By: Kami Krista, CEO and Co-Founder, Elio and Alissa Monk, Head of Sustainability, ten23 health
Introduction
Why Sustainable Procurement Matters in Drug Design
Pharmaceutical companies face growing pressure to design more sustainable drugs and report their impacts in greater detail, while the required data foundation is largely missing. This pressure is increasingly driven by the growing significance that sustainability plays in public procurement processes, that make up 40% of pharmaceutical sales volume.1 In some countries, sustainability criteria already account for 10 to 30% of procurement evaluation (e.g. UK2, France3, Denmark4, Iceland4, Norway4) and the general industry sentiment is that these developments will continue to spread, as legally binding national emission targets across Europe and other parts of the world are translated into their public procurement processes.
Thus, finding ways to reduce the environmental impact of drugs developed is a key future value driver for pharmaceutical companies, as their tender win rate and the price premium they can command are increasingly also influenced by the sustainability impact of the drugs they develop.
Around 45–50% of a drug's final environmental impact comes from its process inputs,5 and since 80% of total impact is determined during process design,6 the majority is set by technical staff during R&D. Influencing this decision process, both the inputs and the suppliers selected, is therefore a key element of eco-design.
The Data Challenge in Sustainable Procurement
Many pharmaceutical companies have made public commitments to reduce their environmental impact, with 75% of the top 20 pharma companies having explicit eco-design goals7. Scope 3 targets are now commonplace across the sector,8 including those validated through the Science Based Targets initiative. Yet despite the ambition of these commitments, only 4% of biopharma companies are on track to halve their emissions by 20309. Broadly there are two challenges:
1. There is little or no data for the options a scientist is considering, at a level of granularity that is meaningful to the decisions being made. 68% say that a lack of data is the primary obstacle to measuring sustainability performance.10
2. Even when the data exists, the tool infrastructure required to act on them at the point of procurement remains largely absent, so considering it consistently becomes too significant of an overhead.
Current Approaches and Their Main Limitations
The dominant approach to Scope 3 measurement is the spend-based method, in which expenditure is multiplied by an industry-average emission factor for a given product category. While administratively simple, this is directionally unreliable at the level of individual purchasing decisions: two products within the same category and at the same price point can carry vastly different environmental footprints, yet receive an identical score. Collecting more granular data directly from suppliers is the obvious alternative, but in practice the administrative burden is prohibitive- data requests yield inconsistent responses, incompatible methodologies, and response rates that rarely exceed a third of those contacted11. Even where carbon data is obtained, it addresses only one dimension of impact, leaving water use, land use, ecotoxicity, and end-of-life entirely unaccounted for. The result is a sector with genuine sustainability ambition but without the data granularity or quality to translate that ambition into procurement decisions.
Out-of-the-box LLMs and chat interfaces help, but have two major drawbacks. First, their hallucination risk means data quality must always be verified manually. Second, the heterogeneity of supplier sustainability evidence is so high that simple direct comparison is impossible; it requires a well-designed scoring matrix that produces a single comparable value.
Beyond the data limitations, even when that it exists it needs to be embedded in daily decision flows to be considered consistently. Today, no tools exist to do that without adding significant overhead.
Motivation
ten23 health aims to become a net-positive business through a weighted, holistic approach to environmental decision-making that looks beyond climate change and midpoint indicators. It is conducting a comprehensive review of consumables across its non-GMP laboratories (with a parallel GMP strategy that acknowledges the regulatory collaboration required), underpinned by a comparability methodology integrating environmental performance with use-level factors such as time and effort and scientific criteria such as sterility.
Methodology
To support product selection, ten23 partnered with Elio to co-develop a tool that identifies alternatives to existing products and ranks them based on environmental performance. Candidates that perform strongly on environmental grounds and meet scientific requirements are then taken forward into a structured comparability assessment.
Sustainability score
Waiting for sufficient supplier-provided sustainability data is not feasible for near-term procurement decisions, as meaningful data coverage is likely still 5–10 years away. At the same time, single-use consumables remain necessary in many pharmaceutical manufacturing contexts, where there are often no viable process alternatives. However, it is still possible to choose more sustainable versions of the consumables already being used.
The sustainability score therefore uses publicly available product information to generate directionally correct assessments that differentiate comparable products from different suppliers. This allows sustainability to be considered in high-frequency purchasing decisions without requiring full supplier LCAs for every product.
For consumables, the score integrates multiple data components into a single normalised, multi-dimensional score for comparison within product categories. It combines an LCA Material Score, based on the generated bill of materials and ecoinvent-matched impact factors; a Claims Score, based on extracted and classified supplier sustainability claims; a Transportation Score, based on distance, transport mode, and ecoinvent transport factors; and a Production Energy Score, based on manufacturing location and local energy mix.
