Portfolio Management & Optimization
Pipeline Physics' assistance with portfolio and pipeline management is unique in three ways: (1) our use of stochastic optimization, (2) our creation of complementary heuristic models and (3) our management of uncertainty, via our annealing process.
Stochastic portfolio optimization (instead of simulation optimization): Before addressing stochastic optimization let's look its cousin, simulation optimization, which a mainstay of current project portfolio management (PPM). Simulation optimization marries portfolio optimization with Monte Carlo analysis, an intense sensitivity analysis, to estimate the range of results that current investments can produce. The total process considers current decisions and final results, but it ignores the interim decisions that occur as events unfold.
Stochastic optimization recognizes that executives make and adjust decisions as events unfold, and it exploits sequences of decisions to create value while limiting downside risk, raising upside potential and achieving strategic goals. Because it considers sequences of decisions, stochastic optimization creates much more value than current PPM practices, including simulation optimization.
Importantly, stochastic optimization is granular, so it can answer important questions, such as:
- Given our phase I investments, what future scenarios might we face for phase IIA? Given our phase II investments, what the future scenarios one might we face for phase III? Will we have enough phase III projects to maintain our stock price? If some key projects fail, how soon will we know and what options will be available?
- How do today's investments affect the strategic situations we might face tomorrow, including one's goals for market access and penetration, brand development and leading the industry in key therapeutic areas?
Additionally, we help our clients by addressing the key issues presented below.
Heuristic models that complement stochastic optimization: Many PPM experts view heuristic models as inferior decision methods, but heuristic models can solve a problem that many optimization models produce. Optimizations models are complex, and the sometimes seem like opaque, black box dispensers of decisions. In contrast, a well-designed heuristic model provides understanding and transparency while achieving near optimal results. Heuristic models are ideal for performing "what if" experiments during meetings and discussions, to illuminate key relationships and results, which helps executives make good decisions.
For its clients, Pipeline Physics builds two models: an optimization model and a heuristic model. Starting with the optimization model we explore the physics of a situation, identify key relationships and provide analyses and recommendations. Then we build a heuristic model to approximate the optimal decisions while presenting the key relationships quickly and clearly.
Managing uncertainty: Current PPM practices manage uncertainty by performing a sensitivity analysis. While we perform sensitivity analyses, Pipeline Physics is pioneering a new approach, called annealing. We estimate the frequency and severity of decision errors caused by a model's imperfect assumptions and errors in its data. We then design our models to minimize the cost of these errors, which makes our models more robust to uncertainty, including unknown-unknowns. At Pipeline Physics we anneal all our models.
Key issues Pipeline Physics addresses with its models
Key issues we address with our models include:
Strategic issues:
- Simultaneously considering risks, financial goals and corporate strategy.
- Balancing investments among early stage trials, late stage trials and current revenue streams.
Pipeline issues:
- The value created when discovery and development resolve uncertainty and risk.
- Compounds' probabilities of technical success and phases' success rates.
- Trials' and programs' cycle-times, because getting medicines to market sooner saves precious patent protected time for sales.
Operational issues:
- Pipeline management: Investing in phase I to create good scenarios for phase II. Investing in phase II to create good scenarios for phase III.
- Back-up (follow-on) compounds: Sequencing backup compounds to maximize value, especially considering cycle-times, the probabilities of technical success and the information produced by the lead compound's trials.
- Sequencing indications: Testing indications in
parallel or in sequence to maximize value. Selecting the right
indication(s) to be first for phase IIA (proof of concept),
phase III, regulatory approval and initial negotiations with
payers. Forming good assumptions about launch date, price, impact
on sales force and probability of success. Creating flexibility
so one can adjust as assumptions change.
- Multiple trials of varying sizes: Maximizing value over all trials, considering both larger trials, which reduce the probabilities of false-positives and false-negatives, and smaller trials, which allow more compounds to be tested while honoring budgets.
- Adaptive phase II trials: Phase II trials decide the dosages, biomarkers and patient populations to be tested in phase III. Yet, they begin with great uncertainty. Adaptive trials create flexibility that increases a drug's likelihood of phase III success while improving its risk-benefit profile and increasing its value.
- Adaptive phase III trials: When key trial issues are still uncertain, such as dosages and biomarkers, adaptive designs can reduce costs and cycle-time, or if needed, extend them, all while increasing a compound's probability of success. Generally, adaptive trials implement a basic principle for managing uncertainty: when uncertainty is resolving, delay decisions until the last responsible moment.
- Coordinating phase II and phase III designs: Phase II has such a strong effect on Phase III's probability of success that jointly designing phase II and phase III trials can create more value than designing the trials separately.
- Dosage levels: Selecting dosages and determining the number of dosages to test in phase II and phase III trials affects a compound's probability of phase III success, its risk-benefit profile and the likelihood and magnitude of reimbursements from payers.
- Confirming biomarkers: Trial designs for testing biomarkers affect costs, cycle-time, the probability of success and the patient populations addressed by an indication. Decisions regarding the number of biomarkers to test and their sequence of testing are similar to those for lead and back-up compounds. Importantly, when making these decisions, one should use a portfolio approach, rather than considering each compound's program independently. For example, in oncology one may have multiple lead candidates and each candidate may have multiple associated indications, backup compounds, primary biomarkers and exploratory biomarkers. Maximizing value, minimizing risk and increasing flexibility, so one has good options if the initial trials fail, requires a portfolio approach.
© 2015 Pipeline Physics. All rights reserved.