Skip to content
Branch: master
Find file Copy path
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
488 lines (360 sloc) 22.8 KB
Display: Figure 7.7 Box plot - Change from Last Baseline to Last Post-baseline by Treatment and Analysis Timepoint, Multiple Studies
White paper: Central Tendency
User Guide:
Macro Library:
Test Data:
Sample Output:
Using this program:
* See USER PROCESSING AND SETTINGS, below, to configure this program for your environment and data
* Program will plot all studies with maximum of 20 boxes on a page (default)
+ see user option MAX_BOXES_PER_PAGE, below, to change limit of 20 boxes per page
* Program separately plots all parameters provided in PARAMCD
* Measurements within each PARAMCD and ATPTN determine precision of statistical results
+ MEAN gets 1 extra decimal, STD DEV gets 2 extra decimals
+ see macro UTIL_VALUE_FORMAT to modify this behavior
* If your treatment names are too long for the summary table, abbreviate them
in the input data, and add a footnote that explains your short Tx codes
+ This program contains custom code to shorted Tx labels in the PhUSE CS test data
+ See "2b) USER SUBSET of data", below
TO DO list for program:
* Complete and confirm specifications (see Outliers & Reference limit discussions, below)
* For annotated RED CIRCLEs outside normal range limits
UPDATE the test data so that default outputs have some IQR OUTLIER SQUAREs that are not also RED.
* The footnote could dynamically describe any normal range lines that appear. See the Fig. 7.1 specifications.
* LABS & ECG - ADaM VS/LAB/ECG domains have some different variables and variable naming conventions.
- What variables are used for LAB/ECG box plots?
- What visits/time-points are relevant to LAB/ECG box plots?
- Handle all of these within one template program?
Or separate them (and accept some redundancy)?
- NB: Currently vars like AVISIT, AVISITN, ATPT, ATPTN are hard-coded in this program
end HEADER ***/
1) REQUIRED - PhUSE CS Utilities macro library.
These templates require the PhUSE CS macro utilities:
User must ensure that SAS can find PhUSE CS macros in the SASAUTOS path (see EXECUTE ONE TIME, below)
2) OPTIONAL - Subset measurement data, to limit resulting plots to specific
- Parameters
- Analysis Timepoints
- Visits
3) REQUIRED - Key user settings (libraries, data sets, variables and box plot options)
M_LB: Libname containing ADaM measurement data, such as ADVS.
WORK by default, since step (2) creates the desired WORK subset.
M_DS: Measuments data set, such as ADVS.
T_VAR: Variable in M_DS with the Treatment Name, such as TRTP, TRTA.
TN_VAR: Variable in M_DS with the Treatment Number (display controls order), such as TRTPN, TRTAN.
C_VAR: Variable in M_DS with change-from-baseline data, such as CHG.
B_VAR: (Leave blank to omit p-value from summary table.) Variable in M_DS with baseline measurements
REF_TRTN: (Leave blank to omit p-value from summary table.) Numeric value of TN_VAR for reference (comparator) treatment
P_FL: Population flag variable. 'Y' indicates record is in population of interest.
A_FL: Analysis Flag variable. 'Y' indicates POST-BASELINE obs that contain:
* A non-missing post-baseline measure (either LAST, MIN or MAX measurement post-baseline)
* including a non-missing baseline val (LAST, MIN or MAX to match the post-baseline measurement)
* and a non-missing chg from baseline (based on the above baseline & post-baseline measurements)
For an example, see the PhUSE CS derivation in ADaM/
But then remove Baseline obs.
