br>'

'and Other Determinations by Gender<br>'

'for Rasmussen Population',

col=1-0, line=3, fontweight=bold, fontsize=12pt);

%move('Men':'Women':'Total':'P (M/W)', line=9, col=2.5, fontweight=bold);

%move('Variables Contributing to Score':

'Other Risk Factors Measured',

col=1, center=2, line=9 24 ,

fontweight=bold, fontstyle=italic);

%move('Number of Patients', col=1, line= 37 , center=2);

%move('Large Artery Elasticity':

'Small Artery Elasticity':

'Resting Systolic BP (mmHg)':

'Resting Diastolic BP (mmHg)':

'Exercise SBP (mmHg)':

'Exercise DBP (mmHg)':

'Change in SBP (mmHg)':

'Change in DBP (mmHg)':

'LVMI (g/m2)':

'Microalbumin (mg/mmoL)':

'BNP (pg/dL)':

'Total Cholesterol (mg/dl)':

'LDL Cholesterol (mg/dl)':

'HDL Cholesterol (mg/dl)':

'Triglycerides (mg/dl)':

'Glucose (mg/dl)':

'CRP (mg/L)':

'Homocysteine (umol/L)':

'Age (y)':

'10-Year CHD Risk (%)':

'Body Mass Index (kg/m2)':

col=1, center=2, line=12L11 25L10 , indent=2);

set out1;

%nmove(med1-med63 lq1-lq63 uq1-uq63, col=2 3 4, combine=y, fchar=4);

set out2;

%nmove(n1-n3, col=2 3 4, line=37, fmt=4.0);

set out3;

%nmove(p1-p21, col=5, line=12L11 25L10, fmt=6.0, specfmt=pvalue6.3);

stop;

run;

%makehtml;

HTML REPORT EXAMPLE 3

Age, Blood Pressure and Additional Risk Factors by Treatment Group for Patients Randomized in CONVINCE
Baseline Variable	Experimental	Standard	Total
Age (y)	65.6 ± 7.4	65.6 ± 7.4	65.6 ± 7.4
Systolic BP (mmHg)	150.1 ± 15.8	150.1 ± 16.0	150.1 ± 15.9
Diastolic BP (mmHg)	86.8 ± 9.8	86.8 ± 9.8	86.8 ± 9.8
Additional Risk Factor (%)
MI	607 (7.5%)	652 (7.9%)	1259 (7.7%)
Stroke	370 (4.5%)	393 (4.8%)	763 (4.7%)
Cigarette Use	1912 (23.5%)	1883 (22.8%)	3795 (23.2%)
Diabetes	1616 (19.9%)	1623 (19.7%)	3239 (19.8%)
LV Hypertrophy	1000 (12.3%)	1019 (12.4%)	2019 (12.3%)
Cholesterol	2540 (31.2%)	2575 (31.2%)	5115 (31.2%)
TIA	184 (2.3%)	162 (2.0%)	346 (2.1%)
Overweight	4150 (51.0%)	4096 (49.6%)	8246 (50.3%)
Vascular Disease	1362 (16.7%)	1387 (16.8%)	2749 (16.8%)
Vascular Bruit	403 (5.0%)	409 (5.0%)	812 (5.0%)
More than 1 condition	4060 (49.9%)	4038 (48.9%)	8098 (49.4%)
Number of Patients	8179	8297	16476

%let clist = agerand sbpbl dbpbl ;

%let blist = rsk001 rsk002 rsk003 rsk004 rsk005 rsk006 rsk007

rsk008 rsk009 rsk010 rskgt1 ;

%breakdn(data=temp,class=group 2T, var=&clist, out=out1);

%breakdn(data=temp,class=group 2T, var=&blist, out=out2);

%wreport(repfile='basechar.table',htmlfile='basechar.html');

%colset(28 13 13 13);

%move('Age, Blood Pressure and Additional Risk Factors<br>'

'by Treatment Group for Patients Randomized in CONVINCE',

col=1-0, line=3,fontweight=bold, fontsize=12pt, background=#D9D9D9);

%move('Experimental':'Standard':'Total', col=2.0, line=9, fontweight=bold);

%move('Baseline Variable':'Additional Risk Factor (%)', col=1, line=9 16, center=2,

fontweight=bold, fontstyle=italic);

%move('Age (y)':

'Systolic BP (mmHg)':

'Diastolic BP (mmHg)',

col=1, center=2, line=12 13 14);

%move('MI':'Stroke':'Cigarette Use':'Diabetes':'LV Hypertrophy':'Cholesterol':

'TIA':'Overweight':'Vascular Disease':'Vascular Bruit':

'More than 1 condition',

col=1, center=2, line=19L11, indent=1);

%move('Number of Patients', col=1, center=2, line=36, fontweight=bold,

fontstyle=italic);

set out1;

%nmove(m1-m9 s1-s9, col=2.4 , line=12L3, combine=y, fchar=1 );

%nmove(n1 n4 n7, col=2.4, line=36, fmt=5.0, fontweight=bold);

set out2;

%nmove(sum1-sum33 m1-m33, line=19L11, col=2.4, combine=y, fchar=2P, fmt=5.0 5.1,

scaler=0.01 1);

stop;

%makehtml;

ADDING SPECIAL CHARACTERS

The following table gives the code to display various mathematical and other special characters. Each starts with &# , followed by a number, followed by a semicolon. Other codes can be found by searching the web for special characters in html.

Placing Special Characters in Tables Produced Using The SAS Macros for Customized Reports
	Code for Character	Result moving to the page via %move
Mathemetical/scientific
Plus or minus	±	±
Less than or equal to	≤	≤
Greater than or equal to	≥	≥
Not equal to	≠	≠
Micro	µ	µ
Greek letters
Alpha	α	α
Beta	β	β
Delta	δ	δ
Epsilon	ε	ε
Mu	μ	μ
Sigma	σ	σ
Tau	τ	τ
Footnote symbols
Dagger	†	†
Double dagger	‡	‡
Section	§	§
Paragraph	¶	¶
Other useful characters
Colon	:	:
Degree	°	°
Bullet	•	•
Copyright	©	©
Smiley face (for good pvalues)	☺	☺
PN=specialcharacters Rundate=07OCT2010

Examples of Text Moves Using Special Characters and HTML Tags
Type of text desired	Example %move statement	Result
Greek letter beta, plus or minus sign	%move('β ± SE',col=3,line=10) ;	β ± SE
Two lines within a cell	%move('Mean &#177; SD <br>or N (%)',col=3,line=12) ;	Mean ± SD or N (%)
Superscript	%move('P-value<sup>1</sup>',col=3,line=14);	P-value^1
Subscript, bold a portion of text only	%move('<b>Viral Load</b> (log<sub>10</sub>)',col=3,line=16) ;	Viral Load (log10)
Footnotes	%move('<b><u>Notes&#58;</u></b> <br> <sup>1</sup> Adjusted for clinical center', col=3,line=18, center=n) ;	Notes: ^1 Adjusted for clinical center
PN=SCandTagExamples Rundate=11OCT2010