Computer-implemented methods for evaluating, summarizing and
Contents
1. File Name Size Date of Creation _bs_sub_alloc_vars rtf 9 KB 4 13 2006 _bs_sub_cale_dd ttf 15 KB 4 13 2006 _bs_sub_cale_pp rtf 24 KB 4 13 2006 _bs_sub_calc_pp_ancova rtf 11 KB 4 13 2006 _bs_sub_cale_pt rtf 25 KB 4 13 2006 _bs_sub_calc_sl rtf 13 KB 4 13 2006 _bs_sub_calc_kr rtf 9 KB 4 13 2006 _bs_sub_calculations rtf 98 KB 4 13 2006 _bs_sub_create_ds_modelways rtf 51 KB 4 13 2006 _bs_sub_create_ds_titlechecks rtf 7 KB 4 13 2006 _bs_sub_create_ds_validmodels rtf 31 KB 4 13 2006 _bs_sub_disp_control rtf 115 KB 4 13 2006 _bs_sub_expand_ds_af rtf 5 KB 4 13 2006 _bs_sub_get_const rtf 3 KB 4 13 2006 _bs_sub_graph rtf 63 KB 4 13 2006 _bs_sub_indata rtf 31 KB 4 13 2006 _bs_sub_indata_attrrtf 22 KB 4 13 2006 _bs_sub_indata_const rtf 9 KB 4 13 2006 _bs_sub_indata_graphs rtf 17 KB 4 13 2006 _bs_sub_indata_groups rtf 46 KB 4 13 2006 _bs_sub_initial rtf 24 KB 4 13 2006 _bs_sub_message rtf 33 KB 4 13 2006 _bs_sub_meta_file rtf 47 KB 4 13 2006 _bs_sub_meta_file_gr rtf 11 KB 4 13 2006 _bs_sub_mod_fune rtf 9 KB 4 13 2006 _bs_sub_model_cale rtf 49 KB 4 13 2006 _bs_sub_out_graph rtf 47 KB 4 13 2006 _bs_sub_out_graph_cal rtf 12 KB 4 13 2006 _bs_sub_parm_check rtf 37 KB 4 13 2006 _bs_sub_placeholder rtf 37 KB 4 13 2006 _bs_sub_prepare_cale rtf 9 KB 4 13 2006 _bs_sub_rep_modeltrans rtf 6 KB 4 13 2006 _bs_sub_rep_zerofill rtf 7 KB 4 13 2006 _bs_sub_report rtf 39 KB 4 13 2006 _bs_sub_resolve_parms rtf 16 KB 4 13 2006 _bs_sub_shrink_datase2. A computer loaded with a general statistical analysis software and a macro integrated with the software such that the computer is capable of performing the computer imple mented method according to claim 16 using the software and macro 34 A computer loaded with a general statistical analysis software and a macro integrated with the software such that the computer is capable of performing the computer imple mented method according to claim 17 using the software and macro 35 The method of claim 1 wherein data is provided from a stability study on one or more potentially stability limiting response variables of the pharmaceutical product over time where the pharmaceutical product of interest is stored in environmentally stable conditions to meet regulatory guide lines over a definable period 36 A computer loaded with a general statistical analysis software and a macro integrated with the software such that the computer is capable of performing the computer imple mented method according to claim 35 using the software and macro
3. BIGSTEP provides a complete statistical analysis of sta bility response data for estimating the shelf life of a new pharmaceutical compound allowing the fitting of various regression functions to characterize the time response rela tionship of a response variable A statistical analysis is pro vided to test the consistency of response among different batches and among the levels of an additional study factor such as product orientation or product packaging Based on user supplied specification limits or acceptance criteria a shelf life is estimated for the best fitted or user defined regres sion function allowing for pooling of batch or additional study factor response data BIGSTEP provides a complete statistical analysis of sta bility response data for confirming or supporting the pro posed shelf life of a marketed pharmaceutical product Fol lowing ICH guidelines batch response is evaluated to determine if the proposed shelf life can be supported If not a statistical analysis similar to that conducted to estimate a shelf life is performed BIGSTEP provides a standardized statistical analysis of data from a stability study The statistical analysis is con trolled by setting various options in the SAS analysis pro gram including the BIGSTEP macro call FIG 1 is a sche matic flow chart of the BIGSTEP analysis There are three sections to a BIGSTEP analysis INPUT Parameter and data verification ANALYSIS Statistical analysi
4. CV CV Row description for coefficient of variation DDS MIN Min Row description for minimum DDS MAX Max Row description for maximum DDS RF Regression function Column header for regression function TRT MDF Model DF Column header for degrees of freedom for model TRT EDF Error DF Column header for degrees of freedom for error TRT SSE SSE Column header for sum of squares error TRT MSE MSE Column header for mean squares error TRT CONV Converge status Column header for convergence status Yes No TRT CONV_YES Yes Column content for convergence status is case of convergence TRT CONV_NO No Column content for convergence status is case of convergence TRT LIN Simple linear TRG Graph title TRT name for linear time response relationship TRG TRT QUAD Quadratic TRG Graph title TRT name for quadratic time response relationship TRG TRT CUB Cubic TRG Graph title TRT name for cubic time response relationship TRG TRT EXP1 Exponential 1 TRG Graph title TRT name for exponential 1 time response relationship TRG TRT EXP2 Exponential 2 TRG Graph title TRT name for exponential 2 time response relationship TRG TRT EXP3 Exponential 3 TRG Graph title TRT name for exponential 3 time response relationship TRG TRT MODEL Model Column header for model SOURCE Source of variability Column header for source in pooling process table if table type is ANCOVA PPT INTERCEPTS Intercepts Column content for main fa
5. No change over time POOLED Pooled Column content for use in MF AF columns if pooling is possible PTT CI TMAX conf level Confidence interval Column header for confidence interval at at tmax maximum extrapolation time SLT SLS CI_PROPSL conf level Confidence interval Column header for confidence interval at proposed at proposed shelf life shelf life SLT SLS PILTMAX conf level Prediction interval Column header for prediction interval at maximum at tmax extrapolation time SLT SLS PI_PROPSL conf level Prediction interval Column header for prediction interval at proposed at proposed shelf life shelf life SLT SLS TI conf level Column header for tolerance interval SLT SLS Tolerance interval coverage coverage SUPPORTED Supported Column header for supported yes no column SLT SUPPORTED_YES Yes Column content for lines where shelf life can be supported SLT SUPPORTED_NO No Column content for lines where shelf life cannot be supported SLT CI TYPE Confidence Text containing only the interval type CI TI_TYPE Tolerance Text containing only the interval type TI PI_TYPE Prediction Text containing only the interval type PI INTERCEPT Intercept Column header for coefficient INTERCEPT SLT BETA beta Column header for coefficient beta SLT GAMMA gamma Column header for coefficient gamma SLT MAXTIME Maximum extrapolation time Column header for maximum extrapolation time tmax time unit SLT SLS PROPTIME Prop
6. Two SAS options are required given in the OPTIONS statement MSTORED is a SAS option that specifies that the macro facility search for stored compiled macros in the SASMACRO catalog of the SAS data library that is refer enced by the SASMSTORE option SASMSTORE specifies the libref ofa SAS data library that contains or will contain a catalog of stored compiled SAS macros This libref can not be WORK which is the SAS default temporary work directory In the example three temporary SAS datasets are used for a BIGSTEP analysis RESPONSES with data from per manent dataset STABILITY_DATA ATTRIBUTES and GROUPS These temporary datasets are defined for the cur rent BIGSTEP analysis in the analysis program and are resi dent in memory only during the current BIGSTEP session A SAS dataset with the stability data and the ATTRIBUTES dataset are necessary and have to be defined by the user The GROUPS dataset is optional and is not always necessary For the three mandatory background SAS datasets CON STANTS DEFAULTS and GRAPHS the permanent default data sets are used for this BIGSTEP analysis These datasets are stored in the same directory as the BIGSTEP macro library and are automatically reference by the BIGSTEP pro gram through LIBNAME BIGSTEP As discussed in the fol lowing sections it is usually not necessary to change them The stability analysis dataset to be statistically analyzed must be a permanent SAS da
7. Sided Level of Significance Selection Model Selection Necessory if No Pooling Process U S Patent May 8 2012 Sheet 1 of 2 US 8 175 843 B2 Input GL IN Ele DATA ATTRIBUTES GROUPS GRAPHS CONSTANTS Output OUT OUTG OUTT 0 Eee s DDA DDG TRC TRG PPC PIC PIG SLC SLG SLS DDS DDT TRT PPT PTS PTT SLT LST DOC RTF PTS SLX FIG 1 U S Patent May 8 2012 Sheet 2 of 2 US 8 175 843 B2 PROCESS RESULT TABLES RESULT GRAPHS Stability Data SAS Data Set Optional DD Individual Data Data Description Tables J Graphs Summary Statistics Additional Factor Tables Graphs Optional OR Time Response Time Response Relationship Tables Relationship Graphs Optional pp Pooling Process Tables Optional Pre Test les If change ETA E Se E a E Optional Further options Confidence or Prediction Intervals One Sided or Two Sided Level of a ie Selection Model Selection Necessary if No Pooling Process Ta Shelf Life Shelf Life Tables Graphs Tables and Graphs FIG 2 US 8 175 843 B2 1 COMPUTER IMPLEMENTED METHODS FOR EVALUATING SUMMARIZING AND PRESENTING DATA ON STABILITY OF DRUG SUBSTANCES AND DRUG PRODUCTS AND SOFTWARE MODIFIED COMPUTERS FOR SUCH METHODS Reference is hereby made to a computer program listing appendix submitted on a compact disc for this application An identical duplicate disc is also submitted A total of two compact discs are filed each of which contains the following files
