This article explores how structured reasoning and real case studies help assess environmental risks before releasing a gene drive organism
This version (Version 1) was created on 28 May 2025
A genetically modified mosquito carrying an engineered gene drive (EGD) might behave one way in a lab cage, but what would happen if it was released into a diverse, wild, and unpredictable ecosystem? In the future, risk assessors must bridge that gap. They do so with limited data, a great amount of uncertainty, and many questions to be answered.
Risk assessments (RAs) for EGDs are the starting point for deciding whether, and under what conditions, field testing may take place. Their purpose is to identify plausible sources of harm, estimate the likelihood of those harms to occur in different contexts, and assess whether the risks can be managed or reduced to an acceptable level (Combs et al., 2023).
When applied to EGDs, undertaking RAs presents different difficulties than those for other technologies.
Gene drives are designed to spread. They can persist in the host populations over multiple generations, even if they reduce the fitness of the organisms that carry them. They may also move beyond intended populations or regions (Romeis et al., 2020). These key features introduce increased complexity and uncertainty. Most EGDs are being designed to be expressed in populations living in wild environments, where conditions vary and are often not well described. This makes outcomes much harder to predict. The risks themselves also depend on several variables: the species targeted, the design of the gene drive, the traits introduced, the release strategy, and the surrounding ecosystem (Devos et al., 2022b).
To date, no EGD has been released into the environment. However, several EGD animals have reached the stage of confined testing in labs or cages (CBD, 2024). In anticipation of future applications, risk assessment strategies are evolving. As expectations rise, so does the need for consistent and science-based approaches to framing and determining risk.
In this article, I explore how recent assessment exercises have used problem formulation to guide the evaluation of gene drive organisms, typically involving additional structured stakeholder engagement and expert consultations. As new evidence emerges, I will update the content to reflect it.
1. Identifying and framing potential hazards
The range of concerns associated with gene drive organisms is broad, and not all of it is easy to categorise. Some risks echo those found in other genetic or biological control technologies. Others are novel, more complex, or not yet well understood. Framing these concerns into testable hypotheses is a necessary step to build a risk assessment that is both systematic and useful.
Across the RA exercises published in the scientific literature, the most frequently cited EGD hazards relate to disease transmission, ecological disruption, and the persistence or unintended spread of the genetic construct. For example, in a hypothetical small-scale release of male-bias mosquitoes in Burkina Faso, four key concerns stood out: changes in vector competence, non-specific ecosystem effects, uncontrolled spread, and vague future risks such as mutation or microbiome alteration. Together, these made up around two-thirds of stakeholder concerns (Hayes & Hosack, 2025).
In this case, some potential hazards were considered implausible or irrelevant. These included behavioural changes unrelated to biting, changes in land use or farming practices, and indirect consequences from insectary procedures (Hayes & Hosack, 2025). This distinction matters. Risk assessments are not meant to address every imaginable outcome. They focus on what is realistically possible under the specific conditions of release.
In broader contexts, especially for population modification strategies, additional concerns have emerged. These include niche replacement, altered vector behaviour, the evolution of resistance to the introduced trait, and effects on non-target organisms through food web disruption or toxic exposure (Devos et al., 2020; Devos et al., 2022a; Devos et al., 2022b). Water quality has also been flagged, given the ecological role of mosquito larvae in aquatic systems (Roberts et al., 2017).
Having concerns and identifying a potential harm however, is not only about what could go wrong. It also depends on the conditions under which harm might arise. Some potential effects are straightforward to measure while others are diffuse or unfold slowly over time. For example, resistance to gene drives is well documented in laboratory settings but predicting how such resistance might evolve or spread in the wild remains difficult (Devos et al., 2020). Other concerns include gene flow to sibling species through hybridisation, or to unrelated organisms through horizontal transfer, both more likely in genetically diverse wild populations (CBD, 2024). Indirect ecological changes, such as shifts in species interactions or longer-term evolutionary responses, also represent forms of harm that are harder to quantify but cannot be dismissed (Hayes et al., 2017).