Raw component scores are mapped to environmental impact categories such as climate change, land use, water use, ecotoxicity, and resource depletion, and weighted based on the impact category weights from the EU Product Environmental Footprint framework 3.1. The final score is then normalised from 0 to 100 (100 being the best) within relevant product categories, based on a pre-defined product category taxonomy.
Comparability assessment
In developing the comparability methodology, ten23 observed that no industry standard currently exists and that the effort required to evaluate sustainable alternatives remains one of the primary barriers to adoption. ten23 is therefore working to establish a shared comparability framework in which verified test data, generated using a standardised methodology across different analytical needs, can be shared across the industry — with the aim of reducing the burden on individual organisations, supporting the transition to more sustainable products, and providing suppliers with meaningful performance insights about their offerings (This work is being progressed as part of a collaboration within the Pharma Eco Design Consortium. Those interested in contributing to or learning more about this initiative are encouraged to reach out to join the discussion).
Demonstrating the approach: Method comparison and case studies/scenarios
Commonly used lab consumables represent the largest environmental footprint by volume and were therefore chosen to be the initial focus for identifying sustainable alternatives that meet both use and performance criteria. In order to illustrate the value of leveraging sustainability scores, this section compares how sustainability information can be gathered and applied to consumable selection. The method comparison contrasts manual search, general AI tools, and Elio across time, data coverage, and limitations. Scenario A then applies this to a direct one-to-one comparison of pipette tips, showing how relative scoring can meaningfully differentiate products. Scenario B extends this to a single-use versus multi-use filtration comparison, illustrating both the value and limitations of relative scoring across process types, and how supplementary data can support conclusions ahead of fully developed absolute impact models.
|
Method Comparison: Manual Search vs General AI Tool vs Elio
|
Comparison element
|
Manual search
|
General AI tool
|
Elio
|
|
Time / product
|
• Comprehensive review: 6–8 hrs (single product)
• Basic review: 1–2 hrs
|
• 10–20 mins data collection and presentation (developing prompts / refinement)
• 1–2 hrs verification of information
|
• ~5 mins (multiple product review)
|
|
Data coverage / quality
|
• Only publicly available / not behind paywalls; in practice very limited to no usable data gathered
• Data points collected in isolation and rarely cross-referenced, given the time cost
|
• Limited to publicly available sources; returns a loose set of (often qualitative) facts
• High hallucination risk — outputs generated with no mechanistic checks against source data
• Retrieves and summarises but does not connect findings into impact chains
• Does not produce a meaningful single score based on an aligned methodology
|
• Enhances publicly available information with additional data points to draw further conclusions
• Uses LLMs largely for extraction of existing data — grounded in retrieved sources — with built-in quality control and validation checks, reducing hallucination risk
• Provides transparency into underlying data sources
• Connects disparate data points into second-order conclusions (e.g. location → grid → emissions) and applies an aligned methodology to produce a single, comparable score
|
|
Limitations
|
• Time commitment makes comprehensive assessment impractical
|
• Supports qualitative understanding and faster research, but not necessarily directionally correct weighting
• May require expert knowledge to correctly interpret outputs
|
• Complex scenarios where many changes are occurring that are not 1:1 replacements are not suitable to evaluate with a relative score (should not be scaled)
|
|
Conclusions
|
• Generally not performed to a high level of data quality because of the time investment; only basic assessments are done, which still require large time investment
|
• A strong alternative to get higher quality information faster than manual collection
• Requires an expert to develop and refine prompts
• Still requires verification
|
• Embedded knowledge — does not require sustainability expertise
|
* For time, factors measured include: locating relevant pages, reading reports, identifying certifications, mapping data to an assessment framework such as EU PEF, and scoring.