8. as product orientation up right inverted or side storage or packaging type bottles or blister packs on sample response In these more involved studies there are two designs that can be statistically ana lyzed through BIGSTEP The two study designs are defined by how the initial or Storage Time 0 sample response data are obtained The first study design conveniently referred to as the independent sampling design has an independent sample being measured at Time 0 For example considering a stability study where product storage orientation is also of interest samples stored upright and samples stored in an inverted orientation are independently measured for response at Time 0 for each batch or lot Alternatively the second study design conveniently referred to as the common sam pling design measures response at Time 0 for both the upright and inverted samples for example for each batch while the pharmaceutical product is still in a common state For example for ease and practicality of handling the samples or to keep the overall sample size to a minimum the Time 0 response is measured from the common batch at batch release It would be a contrivance and technically impos sible to measure a response for upright and inverted samples US 8 175 843 B2 7 at a true Time 0 Thus a single common sample for each batch or lot is measured and recorded for both the upright and inverted Time 0 response The statistical issue di
9. directory for the stability study Let the root directory be C BIGSTEP Study although the root direc tory name can be anything the user prefers Under the root directory for example create three subdirectories named C BIGSTEP Study Data C BIGSTEP Study Program and C BIGSTEP Study Results for the study data analysis pro gram and study results respectively Following the separate directory structure discussed above C BIGSTEP Study Data is the subdirectory contain ing the stability dataset to be statistically analyzed through BIGSTEP The necessary file format for the dataset for BIG STEP is as a permanent SAS dataset Other file structures such as an Excel dataset are acceptable but require additional programming code to read into BIGSTEP The data file can be named any appropriate SAS dataset name for example STABILITY_DATA SAS7BDAT for a Version 8 2 SAS dataset The data file is called through the analysis program option IN_DATA Table 1 is a listing of the variables that can be included in the data file The particularly preferred variables to be included in the data file are ATTRIBUTE STORAGE BATCH TIME and LEVEL The BYVAR variable is included in the data file to allow separate statistical analyses for stability characteristics that are not of interest to be compared statistically such as product packag ing type The variable FACTOR is included in the data file for those stability studies which include an addi
10. evaluation and sum marization of data obtained from one or more stability studies on a pharmaceutical product This invention allows standard as well as non standard statistical analysis steps to be con ducted automatically It transforms the raw data into a useful display of the results which are preferably summarized in Word format for insertion into a statistical or study report Because of the automatic transfer of all output tables and graphs to Word the invention reduces the time needed to evaluate the data The invention also greatly reduces the han dling checking and report preparation efforts The invention results in harmonizing the statistical evaluations which are conducted and also the resulting report tables and graphs The result is a uniform presentation of stability data in stability study reports Such evaluated and presented results are par ticularly useful for submission to regulatory agencies e g the FDA for drug product approval Thus the invention is useful for facilitating and speeding internal processing of stability studies of drug products but also the regulatory agency review thereof The invention can be used for all statistical analyses of stability data including for example testing due to variation on the market authorization such as that required for regis tration purposes and on an ongoing basis for all commercial products It can also be used to support stability testing for different types of
11. is necessary due to a particular stability analysis or specific report requirements they can be changed in the BIGSTEP macro call as in the analysis program shown in Table 4 In case a modified option is to be used repeatedly for various stability analyses it might be more convenient to change the default setting in the dataset DEFAULTS or to use a modified dataset containing the needed defaults The defaults of the DEFAULTS dataset can be changed in 60 four ways The settings in the DEFAULTS dataset are changed and the modified dataset is resaved A second permanent dataset is created and referenced in the IN_SYSTEM_DEFAULTS option through the BIG 65 STEP macro call A temporary dataset is used and referenced in IN_SYS TEM_DEFAULTS US 8 175 843 B2 23 By changing the parameter options in the BIGSTEP macro call Again by changing the contents of that background SAS datasets consistency cannot be maintained Thus this is not preferred The GRAPHS dataset see Table 10 is a SAS dataset required by BIGSTEP By default the GRAPHS dataset is assumed to be stored in the directory referenced by LIB NAME BIGSTEP and named GRAPHS This background 24 dataset listing the variables ELEMENT SYMBOL SYM BOL_FONT SYMBOL_VALUE SYMBOL_HEIGHT LINE LINE_HEIGHT and COLOR see Tables 11 13 defines the defaults for the graph options used by BIGSTEP to produce the graphs in the file format EMF Enhanced Win dows Metafile It
12. level of by group attribute and storage condition instead of including all information as identifying columns in the output result table The invention can be applied to stability data for different combinations of Attributes e g Assay Impurities Additional factor levels e g levels of storage orientation like Upright and Inverted Storage conditions e g 25 C 60 RH 30 C 70 RH By variables combinations of the levels of different by variables are called by groups The main functions of invention are Group DD Presentation of individual data and summary statistics Group TR Comparison of different time response relationships 30 35 40 45 50 55 60 65 possible in the case of many main factor levels and or additional factor levels the Group SL Shelf life calculations with one or two sided confi dence or prediction intervals Support of proposed shelf life The FIG 2 flow chart is an overview of the integrated functionalities The individual data to be evaluated must be given as an SAS dataset A representative evaluation could follow the stages ofthe flow chart After a descriptive analysis DD with individual data tables summary statistic tables and graphs different time response relationships are com pared TR and the user has to decide which time response relationship is used in the following process The pooling process PP selects the most appropriate regressi
13. macro 27 A computer loaded with a general statistical analysis software and a macro integrated with the software such that the computer is capable of performing the computer imple mented method according to claim 10 using the software and macro 28 A computer loaded with a general statistical analysis software and a macro integrated with the software such that the computer is capable of performing the computer imple mented method according to claim 11 using the software and macro 29 A computer loaded with a general statistical analysis software and a macro integrated with the software such that the computer is capable of performing the computer imple mented method according to claim 12 using the software and macro 30 A computer loaded with a general statistical analysis software and a macro integrated with the software such that the computer is capable of performing the computer imple mented method according to claim 13 using the software and macro 31 A computer loaded with a general statistical analysis software and a macro integrated with the software such that the computer is capable of performing the computer imple mented method according to claim 14 using the software and macro 32 A computer loaded with a general statistical analysis software and a macro integrated with the software such that the computer is capable of performing the computer imple mented method according to claim 15 using the software and macro 33
14. S line type used for ELEMENT must be specified LINE_HEIGHT yes num height of line used for ELEMENT must be specified specified in point size COLOR yes char color used for ELEMENT must be specified US 8 175 843 B2 25 The GRAPHS defaults can be changed in three ways The settings in the GRAPHS dataset are changed and the modified dataset is resaved A second permanent dataset with a different name is cre ated and referenced in the IN DATA_GRAPHS option through the BIGSTEP macro call A temporary dataset is used and referenced in IN_DATA_ GRAPHS However it is again preferred to provide consistency that the settings not be changed TABLE 12 5 26 Group PP Model selection by pooling of batches and additional factor levels Use of common release data in pooling process Use of different time response relationships in pooling process Group PT Pre test to test if a formal statistical analysis is necessary Definition of Symbols and Lines Graphical element Comment AC In al the observation identified by AC In al observation identified by INTERVAL Ina INTERVAL SINGLELINE graphs all lines for acceptance criteria are plotted with the defined line layout given in graphs all intervals PI CI or TI are plotted with the defined line layout given in the graphs which plot only one line e g regression line or curve interpolation line this is plotted with the defined line and symbol layout given in the o