Given this diversity of hazards and possible harms, and the fact that many depend heavily on context, how can the risk assessment process be structured to remain transparent, consistent, and fit for real-world decision-making?
2. How problem formulation structures the assessment
Problem formulation is the starting point for assessing the environmental risks of gene drive organisms. It defines what the risk assessment must address and how. For gene drives, where potential impacts may unfold over long timescales and broad geographies, this structured framing step becomes crucial.
Its purpose is to organise concerns into testable questions. Several case studies have applied this method to simulated or hypothetical gene drive releases (Roberts et al., 2017; Connolly et al., 2021; Ogoyi et al., 2024). These exercises are not only technical trials but also practical preparations for regulatory evaluations.
The process typically begins by identifying broad protection goals, such as safeguarding biodiversity, maintaining ecosystem services, or preventing harm to human or animal health. These broad goals are then made operational by specifying exactly what is to be protected, at what scale, for how long, and with what margin of tolerance (Devos et al., 2022a). Figure 1 illustrates the typical components of the risk analysis process for genetically modified organisms.
Once policy goals are defined (regulatory context), assessors devise plausible pathways to harm. Each pathway describes a causal sequence of events by which a gene drive release could lead to an adverse effect. These sequences are constructed based on what is biologically reasonable given the organism, the construct, and the context of release (Connolly et al., 2021; Hayes & Hosack 2025).
The next step is to formulate risk hypotheses for key points in each pathway. These hypotheses help guide the search for evidence and clarify what is still uncertain. An analysis plan is then developed, setting out how those hypotheses could be tested, whether through laboratory data, modelling, literature review, or field trials (Devos et al., 2020; CBD, 2024). This approach, which is graphically summarised in Figure 2, links policy objectives to practical scientific steps and supports clear decision-making.
Problem formulation is not limited to risk identification. It also makes the entire assessment process more transparent. It shows what was included, what was excluded, and why. When potential pathways are deemed implausible, the rationale must be stated, allowing others to judge the robustness of the process (Devos et al., 2022b; CBD, 2024).
Importantly, the process is iterative. As new information becomes available or as stakeholder views shift, the framing can be revised. Engaging stakeholders throughout, especially when setting operational protection goals, can strengthen both the scientific and social relevance of the assessment.
Problem formulation does not remove uncertainty, but it helps structure it in a way that is manageable. It clarifies what is known and what is assumed. It sets the stage for a focused and traceable assessment.
In the next section, I examine how these principles are applied to build actual pathways to harm, using examples from the literature that illustrate both the logic and the complexity of these causal frameworks.
3. Learning from examples in risk assessment
Once protection goals are clearly defined, the next step in environmental risk assessment is to understand how those goals could be compromised. This is done by identifying plausible causal chains, called pathways to harm, that describe how a gene drive release might lead to specific adverse outcomes.
Some are straightforward. Others are longer, more speculative, and sensitive to ecological feedback. They all share a common aim however and that is to clarify what could go wrong, under what circumstances, and how to investigate it.
The following examples illustrate how this approach has been used in structured problem formulation exercises. Each one stems from a different ecological or regulatory context, but all follow the same basic logic: define what could be harmed, construct the causal steps, and identify the evidence needed to assess those risks.
Example 1. Small-scale field study of non-gene drive, genetically modified, male-biased Anopheles mosquitoes in Burkina Faso (Hayes & Hosack, 2025)
In this study, seven potentially harmful outcomes were identified in relation to the proposed release of a non-gene drive, male-biased Anopheles coluzzii mosquito strain. These were transformed into 19 distinct pathways to harm. Each was accompanied by a plan to collect relevant evidence.
Two representative cases are worth highlighting. The first concerns the risk of unintended spread and persistence of the modified mosquitoes. As shown in Figure 3, the authors mapped out four alternative sequences that could lead to this outcome, proposed testable risk hypotheses, and suggested appropriate analysis methods.