C_MODE: Change Mode label. This is merely a label that describes the obs flagged by user-specified A_FL
* LAST: Default. Change from LAST baseline to LAST post-baseline analysis measurement
* MIN: Change from MIN baseline to MIN post-baseline analysis measurement
* MAX: Change from MAX baseline to MAX post-baseline analysis measurement
Maximum number of boxes to display per plot page (see "Notes", above)
Location to write PDF outputs (WITHOUT final back- or forward-slash)
*** user processing and settings ***
%*--- 3) Key user settings ---*;
**Specify ADaM BDS Dataset;
%let DS = ADVS;
%let param = 'DIABP';
%let m_lb = work;
%let m_ds = &ds._&LMM;
%let t_var = trtp_short;
%let tn_var = trtpn;
%let c_var = chg;
%let b_var = base;
%let ref_trtn = 0;
%let p_fl = saffl;
%let lmm = MIN;
*--- C_MODE is a label for &A_FL, which identifies one record for each STUDYID USUBJID PARAMCD ATPT ---*;
%let c_mode = &LMM;
%let a_fl = anl02fl;
%let max_boxes_per_page = 20;
%let outputs_folder = \\\enterprise\Apps\sasdata\StatOpB\CSV\PhUSE\phuse-scripts\whitepapers\WPCT\GB_Test;
OPTIONS sasautos=( "\\\enterprise\Apps\sasdata\StatOpB\CSV\PhUSE\phuse-scripts\whitepapers\utilities"
"\\\enterprise\Apps\sasdata\StatOpB\CSV\PhUSE\phuse-scripts\whitepapers\ADaM" %sysfunc(getoption(sasautos)));
%put WARNING: (WPCT-F.07.07) User must ensure PhUSE CS utilities are in the AUTOCALL path.;
/*** 1) PhUSE CS utilities in autocall paths (see "Macro Library", above)
EXECUTE ONE TIME only as needed
NB: The following line is necessary only when PhUSE CS utilities are NOT in your default AUTOCALL paths
OPTIONS sasautos=("C:\CSS\phuse-scripts\whitepapers\utilities" "C:\CSS\phuse-scripts\whitepapers\ADaM" %sysfunc(getoption(sasautos)));
/*** 2a) REMOTE ACCESS data, by default PhUSE CS test data, and create WORK copy. ***/
/*** NB: If remote access to test data files does not work, see local override, below. ***/
*%util_access_test_data(&ds, folder=cdisc-split)
*--- NB: LOCAL PhUSE CS test data, override remote access by providing a local path ---*;
%util_access_test_data(&ds, local=\\\enterprise\Apps\sasdata\StatOpB\CSV\PhUSE\phuse-scripts\data\adam\cdisc-split\) ;
/*** 2b) USER SUBSET of data, to limit number of box plot outputs, and to shorten Tx labels ***/
data &ds._sub;
set work.&ds;
where (paramcd in (&param) and atptn in (815 817));
attrib trtp_short length=$6 label='Planned Treatment, abbreviated';
select (trtp);
when ('Placebo') trtp_short = 'P';
when ('Xanomeline High Dose') trtp_short = 'X-high';
when ('Xanomeline Low Dose') trtp_short = 'X-low';
otherwise trtp_short = 'UNEXPECTED';
%*--- Use PhUSE CS derivation of LAST, MIN or MAX Post-Baseline measures, with Change from corresponding Baseline ---*;
%derive_lastminmax_measure(&ds._sub, &LMM,
grpvars=studyid usubjid &tn_var paramcd atptn,
incl=trtp_short &p_fl param atpt,
*--- but program requires ONLY POST-BASELINE obs ---*;
data &ds._&LMM;
set &ds._&LMM;
where avisit =: 'Post-baseline';
/*** end USER PROCESSING AND SETTINGS ***********************************
*** RELAX. ***
*** The rest should simply work, or alert you to invalid conditions. ***
Explain to user in case environment or data do not support this analysis
Keep just those variables and records required for this analysis
For details, see specifications at top
options nocenter mautosource mrecall mprint msglevel=I mergenoby=WARN ls=max ps=max;
%let ana_variables = STUDYID USUBJID &p_fl &a_fl &t_var &tn_var PARAM PARAMCD &c_var &b_var ATPT ATPTN;
%*--- Global boolean symbol CONTINUE, used with macro assert_continue(), warns user of invalid environment. Processing should HALT. ---*;
%let CONTINUE = %assert_depend(OS=%str(AIX,WIN,HP IPF),
vars=%str(&m_lb..&m_ds : &ana_variables),
macros=assert_continue assert_var_nonmissing assert_unique_keys util_labels_from_var
util_count_unique_values util_proc_template util_get_var_min_max util_value_format
util_boxplot_block_ranges util_axis_order util_delete_dsets,
symbols=m_lb m_ds t_var tn_var c_var b_var ref_trtn p_fl c_mode a_fl
max_boxes_per_page outputs_folder
%assert_continue(After asserting the dependencies of this script)
