* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download PROC QRPT: Procedure for the Estimation of ED50, Relative Potency and Their Fiducial Limits in Quantal Response Assays
Survey
Document related concepts
Transcript
PROC QRPT: Procedure for the Estimation of EDSO, Relative Potency and Their Fiducial Limits in Quantal Response Assays Ronald G. Fleming Squibb Corporation Sue Huang and Nguyen V. Dat Ortho Phanmaceutical Corporation 1. ABSTRACT The frequency counts must be sorted by preparation and dose (BY VARIABLES also require sorting. if they are being used). PROC QRPT procedure is a SAS procedure which provides estimations for ED50 and Relative Potency and their fiducial limits in a PROC SORT; BY PREP DOSE; convenience of the user in the SAS PROC QRPT; CLASS PREP DOSE; VAR RESP; TOTAL TOT; Quantal Response Assay_ This SAS procedure is written in FORTRAN and is geared to the enviroment. Exampl e I: This program estimates a 50~ effective dose (ED50) and relative potency with their 95% fi duci aI lim; ts wi thi n the framework qf Fieller's Theorem from a joint probit X2 statistics for testing the similarity of the probit regression lines required for Drug JfBm Dose (DOSE) A I attaining parallelism. A A 2 PROC QRPT is very user-oriented and can be tun easily by usi ng SAS data sets. B B B INTR DOUeT! ON Responded Total Animal s RESP TOT (Freg count) (Freq count) 3 I 2 44 24 6 47 34 3 6 50 46 50 50 48 50 where the overa 11 sampl e si ze is 294 We would like to thank Dr. Richard Lam for 2. The preparation values must contributing the original FORTRAN program in be eight characters or less. Dose values ~ be numeric. Which this PROC is founded. If you would like a copy of thi s PROC, please 3. Each VAR variable must have a corresponding TOTALYariable. Ronald Fleming at Squibb Corporation. 'Princeton. NJ. ~ontact 4. Class statement must conta; n two vari abIes only. VAR statement contains 50 variables maximum. The pROCEDURE QRPT The response counts and total counts for each maximum number of dose 5 or dose within preparation (e.g. drug) must be input into PROC QRPT (sooetimes using an output dataset from PROC FREQ and responses (per preparation) are 20 and 50 respectively. transposed). The original progr~ has been altered to reduce the n~ber or required input para~eters. For e~ample, the n~ber of levels of the preparation and dose are detemined within the procedure. However, -the user r.tust coraply wi th certain constraints: "pROCEDURE SYNTAX ,he PROC QRPT stateoent 'PROC QRPT options; The options below may appear in -n'e PROC QRPT statement. STANDARD: 'value' or S: 'value' BOO The STANDARD = option allows the user to specify the standard data set against which all othe~ data sets wi 11 be compared. Th, s value Will be compared agal nst the preparation values (PREP) in the CLASS statement. Note: The STANDARD value and the preparation values both must be eight characters or less. CMEAN = value or eM = Example: PROC SORT; BY PROOECT DRUG DOSE; PROC QRPT; CLASS DRUG DOSE; BY PRnJECT; Where DRUG is the preparation, DOSE is the dose and PRWECT is the BY VARIABLE. 3. VAR variable names; (required) The VAR statement wi 11 defi ne the response variables. These vari abIes contai n the frequency counts for each comb1 nation of level s of the CLASS variables. 4-. TOTAL variable names; (opti onall TOTAL is a statement that defines the total population of each VAR variable. The order of variables 1 n the TOTAL statement corresponds di rectly to that of the variables in the VAR statement as illustrated below. value. This constant; s applicable when a threshold response exists and the control value represent this thre sho 1d. Default val ue i s zero. CctlSTANT = value or C = value. This constant; 5 applicable when a dose equals zero. The log of this value cannot be calculated, therefore a constant should be added. If this constant is not specified, the followi ng -procedure