In contrast, Figure 4 shows a more speculative pathway, e.g. horizontal gene transfer to humans, leading to potential impacts on fertility or health. While biologically unlikely, the pathway was included to demonstrate transparency and caution.
Although this example did not involve a gene drive organism, the framework is directly applicable to EGD mosquitoes, particularly those with similar vector roles or ecological behaviour.
Example 2. Plausible pathways to harm of a population suppression gene drive targeting Anopheles mosquitoes in West Africa (Connolly et al., 2021)
This exercise mapped 46 plausible pathways to harm. Each pathway was categorised by exposure type (e.g. biting, ingestion, ecological interaction) and by the nature of the potential consequence, from disease transmission to ecosystem disruption.
One illustrative case concerns the suppression of A. gambiae and the resulting ecological vacuum. Figure 5 shows how a competitor species could expand into that niche, ultimately reducing a valued species and affecting ecosystem services.
Figure 6 focuses on human health. It outlines how anatomical changes in the transgenic mosquitoes might increase their vectorial capacity for diseases not previously transmitted by A. gambiae, even as their overall population is reduced.
Figure 7 shifts to animal health. Here, the concern was that increased fitness or insecticide resistance in gene drive mosquitoes could enhance disease transmission to livestock.
Example 3. Regional stakeholder consultations in Africa 2016–2018 (Teem et al., 2019)
Four regional workshops brought together diverse participants to apply problem formulation to gene drive mosquito releases. One major concern was consistently raised across all sessions (Figure 8): the suppression of Anopheles might increase transmission of diseases borne by other mosquito species.
Tables 1 to 3 present examples of specific harms linked to human health, biodiversity, and water quality.
The consultations emphasised how stakeholder concerns, even those lacking immediate data, can shape the early stages of risk framing.
Example 4. Agricultural gene drive applications targeting Drosophila suzukii (Romeis et al., 2020)
This example explores the potential environmental impacts of releasing a suppression gene drive in D. suzukii, an agricultural pest. Pathways to harm were drawn for several ecosystem services: biological control, crop yield, decomposition, and pollination. Figure 9 illustrates plausible pathways to harm for those protection goals.
Many of the identified pathways are indirect. For instance, removal of D. suzukii may reduce food sources for beneficial predators, disrupting biological control networks. Hybridisation and gene transfer to other Drosophila species are also considered, reflecting the complexity of ecological interactions in agricultural landscapes.
Example 5. Results from the workshop “Problem Formulation for the Use of Gene Drive in Mosquitoes” in Africa (Roberts et al., 2017)
This paper did not publish formal pathways to harm, but it did record consensus points on potential impacts. For human health, biting remains the key route of concern, particularly for altered transmission dynamics of Plasmodium species. Other mechanisms, such as toxicity from inserted proteins, were deemed unlikely.
In terms of biodiversity, A. gambiae was not seen as ecologically critical, but interactions with threatened or valued species required consideration. Gene flow within the species complex is expected, while horizontal transfer to unrelated species is not.
For livestock, concerns centred on the possible alteration of disease patterns. Again, exposure through non-biting routes was considered negligible.
These five examples show the diversity of harms that could result from gene drive releases, depending on the various contexts. They also show the value of a structured approach that links protection goals to plausible causal chains. Some pathways share common steps or diverge only at later stages, which makes it essential to distinguish between what is theoretically possible and what is actually plausible under given conditions.
As Devos and colleagues point out, the effectiveness and timing of gene drive action, especially suppression, can strongly influence the likelihood and magnitude of harm, with slower or partial suppression potentially increasing ecological exposure and risk (Devos et al., 2020). Each step in the causal sequence provides an opportunity to construct and test risk hypotheses, using models, existing data, or new experiments. This structured reasoning, rooted in policy-relevant definitions of harm, helps make risk assessment both transparent and adaptable (Devos et al., 2019).