/*** Data Prep
1. Restrict analysis to SAFETY POP (&p_fl) and ANALYSIS OBS (&a_fl) -- Confirm NO MISSING change-from-baselines.
2. Plot requires 'Pooled' data with UNIQUE USUBJID for across-study results
3. Plot requires a Study Number variable for the X-Axis, to cluster boxes and stats
data css_safana;
set &m_lb..&m_ds (keep=&ana_variables where=(&p_fl = 'Y' and &a_fl = 'Y'));
%let c_mode = %upcase(&c_mode);
%*--- Expect non-missing change-from-baseline values ---*;
%let CONTINUE = %assert_var_nonmissing(css_safana, &c_var);
%assert_continue(After restricting analysis data - No missing values for "&c_mode" change-from-baseline in %upcase(&c_var))
data css_anadata;
set css_safana
css_safana (in=in_pool);
if in_pool then do;
*--- Leading hex-char 'A0'x forces 'Pooled' results to follow individual studies ---*;
studyid = 'A0'x !! 'Pooled';
substr(usubjid,1,1) = 'P';
*--- Create a Study Number, for plot X-Axis ---*;
proc sort data=css_anadata;
by studyid &tn_var;
data css_anadata;
set css_anadata;
by studyid &tn_var;
if first.studyid then studynum+1;
%*--- Expect exactly 1 obs per U-Subject per parameter, and analysis timepoint ---*;
%assert_unique_keys(css_anadata, studyid usubjid paramcd atptn)
%assert_continue(After creating pooled data - Unique records.)
/*** GATHER INFO for data-driven processing
Collect required information about these measurements:
Number, Names and Labels of PARAMCDs - used to cycle through parameters that have measurements
&PARAMCD_N count of parameters
&PARAMCD_VAL1 to &&&PARAMCD_VAL&PARAMCD_N series of parameter codes
&PARAMCD_LAB1 to &&&PARAMCD_LAB&PARAMCD_N series of parameter labels
%*--- Parameters: Number (&PARAMCD_N), Names (&PARAMCD_VAL1 ...) and Labels (&PARAMCD_LAB1 ...) ---*;
%util_labels_from_var(css_anadata, paramcd, param)
%*--- Number of studys: Set &STDYN from ana variable STUDYID (Include POOLED data) ---*;
%util_count_unique_values(css_anadata, studyid, stdyn)
/*** BOXPLOT for each PARAMETER and ANALYSIS TIMEPOINT in selected data
One box plot for each PARAMETER and ANALYSIS TIMEPOINT.
By Visit and Treatment.
In case of many visits and treatments, each box plot will use multiple pages.
TEMPLATE Positional parameter, the name of the template to compile.
DESIGNWIDTH Default is 260mm, suitable for one full-page landscape Letter/A4 plot.
130mm is suitable for these 2 side-by-side plots.
DESIGNHEIGHT Default is 170mm, suitable for one full-page landscape Letter/A4 plot.
CLEANUP Default is 1, delete intermediate data sets.
Set to 0 (zero) to preserve temp data sets from the final loop.