15. US008175843B2 az United States Patent 10 Patent No US 8 175 843 B2 Kubiak et al 45 Date of Patent May 8 2012 54 COMPUTER IMPLEMENTED METHODS 6 925 391 BL 8 2005 Pesce etal vce 702 21 FOR EVALUATING SUMMARIZING AND 7 469 390 B2 12 2008 Mickle etal 716 103 2003 0158670 Al 8 2003 Hougaard 702 19 PRESENTING DATA ON STABILITY OF 2006 0031022 Al 2 2006 Hougaard 702 19 DRUG SUBSTANCES AND DRUG PRODUCTS 2007 0022142 Al 1 2007 Palmer etal esses 707 200 AND SOFTWARE MODIFIED COMPUTERS FOR SUCH METHODS 75 Inventors Rene Kubiak Assmannshardt DE James Schwenke New Milford CT US Volker Krzykalla Ummendorf DE Hans Juergen Lomp Mittelbiberach DE Cornelia Schepers Mainz DE 73 Assignee Boehringer Ingelheim Pharma GmbH amp Co KG Ingelheim am Rhein DE Notice Subject to any disclaimer the term of this patent is extended or adjusted under 35 U S C 154 b by 653 days 21 Appl No 12 292 288 22 Filed Nov 14 2008 65 Prior Publication Data US 2010 0125434 Al May 20 2010 51 Int Cl GO6F 17 18 2006 01 52 USCh eiris 702 179 702 19 705 2 705 3 703 12 58 Field of Classification Search 0 0 702 19 702 179 705 2 3 703 12 See application file for complete search history 56 References Cited U S PATENT DOCUMENTS 6 532 427 B1 3 2003 Joshi etal o ae 702 84 6 766 319 B1 7 2004 Might 0 ccc 1 1 PROCESS Stability Data SAS Doto S
16. Water 252 C 007 RH Upright 000103 12 5229 Water 25 C 60 RH Inverted 000103 6 5218 Water 25 C 60 RH Upright 000103 18 5347 Water 25 C 60 RH Inverted 000103 9 5327 Water 25 C 60 RH Upright 000103 24 5523 Water 25 C 60 RH Inverted 000103 12 5344 gs Water 25 C 60 RH Inverted 000103 18 5369 Water 25 C 60 RH Inverted 000103 24 5394 o 0 H A 4 w i a T a ria Mics g 5 5 A stability study statistical analysis is conducted by a call ater f o RH prig f Water 25 C 60 RH Upright 000103 6 5186 to the BIGSTEP macro library through a SAS Version 8 2 Water 25 C 60 RH Upright 000103 9 5249 gq program The SAS program is referred to as the analysis Water 25 C 60 RH Upright 000103 12 5229 program The analysis program may be named any appropri Water 23 CI GOW RHS Upright 000103 18 2347 ate SAS program file name It is suggested that the analysis Water 25 C 60 RH Upright 000103 24 5523 Other dataset variables columns can be added in the data file but will not be used in the analysis For example for data 65 program file name extension should be SAS which is the naming structure expected but not mandated by SAS Let the analysis program file name be STABILITY SAS Fol lowing the proposed directory structure the analysis program file is to be stored in C BIGSTEP Study Program US 8 175 843 B2 11 Table 4 is a listing of the SAS code for a basic analysis program for co
17. al statistical analysis software and a macro integrated with the software such that the computer is capable of performing the computer imple mented method according to claim 3 using the software and macro 22 A computer loaded with a general statistical analysis software and a macro integrated with the software such that the computer is capable of performing the computer imple mented method according to claim 5 using the software and macro 23 A computer loaded with a general statistical analysis software and a macro integrated with the software such that the computer is capable of performing the computer imple mented method according to claim 6 using the software and macro 24 A computer loaded with a general statistical analysis software and a macro integrated with the software such that the computer is capable of performing the computer imple mented method according to claim 7 using the software and macro 25 A computer loaded with a general statistical analysis software and a macro integrated with the software such that 20 25 30 35 40 45 50 55 60 30 the computer is capable of performing the computer imple mented method according to claim 8 using the software and macro 26 A computer loaded with a general statistical analysis software and a macro integrated with the software such that the computer is capable of performing the computer imple mented method according to claim 9 using the software and
18. atistical or study report acceptable for submission to a regulatory agency for pharma ceutical product approval 9 The method of claim 1 wherein the macro is imple mented using the same language and commands used with the statistical analysis software 10 The method of claim 1 wherein the macro connects the data to be analyzed and defines particular options for the statistical analysis of the data and defines the output data and graphical presentation requirements 11 The method of claim 1 wherein statistical analysis of data from several response variables and storage conditions are conducted in one run 12 The method of claim 1 wherein the statistical analysis includes analyses of attributes of the pharmaceutical product including assay and impurity attributes additional factor lev US 8 175 843 B2 29 els including levels of storage orientation of the pharmaceu tical product and storage conditions of the pharmaceutical product 13 The method of claim 1 wherein the statistical analysis includes comparing of different time response relationships 14 The method of claim 1 wherein the statistical analysis includes model selection by pooling of batches and additional factor levels use of common or independent release data in a pooling process and use of different time response relation ships in a pooling process 15 The method of claim 1 wherein the statistical analysis and output includes graphical comparison of si
19. bservation identified by SINGLELINE REFLINE In al graphs all reference lines defined in the dataset ATTRIBUTES are plotted with the defined line layout given in the observation identified by REFLINE DATA lt n gt In al graphs that plot more than one line symbol these are plotted with the defined line and or symbol layout given in the observation identified by DATA lt n gt For each line lt n gt one DATA lt n gt is used one by one If there are not enough DATA lt n gt observations available which is macro uses the defined layout values again The BIGSTEP statistical analysis to be conducted and the format of the summary output report are controlled through a series of program options Each option is listed in the DEFAULTS permanent SAS dataset Where appropriate a default is set for many of the options noted in the option definition below The option default can be changed by speci fying the change in either a temporary dataset in the analysis program named DEFAULT or permanent DEFAULT SAS dataset or specified in the BIGSTEP call to the macro library Although the invention provides methods and apparatus which allow more harmonized stability reports the shown default tables and graphs can be modified as described in the additional information lists of this chapter to satisfy user and site specific requirements For example it is possible to pro duce separate tables for each evaluation group
20. ce line in line is displayed in every graph label for IN_REF_LINE1 only relevant if IN_REF_LINE1 is stated additional horizontal reference line in line is displayed in every graph label for IN_REF_LINE2 position for label of reference line only relevant if IN_REF__LINE2 is stated must be defined if IN REF_LINE1 2 and IN_REF_LINE1 2_ DESC are stated valid values LEFT RIGHT ATTRIBUTE presentation order for if not stated order is the same as in the Variable Mandatory Type Description ATTRIBUTE yes same as in analysis dataset IN_LEVEL_UNIT yes char IN_ACCEPT_LL no num IN_ACCEPT_UL no num IN_REF_LINE1 no num graphs IN_REF_LINE1__DESC no char IN__REF_LINE2 no num graphs IN_REF_LINE2__DESC no char IN_REF_POS no char IN_ANALYTICAL_PROCEDURE no char analytical procedure IN_ATTRIBUTE_SORT no num summary results output analysis dataset IN_ATTRIBUTE_DEC yes num number of decimal places in summary valid values 0 lt Integer lt 8 results used for raw data values for summary statistics mean standard deviation confidence limits prediction limits tolerance limits one additional decimal place is used calculations are performed without rounding An optional GROUPS dataset may contain the information about the statistical analysis to be conducted for each attribute or response variable as well as which summary information should be provided such as the regression model to charac terize the response
21. ctor dependent OVERALL_SLOPE Overall slope intercepts in table type ANCOVA PPT Column content for common slope source in table type ANCOVA PPT SLOPES Slopes Column content for independent slope source in table type ANCOVA PPT RESIDUAL Residual Column content for ERROR source in table type ANCOVA PPT DF DF Column header for degrees of freedom PPT PTT FV F value Column header for F value PPT PV p value Column header for p value PPT PTT ALPHA alpha Column header for alpha PPT If possible the greek symbol is used POOLING Pooling Column header for pooling Yes or No PPT POOLING_COMMON common Column content for positive pooling decision in table type ANCOVA PPT POOLING_IND individual Column content for negative pooling decision in table type ANCOVA PPT POOLING_YES Yes Column content for positive pooling decision in table type F tests PPT US 8 175 843 B2 21 22 TABLE 9 continued Options for Dataset CONSTANTS Identification Value Comment POOLING_NO Rejected Column content for negative pooling decision in table a F tests PPT SLOPE Slope Column header for slope PTT SE Standard error Column header for standard error PTT TV t value Column header for t value PTT COT Change over time Column header for change over time Yes No PTT COT_YES Yes Column content for decision Change over time COT_NO No Column content for decision
