Experimental Designs for Drug Combination Studies

Size: px
Start display at page:

Download "Experimental Designs for Drug Combination Studies"

Transcription

1 Experimental Designs for Drug Combination Studies B. Almohaimeed & A. N. Donev First version: 1 October 2011 Research Report No. 8, 2011, Probability and Statistics Group School of Mathematics, The University of Manchester

2 Experimental Designs for Drug Combination Studies B. Almohaimeed and A. N. Donev Abstract The interest in drug combinations is growing rapidly due to the opportunities they create to increase the therapeutic effect and to reduce the frequency or magnitude of undesirable side effects when single drugs fail to deliver satisfactory results. Considerable effort in studying the mechanisms leading to the benefits of the joint action of drugs has been matched by the development of relevant statistical methods and tools for statistical analysis of the data obtained in such studies that allow important statistical assumptions to be taken into account, i.e. the appropriate statistical model and the distribution of the response of interest (e.g. Gaussian, Binomial, Poisson). However, much less attention has been given to the choice of suitable experimental designs for such studies, while only high quality data can ensure that the objectives of the studies will be fulfilled. We propose methods for construction of such experimental designs which are economical and make most efficient use of the available resources. We also propose simple but flexible experimental designs, which we call ray-contour designs, which are particularly useful when the use of low or high doses is undesirable and hence a standard statistical analysis of the data is not possible. Keywords: additivity; antagonism; ray design; ray-contour design; synergy. 1 Introduction Combinations of drugs have been known to be highly successful in treating different diseases for many years [1]. They provide additional opportunities over single drug therapies to achieve sufficient therapeutic effect at a lower 1

3 and potentially safer dose. Extensive discussions of the scientific background of such studies, include [1, 2, 3, 4, 5]. There are many examples of successful combination studies in the literature, e.g. [6, 7, 8, 9, 10, 11]. Recently there has been substantial interest in the appropriate statistical methods for analysis of data obtained in combination studies. These publications, including [9, 12, 13, 14, 15, 16, 17, 18, 19], discuss typical methods for statistical analysis of combination studies. Some of these methods have been implemented in statistical packages. Software developed for design and analysis of data from bioassay is also useful. Packages recently developed using the free statistical software R [20] include drc, grofit, dosefinding [21, 22, 23, 24, 25, 26]. They are regularly improved and updated. Common feature of these packages is that they allow for a variety of statistical models to be estimated and different assumptions about the distribution of the response of interest (e.g. Gaussian, Binomial, Poisson) to be made. In Section 2 we summarize the common statistical models used in the statistical analysis of data collected in a bioassay and present ways of assessing useful statistical properties of the results that depend on the choice of the experimental design. Combination studies are a great deal more complex than studying the effect of individual drugs separately. They typically take up longer time and substantial resources compared to studies of single drugs. Therefore, it is important that the experimental designs for such studies make use of all relevant available information in order to ensure statistically viable results and most efficient use of the available resources. Frequently used experimental designs require to collect data for the response of interest at different combinations (rays) of doses of the studied drugs. These ray designs [27], are easy to implement and the results obtained using them are easy to interpret. We discuss how economical but efficient ray designs can be constructed using information obtained in previous studies of the individual components in Section 3. We propose an approach for design of combination studies that takes into account the statistical analysis that will be carried out, i.e. the type of the model that will be estimated as well as the distribution of the experimental errors: we consider the cases when these distribution can be Gaussian, inverse Gaussian, Binomial, Poisson and Gamma. Designs constructed this way can be useful for many experimental situations and are particularly useful for preclinical studies. Finally, in Section 4 we provide an extension of the ideas discussed previously and propose the use of raycontour designs that permit a variety of combinations studies to be designed in such a way that low or high doses are possible to avoid and the statistical 2

4 analysis can be simplified. For example, ray-contour designs are particulary useful for combination studies where small doses should be avoided due to ethical considerations, e.g. in animal studies and in early studies in men when the use of higher doses may not be desirable because of a concern for their safety. We illustrate some benefits of using ray-contour in an example based on the results of a pilot studied carried out at the Paterson Institute for Cancer Research, Manchester, UK. All discussions in the paper are limited to cases where the combination of 2 drugs is required. However, the presented methodology can be easily extended to more general situations. Computer programs implementing all methods proposed in this paper have been written using the computer language R and are available at adonev/combinations. 2 Statistical analysis of combination studies When a ray design has been used, the statistical analysis of the data involves the estimation of the model relating the response of interest Y ij and the doses x ij, δ γ Y ij = η(x ij, θ) = γ + ( ) (x ij α i,)β λi + ɛ ij, (1) i where i, i = 1,..., r, denotes the drug combination (ray), j, j = 1,..., c, denotes the number of dose level at which the response Y ij is measured, x ij is the dose, while θ = (α i β i γ δ λ i ), i = 1,..., r, is a vector of all model parameters. The dose x ij is obtained by combining the doses d ij1 and d ij2 of the studied drugs. We denote R i = d ij1 d ij2 the ratios of the drugs in the combinations and p i = d ij2 d ij1 +d ij2 the corresponding proportions of drug 2. They both are the same for all rays. The model parameters have a useful and easy interpretation: α i represents the logarithm to base 10 of the dose leading to a response which is the average of the minimum and maximum possible response γ and δ, respectively. This dose is also known as the logic50, or logec50. In (1) β i is related to the rate at which the response changes with the dose and it is called Hill s slope. If λ i = 1, the relationship between the response and the dose is symmetric around the dose α i. However, including λ i in the model, allows it to capture departure from such symmetry. For simplicity, in this 3

5 paper we do not consider the case when one of the drugs or any drug combinations are not able to achieve 100% inhibition, but the extension of our results in such cases is straightforward. In (1) ɛ ij is the experimental error which we assume to be additive. Assumptions about the distribution of ɛ ij are usually made based on the knowledge of the nature of the response variable. When the response is continuous a common assumption is that ɛ is normally distributed with zero mean and a variance that may depend on the response. When the experimental errors are expected to be asymmetric around zero, a more common assumption is that their distribution is inverse Gaussian or Gamma. When the response of interest is counts, more suitable assumption for ɛ ij is that it has a Poison distribution, while if it is binary ɛ ij has a Binomial distribution. Often existing knowledge about the studied response allows for simpler models to be used. When λ is assumed to be one, Hill model [28] is obtained. In many cases it is possible to make further assumptions that γ = 0, or that γ = 0 and δ = 1, resulting in simpler models. Thus, models obtained by simplifying (1) that have 2, 3, or 4 parameters may be used. Depending on the distributions of the experimental errors model (1) is nonlinear or generalized nonlinear model and the estimation of its parameters θ can be done using available software. More complicated analysis required when more factors affect the results is also possible but is not discussed in this paper. A case when it may impractical or not possible to collect data to fit model (1) or a 2, 3 or 4 parameter simplification, is considered in Section 4. We use Loewe s [29] definition for additivity. Two compounds are considered to have additive joint action when the combination index CI = d 1 + d 2 D y0,1 D y0,2 (2) is equal to 1, where D y0,1 and D y0,2 are doses of compounds 1 and 2, respectively, required to achieve a defined level of response y 0, or equivalently an inhibition level I, 0 I 100. In (2) d 1 and d 2 denote doses of the two compounds which when used in combination lead to the same level of the response. For example, y 0 can be chosen to correspond to the IC50 (or EC50) dose, when I = 50, i.e. the dose required to obtain 50% of the maximum possible effect. In a study looking for an improvement of efficacy, it is desirable for the two compounds to have synergistic effect. Then CI < 1. 4