%macro boxplot_each_param_tp(plotds=css_anadata, cleanup=1);
%local pdx tdx css_pval_a css_pval_p;
%do pdx = 1 %to &paramcd_n;
/*** LOOP 1 *****************************************************
*** Loop through each PARAMETER, working with ALL TIMEPOINTS ***
data css_nextparam;
set &plotds (where=(paramcd = "&&paramcd_val&pdx"));
%*--- Analysis Timepoints for this parameter: Num (&ATPTN_N), Names (&ATPTN_VAL1 ...) and Labels (&ATPTN_LAB1 ...) ---*;
%util_labels_from_var(css_nextparam, atptn, atpt)
%*--- Y-AXIS alternative: Fix Y-Axis MIN/MAX based on all timepoints for PARAM. See Y-AXIS DEFAULT, below. ---*;
%*--- NB: EXTRA normal range reference lines could expand Y-AXIS range. ---*;
%* %util_get_var_min_max(css_nextparam, &c_var, aval_min_max) *;
%do tdx = 1 %to &atptn_n;
/*** LOOP 2 ********************************************************************
*** Loop through each TIMEPOINT for this parameter, working with ALL VISITS ***
*** NB: PROC SORT here is REQUIRED in order to merge on STAT details, below ***
proc sort data=css_nextparam (where=(atptn = &&atptn_val&tdx))
by studyid &tn_var;
%*--- Y-AXIS DEFAULT: Set Y-Axis MIN/MAX based on this timepoint. See Y-AXIS alternative, above. ---*;
%util_get_var_min_max(css_nexttimept, &c_var, aval_min_max)
%*--- Create format string to display MEAN and STDDEV to default sig-digs: &UTIL_VALUE_FORMAT ---*;
%util_value_format(css_nexttimept, &c_var)
%*--- Create macro variable BOXPLOT_BLOCK_RANGES, to subset visits into box plot pages ---*;
%util_boxplot_block_ranges(css_nexttimept, blockvar=studynum, catvars=&tn_var);
*--- Calculate summary statistics, KEEP STUDYNUM and TRT LABELS for plotting, below ---*;
proc summary data=css_nexttimept noprint;
by studyid &tn_var studynum &t_var;
var &c_var;
output out=css_stats (drop=_type_ _freq_)
n=n mean=mean std=std median=median min=datamin max=datamax q1=q1 q3=q3;
%*--- Add ANCOVA p-values for Endpoint: CHG = BASE + TRT + STUDY (if user specified a reference arm) ---*;
%if %length(&b_var) > 0 and %length(&ref_trtn) > 0 %then %do;
%let css_pval_a = css_pval_stdy;
%let css_pval_p = css_pval_pool;
*--- For the INDIVIDUAL STUDY DATA, model by STUDYNUM ---*;
ods select parameterestimates;
ods output parameterestimates = &css_pval_a;
proc glm data=css_nexttimept;
class &tn_var (ref="&ref_trtn");
by studynum;
where studyid ^= 'A0'x !! 'Pooled';
model &c_var = &b_var &tn_var / solution;
run; quit;
*--- For the POOLED DATA, include STUDYNUM as an independent effect ---*;
ods select parameterestimates;
ods output parameterestimates = &css_pval_p;
proc glm data=css_nexttimept;
class &tn_var (ref="&ref_trtn");
where studyid = 'A0'x !! 'Pooled';
model &c_var = &b_var &tn_var studynum / solution;
run; quit;
*--- Combine Indiv. Study and Pooled p-values. Keep only results for &TN_VAR parameters ---*;
data &css_pval_a;
set &css_pval_a (where=(parameter=:"%upcase(&tn_var)"))
&css_pval_p (where=(parameter=:"%upcase(&tn_var)") in=in_pool);
if in_pool then studynum = &stdyn;
*--- UPDATE CSS_STATS with p-values for active arms, and for Pooled ---*;
data temp;
*--- We simply need the structure of these vars, for subsequent merge ---*;
set css_stats (keep=studynum &tn_var);
data &css_pval_a;
set temp &css_pval_a (keep=studynum parameter probt
label pval="GLM ANCOVA p-value: Reference is %upcase(&tn_var) = &ref_trtn";
&tn_var = input(scan(parameter,-1,' '), best8.);
proc sort data=&css_pval_a;
by studynum &tn_var;
data css_stats;
merge css_stats &css_pval_a (keep=studynum &tn_var pval);
by studynum &tn_var;
STACK statistics (do NOT merge) BELOW the plot data, one obs per TREATMENT/VISIT.