22. drugs e g drug substances or drug prod ucts as well as the status of the drugs life cycles e g first marketing authorization application or changes post approval It can be used for example to establish a retest period or to support or extend a proposed retest period for a drug substance or a shelf life for a drug product The system provided by the invention is a validated system that can be used for several stability study designs and time US 8 175 843 B2 3 response relationships and meets all regulatory requirements Thus not only standard designs and approaches are imple mented to be analyzed comfortably and quickly but also solutions which are not described in the literature for non standard designs The macro programming of the invention is designed to be user friendly with a standard statistical analysis software pro gram e g SAS Version 8 2 The language and commands used with the statistical analysis software are sufficient to control the macro of the invention The macro of the invention connects the data to be analyzed and defines particular options for statistical stability analyses as well as table and graph output requirements Thus various analysis output and layout options can be specified allowing for a variety of stability studies and layout designs to analyze data from sev eral response variables and storage conditions with one run Examples of the general advantages of the invention include
23. e a Time 0 response is measured independently for each storage orien tation from each batch Table 3 is an example of a common release design where the Time 0 measurement was obtained from batch release information and used as the Time 0 response for both the inverted and upright MDI canisters for that batch datasets 35 TABLE 3 Table 2 Is an examp le sho mga partial listing ofa dataset Example Dataset Format for Common Release Stability Study for a stability study for MDI Metered Dose Inhaler canisters stored at 25 C 60 RH The water content of the canister ATTRIB contents in parts per million was measured for several UTE STORAGE FACTOR BATCH TIME LEVEL batches over a 24 month study duration Two canister storage orientations upright and inverted were studied using are samples ieomeach batch Water 25 C 60 RH COMMON 000103 0 5049 P Water 25 C 60 RH Inverted 000103 3 5304 Water 25 C 60 RH Inverted 000103 6 5218 TABLE 2 45 Water 25 C 60 RH Inverted 000103 9 5327 Example Dataset Format for Independent Release Stability Study Water 25 C00 RH pier 000103 le 534 Water 25 C 60 RH Inverted 000103 18 5369 FAC Water 25 C 60 RH Inverted 000103 24 5394 ATTRIBUTE STORAGE TOR BATCH TIME LEVEL Water 25 C 60 RH Upright 000103 3 5124 50 Water 25 C 60 RH Upright 000103 5186 e o 0 i Water 25 C 60 RH Inverted 000103 0 5049 Water 2 ets ee OOOO ihe 3920 Water 25 C 60 RH Inverted 000103 3 5304
24. e valid If possible parameter checks are performed only if they are necessary Therefore the parameter GL_DISPLAYS must be checked very early since the value of this parameter defines whether other parameters are conditional mandatory or not The parameter check handles as many parameters as pos sible If several parameters with invalid values are handed over to the macro various enor messages appear If the check 20 30 40 45 55 60 65 28 ofa main parameter such as GL_DISPLAYS fails the macro cannot check all given parameters because many parameters are dependent on the chosen displays In case of severe errors as for example invalid values for parameters or missing information in datasets the macro stops with error message s BRIEF DESCRIPTION OF THE DRAWINGS FIG 1 is a schematic flow chart of the BIGSTEP analysis FIG 2 flow chart is an overview of the integrated function alities of the program We claim 1 A computer implemented method using a macro inte grated into statistical analysis software for statistical analysis and summarization of a stability study on a pharmaceutical product which comprises inputting and summarizing data observed for a stability study on a pharmaceutical product statistically analyzing the data including using at least one regression function to characterize the time response relationship of at least one stability response variable and to estimate a shelf life of
25. efined by parameter values are not deleted IN_LEVEL_UNIT The analysis dataset variable names listed in Table 1 BYVAR ATTRIBUTE STORAGE FACTOR BATCH TIME and LEVEL are the default for a BIGSTEP analysis User defined variable names can be used in the analysis dataset However if variable names other than the default names are used the user defined variable names must be designated in the analysis program through the IN_VARS option and the user defined names must be used consistently throughout the analysis program Labels can be assigned to the variables through the analysis dataset BIGSTEP does not use any variable formats defined directly to the input dataset Formats are defined by the user through the VALUE statement in PROC FORMAT The for mat name is then specified in the analysis program through the option IN_FORMATS Formats specified in the IN_FOR MATS statement are used for the summary output and output 20 25 30 automatically recording purposes a column for sample replicate number could be added if replicate measurements were recorded for each sample As discussed above there are two basic stability study designs for studies involving an additional study factor inde pendent release designs and common release designs The different study designs are differentiated by how the initial or Time 0 response is measured for each batch Table 2 is an example of an independent release design wher
26. empty A second permanent dataset with a different name is cre ated and referenced in the IN SYSTEM_CONSTANTS option through the BIGSTEP macro call A temporary dataset is used and referenced in IN_SYS TEM_CONSTANTS However BIGSTEP was designed to be a globally harmo nized stability analysis program providing a consistent shelf US 8 175 843 B2 19 20 life analysis and summary report By changing the contents of that background SAS datasets that consistency cannot be maintained and is thus not preferred TABLE 9 Options for Dataset CONSTANTS Identification Value Comment AC Acceptance criteria Column header for acceptance criteria Different tables AP Analytical procedure Column header for analytical procedure DDT DDS sc Storage condition Column header for storage condition Different tables OBS Obs Column header for observation number DDT TIME Time time unit Common column header for all time columns DDT MISSVALUE n d Content for missing values in input dataset DDT STATS Summary statistics Column header for names of statistics N Mean DDS N N Column header row description for N DDS MEAN Mean Column header row description for mean DDS SD SD Column header row description for standard deviation DDS NC not calc If statistics cannot be calculated because N lt 1 this text is displayed DDS If data are constant and shelf life cannot be calculated this text is displayed SLS
27. et RESULT TABLES FOREIGN PATENT DOCUMENTS JP 2000010793 A 1 2000 OTHER PUBLICATIONS Krzykalla Volker et al BIGSTEP Boehringer Ingelheim Global Stability Testing Evaluation Program User s Manual Version 1 1 Boehringer Ingelheim 2005 1 138 cited by examiner Primary Examiner Carol Tsai 74 Attorney Agent or Firm Michael P Morris Mary Ellen M Devlin 57 ABSTRACT Computer implemented methods for statistical analysis and summarization of a stability study on a pharmaceutical prod uct using of a macro integrated into statistical analysis soft ware The method includes inputting and summarizing data observed for a stability study on a pharmaceutical product statistically analyzing the data including using at least one regression function to characterize the time response rela tionship of at least one stability response variable to estimate a shelf life of the pharmaceutical product or confirm the shelf life of an existing pharmaceutical product and providing standardized output data and graphical presentations of observed stability response data Also a computer loaded with a general statistical analysis software and a macro inte grated with the software such that the computer is capable of performing these computer implemented methods using the software and macro 36 Claims 2 Drawing Sheets RESULT GRAPHS Further options Confidence or Prediction Interva s One Sided or Two
28. fferentiating between the indepen dent and common sampling designs is one of sample size and thus degrees of freedom Considering the example where storage orientation is an additional factor to be considered in a stability study the independent sampling stability design would have two samples observed at Time 0 for each batch or lot corresponding to the upright and inverted storage orien tations However the common sampling stability design would have only a single sample observed at Time 0 for each batch with the single sample to be simultaneously used as the Time 0 response for both the upright and inverted storage orientations For stability study designs where an additional study factor is considered independent or common sampling designs are differentiated for a BIGSTEP analysis through appropriately defining the analysis data file The statistical analysis of stability study data is based on regression analysis methodology In regression analysis a regression model or function is defined to adequately char acterize the observed data for a particular stability response variable Four regression models which have been proven useful for the statistical analysis of stability data are available through BIGSTEP simple linear quadratic first order expo nential and second order exponential models The details of the statistical analyses conducted through BIGSTEP are dis cussed in the Users Manual which is incorporated by refer e