6 Alternatively, when the response is related to the drugs safety (e.g. existence, number or severity of undesirable side effects) it is hoped that the joint action of the drugs will be antagonistic, hence CI > 1. Desirable effect may exist only for some drug combinations. The estimation of the combination index and the evaluation of its statistical and biological interpretation plays an important role in combination studies. The values of the parameters α j in (1) are used to calculate the combination indices (2) for all studied drug combinations. [30, 31] show that when ɛ N(0, σ 2 ), the distribution of the logarithm of the combination index has approximately a normal distribution. This makes the interpretations of the results easy: a test for the statistical hypothesis that the combination ratio for a particular drug combination is zero is equivalent to testing whether or not the joint action of the drugs in that combination is additive. The accuracy of the estimate of the combination index depends on the experimental design used to collect the data and on the variability in the study. The Fisher s information matrix for the model parameters holds useful information about the statistical properties of the estimated model. Firstorder approximation of this matrix is given by M(θ) = X T W X, (3) where X is the n p design matrix, n is the total number of observations and p is the number of the model parameters. The columns of M are given by the partial derivatives of the model with resect to the model parameters evaluated at the observations. Appendix B lists the formulas for these calculations when model (1) is used. The information matrix (3) depends on the matrix W. When the observations are independent, W is diagonal with diagonal elements the variances of the observations, i.e. W = diag ( w 1 w 2 w n ). (4) The ith diagonal element of the matrix W depends on the distribution of the experimental errors and is given by ( ) dµ w i = V 1 (µ), (5) dη where µ is the mean of Y evaluated for the model parameters values. Table 5 in Appendix A provides the formulas for these entries when the distribution of ɛ is assumed to be Gaussian, inverse Gaussian, Binomial, Poisson or 5

7 Gamma. When simpler models with 2, 3 and 4 parameters are needed, the corresponding elements of the matrix W are obtained by setting γ = 0, δ = 1 and λ = 1 as necessary. The approximate standardized prediction variance at a dose x ij is d(x ij, θ) = nw 1 x ij f T (x ij, θ)m 1 (θ) f(x ij, θ), (6) where f(x ij, θ) for model (1) is defined by (14) given in Appendix B. An experimental design that minimizes the maximum of (6) over the design region is called G-optimum. The volume of the confidence region for the estimates of the model parameters is proportional to the inverse of the determinant of the Fisher s information matrix (3). The criterion requiring minimization of this volume, and hence the maximization of the determinant of (3), is called D-optimality. However, as the elements of (3) depend on the true values of θ, the determinant can only be calculated for specific values of θ. The same extends also to d(x ij, θ). This creates difficulties for comparing and constructing experimental designs with respect to these useful criteria. We discuss how they can be overcome in the next section. For a general review of the theory of design optimality, see [32]. 3 Experimental designs We consider a number of typical experimental scenarios and present statistical methodology and tools for the construction of suitable designs for combination studies. However, the choice of experimental designs for combination studies is affected by many considerations. The final choice of a design may be influenced by various additional scientific and practical considerations. We divide the experimental designs in three groups, each suitable for specific experimental situations: serial dilutions ray designs, small efficient ray designs, ray-contour designs. Ray-contour designs are the subject of the next section. A serial dilution design is specified by the maximum dose (MD) that is used, the number 6

8 of doses (ND= c) and the dilution factor (DF). In the case of combination studies, the dose is the total of the doses of the ingredients of the drug combination. These designs ensure good coverage of the different doses of the studied combinations and excellent statistical properties when the parameters of the design are suitably chosen. [33] provide advice about how MD, ND and DF can be chosen so that the required resources are minimized. They also provide computer code for their construction. However, the use of such experimental designs for combination studies is expensive. Also, these designs lack the sophistication to fully take into account the relevant model assumptions (e.g. model and distribution of experimental errors) and therefore may use badly the experimental resources. We devote the rest of this section to the construction of small and efficient ray designs which can be obtained using a specific design criterion for their construction. Similar designs have been used in monotherapeutic studies. For example, [34] propose the use of an optimality criterion for the construction of such designs and the observations are normally distributed with homogenous variance. As it was shown in Section 2 the elements of the information matrix depend on the unknown parameter values θ. One possibility to construct a D-optimum design for a specific combination study is to use plausible values for the model parameters in order to calculate the design criterion. The resulting designs are called locally D-optimum experimental designs. Typically locally D-optimum designs for estimating nonlinear models are saturated, i.e. require as many distinct observations as the number of the estimated parameters. In the case of a combination study with r rays and c = p doses used for each ray, c r drug combinations need to be tested, including a suitable number of positive and negative controls, i.e. measurements of the response when no drug is used and when a maximum effect can be seen. Clearly such designs are very economical and at the same time they make most efficient use of the experimental resources. We extend this idea to the cases when generalized nonlinear models are needed and consider the specific challenges that the combinations studies pose. A combination study will typically follow a careful study of the individual affects of the components of the studied drug combinations. As [35] point out, the estimates of the parameter values α 1 and α 2 obtained at that stage can be used to calculate the values α i, i = 3,..., r, for all studied drug 7

9 combinations. If the joint action of the drugs is additive, ˆα i = α 1α 2 (1 + p r ) α 1 + p r α 2, i = 3,..., r. (7) In order to estimate the remaining corresponding parameters in the case of additivity a functional relationship has to be assumed about how they change with the ratio of the components. In our experience, the values for β i and α i, i = 3,..., r can be successfully estimated using a linear interpolation between the corresponding values for the individual drugs, while γ, δ and λ i can be chosen in the same way as in the study of the individual drugs. However, it is clear that the parameter values used to obtain locally D- optimum designs may not be accurate and therefore, the optimum will only be approximately optimum. An alternative approach to construct small efficient ray designs is to choose the doses in such a way that the designs are approximately D-optimum for ranges of parameter values, or for their prior distribution. Such designs are called pseudo-bayesian D-optimum experimental designs. In general they may require a larger number of doses than the local D-optimum designs but they are more robust against possible discrepancy between the assumed model parameter values and those observed in the experiment. Pseudo- Bayesian D-optimum experimental designs are usually also approximately D-optimum. They are generally considered more robust against misspecification of the model parameters. Whatever choice is made, i.e. to use locally or pseudo-bayesian D- optimum designs, the solution of the resulting optimization problem is helped by the result of the General Equivalent Theorem of Kiefer and Wolfowitz [36] as it provides a way to verify whether the constructed design is optimum with respect to the chosen criterion of optimality for each studied drug combination. The theorem states that when the design is D-optimum, the maximum of the standardized prediction variance (6) is equal to the number of the model parameters, p and it is attained at the tested doses. In this case the experimental design is both D- and G-optimum. 3.1 Example 1. Locally D- and G-optimum designs We illustrate the construction of locally and pseudo-bayesian D-optimum experimental designs for combination studies for a variety of statistical models and possible error distributions with examples. Table 2 lists the locally 8

10 D- and G-optimum designs for model 1 and the values of the model parameters as given in Table 1 under an assumption for additive joint action of two drugs when Gaussian, Poisson, Binomial, Gamma, and inverse Gaussian distribution for the experimental errors are assumed. The ratios of the drug combinations for seven rays are chosen to be (0, 0.2, 0.3, 0.5, 0.7, 0.8, 1), respectively. The dose range was chosen to include doses up to 10 units. The designs can be easily recalculated for other choices. In all cases the maximum available dose was included in the design. Therefore, max is used in the tables of results to indicate that the optimum design point is located at the maximum available dose. Figure 1 shows the standardized prediction variance of the response 6 over the design region. In order to save space, this is done only for one of the rays (Ray 4). The plots for the remaining rays are similar. The standardized prediction variance do not exceed the number of model parameters, and it achieves its maximum at the design points, indicated by the dots ( ). Therefore, based on the result of the General Equivalence Theorem it is clear that the designs are both locally D- and G-optimum. 3.2 Example 2. Pseudo Bayesian D- and G-optimum designs Similarly to Example 1, Table 3 lists the pseudo Bayesian D- and G-optimum designs for model (1). The values of the model parameters used to construct the designs are given in Table 1, while Figure 2, shows the corresponding standardized prediction variances of the response for Ray 4. For the chosen design regions, the pseudo-bayesian D-optimum designs are saturated, but this may not be the case or other design regions and model parameters values. 4 Ray-contour designs The aim of these designs is to allow for local synergy or antagonism to be detected; avoid the use of too small or too high doses if necessary; simplify the statistical analysis of the data. 9

11 Table 1: Parameters values for Example 1 and Example 2. Parameters α 1 α 2 β 1 β 2 γ δ λ σ 2 κ υ n Example p Example 2 [1.31,3.01] [1.12,2.81] [0.50,1.15] [0.30,1.11] p Standardized prediction variance Gaussian errors Standardized prediction variance Poisson errors Dose Dose Standardized prediction variance Binomial errors Standardized prediction variance Gamma errors Dose Dose Standardized prediction variance Inverse Gaussian errors Dose Figure 1: Standardized variance for local D- and G-optimum design for combination ray 4. 10

