A question based approach to drug development
Visser, S.J. de
Citation
Visser, S. J. de. (2003, September 10). A question based approach to drug development.
Retrieved from https://hdl.handle.net/1887/28222
Version:
Corrected Publisher’s Version
License:
Licence agreement concerning inclusion of doctoral thesis in the
Institutional Repository of the University of Leiden
Downloaded from:
https://hdl.handle.net/1887/28222
Cover Page
The handle
http://hdl.handle.net/1887/28222
holds various files of this Leiden University
dissertation
Author: Visser. Samuel Jacob de
The value of research on biomarkers
Estimating the hypothetical additional value of reviews on biomarkers assumes that useful knowledge about biomarkers for the effects of cns drugs in healthy volunteers allows better selection of the appropriate marker for answering a relevant question. In the traditional drug development plan, this would lead to a higher probability of success to enter phase ii (or a lower probability the drug will be abandoned after phase i). To investigate the impact of the phase i to phase ii transition probability, a general drug development plan is constructed usinghistorical data. This development plan has a limited amount of options during the execution of the plan: drugs can ‘fail’ or ‘succeed’ in phase i, phase ii and phase iii. Success in phase iii means payoff of the estimated market value.
The marketvalue is arbitrarily chosen as M¤ 400 making the overall project value positive (otherwise, the drug will not be developed at all). Failure in any of the phases obviously implies no payoff and a negative outcome (equal to the cumulative costs made in the phases until failure). An historical probability is assigned to each of the phase transitions. The defined costs throughout the three clinical phases are also based on historical data [1-2]. The input parameters for this general drug development program are listed in table 1.
ta b l e 1 Hypothetical input parameters of the classic development plan
Parameter Value
Success probability phase i 70% Success probability phase ii 50% Success probability phase iii 85% Costs phase IM¤ 15
Costs phase ii M¤ 60
Costs phase iii M¤ 100
Market value M¤ 400
a value defined by the probability it will happen multiplied by the net cash flow at that point. The estimated project value is the sum of these probability-corrected cash flows of all possible events. In the presented example, the estimated value of the program equals M¤ 27. This is represented in table 2.
ta b l e 2 Value estimation of all possible outcomes of the classic development plan
Possible development Probability Costs Payoff Profit Profit*
outcomes (M¤) (M¤) (M¤) Probability (M¤)
Successful development 0.2975 (70%*50%*85%) 175 400 225 66.9375 Abandoned after phase iii 0.0525 (70%*50%*15%) 175 0 -175 -9.1875
Abandoned after phase ii 0.35 (70%*50%) 75 0 -75 -26.25
Abandoned after phase I0.3 (30%) 15 0 -15 -4.5
Total: 1 (100%) 440 400 -40 27
Another way to represent this is by using a decision tree (generated by decision tools®) showing that the positive project value makes the decision to go ahead with the development of the drug is ‘true’ (Figure 1).
The fact that drug development is a risky process [4-6] can be reflected in the risk profile of a decision tree [7]. The risk profile of this example is presented in Figure 2. The graph is a graphical display of table : all possible ‘profits’ are displayed on the x-axis and the probability this will occur is represented on the y-axis. From this graph it can be concluded that it is most likely the development of the drug will be abandoned after phase ii losing M¤ 75 (35%).
Assuming that knowledge on biomarkers for the effects of cns active drugs eventually has impact on the probability of success from phase i to phase ii, sensitivity analysis on this input value quantitatively estimates the effect of each change of 1% success probability. Varying the success probability phase i 10% around the initial 70% and recalculating the project value yields a linear relationship between the two parameters (Figure 3).
70.0% Phase II -15 M¤45 TRUE Result Phase I 0 M¤27 30.0% 0.3 -M¤15 -15 Phase I M¤27 FALSE 0 M¤0 0 Drug development Continue Quit Success Failure Continue Quit
f i g u r e 1 Decision tree for a classical drug development program
f i g u r e 2 Risk profile of a classical drug development program
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 -250 -200 -150 -100 -50 0 50 100 150 200 300 Value (M¤)
85.0% 0.2975
M¤300 225
TRUE Result Phase III
0 M¤165 15.0% 0.0525 -M¤100 -175 50.0% Phase III -60 M¤165 FALSE 0 M¤0 -75 TRUE Result Phase II 0 M¤45 50.0% 0.35 -M¤60 -75 FALSE 0 M¤0 -15 Success Failure Continue Quit Success Failure
f i g u r e 3 Sensitivity analysis of the success in phase i probability on the project value
25.5 26 26.5 27 27.5 28 28.5 0.62 0.64 0.66 0.68 0.7 0.72 0.74 0.76 0.78
Probability of success phase I
Project Value (M¤)
A similar decision tree can be constructed using the question-based approach. The same overall initial values are used (i.e. the probability the hypothetical compound will reach market launch remains 30%, the overall costs remain M¤175 and the market value payoff remains M¤ 400; table 3). The combination of the probabilities and costs are chosen in a way that the project value estimate is M¤ 27 (similar to the ‘classic’ phase development project value). The estimated probabilities that the drug will successfully answer the question is assumed to be based on consensus between experts and are valid for a new cns active drug (e.g. a new antipsychotic or anxiolytic or antidepressant) [8]. The purpose of replacing all the phases by questions is to try and clarify the real issues that need to be solved throughout the development (see introduction).