29. gnificant change in response over storage time is not detected a simple scatter plot of the response data is produced The first step in a stability statis tical analysis is to determine the best fitted regression model to be used to characterize response Depending on the type of an 5 25 30 40 45 50 55 6 regression model selected by the user a statistical Pooling Process PP is conducted to determine if differences among batch or lot response or differences among the categories of the additional study factor if an additional study factor was considered in the stability study can be detected If differ ences in response cannot be detected the data among batches and or the additional study factor are pooled together and characterized by a more simple regression model The stabil ity statistical analysis is concluded by estimating a Shelf Life SL based on the appropriate regression model The results of the stability statistical analysis are reported through the output phase of the BIGSTEP analysis The for mat and content of the various summary tables and graphical displays are controlled through various options and param eters set in the BIGSTEP analysis program Summary results can be displayed on the computer screen or saved as files in either list RTF or Word formats Results saved in either RTF or Word formats are suitable for insertion into formal stability study summary reports The basic stabili
30. he format of the ATTRIBUTES dataset follows conventional SAS dataset formatting An example of an ATTRIBUTES dataset is given in Table 5 A LIBNAME needs to be created in the analysis program or one of the existing LIBNAMEs can be used if the permanent ATTRIBUTES dataset is stored in the appropriate directory The permanent dataset ATTRIBUTES is then referenced in the analysis program in a dataset statement similar to the data statement in the example analysis program in Table 4 Similar to the example analysis program in Table 4 additional ATTRIBUTES options can be defined in the dataset state ment US 8 175 843 B2 13 TABLE 4 14 Example Dataset Format for ATTRIBUTES ATTRIBUTE IN _LEVEL_UNIT IN_ACCEPT_LL IN _ACCEPT_UL IN ANALYTICAL PROCEDURE ASSAY 90 105 HPLC IMP an 0 5 WATER pap 25 37 ATTRIBUTE IN_ATTRIBUTE_SORT IN_ATTRIBUTE_DEC ASSAY 2 2 IMP 1 3 WATER 3 0 TABLE 5 Options for Dataset ATTRIBUTES name of attribute to be evaluated upper acceptance criterion for attribute Comments values must not be missing values must be the same as in RESPONSES dataset attribute measurement unit lower acceptance criterion for attribute IN__ACCEPT_LL or IN must be defined if IN_ACCEPT_LL and IN_ACCEPT_UL are stated LL must be less than UL IN__ACCEPT_LL or IN_ACCEPT_UL must be defined if IN_ACCEPT_LL and IN_ACCEPT_UL are stated LL must be less than UL ACCEPT_UL additional horizontal referen
31. hing is defined 1 minor tick is the default The variable is only relevant for graphs The CONSTANTS dataset see Table 8 is a SAS dataset required by BIGSTEP By default the CONSTANTS dataset is assumed to be stored in the directory referenced by LIB NAME BIGSTEP and named CONSTANTS This back ground dataset listing the three variables ID VALUE and COMMENT see Table 9 defines the values for constant texts like table headers or line descriptions in graphs used by BIGSTEP and should not be changed TABLE 8 Variables of Dataset CONSTANTS Variable Mandatory Type Description ID Identifier yes Char Identification of the constant VALUE yes Char Value which is substituted for the constant COMMENT no Char Comment with description of relevance Ifa change of constants is necessary the corresponding text in variable VALUE must be changed For texts defining a column header or column content the sign can be used for indicating line breaks The variable ID is used as identifica tion of the constant and must not be changed Only the defined IDs are used by the macro If an expected ID as for example AC cannot be found in the datasets the macro uses the 40 default ID instead of the VALUE and prints a message in the LOG The CONSTANTS defaults can be changed in three ways The settings in the CONSTANTS dataset are changed and the modified dataset is resaved Value check Must not be empty Must not be
32. if a formal statistical analysis is neces sary can be conducted Four non linear functions Simple Linear Quadratic Two exponential Functions can be used for statistical evalu ation a model selection pooling process if an additional factor should be considered for pooling can be defined on basis of analysis of covariance techniques and imple mented common or individual release data can be considered in statistical model two different model selection approaches on full or reduced model can be used for standard or non standard evaluations The following description is provided as an example for installing and running a specific macro called BIGSTEP into a SAS statistical analysis program to provide a system for carrying out the invention A copy of the User s Manual for BIGSTEP is filed with the application and is incorporated by 0 a 5 20 25 35 40 45 50 55 60 4 reference into this application Further the computer code for the BIGSTEP macro is provided as a computer program listing appendix in the discs as discussed above It is to be understood however that this is only an example of one embodiment of the invention and the invention is not limited to this specific description For example according to the invention a different macro for performing substantially the same steps could be provided for performing the invention in connection with a statistical analysis program incl
33. ine polled for slopes and intercepts Before a stability data evaluation is started several data and parameter checks to avoid an incorrect use are performed If the data are not appropriate the parameter settings inconsis tent or incomplete the program according to the invention can produce detailed ERROR and WARNING messages which are described in this chapter An ERROR message usually stops the macro whereas a WARNING does not If an ERROR message appears the detailed information given helps to find the cause of the problem After solving the problem by modification of the data or parameter setting the program can be started again For parameter check the macro performs the following steps 1 Setting of all empty parameters to default 2 Setting of all parameters with value NONE to empty 3 Start of parameter check and processing Thus if one wants to set a parameter to empty NONE has to be used For example DDS_TITLE3 should be set to empty the syntax DDS_TITLE3 NONE instead of DDS_TITLE3 must be defined Otherwise BIGSTEP uses the default setting Usually the parameters are upcased and dequoted Titles footnotes and filenames are typical examples for case sensi tive parameters The parameter checks are 1 Check whether all mandatory parameters are filled 2 Check whether all conditional mandatory parameters are filled if the condition is true 3 Check whether parameter contents and parameter com binations ar
34. is intended that this dataset is to be modi fied to accommodate the requirements of the statistical analy sis and study report TABLE 10 Default GRAPHS Dataset ELEMENT SYMBOL SYMBOL_FONT SYMBOL_VALUE SYMBOL_HEIGHT LINE LINE_HEIGHT COLOR AC 21 1 red REFLINE 1 green INTERVAL 4 1 black SINGLELINE DOT 4 1 ii blue DATA1 WINGDINGS 8C x 9 1 t blue DATA2 WINGDINGS 8D x 9 1 1 green DATA3 WINGDINGS 8E x 9 1 1 red DATA4 WINGDINGS 8F x 9 1 1 black DATAS WINGDINGS 90 x 9 1 1 gray DATA6 WINGDINGS 91x 9 1 1 cyan DATA7 WINGDINGS 92 x 9 1 1 red TABLE 11 Variable Definitions for Dataset GRAPHS Variable Mandatory Type Description Comments ELEMENT yes char graph element to be formatted the following elements rows must be defined AC REFLINE INTERVAL SINGLELINE additional elements are DATA lt n gt SYMBOL yes char SAS symbol used for ELEMENT values can be DOT CIRCLE etc SYMBOL_FONT yes char font and value used for ELEMENT if SYMBOL is specified SYMBOL_VALUE SYMBOL_FONT and SYMBOL_VALUE must both be missing SYMBOL_FONT and SYMBOL_VALUE must both be missing or both be specified SYMBOL_HEIGHT _ yes num height of symbol used for ELEMENT must not be missing if SYMBOL or SYMBOL_FONT is specified specified in point size symbol size is used for mean values and individual data if a graph contains mean values and individual data the individual data are displayed smaller LINE yes num SA
35. levels the most parsimonious system of equations to characterize change in response Starting with a model fitting US 8 175 843 B2 27 individual coefficients for each batch or additional factor most complex model the system checks step by step the poolability of coefficients across batches or additional fac tors The invention is amenable to pooling strategies for several situations where the number of models included in the pool ing process is defined by the design and evaluation strategy of the study Number of considered models in the pooling process is defined by two study design questions Is an additional factor to batch considered for pooling or not Defined by parameter IN_ ADD_FACTOR Are common release data used or not Defined by parameter GL_COMMON_RELEASE Number of considered models in the pooling process is defined by two study evaluation strategy questions Are slopes tested before intercepts pooling without order or not Defined by parameter PPC_SLOPES_FIRST Is the additional factor considered for pooling before batch Defined by parameter PPC_AF_FIRST The different combinations of these criteria define the number of models considered in the pooling process The sections charts and plots in the User s Manual depict the strategy for the simple linear approach indicating the possible paths for the pooling process from the highest most complex model to the lowest model with just one regression l