12 Table 2: Locally D- and G-optimum designs for combination studies (r = c = 7). Doses Distribution Ray (max) (max) (max) Gaussian (max) (max) (max) (max) (max) (max) (max) Poisson (max) (max) (max) (max) (max) (max) (max) Binomial (max) (max) (max) (max) (max) (max) (max) Gamma (max) (max) (max) (max) (max) (max) (max) Inverse Gaussian (max) (max) (max) (max)

13 Table 3: Bayesian D- and G-optimum designs for combination studies (r = 7). Doses Distribution Ray (max) (max) (max) Gaussian (max) (max) (max) (max) (max) (max) (max) Poisson (max) (max) (max) (max) (max) (max) (max) Binomial (max) (max) (max) (max) (max) (max) (max) Gamma (max) (max) (max) (max) (max) (max) (max) Inverse Gaussian (max) (max) (max) (max)

14 Standardized prediction variance Gaussian errors Standardized prediction variance Poisson errors Dose Dose Standardized prediction variance Binomial errors Standardized prediction variance Gamma errors Dose Dose Standardized prediction variance Inverse Gaussian errors Dose Figure 2: Standardized variance for Bayesian D- and G-optimum design for combination 4. 13

15 This is achieved by including in the experimental design doses in such a way that if the joint action of the drugs is additive, the expected response to the j th dose of each ray is the same and equal to a predetermined inhibition level I i %, i = 1,..., r. A similar suggestion and a discussion of the possible benefits of using such an experimental design is given in [37]. However, a simple situation is only considered. The doses of a ray-contour design can be calculated by first calculating the expected values of ˆα i, i = 1,......, r, using equation (7). Then, the combined dose for each drug combination can be calculated as IC(I j ) = (I j /(100 I j )) β j 10 α j, (8) where β i, i = 1,..., r, are estimated by interpolation between β 1 and β 2 obtained in previous studies. The individual doses are obtained using the chosen proportions p i of the drugs in the combinations. Therefore, the rays of a ray-contour design are specified by the proportions p i, i = 1,..., r, of the studied drug combinations, while the contours are specified by the selected inhibition levels. The choice of the inhibition levels I i %, i = 1,..., c, as well as their number, can be made by taking into account practical considerations and research objectives. For example, in an in vitro study, a ray contour design would allow the experimenter to see whether synergy or antagonism is exhibited only at specific dose ranges. Also, these designs will be useful in early animal combination studies where in the absence of reliable information about the suitable dose range, the experimenter may want to avoid using too high doses. This approach also provides an easy way to design dose escalating clinical trials for combination studies. In other clinical trials, for example when the drug-combination is an anaesthetic, the experimenter may prefer to avoid unethically low doses. In this case the contours can be chosen to correspond to relatively high inhibition levels. No existing experimental designs are suitable in these cases. The statistical analysis of data obtained using a ray-contour design is simple and does not require fitting model (1). As a result of the construction of the experimental design, when the joint action is additive the expectations of the observations of the response taken at each contour (i.e. inhibition level) should be the same for all studied drug combinations. Therefore, the analysis of variance of the data (ANOVA) can be used to the test for synergy or antagonism at each inhibition level. ANOVA is a method well described in 14

16 many textbooks; see for example [38]. It is also implemented by all statistical packages. Example 3. Cancer Study. The aim of the study was to investigate the combined efficacy of a conventional cytotoxic drug (drug 1) and a novel mechanism based agent (drug 2) in colorectal cancer cell lines, with the ultimate intension of combining the two agents in the clinic. The ratios of the doses of Drug 2 to Drug 1 used in the combinations were: 0.5, 1.0, 2.0, 4.1 and 8.2 (these rays were numbered in reverse order), while the percentages inhibition to be studied were chosen to be 20, 35, 50, 60, 70, and 80. Data were also collected for the individual drugs (Ray 1 and Ray 2). The response was measured also at positive and negative control. The ray-contour experimental design that was used is given in Table 4 and was obtained as described above using the results of previous studies where the IC50 and Hill slopes were estimated as 4.62 µm and 9.47 µm, and and -3.82, for Drug 1 and Drug 2, respectively. The design was replicated on 3 plates. A plot of the means of the observations is shown in Figure 3. The benefits of using a ray-contour design are obvious - one does not need to perform a complicated statistical analysis in order to see that antagonism is observed at Ray 3 where Drug 2 is used 8.2 time more than Drug 1 in the combination. The synergy reduces and then diminishes as the ratio reduces. Many research laboratories that use ray designs with just a single drug combination been defined by the ratio of the IC50 values of the two drugs would have missed this useful result. Clearly, the choice of the drug combinations is very important to the success of the study. Usually it is not known in advance what drug ratio may possess useful properties. Therefore studying as large as possible number of ratios appears to be desirable. However, the experimental resources are usually limited. A compromise could be reached by keeping the number of rays large, while reducing the number of contours (i.e. the doses on each ray). Note that the contour corresponding to 50% inhibition ensures largest sensitivity to detect non-additive joint drug action. We trust the ray-contour designs will find many useful traditional and new applications that will make them popular with experimenters. For example, they will be well suited to new research devoted to identifying interaction thresholds [39]. 15

17 Inhibition Ray Figure 3: Means with error bars at one standard deviation for the observations at different inhibition levels. 16

18 Table 4: Ray-contour design for Example 3. Ray Drug ratio Inhibition (%) Dose 1 Dose inf inf inf inf inf inf inf inf

19 Acknowledgements. The authors are grateful to Dr. Christopher Morrow, Paterson Institute for Cancer Research, Manchester, UK, for providing the data of Example 3. References [1] Chou, T.C. Theoretical Basis, Experimental Design, and Computerized Simulation of Synergism and Antagonism in Drug Combination Studies. Pharmacological Reviews 2006; 58: [2] Berenbaum MC. What is Synergy?. Pharmacological Reviews 1989; 41: [3] Chou, T. C., and D. C. Rideout. Synergism and antagonism in chemotherapy. Academic Press, New York, N.Y [4] Greco, W. R.,Bravo G. and Parsons, J. C. The Search for Synergy: A Critical Review from a Response Surface Perspective. Pharmacological Reviews 1995, 47: [5] Tallarida, R. J. Drug Synergism and Dose-Effect Data Analysis. Chapman, Hall/CRC, [6] Berenbaum MC. The Expected Effect of a Combination of Agents: The General Solution. Journal of Theoretical Biology 1985; 114: [7] Faessel,H.M., Slocum,H. K., Jackson, R. C., Boritzki,T. J., Rustum,Y. M., Nair, M. G., and W. R. Greco: Super in Vitro Synergy between Inhibitors of Dihydrofolate Reductase and Inhibitors of Other Folaterequiring Enzymes: The Critical Role of Polyglutamylation. Cancer Research : [8] Loewe, S. Antagonisms and Antagonists. Pharmacological Reviews 1957;9, [9] Straetemans R. and Bijnens L. Application and review of the separate ray model to investigate interaction effects. Front Biosci 2010;E2: [10] Fitzgerald, J. B., Schoeberl, B., Nielsen, U. B. and Sorger, P. Systems biology and combination therapy in the quest for clinical efficacy. Nature Chemical Biology 2006;2(9):

20 [11] Kong M, Lee JJ. A General Response Surface Model with Varying Relative Potency for Assessing Drug Interactions. Biometrics 2006;62(4): [12] Kong M. and Lee JJ. Applying Emax model and bivariate thin plate splines to assess drug interactions. Front Biosci 2010;E2: [13] Lee JJ, Lin H.Y, Liu DD and Kong M. Emax model and interaction index for assessing drug interaction in combination studies. Front Biosci 2010;E2: [14] Peterson JJ. A review of synergy concepts of nonlinear blending and dose-reduction profiles. Front Biosci 2010; S2: [15] Donev, A. N. Comparison of methods for statistical analysis of combination studies. Front Biosci (Elite Ed) 2010;2: [16] Brun YF. and Greco W.R. Characterization of a three-drug nonlinear mixture response model. Front Biosci 2010; S2: [17] Palomares, I.Rodea, A.L. Petre, K. Boltes, F. Legans, J.A. Perdign- Meln, R. Rosal, F. Fernndez-Pinas. Application of the combination index (CI)-isobologram equation to study the toxicological interactions of lipid regulators in two aquatic bioluminescent organisms. Water Research 2010;44(2): [18] Yan Han, Bo Zhang, Shao Li and Qianchuan Zhao. A formal model for analyzing drug combination effects and its application in TNF-α-induced NFkB pathway. BMC Syst Biol 2010;4:50. [19] Fujimoto Junya, Kong Maiying, Lee J. Jack, Hong Waun Ki, and Lotan Reuben. Validation of a Novel Statistical Model for Assessing the Synergy of Combined-Agent Cancer Chemoprevention. Cancer Prev Res 2010, / CAPR [20] R Development Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing 2011;Vienna, Austria URL: [21] Ritz, C. and Strebig, J. Bioassay Analysis using R. Journal of Statistical Software 2005; 5:12. URL: 19