NB: We need exactly ONE obs per timepoint and trt: AXISTABLE defaults to a SUM function
data css_plot;
set css_nexttimept
format mean %scan(&util_value_format, 1, %str( )) std %scan(&util_value_format, 2, %str( ));
*--- Graphics Settings - Set defaults for all graphs. Print missing P-VALUES as a space (not a dot) ---*;
options orientation=landscape missing=' ';
goptions reset=all;
ods graphics on / reset=all;
ods graphics / border=no attrpriority=COLOR;
title justify=left height=1.2 "Box Plot - &&paramcd_lab&pdx Change from &c_mode Baseline to &c_mode Post-Baseline Measure for Multiple Studies and Analysis Timepoint &&atptn_lab&tdx";
footnote1 justify=left height=1.0 'Box plot type is schematic: the box shows median and interquartile range (IQR, the box height); the whiskers extend to the minimum and maximum data points';
footnote2 justify=left height=1.0 'within 1.5 IQR of the lower and upper quartiles, respectively. Values outside the whiskers are shown as outliers. Means are marked with a different symbol';
footnote3 justify=left height=1.0 'for each treatment. P-value is for the treatment comparison from ANCOVA model Change = Baseline + Treatment (+ Study for "Pooled").';
%let y_axis = %util_axis_order( %scan(&aval_min_max,1,%str( )), %scan(&aval_min_max,2,%str( )) );
*--- ODS PDF destination (Traditional Graphics, No ODS or Listing output) ---*;
ods listing close;
ods pdf file="&outputs_folder\WPCT-F.07.07_Box_plot_&&paramcd_val&pdx.._change_&c_mode._base_post_by_study_for_timepoint_&&atptn_val&tdx...pdf"
notoc bookmarklist=none dpi=300
author="(&SYSUSERID) PhUSE CS Standard Analysis Library"
subject='PhUSE CS Measures of Central Tendency'
title="Boxplot of &&paramcd_lab&pdx Change from &c_mode Baseline to &c_mode Post-baseline Measure for Multiple Studies and Analysis Timepoint &&atptn_lab&tdx"
/*** LOOP 3 - FINALLY, A Graph ****************************
*** - Multiple pages in case of many visits/treatments ***
%local vdx nxtvis;
%let vdx=1;
%do %while (%qscan(&boxplot_block_ranges,&vdx,|) ne );
%let nxtvis = %qscan(&boxplot_block_ranges,&vdx,|);
proc sgrender data=css_plot (where=( %unquote(&nxtvis) )) template=PhUSEboxplot ;
_MARKERS = "&t_var"
_BLOCKLABEL = 'studyid'
_XVAR = 'studynum'
_YVAR = "&c_var"
_YLABEL = "&&paramcd_lab&pdx"
_YMIN = %scan(&y_axis, 1, %str( ))
_YMAX = %scan(&y_axis, 3, %str( ))
_YINCR = %scan(&y_axis, 5, %str( ))
_N = 'n'
_MEAN = 'mean'
_STD = 'std'
_DATAMIN = 'datamin'
_Q1 = 'q1'
_MEDIAN = 'median'
_Q3 = 'q3'
_DATAMAX = 'datamax'
_PVAL = 'pval'
%let vdx=%eval(&vdx+1);
%end; %* --- LOOP 3 - Pages of box plots, VDX ---*;
*--- Release the PDF output file! ---*;
ods pdf close;
ods listing;
%end; %*--- LOOP 2 - Time Points, TDX ---*;
%end; %*--- LOOP 1 - Parameters, PDX ---*;
*--- Clean up temp data sets required to create box plots ---*;
%if &cleanup %then %util_delete_dsets(css_safana css_nextparam css_nexttimept &css_pval_a &css_pval_p css_stats css_plot);
%mend boxplot_each_param_tp;
/*** END boxplotting ***/
You can’t perform that action at this time.