ta b l e 3 Hypothetical input parameters question based development plan
Parameter Value
Success action site 75%
Success pharmacological effect 65%
Success clinical efficacy 80%
Success therapeutic window 85%
Success population 90%
Costs action site M¤ 25
Costs Pharmacological effect M¤ 25
Costs Clinical effect M¤ 60
Costs Clinical window M¤ 30
Costs Population M¤ 35
Estimated marketvalue M¤ 400
f i g u r e 4 Risk profile for the question based development tree for the first choices only
This graph shows the first decision has a substantial effect on the risk profile of the development plan. All first choices have different negative outcome values if the drug is abandoned at some later stage. If all questions are successfully answered, there is always a 30% probability the profit will be M¤ 225 independent of the sequence. However (similar to the example shown in table 2), the expected project value also incorporates the probabilities and costs if the drug development is eventually abandoned. Obviously, this reckoning varies for each sequence. For instance, table 4 compares the possible outcomes for the (optimal) sequences starting with ‘pharmacology’ and ‘clinical’.
ta b l e 4 Comparison of the possible outcomes (M¤) in the question based development
for the first questions ‘pharmacology’ and ‘clinical’
1st question: Pharmacology 1st question: Clinical
Outcome Probability Outcome Outcome Probability Outcome*
*Probability Probability -175 3% -6 -175 3% -6 -140 8% -12 -140 6% -8 -80 7% -6 -110 13% -14 -50 16% -8 -85 28% -24 -25 35% -9 -60 20% -12 225 30% 67 225 30% 67 Total: 27 Total: 3 0 0.2 0.4 -250 -200 -150 -100 -50 0 50 100 150 200 250 300 Value (M¤)
Probabilities of possible outcomes
This table shows that if the first question is ‘pharmacology’, the estimated optimal project value is much higher than the estimated optimal project value if the first question is ‘clinical’. Similarly, the optimal choice of the remaining questions can be analysed. From the risk profiles and decision analysis on all sequence options the unique optimal sequence of events can be determined which is given in table 5. This table represents the optimal path among the 120 different successful options of the all-sequence-decision tree and the only sequence with an estimated project value of M¤ 27. All other sequences have lower estimated project values.
ta b l e 5 Optimal path in the question based development tree
Priority ranking Question
1 Pharmacology
2 Site
3 Window
4 Clinical
5 Population
Subsequent sensitivity analyses of the success probabilities on the estimated project value highlight the impact of changed success probabilities on the project value of the question based development tree (Figure 5). Furthermore, sensitivity analyses can serve as a check of the initial model estimations by showing their relevant contributions to the project value.
The reviews presented in this section affect both the ‘clinical’ and ‘pharmacology’ question by increasing the knowledge on available biomarkers to answer these questions in the development of new drugs. Therefore, two-way sensitivity analyses were performed to show how the project value changes with combined changes in the success probabilities for these questions (Figure 6). The data used to construct Figure 6 is represented in the break-even table 6. This table shows the estimated project value for different sets of success probability estimations.
ta b l e 6 Break-even table for changes in the estimated success probability for ‘clinical’ and ‘pharmacology’ on the project value (M¤)
Success clinical Success 72% 74% 75% 77% 78% 80% 82% 83% 85% 86% 88% pharmacology 59% 12 14 16 18 20 22 24 26 28 30 32 60% 13 15 17 19 21 23 25 27 29 31 33 61% 14 16 18 20 22 24 26 28 30 32 35 62% 15 17 19 21 23 25 27 29 31 34 36 64% 15 18 20 22 24 26 28 30 33 35 37 65% 16 18 21 23 25 27 29 31 34 36 39 66% 17 19 21 24 26 28 30 32 35 37 40 68% 18 20 22 25 27 29 31 34 36 39 41 69% 19 21 23 26 28 30 32 35 37 40 42 70% 20 22 24 27 29 31 33 36 38 41 44 72% 20 23 25 27 30 32 35 37 40 42 45
question by 1% does not negatively affect the overall project value. Similarly, for each 1% increase in the estimated success on the ‘clinical’ question, M¤ 1.4 is gained in project value for each review.
Assuming that the knowledge obtained on the available methods for evaluating three cns drugs presented in the three reviews each increase the probability of success on both ‘pharmacology’ and ‘clinical’ with only 1%, the increase in project value in each case is M¤ 0.8 + M¤ 1.4 = M¤ 2.2. If an antipsychotic drug, a
f i g u r e 5 Spider graph after sensitivity analysis on the success probabilities in the question based drug development program
f i g u r e 6 Two-way sensitivity analysis of the success of ‘pharmacology’‚ probability and
success of ‘clinical’‚ on the project value 15 20 25 30 35 40 -10.0% -5.0% 0.0% 5.0% 10.0% 15.0%
% Change from Base Value
Value (M¤) Success site Success pharmacology Success clinical Success window Success population 0.585 0.598 0.611 0.624 0.637 0.65 0.663 0.676 0.689 0.702 0.715 0.72 0.768 0.816 0.864 10 15 20 25 30 35 40 45 Project value (M¤)
Probability of success 'pharmacology'
r e f e r e n c e s
1 Nefarma. De ontwikkeling van een geneesmiddel; van molecuul tot medicijn. Farma Feiten 2001;1-3 2 EFPIA. 2000-2001 The year in review. 1-34. 2001 3 Clemen RT, Reilly T. Making hard decisions, 2nd
Edition, Duxbury Thomson Learning, 2001: 1-733 4 DiMasi JA, Hansen RW, Grabowski HG, Lasagna L.
Cost of innovation in the pharmaceutical industry. J.Health Econ. 1991; 10:107-142 5 DiMasi JA. New drug development in the United
States from 1963 to 1999. Clin.Pharmacol.Ther. 2001; 69:286-296
6 DiMasi JA. Risks in new drug development: approval success rates for investigational drugs. Clin.Pharmacol.Ther. 2001; 69:297-307 7 Winston WL, Albright SC. Practical Management
Science, 2nd Edition, Duxbury Thomson Learning, 2001: 1-953