21 [22] Ritz, C. and Strebig, J. R package, Analysis of dose-response curves. URL: [23] Kahm, M., Hasenbrink, G., Lichtenberg-Frate, H., Ludwig, J. and Kshischo, M. grofit : fitting biological growth curves with R. Journal of Statistical Software 2010; 33, 121. [24] Boik J.C., Narasimhan N. An R Package for Assessing Drug Synergism/Antagonism. Journal of Statistical Software 2010; 34(6), 118. [25] Bornkamp, B., Pinheiro, J., Bretz, F. MCPMod : An R Package for the Design and Analysis of Dose-Finding Studies. Journal of Statistical Software 2009; 29, 123. [26] Bornkamp, B., Pinheiro, J., Bretz, F. DoseFinding : Planning and Analyzing Dose Finding experiments Journal of Statistical Software 2011; URL: [27] Mantel, N. An Experimental Design in Combination Chemotherapy. Ann. New York Acad. Sci. 1958;76, [28] Hill, A. V. The Combinations of Haemoglobin with Oxygen and with Carbon Monoxide. Biochem 1913, 7: [29] Loewe, S. The problem of synergism and antagonism of combined drugs. Arzneim. Forsch 1953; 3: [30] Lee JJ, Kong M, Ayers GD, Lotan R. Interaction Index and Different Methods for Determining Drug Interaction in Combination Therapy. Journal of Biopharmaceutical Statistics 2007; 17: [31] Lee J.J. and Kong M. Confidence intervals of interaction index for assessing multiple drug interaction. Statistics in Biopharaceutical Research 2009;1(1): 417. [32] Atkinson, A.C., Donev, A.N. and Tobias, R.D. Optimum Experimental Designs, with SAS. Oxford University Press, [33] Donev, A. N. and Tobias, R. Optimal Serial Dilutions Designs For Drug Discovery Experiments. Journal of Biopharmaceutical Statistics 2011, 21:

22 [34] Abdelbasit, K.M. and Plackett, R. L. Experimental Design for Joint Action. Biometrics 1982; 38: [35] Straetemans, R., OBrien, T., Wouters, L., Dun, J. V., Janicot, M., Bijnens, L., Burzykowski, T., Aerts, M. Design and analysis of drug combination experiments. Biometrical Journal 2005;47(3): [36] Kiefer, J. and Wolfowitz, J. The equivalence of two extremum problems. Canadian Journal of Mathematics 1960, 12: [37] Gennings, C. An efficient experimental design for detecting departure from additivity in mixtures of many chemicals. Toxicology 1995; 105: [38] Clarke, G.M. and Cooke, D. A Basic Course in Statistics, fourth edition. Arnold, [39] Hamm, A. K., Carter Jr., W. H. and Gennings, C. Analysis of an interaction threshold in a mixture of drugs and/or chemicals. Statistics in Medicine 2005; 24:

23 Appendix A. Table 5: The i th,i = 1,..., n, diagonal element of the W matrix for model (1). Distribution Normal with mean µ and variance σ 2, Expression i.e. w i = σ 2 Y i N (µ, σ 2 ) Poisson with mean µ, Y i Pois (µ) w i = e η(x i,θ) Binomial with n trials, and p probability of success, i.e. Y i B (n, p) and the variance is Var (Y i ) = np (1 p) w i = ne η(x i,θ) Gamma with scale ϑ and shape κ parameters i.e. Y i Γ (κ, ϑ) and variance Var (Y i ) = κϑ 2 w i = κη(x i, θ) 2 Inverse-Gaussian with mean µ and shape parameter υ i.e. Y i IG (µ, υ) w i = υ 4 (η(x i, θ)) 3/2 and variance Var (Y i ) = µ3 υ 22

24 Appendix B. The derivatives of model (1) with respect to the parameters (α i, β i, γ, δ and λ i, respectively) are η k (x ij, θ) α i = (δ γ) λ i 10 (x ij α i )β i β i ln (10) ( (x ij α i )β i ) λi ( (x ij α i )β i ), (9) η k (x ij, θ) β i = (δ γ) λ i10 (xij αi)βi (x ij α i ) ln (10) ( (x ij α i )β i ) λi ( (x ij α i )β i ), (10) η k (x ij, θ) δ = ( ( (x ij α i )β i ) λi ) 1, (11) and η k (x ij, θ) γ η k (x ij, θ) λ i = 1 ( ( (x ij α i )β i ) λi ) 1 (12) = ( ) (δ γ) ln (x ij α i )β i ( ) (x ij α i )β λi. (13) i Therefore f T (x ij, θ) = ( ηk (x ij,θ) α i η k (x ij,θ) β i η k (x ij,θ) δ η k (x ij,θ) γ ) η k (x ij,θ) λ i. (14) 23

Design of preclinical combination studies

Design of preclinical combination studies University of Manchester, UK Talk Outline 1 Preliminaries What is a drug combination? Data structures. Ray design 2 Combination index Sources of variability 3 Two drugs Analysis of data 4 Types of studies

More information

Using R For Flexible Modelling Of Pre-Clinical Combination Studies. Chris Harbron Discovery Statistics AstraZeneca

Using R For Flexible Modelling Of Pre-Clinical Combination Studies. Chris Harbron Discovery Statistics AstraZeneca Using R For Flexible Modelling Of Pre-Clinical Combination Studies Chris Harbron Discovery Statistics AstraZeneca Modelling Drug Combinations Why? The theory An example The practicalities in R 2 Chris

More information

Generalizing the MCPMod methodology beyond normal, independent data

Generalizing the MCPMod methodology beyond normal, independent data Generalizing the MCPMod methodology beyond normal, independent data José Pinheiro Joint work with Frank Bretz and Björn Bornkamp Novartis AG ASA NJ Chapter 35 th Annual Spring Symposium June 06, 2014 Outline

More information

MULTIPLE-OBJECTIVE DESIGNS IN A DOSE-RESPONSE EXPERIMENT

MULTIPLE-OBJECTIVE DESIGNS IN A DOSE-RESPONSE EXPERIMENT New Developments and Applications in Experimental Design IMS Lecture Notes - Monograph Series (1998) Volume 34 MULTIPLE-OBJECTIVE DESIGNS IN A DOSE-RESPONSE EXPERIMENT BY WEI ZHU AND WENG KEE WONG 1 State

More information

Generalizing the MCPMod methodology beyond normal, independent data

Generalizing the MCPMod methodology beyond normal, independent data Generalizing the MCPMod methodology beyond normal, independent data José Pinheiro Joint work with Frank Bretz and Björn Bornkamp Novartis AG Trends and Innovations in Clinical Trial Statistics Conference

More information

NIH Public Access Author Manuscript Front Biosci (Elite Ed). Author manuscript; available in PMC 2010 November 5.

NIH Public Access Author Manuscript Front Biosci (Elite Ed). Author manuscript; available in PMC 2010 November 5. NIH Public Access Author Manuscript Published in final edited form as: Front Biosci (Elite Ed). ; 2: 582 601. E max model and interaction index for assessing drug interaction in combination studies J.