36. nce herein Finally the analysis program file calls the BIGSTEP SAS macro library of programs the SAS dataset to be analyzed and provides the appropriate options for the statis tical analysis The SAS analysis program file can be named any acceptable SAS filename and can be saved in any direc tory convenient for the user The format of the analysis pro gram file is discussed extensively throughout the User s Manual To install BIGSTEP for example a directory is created for the BIGSTEP SAS programs and Word template C BIGSTEP Program Files System To conduct a BIGSTEP 10 15 20 25 30 35 40 8 analysis ofa stability study it is recommended that a separate directory structure be created for the study data analysis program and summary results for each study This is because any number of stability studies can be statistically analyzed with one installation of BIGSTEP The separate directory structure is a pragmatic suggestion for maintaining the integ rity of the different study data analysis and result files One option is to create one additional directory to store the data analysis and result files together To demonstrate the options available through BIGSTEP three additional directories can be created to store the data analysis and result files separately Following the suggestion to create separate directories for the stability study data analysis program and analysis results create a root
37. nducting a complete stability statistical analy sis with the objective of estimating a product shelf life The data are from a 24 month stability study of Atrovent HFA metered dose inhalers For mere exemplification the pre sented analysis is being limited to four response variables each of which have potentially stability limiting characteris tics canister weight loss total canister assay canister citric acid content canister water content In this stability study MDI canisters were stored under two storage conditions 25 C 60 RH and 30 C 70 RH For each storage condition responses to the canister being stored in an upright and inverted orientation were studied Response was measured at 0 3 6 9 12 18 and 24 months of storage Because of the complexity of defining samples for canisters stored at a storage condition and orientation at Time 0 batch release measurements were used instead Thus the response at Time 0 for each batch is the sample recording for both the upright and inverted storage conditions This defines the com mon release stability study A partial data listing for this study is given in Table 3 The analysis program file STABILITY SAS begins by defining two LIBNAMEs A LIBNAME is a SAS state ment which sets a path to a specific directory LIBNAME DATA points to the directory where the stability data file is stored LIBNAME BIGSTEP points to the directory where the BIGSTEP macro library is stored
38. nse variables RESPONSES is then specified as the analysis dataset through the IN_DATA option in the BIGSTEP call as IN_DATA RESPONSES The ATTRIBUTES dataset that has to be defined by the user contains the information about the characteristics of the attributes or response variables such as the unit of measure ment upper and lower acceptance criteria and the order the attributes are to be presented in the stability summary report A complete list of the preferred and other available options through the ATTRIBUTES dataset is given in Table 6 The attribute information is passed to BIGSTEP as the value of SAS variables in the ATTRIBUTES dataset through the IN_DATA_ATTRIBUTES option in the BIGSTEP call to the macro library The dataset ATTRIBUTES can be defined in two ways either as part of the analysis program or as a permanent SAS dataset referenced in the analysis program As an example the ATTRIBUTES dataset is defined in the analysis program given in Table 4 The names used for the default variables ATTRIBUTE and STORAGE in the RESPONSES dataset must be the same used in the ATTRIBUTES dataset Not all options available through the ATTRIBUTES dataset need to be specified Variable labels and formats cannot be set in the ATTRIBUTES dataset Alternatively for those options that are to be used repeat edly for several stability analyses ATTRIBUTES can be defined as a permanent SAS dataset stored in a convenient directory T
39. on model With respect to this model a pre test PT can be conducted If the pre test is chosen the system decides dependent on the change over time to stop the formal evaluation or to move on If either the result of the pre test is that there is a change over time or if otherwise the pre test is not chosen the shelf life SL is calculated All results can be presented in tables and graphs and transferred to e g a Word file If the aim is to support a given shelf life the program first checks whether the pre defined proposed shelf life can be supported with the start model individually per batch and additional factor This check is performed with pre test and shelf life calculation without pooling Ifthe shelf life cannot be supported pooling is performed pooling can be limited to a given level and pre test and shelf life calculation are cal culated again with the pooled model Since the estimated shelf lives are usually longer when the data basis is broader the aim of the pooling process is to combine data from different batches main factor levels and or different levels of the additional factor for example stor age orientation or package material The storage conditions are not considered in the pooling process the analyses are always performed separately for the different storage condi tions The pooling process is a hierarchical process in order to identify for an attribute across main factor levels or additional factor
40. or DDA_SELECT groups related display is selected DDG_SELECT If for example TRC_SELECT is switched to The values must be YES NO Y N DDS_SELECT NO for all groups attributes by groups of case insensitive or empty Only TRC_SELECT storage condition 5 the time response NONN has the deselection effect PPC_SELECT displays are not created for these storage for the considered group PTC_SELECT condition groups YES Y has the same meaning SLC_SELECT All displays selected by parameters as missing GL_DISPLAYS and GL_SHORT_REPORT are created despite of the displays deselected here GL_FUNCTION no char User defined types of regression functions for Same valid values as the parameter pooling process pre test and shelf life calculations overwrites homonymous parameter for the considered group GL_MODEL no char User defined model for shelf life calculations Same valid values as the parameter overwrites homonymous parameter for the considered group GL_EXTRAPOLATION_LIMIT no num Extrapolation limit Same valid values as the parameter overwrites homonymous parameter for the considered group US 8 175 843 B2 17 18 TABLE 7 continued Options for Dataset GROUPS Variable Mandatory Type Description Comments SLC_PROPOSED_SL no num Proposed shelf life If the general parameter overwrites homonymous parameter for the SLC_PROPOSED_SL is not set considered group either all or none of the values must be set For all gro
41. osed shelf life time Column header for proposed shelf life time SLT unit SLS SL Shelf life time unit Column header for calculated shelf life SLT NOCOT No change over time Text displayed in shelf life column if pretest indicated no change over time SLS CONSTANT constant Text displayed in model column if SD lt 10e 6 GRAPH_AC_LL GRAPH_AC_DL LANDSCAPE _MARGIN_UPPER LANDSCAPE MARGIN_LOWER LANDSCAPE MARGIN_LEFT LANDSCAPE MARGIN_RIGHT PORTRAIT_MARGIN_UPPER PORTRAIT_MARGIN_LOWER PORTRAIT_MARGIN_LEFT PORTRAIT_MARGIN_RIGHT PAPER _ FORMAT Acceptance criterion Acceptance criterion 2 54 2 54 3 42 3 42 3 42 3 42 2 54 2 54 A4 SLS By change of these constants the labeling of the acceptance criterion line can be influenced relevant for all graphs Margins depending on paper orientation defined by parameter OUT_ORIENTATION unit is cm All values NE A4 are interpreted as letter A4 format is 21 29 7 cm Letter format is 21 59 27 94 cm By use of the paper format and the margins the size of the graphs is calculated 55 The DEFAULTS dataset is a SAS dataset required by BIGSTEP By default the DEFAULTS dataset is assumed to be stored in the directory referenced by LIBNAME BIGSTEP and named DEFAULTS This background dataset listing the three variables PARAMETER DEFAULT and COMMENT defines the default settings for every BIGSTEP analysis and output option If a change of the parameter setting