will be used: S2 is the second lowest dose and 53 is the third lowest dose. CctlSTANT=(S2-0 )X(S2/S3)' DilTA = Use DATA= to give the name of the EXAWLE 2: Dose 0 or .1 or .01 is assigned at random to 70 anfmals. Tne response categories are eye. nose and ear. The response is either yes or no to whether or not an i nflar.mation occurred. The tota 1 s may look like RSTI - RST3 for the response variables in the follOwing data set. SAS data set to be used by QRPT. If it is omitted, the most recently created data set wi 11 be used. Statements used with QRPT 1.. BY variable names; (optional) 2. CLASS preparation name dose ~ame; (required) The CLASS statement indicates the names of the preparati on and dose variables. in that order. The data must be sorted by "BY VARIABLES' (if any), preparation and dose. 80' i. I'ROC SORT; BY PROJECT DRUG DOSE PROC QRPT: CLASS DRUG DOSE: BY PROJECT: VAR EYE NOSE EAR: TOTAL RSTl RST2 RST3; PROJECT DRUG DOSE EYE (freq) 1 1 1 2 1 2 2 2 10 7 9 8 9 2 3 2 EAR RST1 NOSE RST2 (freq) (freq) (freq) (freq) 1 2 5 3 5 3 10 10 3 5 9 8 10 1 3 8 3 7 2 5 5 9 10 10 RST3 (freq) 10 8 9 10 9 2 2 9 Area of Application The probit line is then used to estimate the E050: Quantal response assay is a qualitative indirect bioassay. It is examp1ified by experiements in which a dose of a dru9 (or preparation) is applied to a total of n animals and r of them response to the drug. if yl=probit (.5)=0 then m, the estimator Of log E050, is: Investigators may be interested in estimating the 50% effective dose (ED50) of a ----r-a _-- a m - yI preparation and its confidence limits. - ~ The estimated E050 f s the antilog of m. Confi dence 1imit for thf s estimate can be A way to estimate the E050 ; s based on the probit transformation which transforms p=r/n into its probit Y, Y=(probit(p», Y is then used to regress a line on the dose or its log calculated using the weighted standard error of the esti~ator m: transformation (X=log dose): Y = probit (p) = a + S =! 1 bx. m Where a and b are estimated us; n9 the least square method; b_ - where 1 tXY - z L..J.rl pq d EX LY 1\ = J:nwX tow ZX2 -} (u)2 a= ~ w= b tnw (1: Y - b EX) where dis number of doses. 802 +m "_-_-"'X_.....,,. Enw(X _ 10 2 (f(y) = z 1 e -Y ) ff 2' When one or more (KI preparations are being compared with. standard one, parallel probit lines are assumed. A coman slope is estimated for overall data: REFERENCES; 1. Estimation. ci Zj i=l Where Zi is the Indicator variables: Yj = a + bX + 1: k Z i = {I otherwi preparation I se. if Fie1ler, E. C., M. A. Creasy and S. r. David (1954). Symposium on Interval 16: 2. J. Roy. Finney, D. J., (1971). Probit Analysis, 3rd Ed. New York: Cambridge U. Press. =j 0 3. Regular F-test is used to test the non-parallel co~ponents (the test is equivalent to the test of Interactions Finney, D. J., (1978). Statistical Method in Biological Assay, 3rd Ed. London, Charles Griffin & Co. between Zi 's and Xl. 95% confidence limits for the slope are al so given using the regular least square oethod to calculate the standard error of the slope: A A S(Y - a - b XI 2 (d - 2 I 1: (X - )0 2 The relative potency is calculated as the ratio of the ED50 values. Confidence limits of the relative potency is evaluated using Fieller's Theorem as applied by Finney (1978). SAf4?LE OUTPUT See data from Example 1. TH, CONTROL •.z.ooocooo 1.0000001) 3.00000()O THE B NU:~9ER !-IE.\~j TH' 0.0 50 OF ••50 NUf16Eit OF .• Z4 ., ~OSES IS: ) .0 1.0000000 Z.OI)OOOOO 3 ••JOOOJ03 [s Doses•• IS: 3 47 3• 18 REGRESSl'JN ANALYSIS EFFECT IJF CHI-SOU, ~O~I-II 1 Z 0.6115 2.601 PESIO. SLOPE -4.4zal 95 PC VA~IANCC A B ~ELATlvE 0.0 PROBIF) 0.0 LI~ITS -).4515 ~=IT¢S:,JRTCVbll¢eZ/Se*z", POTENCY F-RATIO 0.0 CQ~FIOENCE -5.404S statist. Soc. B. 175-222 D.04S64S0 TO A QS pc CO/l;f:IOEIIICE LI"'tITS POTENCY Q5?C CONFIOENCE LIMIT 0.1861D+01 0.1642D+Ol 0.21050.01 O.ZS31~+Ol 0.22390+0t 0.29200+01 0.73310+00 0.60570+00 0.37290+00 803