More information

Non-Gaussian Berkson Errors in Bioassay

Non-Gaussian Berkson Errors in Bioassay Non-Gaussian Berkson Errors in Bioassay Alaa Althubaiti & Alexander Donev First version: 1 May 011 Research Report No., 011, Probability and Statistics Group School of Mathematics, The University of Manchester

More information

Bayesian concept for combined Phase 2a/b trials

Bayesian concept for combined Phase 2a/b trials Bayesian concept for combined Phase 2a/b trials /////////// Stefan Klein 07/12/2018 Agenda Phase 2a: PoC studies Phase 2b: dose finding studies Simulation Results / Discussion 2 /// Bayer /// Bayesian

More information

Pubh 8482: Sequential Analysis

Pubh 8482: Sequential Analysis Pubh 8482: Sequential Analysis Joseph S. Koopmeiners Division of Biostatistics University of Minnesota Week 10 Class Summary Last time... We began our discussion of adaptive clinical trials Specifically,

More information

Paper CD Erika Larsen and Timothy E. O Brien Loyola University Chicago

Paper CD Erika Larsen and Timothy E. O Brien Loyola University Chicago Abstract: Paper CD-02 2015 SAS Software as an Essential Tool in Statistical Consulting and Research Erika Larsen and Timothy E. O Brien Loyola University Chicago Modelling in bioassay often uses linear,

More information

Bootstrap Procedures for Testing Homogeneity Hypotheses

Bootstrap Procedures for Testing Homogeneity Hypotheses Journal of Statistical Theory and Applications Volume 11, Number 2, 2012, pp. 183-195 ISSN 1538-7887 Bootstrap Procedures for Testing Homogeneity Hypotheses Bimal Sinha 1, Arvind Shah 2, Dihua Xu 1, Jianxin

More information

Drug Combination Analysis

Drug Combination Analysis Drug Combination Analysis Gary D. Knott, Ph.D. Civilized Software, Inc. 12109 Heritage Park Circle Silver Spring MD 20906 USA Tel.: (301)-962-3711 email: csi@civilized.com URL: www.civilized.com abstract:

More information

D-optimal Designs for Factorial Experiments under Generalized Linear Models

D-optimal Designs for Factorial Experiments under Generalized Linear Models D-optimal Designs for Factorial Experiments under Generalized Linear Models Jie Yang Department of Mathematics, Statistics, and Computer Science University of Illinois at Chicago Joint research with Abhyuday

More information

ADAPTIVE EXPERIMENTAL DESIGNS. Maciej Patan and Barbara Bogacka. University of Zielona Góra, Poland and Queen Mary, University of London

ADAPTIVE EXPERIMENTAL DESIGNS. Maciej Patan and Barbara Bogacka. University of Zielona Góra, Poland and Queen Mary, University of London ADAPTIVE EXPERIMENTAL DESIGNS FOR SIMULTANEOUS PK AND DOSE-SELECTION STUDIES IN PHASE I CLINICAL TRIALS Maciej Patan and Barbara Bogacka University of Zielona Góra, Poland and Queen Mary, University of

More information

AP-Optimum Designs for Minimizing the Average Variance and Probability-Based Optimality

AP-Optimum Designs for Minimizing the Average Variance and Probability-Based Optimality AP-Optimum Designs for Minimizing the Average Variance and Probability-Based Optimality Authors: N. M. Kilany Faculty of Science, Menoufia University Menoufia, Egypt. (neveenkilany@hotmail.com) W. A. Hassanein

More information

Compound Optimal Designs for Percentile Estimation in Dose-Response Models with Restricted Design Intervals

Compound Optimal Designs for Percentile Estimation in Dose-Response Models with Restricted Design Intervals Compound Optimal Designs for Percentile Estimation in Dose-Response Models with Restricted Design Intervals Stefanie Biedermann 1, Holger Dette 1, Wei Zhu 2 Abstract In dose-response studies, the dose

More information

Chapter 1 Statistical Inference

Chapter 1 Statistical Inference Chapter 1 Statistical Inference causal inference To infer causality, you need a randomized experiment (or a huge observational study and lots of outside information). inference to populations Generalizations

More information

On the efficiency of two-stage adaptive designs

On the efficiency of two-stage adaptive designs On the efficiency of two-stage adaptive designs Björn Bornkamp (Novartis Pharma AG) Based on: Dette, H., Bornkamp, B. and Bretz F. (2010): On the efficiency of adaptive designs www.statistik.tu-dortmund.de/sfb823-dp2010.html

More information

Relative Potency Estimations in Multiple Bioassay Problems

Relative Potency Estimations in Multiple Bioassay Problems Relative Potency Estimations in Multiple Bioassay Problems Gemechis Dilba Institute of Biostatistics, Leibniz University of Hannover, Germany 5 th International Conference on Multiple Comparison Procedures

More information

Model Selection for Semiparametric Bayesian Models with Application to Overdispersion

Model Selection for Semiparametric Bayesian Models with Application to Overdispersion Proceedings 59th ISI World Statistics Congress, 25-30 August 2013, Hong Kong (Session CPS020) p.3863 Model Selection for Semiparametric Bayesian Models with Application to Overdispersion Jinfang Wang and

More information

CHL 5225H Advanced Statistical Methods for Clinical Trials: Multiplicity

CHL 5225H Advanced Statistical Methods for Clinical Trials: Multiplicity CHL 5225H Advanced Statistical Methods for Clinical Trials: Multiplicity Prof. Kevin E. Thorpe Dept. of Public Health Sciences University of Toronto Objectives 1. Be able to distinguish among the various

More information

Online publication date: 22 March 2010

Online publication date: 22 March 2010 This article was downloaded by: [South Dakota State University] On: 25 March 2010 Access details: Access Details: [subscription number 919556249] Publisher Taylor & Francis Informa Ltd Registered in England

More information

A Two-Stage Response Surface Approach to Modeling Drug Interaction

A Two-Stage Response Surface Approach to Modeling Drug Interaction This article was downloaded by: [FDA Biosciences Library] On: 27 October 2012, At: 12:44 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office:

More information

PubH 7405: REGRESSION ANALYSIS INTRODUCTION TO LOGISTIC REGRESSION

PubH 7405: REGRESSION ANALYSIS INTRODUCTION TO LOGISTIC REGRESSION PubH 745: REGRESSION ANALYSIS INTRODUCTION TO LOGISTIC REGRESSION Let Y be the Dependent Variable Y taking on values and, and: π Pr(Y) Y is said to have the Bernouilli distribution (Binomial with n ).

More information

Research Article A Nonparametric Two-Sample Wald Test of Equality of Variances

Research Article A Nonparametric Two-Sample Wald Test of Equality of Variances Advances in Decision Sciences Volume 211, Article ID 74858, 8 pages doi:1.1155/211/74858 Research Article A Nonparametric Two-Sample Wald Test of Equality of Variances David Allingham 1 andj.c.w.rayner

More information

Bivariate Weibull-power series class of distributions

Bivariate Weibull-power series class of distributions Bivariate Weibull-power series class of distributions Saralees Nadarajah and Rasool Roozegar EM algorithm, Maximum likelihood estimation, Power series distri- Keywords: bution. Abstract We point out that

More information

Optimal Experimental Designs for the Poisson Regression Model in Toxicity Studies

Optimal Experimental Designs for the Poisson Regression Model in Toxicity Studies Optimal Experimental Designs for the Poisson Regression Model in Toxicity Studies Yanping Wang Dissertation submitted to the Faculty of the Virginia Polytechnic Institute and State University in partial

More information

A SAS/AF Application For Sample Size And Power Determination

A SAS/AF Application For Sample Size And Power Determination A SAS/AF Application For Sample Size And Power Determination Fiona Portwood, Software Product Services Ltd. Abstract When planning a study, such as a clinical trial or toxicology experiment, the choice

More information

As mentioned in the introduction of the manuscript, isoboles are commonly used to analyze

As mentioned in the introduction of the manuscript, isoboles are commonly used to analyze Appendix 1: Review of Common Drug Interaction Models As mentioned in the introduction of the manuscript, isoboles are commonly used to analyze anesthetic drug interactions. Isoboles show dose combinations

More information

Duke University. Duke Biostatistics and Bioinformatics (B&B) Working Paper Series. Randomized Phase II Clinical Trials using Fisher s Exact Test

Duke University. Duke Biostatistics and Bioinformatics (B&B) Working Paper Series. Randomized Phase II Clinical Trials using Fisher s Exact Test Duke University Duke Biostatistics and Bioinformatics (B&B) Working Paper Series Year 2010 Paper 7 Randomized Phase II Clinical Trials using Fisher s Exact Test Sin-Ho Jung sinho.jung@duke.edu This working

More information

Efficient algorithms for calculating optimal designs in pharmacokinetics and dose finding studies

Efficient algorithms for calculating optimal designs in pharmacokinetics and dose finding studies Efficient algorithms for calculating optimal designs in pharmacokinetics and dose finding studies Tim Holland-Letz Ruhr-Universität Bochum Medizinische Fakultät 44780 Bochum, Germany email: tim.holland-letz@rub.de

More information

Optimum Designs for the Equality of Parameters in Enzyme Inhibition Kinetic Models

Optimum Designs for the Equality of Parameters in Enzyme Inhibition Kinetic Models Optimum Designs for the Equality of Parameters in Enzyme Inhibition Kinetic Models Anthony C. Atkinson, Department of Statistics, London School of Economics, London WC2A 2AE, UK and Barbara Bogacka, School

More information

Approximate and Fiducial Confidence Intervals for the Difference Between Two Binomial Proportions

Approximate and Fiducial Confidence Intervals for the Difference Between Two Binomial Proportions Approximate and Fiducial Confidence Intervals for the Difference Between Two Binomial Proportions K. Krishnamoorthy 1 and Dan Zhang University of Louisiana at Lafayette, Lafayette, LA 70504, USA SUMMARY

More information

Paper Equivalence Tests. Fei Wang and John Amrhein, McDougall Scientific Ltd.

