Doctoral thesis · ETH Zurich · 2026
Evidence gap, priorities, and a prespecified simulation framework for ranking under uncertainty. Department of Chemistry and Applied Biosciences. Supervised by Prof. Roger Schibli and Prof. Alessio Figalli, with Prof. Frank Bretz as External Institute advisor.
Five build on one another. The sixth stands on its own.
Each sustainability pillar was operationalised as a single measured criterion.
Eight operational elements of published guidance, examined against each sustainability pillar. No sector-specific guidance on trial carbon was found under the carbon-based coding rule. Source: Singh et al., Clinical Pharmacology & Therapeutics, 2023.
Four hundred and forty-seven people consented, and in every pillar the largest single group ranked nothing at all. That is a finding about the state of the field, not a defect in the survey: the priorities exist, but they are held by a minority of the people asked.
Percentages use a denominator of all 447 participants, so the ranked shares and the not-ranked share describe the whole cohort. In every pillar, more participants declined to rank than chose any single criterion.
Not ranked: 160 of 447 (35.8%)
Not ranked: 189 of 447 (42.3%)
The published table carries no Other column for this pillar, though the article’s own text quotes other at 3.4 per cent. Every economic row is short by exactly that column: 15 responses overall.
Not ranked: 194 of 447 (43.4%)
The article’s Methods name three social criteria. The published table has a column for two of them, and none for trial participants living far from the site.
Source: Singh, Sverdlov, Beckman and Burden, Clinical and Translational Science 18:e70241, 2025, reproduced in my thesis at Table 3.3. The two notes above are my thesis’s, not the article’s.
| Pillar | Ranked first | n | Per cent of 447 |
|---|---|---|---|
| Environmental | Greenhouse gas | 100 | 22.37 |
| Environmental | IMP disposal | 63 | 14.09 |
| Environmental | Device and technology recycling | 49 | 10.96 |
| Environmental | Other | 38 | 8.50 |
| Environmental | Two or more criteria | 37 | 8.28 |
| Environmental | Not ranked | 160 | 35.79 |
| Economic | Trial probability of success | 67 | 14.99 |
| Economic | Time to market | 53 | 11.86 |
| Economic | R&D costs | 42 | 9.40 |
| Economic | Regulator’s probability of success | 40 | 8.95 |
| Economic | Two or more criteria | 41 | 9.17 |
| Economic | Not ranked | 189 | 42.28 |
| Social | Patient convenience | 104 | 23.27 |
| Social | Diverse population | 87 | 19.46 |
| Social | Two or more criteria | 56 | 12.53 |
| Social | Other | 6 | 1.34 |
| Social | Not ranked | 194 | 43.40 |
The framework is a research instrument, not a tool a company could deploy today. Calibration against a real portfolio, an independent implementation, and comparison against realised outcomes all remain outstanding.
Checks on the ranking put the framework short of the precision a real funding decision would need. Under the literature-informed evidence position it returns a hold at five of the thirteen named weighting landmarks.
Under the literature-informed evidence position, the framework returns a hold at five of the thirteen named weighting landmarks. All five place between 0.400 and 0.600 of the weight on the economic criterion.
Weight on the economic criterion
At the other eight landmarks the rank-1 acceptability of the leading portfolio runs from 0.5375 to 0.6096. The evidence position was set from clinical-development experience, and changing it changes the decision status. Individual landmark positions are not plotted; only the band the five holds fall within is. Source: my thesis, Chapter 6.
The analytical result stands without any of that. A company can examine a sustainability limit it has already set, from its own performance, and ask the three questions directly.
Why the limits are on this page. A framework presented without its boundaries invites use it cannot support. Keeping verified results and illustrative ones visibly distinct is not modesty — it is the reason the verified ones are worth anything.
Twenty-three peer-reviewed articles across sixteen years — trial reports first, then estimand and probability-of-success methodology. The last two are first-author papers on sustainability in portfolio and trial decisions, and they are what the doctorate builds on.
Six in 2015 alone. The last two are the first-author sustainability papers the doctorate builds on.
Peer-reviewed journal articles only. One book chapter, Bayesian methods in pharmaceutical research (2020), is counted in the totals above but not plotted. Complete record at ORCID 0009-0001-2092-007X.
| Year | Articles | Of which sustainability |
|---|---|---|
| 2009 | 3 | 0 |
| 2010 | 1 | 0 |
| 2011 | 1 | 0 |
| 2012 | 0 | 0 |
| 2013 | 0 | 0 |
| 2014 | 0 | 0 |
| 2015 | 6 | 0 |
| 2016 | 3 | 0 |
| 2017 | 0 | 0 |
| 2018 | 0 | 0 |
| 2019 | 3 | 0 |
| 2020 | 0 | 0 |
| 2021 | 1 | 0 |
| 2022 | 1 | 0 |
| 2023 | 2 | 1 |
| 2024 | 1 | 0 |
| 2025 | 1 | 1 |
First author on both sustainability papers. Twenty-three peer-reviewed publications and one book chapter in total — complete record at ORCID 0009-0001-2092-007X.