42. s OUTPUT Output of result tables and graphs In the input or verification phase of a BIGSTEP analysis internal program checks are made on the analysis program and the DEFAULTS dataset If an error is detected the analy sis is stopped and an error message is reported The analysis dataset here called STABILITY for convenience in addition to the BIGSTEP system datasets CONSTANTS and GRAPHS and the system input datasets ATTRIBUTES and GROUPS are verified Once the input datasets are verified the analysis phase begins Through the setting of various analysis options and parameters in the BIGSTEP analysis program several differ ent analysis objectives can be achieved There are five major analysis components to the BIGSTEP analysis DD Data Description TR Time Response regression model PP Pooling Process for regression model PT Pre Test evaluation SL Shelf Life estimation The Data Description DD component provides for data listings summary statistics and plots of the observed data The Time Response TR component allows for investigating different regression model fits to the observed response data Summary information is provided through tables and graphi cal displays The Pre Test PT is a statistical evaluation to determine if there is a significant change in response over storage time If a significant change in response over storage time is detected a full stability analysis is then conducted through BIGSTEP If a si
43. s application Pharmaceutical products must be intensively tested to guarantee high safety for patients Quantifying the stability of a drug to define a shelf life is one of the objectives A statis tical analysis of stability studies on pharmaceutical products can be conducted using analyses of covariance techniques by fitting confidence intervals for mean responses represented by regression lines or curves Several time consuming steps are involved in these analyses including the preparation of result tables and graphs From the ICH International Con ference on Harmonization June 2004 guideline Q1E sets forth the needed statistical analyses for stability studies and the invention is particularly provided as a means for comply ing with this guideline The invention provides a validated system of computer implementable programming for use together with statistical analysis software For example the programming can be used with SAS Version 8 2 system software higher versions of this software or software which performs essentially the same functions as this software Validated means that checks and tests were performed to be sure that everything works as it was defined in the user requirements document The pro gramming is preferably in the form of a macro that is usable with the base statistical analysis software A computer loaded with the base statistical analysis software and a macro accord ing to the invention can be used for the
44. statement in the example analysis program 15 US 8 175 843 B2 statement TABLE 6 16 in Table 4 Similar to the example analysis program in Table 4 additional GROUPS options can be defined in the dataset Example Dataset Format for GROUPS ATTRIBUTE STORAGE GL_FUNCTION GL_MODEL GL_EXTRAPOLATION_LIMIT ASSAY 25 C 60 RH EXP1 ASSAY 30 C 70 RH EXP1 ASSAY 40 C 75 RH EXP1 24 IMP 25 C 60 RH IMP 30 C 70 RH IMP 40 C 75 RH M_1_ 24 ASSAY 25 C 60 RH EXP1 ASSAY 30 C 70 RH EXP1 ASSAY 40 C 75 RH EXP1 24 IMP 25 C 60 RH IMP 30 C 70 RH IMP 40 C 75 RH M_1_1 24 ATTRIBUTE OUTG_XAXIS_END OUTG_XAXIS_BY OUTG_XAXIS_MINOR ASSAY ASSAY ASSAY 25 2 IMP IMP IMP 25 2 ASSAY ASSAY ASSAY 25 2 IMP IMP IMP 25 2 TABLE 7 Options for Dataset GROUPS Variable Mandatory Type Description Comments lt BYVAR gt yes same as levels of variable to be analyzed separately defines data group to be evaluated if analysis defined by parameter IN_BY_VARS values must be the same as in BYVAR dataset RESPONSES dataset is used in analysis dataset ATTRIBUTE yes same as name of attribute to be evaluated defines data group to be evaluated analysisdataset values must be the same as in RESPONSES dataset STORAGE yes char storage condition defines data group to be evaluated values must be the same as in RESPONSES dataset DDT_SELECT no char Selection of output displays for single Only relevant if the respective
45. t rtf 5 KB 4 13 2006 _bs_sub_single_check rtf 64 KB 4 13 2006 _bs_sub_start_vals rtf 22 KB 4 13 2006 _bs_sub_system_defaults rtf 10 KB 4 13 2006 _bs_sub_title_check rtf 24 KB 4 13 2006 _bs_sub_value_in_list rtf 11 KB 4 13 2006 _bs_sub_y_scaling rtf 11 KB 4 13 2006 _xx_entimo_check_file_exist 1tf 6 KB 4 13 2006 _xx_entimo_check_file_read rtf 5 KB 4 13 2006 _xx_entimo_check_fileref rtf 5 KB 4 13 2006 _xx_entimo_check_library_exist rtf 4 KB 4 13 2006 _xx_entimo_check_num_value rtf 10 KB 4 13 2006 _xx_entimo_check_quoted tf 6 KB 4 13 2006 _xx_entimo_check_sas_aspects rtf 61 KB 4 13 2006 _xXx_entimo_cnt_quoted_items rtf 4 KB 4 13 2006 _xx_entimo_delete_in_library rtf 3 KB 4 13 2006 _xx entimo descr stats rtf 41 KB 4 13 2006 _xx_entimo_export_data rtf 8 KB 4 13 2006 _xXx_entimo_get_extract rt 4 KB 4 13 2006 _xx_entimo_get_gitem rtf 6 KB 4 13 2006 _xx_entimo_get_unique_list rtf 5 KB 4 13 2006 _xx_entimo_import_data rtf 7 KB 4 13 2006 _xx_entimo_print_parm rt 6 KB 4 13 2006 _xx_entimo_sel_excl rt 14 KB 4 13 2006 _xx_entimo_start_stop rtf 8 KB 4 13 2006 _xx_entimo_string_with_delim rtf 7 KB 4 13 2006 _Xx_entimo_unquote rt 4 KB 4 13 2006 5 10 20 25 30 35 40 45 50 55 60 65 continued File Name Size Date of Creation _xx_entimo_update_string rtf 9 KB 4 13 2006 _xx_entimo_wordcount rtf 7 KB 4 13 2006 Version rtf 1 KB 4 13 2006 The material in the compact disc is hereby incorporated by reference into thi
46. taset stored in a convenient directory If the analysis dataset requires no further data man agement it can be specified directly through the IN_DATA and IN_DATA_PATH options in the BIGSTEP call to the macro library for example IN_DATA_PATH C BIGSTEP STUDY DATA and IN_DATA STABILITY_DATA following the example analysis dataset name defined above Of course the analysis dataset can be named any acceptable SAS dataset name 20 25 30 35 40 45 50 55 60 65 12 Alternatively the directory the analysis dataset is stored in can be referenced through e g LIBNAME DATA with the IN_DATA option in the BIGSTEP call being IN_DATA DATA STABILITITY_DATA If further data management is required prior to analysis such as setting variable labels or partitioning the dataset the analysis dataset can be defined through a data step The tem porary analysis dataset defined through the data step can be named any acceptable SAS dataset name A data step is used to define the analysis dataset RESPONSES in the example analysis program listed in Table 4 Using the study analysis dataset STABILITY_DATA labels are defined for the dataset variables to be used in the BIGSTEP analysis Further data MANAGEMENT may be conducted through the data step For example because BIGSTEP analyzes all response variables included in the analysis dataset the study analysis dataset may be partitioned to include only a selection of the respo
47. the following performing all standard statistical analyses need for drug stability studies including attributes e g assay impu rities additional factor levels e g levels of storage orientation like Upright and Inverted Storage con ditions e g 25 C 60 RH 30 C 70 RH and By variables combinations of the levels of different by variables are called by gro presentation of individual data and summary statistics comparing of different time response relationships model selection by pooling of batches and additional factor levels use of common release data in pooling process and use of different time response relationships in pool ing process pre testing to determine if a formal statistical analysis is necessary shelf life calculations with one or two sided confidence or prediction intervals and support of proposed shelf life transfer of all result tables and graphs automatically into MS Word format defining of all details for table layout graph layout and statistical evaluation by the user short runtime and a validated system Other optionally applicable advantages include the following non standard statistical analyses six non linear time response functions can be compared graphically and in a table that lists the functions ranked by the best fitting in addition to a three model approach a theoretical possible fourth model can be included into the analysis a pre test to check
48. the pharmaceutical prod uct or confirm the shelf life of an existing pharmaceuti cal product and providing standardized output data and graphical presen tations of observed stability response data 2 The method of claim 1 wherein the standardized output data and graphical presentations are output in a format suit able for standard word processing documents 3 The method of claim 1 further comprising performing a statistical analysis to test the consistency of response among different batches and or among the levels of at least one additional stability factor variable 4 The method of claim 3 wherein the at least one addi tional stability factor variable is product orientation or prod uct packaging 5 The method of claim 1 wherein the shelf life is estimated based on user supplied specification limits or acceptance criteria and using an automated best fitted or user defined regression function 6 The method of claim 1 further comprising automated pooling of batch or additional study factor response data following a pre defined pooling strategy 7 The method of claim 1 wherein the statistical analysis is conducted using analyses of covariance techniques by fitting any desired confidence or prediction interval for mean responses represented by regression lines or curves 8 The method of claim 1 wherein the standardized output data and graphical presentations are output in a format suit able for direct insertion into a st