Paper Equivalence Tests. Fei Wang and John Amrhein, McDougall Scientific Ltd. Paper 11683-2016 Equivalence Tests Fei Wang and John Amrhein, McDougall Scientific Ltd. ABSTRACT Motivated by the frequent need for equivalence tests in clinical trials, this paper provides insights into

More information

A nonparametric two-sample wald test of equality of variances

A nonparametric two-sample wald test of equality of variances University of Wollongong Research Online Faculty of Informatics - Papers (Archive) Faculty of Engineering and Information Sciences 211 A nonparametric two-sample wald test of equality of variances David

More information

Best Linear Unbiased Prediction: an Illustration Based on, but Not Limited to, Shelf Life Estimation

Best Linear Unbiased Prediction: an Illustration Based on, but Not Limited to, Shelf Life Estimation Libraries Conference on Applied Statistics in Agriculture 015-7th Annual Conference Proceedings Best Linear Unbiased Prediction: an Illustration Based on, but Not Limited to, Shelf Life Estimation Maryna

More information

Pooling Experiments for High Throughput Screening in Drug Discovery

Pooling Experiments for High Throughput Screening in Drug Discovery Pooling Experiments for High Throughput Screening in Drug Discovery Jacqueline M. Hughes-Oliver hughesol@stat.ncsu.edu Department of Statistics North Carolina State University Spring Research Conference,

More information

What is Experimental Design?

What is Experimental Design? One Factor ANOVA What is Experimental Design? A designed experiment is a test in which purposeful changes are made to the input variables (x) so that we may observe and identify the reasons for change

More information

Dose-response modeling with bivariate binary data under model uncertainty

Dose-response modeling with bivariate binary data under model uncertainty Dose-response modeling with bivariate binary data under model uncertainty Bernhard Klingenberg 1 1 Department of Mathematics and Statistics, Williams College, Williamstown, MA, 01267 and Institute of Statistics,

More information

Society for Biomolecular Screening 10th Annual Conference, Orlando, FL, September 11-15, 2004

Society for Biomolecular Screening 10th Annual Conference, Orlando, FL, September 11-15, 2004 Society for Biomolecular Screening 10th Annual Conference, Orlando, FL, September 11-15, 2004 Advanced Methods in Dose-Response Screening of Enzyme Inhibitors Petr uzmič, Ph.D. Bioin, Ltd. TOPICS: 1. Fitting

More information

Enquiry. Demonstration of Uniformity of Dosage Units using Large Sample Sizes. Proposal for a new general chapter in the European Pharmacopoeia

Enquiry. Demonstration of Uniformity of Dosage Units using Large Sample Sizes. Proposal for a new general chapter in the European Pharmacopoeia Enquiry Demonstration of Uniformity of Dosage Units using Large Sample Sizes Proposal for a new general chapter in the European Pharmacopoeia In order to take advantage of increased batch control offered

More information

Sample size determination for logistic regression: A simulation study

Sample size determination for logistic regression: A simulation study Sample size determination for logistic regression: A simulation study Stephen Bush School of Mathematical Sciences, University of Technology Sydney, PO Box 123 Broadway NSW 2007, Australia Abstract This

More information

Pairwise rank based likelihood for estimating the relationship between two homogeneous populations and their mixture proportion

Pairwise rank based likelihood for estimating the relationship between two homogeneous populations and their mixture proportion Pairwise rank based likelihood for estimating the relationship between two homogeneous populations and their mixture proportion Glenn Heller and Jing Qin Department of Epidemiology and Biostatistics Memorial

More information

Design of Screening Experiments with Partial Replication

Design of Screening Experiments with Partial Replication Design of Screening Experiments with Partial Replication David J. Edwards Department of Statistical Sciences & Operations Research Virginia Commonwealth University Robert D. Leonard Department of Information

More information

Robust design in model-based analysis of longitudinal clinical data

Robust design in model-based analysis of longitudinal clinical data Robust design in model-based analysis of longitudinal clinical data Giulia Lestini, Sebastian Ueckert, France Mentré IAME UMR 1137, INSERM, University Paris Diderot, France PODE, June 0 016 Context Optimal

More information

SAMPLE SIZE RE-ESTIMATION FOR ADAPTIVE SEQUENTIAL DESIGN IN CLINICAL TRIALS

SAMPLE SIZE RE-ESTIMATION FOR ADAPTIVE SEQUENTIAL DESIGN IN CLINICAL TRIALS Journal of Biopharmaceutical Statistics, 18: 1184 1196, 2008 Copyright Taylor & Francis Group, LLC ISSN: 1054-3406 print/1520-5711 online DOI: 10.1080/10543400802369053 SAMPLE SIZE RE-ESTIMATION FOR ADAPTIVE

More information

Statistics in medicine

Statistics in medicine Statistics in medicine Lecture 3: Bivariate association : Categorical variables Proportion in one group One group is measured one time: z test Use the z distribution as an approximation to the binomial

More information

Locally optimal designs for errors-in-variables models

Locally optimal designs for errors-in-variables models Locally optimal designs for errors-in-variables models Maria Konstantinou, Holger Dette Ruhr-Universität Bochum Fakultät für Mathematik 44780 Bochum, Germany maria.konstantinou@ruhr-uni-bochum.de holger.dette@ruhr-uni-bochum.de

More information

Bayesian Applications in Biomarker Detection. Dr. Richardus Vonk Head, Research and Clinical Sciences Statistics

Bayesian Applications in Biomarker Detection. Dr. Richardus Vonk Head, Research and Clinical Sciences Statistics Bayesian Applications in Biomarker Detection Dr. Richardus Vonk Head, Research and Clinical Sciences Statistics Disclaimer The views expressed in this presentation are the personal views of the author,

More information

RESPONSE SURFACE MODELLING, RSM

RESPONSE SURFACE MODELLING, RSM CHEM-E3205 BIOPROCESS OPTIMIZATION AND SIMULATION LECTURE 3 RESPONSE SURFACE MODELLING, RSM Tool for process optimization HISTORY Statistical experimental design pioneering work R.A. Fisher in 1925: Statistical

More information

Open Problems in Mixed Models

Open Problems in Mixed Models xxiii Determining how to deal with a not positive definite covariance matrix of random effects, D during maximum likelihood estimation algorithms. Several strategies are discussed in Section 2.15. For

More information

Adaptive Prediction of Event Times in Clinical Trials

Adaptive Prediction of Event Times in Clinical Trials Adaptive Prediction of Event Times in Clinical Trials Yu Lan Southern Methodist University Advisor: Daniel F. Heitjan May 8, 2017 Yu Lan (SMU) May 8, 2017 1 / 19 Clinical Trial Prediction Event-based trials:

More information

DIAGNOSTICS FOR STRATIFIED CLINICAL TRIALS IN PROPORTIONAL ODDS MODELS

DIAGNOSTICS FOR STRATIFIED CLINICAL TRIALS IN PROPORTIONAL ODDS MODELS DIAGNOSTICS FOR STRATIFIED CLINICAL TRIALS IN PROPORTIONAL ODDS MODELS Ivy Liu and Dong Q. Wang School of Mathematics, Statistics and Computer Science Victoria University of Wellington New Zealand Corresponding

More information

Binomial Distribution Sample Confidence Intervals Estimation 8. Number Needed to Treat/Harm

Binomial Distribution Sample Confidence Intervals Estimation 8. Number Needed to Treat/Harm Binomial Distribution Sample Confidence Intervals Estimation 8. Number Needed to Treat/Harm Iuliu Haţieganu University of Medicine and Pharmacy, Cluj-Napoca, Romania sbolboaca@umfcluj.ro Abstract Nowadays,

More information

Inverse Sampling for McNemar s Test

Inverse Sampling for McNemar s Test International Journal of Statistics and Probability; Vol. 6, No. 1; January 27 ISSN 1927-7032 E-ISSN 1927-7040 Published by Canadian Center of Science and Education Inverse Sampling for McNemar s Test

More information

Sample Size Determination

Sample Size Determination Sample Size Determination 018 The number of subjects in a clinical study should always be large enough to provide a reliable answer to the question(s addressed. The sample size is usually determined by

More information

Adaptive Designs: Why, How and When?