|
|
Scenario A – Pipette Tips
| | Pipette 1 | Pipette 2 | Pipette 3 |
|---|
| Combined Score |
Score: 59 |
Score: 58 |
Score: 36 |
| Base evidence used to assess impact for all categories |
• BoM weighting (virgin polypropylene)
• Likely manufacturing location: Helsinki, Finland → road transport Finland to gate (example: Vienna) → production energy emissions intensity in Finland
|
• BoM weighting (virgin polypropylene)
• Likely manufacturing location: Wertheim, Germany → road transport Germany to gate (example: Vienna) → production energy emissions intensity in Germany
|
• BoM weighting (virgin polypropylene)
• No manufacturing location found for the manufacturer
|
| Climate Change – Score & brief summary of evidence |
Score: 54
Production energy emissions intensity of 0.142 kg CO₂-Eq
|
Score: 55
Production energy emissions intensity of 0.425 kg CO₂-Eq
|
Score: 32
Claims by manufacturer: "Environmental benefits include: Renewable Energy"
|
| Water Usage – Score & brief summary of evidence |
Score: 47
Production energy water usage intensity of 0.177 m³
|
Score: 61
Production energy water usage intensity of 0.087 m³
|
Score: 34
(based only on BoM)
|
| Ecotoxicity – Score & brief summary of evidence |
Score: 74
Claims by manufacturer: "Non-toxic"
Production energy water usage intensity of 0.426 CTUe
|
Score: 49
Production energy water usage intensity of 1.316 CTUe
|
Score: 35
(based only on BoM)
|
| Resource Depletion – Score & brief summary of evidence |
Score: 77
Claims by manufacturer: "Packaging and tips are 100% recyclable or can be incinerated." "Reusable and fully autoclavable at 121°C for 20 minutes."
|
Score: 64
Claims by manufacturer: "Environmentally friendly Tip Refills reload empty TipBoxes, and have 20% less waste."
|
Score: 43
(based only on BoM)
|
| Land Usage – Score & brief summary of evidence |
Score: 44
Production energy land usage of 2.076 Pts
|
Score: 58
Production energy land usage of 1.105 Pts
|
Score: 34
(based only on BoM)
|
• The chosen gate is Vienna, Austria, which adapts to the user.
• We assume that all products in a given category require the same absolute amount of energy, unless otherwise specified by the manufacturer. The energy source is therefore what distinguishes them on this element, based on the manufacturing location.
|
|
Scenario B – Multi-use vs Single-use: Buffer Preparation
Introduction
Filtration is required across many laboratory analyses. Where sterile conditions are necessary, single-use filtration systems are mandated. For applications such as buffer preparation for certain conditions, sterility is not required and multi-use alternatives are therefore an option. Multi-use is uncommon in modern lab settings, but viable.
Method
Elio was used to compare single-use filtration systems for buffer preparation, and subsequently extended to include a multi-use alternative in which only the filtration paper is single-use and the funnel and vessel are washed between uses.
In an ideal scenario, whole-process impact would be the preferred unit of assessment; however, as relative scoring does not readily accommodate this, a supplementary assessment was conducted that incorporated the environmental impact of process steps such as washing. In-use factors including time, effort, and scientific performance were assessed using the comparability methodology described early in the article.
Results
* In this scenario, filters were tested for performance for buffer preparation using sub-visible particle testing.
Conclusion
Using Elio, a more sustainable single-use filtration system was identified for applications where sterility is required. For applications where sterility is not required, the multi-use system was shown -through supplementary environmental assessment- to meet scientific requirements while substantially reducing both the environmental and cost impact of filtration.
|
Discussion
Decision-making should not wait for perfect data. In most high-frequency decisions the risk of a significantly worse choice based on directional data is low; the more likely outcome is an equivalent option, so the cost of delay must be weighed. Leveraging AI to scale data collection, preparation, and classification under a clear scoring methodology lets options be weighed in an easily interpretable manner, something a chat-based LLM interface alone cannot deliver.
Relative scoring provides a robust and practical basis for direct one-to-one product comparisons, enabling meaningful differentiation between alternatives within the same use category in a way that is both transparent and reproducible. However, a key limitation emerges when attempting to extend this approach across fundamentally different process types. As illustrated in Scenario B, a single-use product cannot be directly compared to a multi-use system through relative scoring alone, as the processes are not equivalent in scope or structure. To bridge this gap, supplementary information (such as the addition of process-level factors) can be incorporated alongside the relative score to allow informed conclusions to be drawn, even in the absence of a fully integrated model. This pragmatic approach reflects the current state of the field; while absolute impact modelling would ultimately provide the most complete picture, it remains resource-intensive and is not yet routinely accessible for this type of procurement decision.
We also acknowledge that for many companies there is likely a lack of fully integrated methodology for procurement to include sustainability considerations for product decisions and that when performed, is generally done in an ad hoc basis using a brief, max 1-2 hour manual scan. This scan can identify the broad ranking of products correctly but would not reliably separate products within those tiers. The key risk is conflating group-level corporate sustainability credentials with product-specific environmental data- which ultimately is the most realistic measure of actual impact. The product level distinction only emerges from more time and knowledge intense information extraction which is critical to get product level information. Certification scope is easily misread at speed; labels that appear to cover a product may on closer inspection apply only to a specific packaging format. And finally, for a robust ranking, sufficient time must be allowed to trace each claim back to its actual scope of coverage.