49. tional study factor that is of interest to compare statistically such as prod uct storage orientation TABLE 1 Analysis Dataset Variable Definitions Variable Mandatory Type Description Comments BYVAR no any levels of variable to be values must not be missing analyzed separately defined by parameter IN_BY_VARS ATTRIBUTE yes any name of attribute to be values must not be missing evaluated i e ASSAY or WATER STORAGE yes char storage condition values must not be missing i e 25 C 60 RH FACTOR no char additional study factor ues must not be missing defined by parameter IN_ADD_FACTOR must be consistent with GL_COMMON_RELEASE is YES the content of the additional factor of time 0 has to be COMMON if GL_COMMON_RELEASE is NO the content of the additional factor of time 0 must to be COMMON parameter GL_COMMON_RELEASE US 8 175 843 B2 9 10 TABLE 1 continued Analysis Dataset Variable Definitions Variable Mandatory Type Description Comments BATCH yes any batch or lot identifier values must not be missing TIME yes num study sampling times values must not be missing sampling time unit is defined check for at least two by parameter different values for each IN_TIME_UNIT combination of by var attribute storage condition and additional factor LEVEL yes num value of attribute sample no data check response data may be missing unit of measurement is observations with missing d
50. tions DEFAULTS SAS7BDAT CONSTANTS SAS7BDAT and GRAPHS SAS7BDAT are SAS datasets which con tain background settings as further explained below The Word template file BIGSTEP DOT contains the document settings for transfers of all BIGSTEP output to Word using A4 or Letter sized paper The next steps are to create an analysis environment the stability data file and the stability analysis program file to conduct a statistical analysis of a stability study as detailed below Through a user defined analysis SAS program file stability data are processed as a SAS dataset a variety of statistical analyses are performed for different study designs and response variables and summary output tables and graphs are produced in Word format The analysis program file is submitted to SAS in either batch mode or in an interactive session BIGSTEP provides a standardized methodology and pre sentation format for the statistical analysis and summariza tion of a stability study BIGSTEP has three major purposes summarize data observed for a stability study provide the statistical support for estimating a shelf life of a new drug provide the statistical support for confirming the shelf life of an existing drug BIGSTEP provides standardized output data listings and graphical presentations of observed stability response data Summary tables and graphs can be output in a format suitable for Word documents US 8 175 843 B2 5
51. ty design follows one or more potentially stability limiting response variables i e assay of active ingredient water content impurities over time where the pharmaceutical product of interest is being stored in environ mentally stable conditions to meet regulatory guidelines The duration of a stability study is typically between 6 months to 36 months Usually 3 to 6 batches or lots of the product are included in the stability study with a pre defined sampling scheme used to observe response to storage time The overall goal of a stability study is to either estimate or confirm a pharmaceutical product s shelf life A stability study may involve more than one environmental storage condition being considered simultaneously How ever common practice has the response data from each stor age condition being statistically analyzed separately This is because the typical objective of a stability study is not to compare response among storage conditions but rather to characterize response for each storage condition to estimate or confirm a shelf life It will be assumed throughout this disclosure that the statistical analysis of one storage condition is being discussed As noted above the basic stability study follows the effect of storage time on one or more potentially stability limiting response variables for different batches A more involved stability study also considers the effect of an additional study factor or treatment factor such
52. uding pro grams other than the SAS program The macro for carrying out the invention can be installed by simply copying a few files to a directory or subdirectory ofthe user s choice In addition to the five system and system parameter default files the macro is called and run through the statistical analysis program SAS which includes the option statements required for a particular stability analysis In particular this SAS product contains Base SAS SAS ASSIST SAS ETS SAS GRAPH SAS IML and SAS STAT For example SAS Version 8 2 provides these com ponents and higher versions of this software or software which performs essentially the same functions as this soft ware could also be used Finally a dataset is used which is to include the stability dataset to be statistically analyzed Five SAS files that contain the BIGSTEP macro program code and system default options are to be copied into the directory in which the program is provided These files are SASMACR SAS7BCAT DEFAULTS SAS7BDAT CONSTANTS SAS7BDAT GRAPHS SAS7BDAT BIGSTEP DOT SASMACR SAS7BCAT is a SAS catalog that contains the SAS code defining the BIGSTEP program This file should not be opened or altered by the user for any reason The BIGSTEP program is written as a series of macros which are SAS subroutines or programs that are called through a macro call in an analysis program by specifying the BIG STEP statistical analysis and summary output op
53. ups SLC_PROPOSED_SL lt GL_EXTRAPOLATION_LIMIT must be satisfied Same valid values as the parameter OUTG_XAXIS_END no num X axis end Same valid values as the parameter overwrites homonymous parameter for the considered group The variable is only relevant for graphs OUTG_XAXIS_BY no num Distance between major ticks of x axis Same valid values as the parameter overwrites homonymous parameter for the considered group The variable is only relevant for graphs OUTG_XAXIS_ MINOR no num Number of minor ticks for x axis Same valid values as the parameter overwrites homonymous parameter for the considered group The variable is only relevant for graphs OUTG_YAXIS_START no num Y axis start If any graph display is selected user defined Usually y axis start is OUTG_YAXIS_START calculated automatically with use of response OUTG_YAXIS_END and values and lower acceptance criterion if OUTG_YAXIS_BY must all be present set or all be empty The variable is only relevant for graphs OUTG_YAXIS_START must be OUTG_YAXIS_END no num Y axis end less than OUTG_YAXIS_END user defined Usually y axis end is calculated automatically with use of response values and upper acceptance criterion if present The variable is only relevant for graphs OUTG_YAXIS_BY no num Distance between major ticks of y axis The variable is only relevant for graphs OUTG_YAXIS_MINOR no num Number of minor ticks for y axis Same valid values as the parameter If not
54. variable or if the pre test results on the response trend in the data are to be reported The GROUPS dataset is only necessary if the user wants to define different settings for different attributes for example different func tions should be used A complete list of the options available through the GROUPS dataset is given in Table 6 The analysis information for each response variable is passed to BIGSTEP as the value of SAS variables in the GROUPS dataset through the IN_DATA_GROUPS option in the BIGSTEP call to the macro library The dataset GROUPS can be defined in two ways either as ee part of the analysis program or as a permanent SAS dataset referenced in the analysis program The GROUPS dataset is defined in the example analysis program given in Table 4 Not all options available through the GROUPS need to be speci fied Alternatively for those options that are to be used repeat edly for several stability analyses GROUPS can be defined as a permanent SAS dataset stored in a convenient directory The format of the GROUPS dataset follows conventional SAS dataset formatting An example of aGROUPS dataset is given in Table 7 A LIBNAME needs to be created in the analysis program or one of the existing LIBNAMEs can be used if the permanent GROUPS dataset is stored in the appro 60 65 priate directory The permanent dataset GROUPS is then referenced in the analysis program in a dataset statement similar to the data
55. x non linear time response functions in a table with a listing of the func tions ranked by the best fitting 16 The method of claim 1 wherein the statistical analysis uses the following four linear and non linear functions simple linear quadratic first order exponential and second order exponential for the statistical analysis of stability data 17 The method of claim 1 which includes providing or performing data description a time response regression model a pooling process for the regression model a pre test evaluation and or shelf life estimation 18 The method of claim 17 wherein the data description includes providing data listings summary statistics and plots of the observed data the time response regression model includes analysis of different regression model fits to the observed response data and the pre test evaluation includes a determination if there is a significant change in response over storage time 19 A computer loaded with a general statistical analysis software and a macro integrated with the software such that the computer is capable of performing the computer imple mented method according to claim 1 using the software and macro 20 A computer loaded with a general statistical analysis software and a macro integrated with the software such that the computer is capable of performing the computer imple mented method according to claim 2 using the software and macro 21 A computer loaded with a gener
Download Pdf Manuals
Related Search