Adaptive Designs: Why, How and When? Adaptive Designs: Why, How and When? Christopher Jennison Department of Mathematical Sciences, University of Bath, UK http://people.bath.ac.uk/mascj ISBS Conference Shanghai, July 2008 1 Adaptive designs:

More information

Model selection and comparison

Model selection and comparison Model selection and comparison an example with package Countr Tarak Kharrat 1 and Georgi N. Boshnakov 2 1 Salford Business School, University of Salford, UK. 2 School of Mathematics, University of Manchester,

More information

Randomized dose-escalation design for drug combination cancer trials with immunotherapy

Randomized dose-escalation design for drug combination cancer trials with immunotherapy Randomized dose-escalation design for drug combination cancer trials with immunotherapy Pavel Mozgunov, Thomas Jaki, Xavier Paoletti Medical and Pharmaceutical Statistics Research Unit, Department of Mathematics

More information

Model identification for dose response signal detection

Model identification for dose response signal detection SFB 823 Model identifi fication for dose response signal detection Discussion Paper Holger Dette, Stefanie Titoff, Stanislav Volgushev, Frank Bretz Nr. 35/2012 Model identification for dose response signal

More information

Lawrence D. Brown* and Daniel McCarthy*

Lawrence D. Brown* and Daniel McCarthy* Comments on the paper, An adaptive resampling test for detecting the presence of significant predictors by I. W. McKeague and M. Qian Lawrence D. Brown* and Daniel McCarthy* ABSTRACT: This commentary deals

More information

Many natural processes can be fit to a Poisson distribution

Many natural processes can be fit to a Poisson distribution BE.104 Spring Biostatistics: Poisson Analyses and Power J. L. Sherley Outline 1) Poisson analyses 2) Power What is a Poisson process? Rare events Values are observational (yes or no) Random distributed

More information

Algorithmisches Lernen/Machine Learning

Algorithmisches Lernen/Machine Learning Algorithmisches Lernen/Machine Learning Part 1: Stefan Wermter Introduction Connectionist Learning (e.g. Neural Networks) Decision-Trees, Genetic Algorithms Part 2: Norman Hendrich Support-Vector Machines

More information

Comment: Some Statistical Concerns on Dimensional Analysis

Comment: Some Statistical Concerns on Dimensional Analysis COMMENT: SOME STATISTICAL CONCERNS ON DA 28 Comment: Some Statistical Concerns on Dimensional Analysis DennisK.J.LIN and Weijie SHEN Department of Statistics The Pennsylvania State University University

More information

Error analysis for efficiency

Error analysis for efficiency Glen Cowan RHUL Physics 28 July, 2008 Error analysis for efficiency To estimate a selection efficiency using Monte Carlo one typically takes the number of events selected m divided by the number generated

More information

In Silico Investigation of Off-Target Effects

In Silico Investigation of Off-Target Effects PHARMA & LIFE SCIENCES WHITEPAPER In Silico Investigation of Off-Target Effects STREAMLINING IN SILICO PROFILING In silico techniques require exhaustive data and sophisticated, well-structured informatics

More information

Bayesian methods for sample size determination and their use in clinical trials

Bayesian methods for sample size determination and their use in clinical trials Bayesian methods for sample size determination and their use in clinical trials Stefania Gubbiotti Abstract This paper deals with determination of a sample size that guarantees the success of a trial.

More information

PubH 7470: STATISTICS FOR TRANSLATIONAL & CLINICAL RESEARCH

PubH 7470: STATISTICS FOR TRANSLATIONAL & CLINICAL RESEARCH PubH 7470: STATISTICS FOR TRANSLATIONAL & CLINICAL RESEARCH From Basic to Translational: INDIRECT BIOASSAYS INDIRECT ASSAYS In indirect assays, the doses of the standard and test preparations are are applied

More information

Using Historical Experimental Information in the Bayesian Analysis of Reproduction Toxicological Experimental Results

Using Historical Experimental Information in the Bayesian Analysis of Reproduction Toxicological Experimental Results Using Historical Experimental Information in the Bayesian Analysis of Reproduction Toxicological Experimental Results Jing Zhang Miami University August 12, 2014 Jing Zhang (Miami University) Using Historical

More information

Now consider the case where E(Y) = µ = Xβ and V (Y) = σ 2 G, where G is diagonal, but unknown.

Now consider the case where E(Y) = µ = Xβ and V (Y) = σ 2 G, where G is diagonal, but unknown. Weighting We have seen that if E(Y) = Xβ and V (Y) = σ 2 G, where G is known, the model can be rewritten as a linear model. This is known as generalized least squares or, if G is diagonal, with trace(g)

More information

Statistical Inference of Covariate-Adjusted Randomized Experiments

Statistical Inference of Covariate-Adjusted Randomized Experiments 1 Statistical Inference of Covariate-Adjusted Randomized Experiments Feifang Hu Department of Statistics George Washington University Joint research with Wei Ma, Yichen Qin and Yang Li Email: feifang@gwu.edu

More information

Analysis of a Large Structure/Biological Activity. Data Set Using Recursive Partitioning and. Simulated Annealing

Analysis of a Large Structure/Biological Activity. Data Set Using Recursive Partitioning and. Simulated Annealing Analysis of a Large Structure/Biological Activity Data Set Using Recursive Partitioning and Simulated Annealing Student: Ke Zhang MBMA Committee: Dr. Charles E. Smith (Chair) Dr. Jacqueline M. Hughes-Oliver

More information

A Generalization for Stable Mixed Finite Elements

A Generalization for Stable Mixed Finite Elements A Generalization for Stable Mixed Finite Elements Andrew Gillette joint work with Chandrajit Bajaj Department of Mathematics Institute of Computational Engineering and Sciences University of Texas at Austin,

More information

CHOOSING AMONG GENERALIZED LINEAR MODELS APPLIED TO MEDICAL DATA

CHOOSING AMONG GENERALIZED LINEAR MODELS APPLIED TO MEDICAL DATA STATISTICS IN MEDICINE, VOL. 17, 59 68 (1998) CHOOSING AMONG GENERALIZED LINEAR MODELS APPLIED TO MEDICAL DATA J. K. LINDSEY AND B. JONES* Department of Medical Statistics, School of Computing Sciences,

More information

Computationally Efficient Estimation of Multilevel High-Dimensional Latent Variable Models

Computationally Efficient Estimation of Multilevel High-Dimensional Latent Variable Models Computationally Efficient Estimation of Multilevel High-Dimensional Latent Variable Models Tihomir Asparouhov 1, Bengt Muthen 2 Muthen & Muthen 1 UCLA 2 Abstract Multilevel analysis often leads to modeling

More information

D-optimal Designs for Multinomial Logistic Models

D-optimal Designs for Multinomial Logistic Models D-optimal Designs for Multinomial Logistic Models Jie Yang University of Illinois at Chicago Joint with Xianwei Bu and Dibyen Majumdar October 12, 2017 1 Multinomial Logistic Models Cumulative logit model:

More information

Gravity Models, PPML Estimation and the Bias of the Robust Standard Errors

Gravity Models, PPML Estimation and the Bias of the Robust Standard Errors Gravity Models, PPML Estimation and the Bias of the Robust Standard Errors Michael Pfaffermayr August 23, 2018 Abstract In gravity models with exporter and importer dummies the robust standard errors of

More information

Adaptive designs beyond p-value combination methods. Ekkehard Glimm, Novartis Pharma EAST user group meeting Basel, 31 May 2013

Adaptive designs beyond p-value combination methods. Ekkehard Glimm, Novartis Pharma EAST user group meeting Basel, 31 May 2013 Adaptive designs beyond p-value combination methods Ekkehard Glimm, Novartis Pharma EAST user group meeting Basel, 31 May 2013 Outline Introduction Combination-p-value method and conditional error function