Conclusion
Sustainable procurement is becoming a material value driver in pharmaceutical drug design, yet the data foundation needed to act on it consistently does not yet exist. As this article has shown, waiting for comprehensive supplier data is not a viable strategy when meaningful coverage remains years away, and procurement decisions are being made every day. The more practical path is to make directionally correct choices now, using publicly available information structured through a transparent and reproducible scoring methodology.
The work with ten23 health demonstrates that AI, when applied with a clear scoring framework rather than as a standalone chat interface, can extract, classify, and connect disparate data points into a single comparable score at a fraction of the time cost of manual or general AI-assisted approaches. This makes sustainability a practical input to high-frequency purchasing decisions rather than an occasional, ad hoc exercise. Relative scoring provides a robust basis for one-to-one comparisons within a product category, while supplementary process-level assessment extends its usefulness to cases, such as single-use versus multi-use systems, where relative scoring alone falls short.
The approach has clear limitations. Relative scores cannot be meaningfully scaled across fundamentally different process types, public data carries inherent uncertainty. However, these are not reasons to defer action for high-frequency choices being made every day, but markers of where the field must develop.
Outlook
The limitations of relative scoring can ultimately be overcome through absolute impact modelling, which would provide a complete picture of environmental impact across both comparable and non-comparable decisions. Historically, this has been a very resource-intensive process, requiring weeks to months of manual work. While not yet developed for consumables, Elio has developed a computational engine based on similar principles to the relative score, but extended to provide absolute product carbon footprints aligned with ISO 14067 by recursively reconstructing the value chain activities of complex chemicals. This can provide such values within a business day or less for chemical compounds not otherwise covered today, promising to bring rigorous, absolute environmental assessment within reach of routine drug design decisions. While developed specifically for chemical compound inputs, this approach could in theory also be extended to consumables in the future, overcoming the limitations of the current relative score.
References
- Gawronski M, Troein P, Newton M. From Regulated Prices to Prices Set in Tenders: Tendering Landscape in Europe. IQVIA; 2019. Accessed June 30, 2026.
- NHS England. Evergreen sustainable supplier assessment. Accessed June 30, 2026.
- Lenaghan M, Merbler C. Understanding the French carbon score for medicines. Anthesis. March 12, 2026. Accessed June 30, 2026.
- Baltruks D, Sowa M, Voss M. Strengthening Sustainability in the Pharmaceutical Sector. Centre for Planetary Health Policy; 2023. Policy Brief 01-2023. doi:10.5281/zenodo.7682082
- Around 70% of a drug product’s CO2e impact comes from its active pharmaceutical ingredient (API) [a] and ~70% of that stems from intermediary pharmaceutical ingredients (IPI) [b]
[a] Diorazio LJ, Mullen A. Engaging scientists in a sustainability culture. Curr Res Green Sustain Chem. 2022;5:100279. Accessed June 30, 2026.
[b] Verlinden A, Boone L, De Soete W, Dewulf J. Environmental impacts of drug products: the effect of the selection of production sites in the supply chain. Curr Res Green Sustain Chem. 2021;4:100174. Accessed June 30, 2026.
- Lovsin Barle E, Melton T, Judge E. Sustainability by design for pharmaceutical products. Pharm Eng. March/April 2023. Accessed June 30, 2026.
- Analysis by authors
- “More than 70% of our companies have long-term targets for reducing scope 3 GHG emissions.” European Federation of Pharmaceutical Industries and Associations. EFPIA White Paper: Climate Change Q and A. EFPIA; 2022. Accessed June 30, 2026.
- Cytiva. Global Biopharma Sustainability Review 2024. Cytiva; 2024. Accessed June 30, 2026.
- Cytiva. Global Biopharma Sustainability Review 2024. Cytiva; 2024. Accessed June 30, 2026.
- ten23 health internal analysis
About the Authors
Kami Krista is CEO and co-founder of Elio. He has advised large pharmaceutical companies on multi-billion-dollar M&A deals and conducted drug development research at MIT while studying bioengineering at Harvard. Over two decades in sustainability, he has served as a UN youth delegate, advised corporates including IKEA, and joined the steering committee of the CPHI Sustainability Collective.
Alissa Monk is Head of Sustainability at ten23 health, previously Environmental Sustainability Lead at Novartis, with nearly 20 years in environmental science and corporate sustainability. She leads multiple collaborative initiatives, including as an executive member of Go Circular in Life Sciences and steering committee member for My Green Lab, driving systemic change through measuring environmental impact and co-building tools to accelerate sustainable action.
Subscribe to our e-Newsletters
Stay up to date with the latest news, articles, and events. Plus, get special
offers from American Pharmaceutical Review delivered to your inbox!
Sign up now!