More information

Sequential Procedure for Testing Hypothesis about Mean of Latent Gaussian Process

Sequential Procedure for Testing Hypothesis about Mean of Latent Gaussian Process Applied Mathematical Sciences, Vol. 4, 2010, no. 62, 3083-3093 Sequential Procedure for Testing Hypothesis about Mean of Latent Gaussian Process Julia Bondarenko Helmut-Schmidt University Hamburg University

More information

Randomisation, Replication, Response Surfaces. and. Rosemary

Randomisation, Replication, Response Surfaces. and. Rosemary Randomisation, Replication, Response Surfaces and Rosemary 1 A.C. Atkinson a.c.atkinson@lse.ac.uk Department of Statistics London School of Economics London WC2A 2AE, UK One joint publication RAB AND ME

More information

Nonparametric predictive inference with parametric copulas for combining bivariate diagnostic tests

Nonparametric predictive inference with parametric copulas for combining bivariate diagnostic tests Nonparametric predictive inference with parametric copulas for combining bivariate diagnostic tests Noryanti Muhammad, Universiti Malaysia Pahang, Malaysia, noryanti@ump.edu.my Tahani Coolen-Maturi, Durham

More information

Oikos. Appendix 1 and 2. o20751

Oikos. Appendix 1 and 2. o20751 Oikos o20751 Rosindell, J. and Cornell, S. J. 2013. Universal scaling of species-abundance distributions across multiple scales. Oikos 122: 1101 1111. Appendix 1 and 2 Universal scaling of species-abundance

More information

Local Likelihood Bayesian Cluster Modeling for small area health data. Andrew Lawson Arnold School of Public Health University of South Carolina

Local Likelihood Bayesian Cluster Modeling for small area health data. Andrew Lawson Arnold School of Public Health University of South Carolina Local Likelihood Bayesian Cluster Modeling for small area health data Andrew Lawson Arnold School of Public Health University of South Carolina Local Likelihood Bayesian Cluster Modelling for Small Area

More information

McGill University. Faculty of Science MATH 204 PRINCIPLES OF STATISTICS II. Final Examination

McGill University. Faculty of Science MATH 204 PRINCIPLES OF STATISTICS II. Final Examination McGill University Faculty of Science MATH 204 PRINCIPLES OF STATISTICS II Final Examination Date: 20th April 2009 Time: 9am-2pm Examiner: Dr David A Stephens Associate Examiner: Dr Russell Steele Please

More information

Type I error rate control in adaptive designs for confirmatory clinical trials with treatment selection at interim

Type I error rate control in adaptive designs for confirmatory clinical trials with treatment selection at interim Type I error rate control in adaptive designs for confirmatory clinical trials with treatment selection at interim Frank Bretz Statistical Methodology, Novartis Joint work with Martin Posch (Medical University

More information

Designs for Generalized Linear Models

Designs for Generalized Linear Models Designs for Generalized Linear Models Anthony C. Atkinson David C. Woods London School of Economics and Political Science, UK University of Southampton, UK December 9, 2013 Email: a.c.atkinson@lse.ac.uk

More information

Compare Predicted Counts between Groups of Zero Truncated Poisson Regression Model based on Recycled Predictions Method

Compare Predicted Counts between Groups of Zero Truncated Poisson Regression Model based on Recycled Predictions Method Compare Predicted Counts between Groups of Zero Truncated Poisson Regression Model based on Recycled Predictions Method Yan Wang 1, Michael Ong 2, Honghu Liu 1,2,3 1 Department of Biostatistics, UCLA School

More information

On Fitting Generalized Linear Mixed Effects Models for Longitudinal Binary Data Using Different Correlation

On Fitting Generalized Linear Mixed Effects Models for Longitudinal Binary Data Using Different Correlation On Fitting Generalized Linear Mixed Effects Models for Longitudinal Binary Data Using Different Correlation Structures Authors: M. Salomé Cabral CEAUL and Departamento de Estatística e Investigação Operacional,

More information

ON THE FAILURE RATE ESTIMATION OF THE INVERSE GAUSSIAN DISTRIBUTION

ON THE FAILURE RATE ESTIMATION OF THE INVERSE GAUSSIAN DISTRIBUTION ON THE FAILURE RATE ESTIMATION OF THE INVERSE GAUSSIAN DISTRIBUTION ZHENLINYANGandRONNIET.C.LEE Department of Statistics and Applied Probability, National University of Singapore, 3 Science Drive 2, Singapore

More information

Fall 2017 STAT 532 Homework Peter Hoff. 1. Let P be a probability measure on a collection of sets A.

Fall 2017 STAT 532 Homework Peter Hoff. 1. Let P be a probability measure on a collection of sets A. 1. Let P be a probability measure on a collection of sets A. (a) For each n N, let H n be a set in A such that H n H n+1. Show that P (H n ) monotonically converges to P ( k=1 H k) as n. (b) For each n

More information

Linear, Generalized Linear, and Mixed-Effects Models in R. Linear and Generalized Linear Models in R Topics

Linear, Generalized Linear, and Mixed-Effects Models in R. Linear and Generalized Linear Models in R Topics Linear, Generalized Linear, and Mixed-Effects Models in R John Fox McMaster University ICPSR 2018 John Fox (McMaster University) Statistical Models in R ICPSR 2018 1 / 19 Linear and Generalized Linear

More information

ABC methods for phase-type distributions with applications in insurance risk problems

ABC methods for phase-type distributions with applications in insurance risk problems ABC methods for phase-type with applications problems Concepcion Ausin, Department of Statistics, Universidad Carlos III de Madrid Joint work with: Pedro Galeano, Universidad Carlos III de Madrid Simon

More information

Likelihood and p-value functions in the composite likelihood context

Likelihood and p-value functions in the composite likelihood context Likelihood and p-value functions in the composite likelihood context D.A.S. Fraser and N. Reid Department of Statistical Sciences University of Toronto November 19, 2016 Abstract The need for combining

More information

On Multiple-Objective Nonlinear Optimal Designs

On Multiple-Objective Nonlinear Optimal Designs On Multiple-Objective Nonlinear Optimal Designs Qianshun Cheng, Dibyen Majumdar, and Min Yang December 1, 2015 Abstract Experiments with multiple objectives form a staple diet of modern scientific research.

More information

Information geometry for bivariate distribution control

Information geometry for bivariate distribution control Information geometry for bivariate distribution control C.T.J.Dodson + Hong Wang Mathematics + Control Systems Centre, University of Manchester Institute of Science and Technology Optimal control of stochastic

More information

The Instability of Correlations: Measurement and the Implications for Market Risk

The Instability of Correlations: Measurement and the Implications for Market Risk The Instability of Correlations: Measurement and the Implications for Market Risk Prof. Massimo Guidolin 20254 Advanced Quantitative Methods for Asset Pricing and Structuring Winter/Spring 2018 Threshold

More information

MODELING AND DESIGN TO DETECT INTERACTION OF INSECTICIDES, HERBICIDES AND OTHER SIMILAR COMPOUNDS

MODELING AND DESIGN TO DETECT INTERACTION OF INSECTICIDES, HERBICIDES AND OTHER SIMILAR COMPOUNDS MODELING AND DESIGN TO DETECT INTERACTION OF INSECTICIDES, HERBICIDES AND OTHER SIMILAR COMPOUNDS 1. Introduction Timothy E. O Brien Loyola University Chicago, Department of Mathematics and Statistics

More information

Efficient Experimental Design Strategies in Toxicology and Bioassay

Efficient Experimental Design Strategies in Toxicology and Bioassay STATISTICS, OPTIMIZATION AND INFORMATION COMPUTING Stat., Optim. Inf. Comput., Vol. 4, June 2016, pp 99 106. Published online in International Academic Press (www.iapress.org) Efficient Experimental Design

More information

Statistical Data Analysis Stat 3: p-values, parameter estimation

Statistical Data Analysis Stat 3: p-values, parameter estimation Statistical Data Analysis Stat 3: p-values, parameter estimation London Postgraduate Lectures on Particle Physics; University of London MSci course PH4515 Glen Cowan Physics Department Royal Holloway,

More information

Liang Li, PhD. MD Anderson

Liang Li, PhD. MD Anderson Liang Li, PhD Biostatistics @ MD Anderson Behavioral Science Workshop, October 13, 2014 The Multiphase Optimization Strategy (MOST) An increasingly popular research strategy to develop behavioral interventions

More information