Of course, identifying pathways to harm is only the beginning. Once these pathways are defined and prioritised, attention must shift to how they will be examined in practice. What indicators should be tracked? Which methods are suitable before and after a release? And how can monitoring remain meaningful in the face of ecological complexity and limited baseline data?
4. Lessons from the literature and future needs
Across the literature, there is wide recognition that environmental risk assessment for gene drive organisms requires stronger, more targeted guidance. Some recommendations are technical. Others relate to governance, process, or stakeholder inclusion. Together, they reflect a field that is still finding its footing.
One recurring theme is the need for practical, fit-for-purpose guidance. Authors argue that existing frameworks, though helpful, are not always sufficient for the particular challenges posed by gene drives. A phased, modular approach has been suggested as a way to manage uncertainty over time and allow decisions to adapt as evidence accumulates (Devos et al., 2021; Devos et al., 2022a). This guidance should be usable not only by regulators but also by scientists and institutions at the national level.
A response to this need can be found in the 2024 guidance materials produced under the Convention on Biological Diversity (CBD). These voluntary documents aim to support case-by-case risk assessments of engineered gene drive organisms, offering a structure to frame risk hypotheses, identify relevant information, and assess harms across multiple pathways. While not binding, the CBD materials are a significant step towards a shared understanding of best practices (CBD, 2024).
Stakeholder engagement is another major focus. Several sources stress the value of involving a broad mix of voices early in the process, during the problem formulation stage. This includes scientific experts, regulatory bodies, health authorities, and members of the communities most affected by the release of the target organism (Connolly et al., 2022; Devos et al., 2022b). Engagement is not just a procedural add-on. It helps ensure that assessments reflect real-world values and that protection goals are meaningful to those most impacted.
Monitoring after a release is just as important as the risk assessment before it. It provides the evidence needed to evaluate whether predictions were accurate and whether unanticipated effects are emerging. Authors have emphasised the need for baseline data and robust systems to track changes in ecological indicators over time. In the African context, specific post-release monitoring concerns have already been mapped out by expert groups, reflecting a variety of protection goals such as water quality, biodiversity and health (Ogoyi et al., 2024).
In parallel, modelling plays a central role in formulating and testing risk hypotheses. It helps visualise long-term outcomes, assess uncertainty and compare intervention strategies. Models are particularly useful for gene drives, which often act across generations and landscapes (Combs et al., 2023).
Tools and methods also need updating. Proposals include the use of ecological network analysis to better reflect complex interactions, testing risk hypotheses against realistic worst-case scenarios and considering multiple comparators to put findings into context (Connolly et al., 2022). Transparent, well-documented models are essential, particularly when regulators must rely on simulations to understand long-term outcomes. Greater investment in open modelling tools would also help to reduce conflicts of interest and improve reproducibility (Combs et al., 2023).
In light of these discussions, I would like to add a few observations.
First, there is clear value in systematically publishing problem formulation exercises. These structured assessments, often developed through expert consultations or stakeholder workshops, help build a shared body of evidence. While most current examples focus on Anopheles mosquitoes and malaria control, the approach is equally relevant to other species. These include Aedes aegypti, Culex, Ceratitis, Drosophila, Plutella, and even non-insect species such as mice (Mus musculus) or fungi (Fusarium), which are all in development processes.
Second, the geographic focus must expand. Africa has hosted many of the most advanced consultations and pilot assessments to date, and rightly so. However, Latin America, Southeast Asia, and island nations in the Pacific also face challenges that gene drive technology aims to address. These regions are not yet meaningfully represented in scholarly literature. That needs to change.
Gene drives will eventually move from controlled environments to proposed field trials. When that happens, decisions about risk will carry weight far beyond the lab. With this article, I wanted to show that problem formulation, when applied transparently and thoughtfully, offers a strong foundation for those decisions. The work doesn’t stop at framing the problem, though. It also requires ongoing investment in modelling, monitoring, engagement and shared learning, so that future assessments are not only better, but also more inclusive and more trusted.
For a quick overview, in the following infographic, I highlight the central themes from this article in four core areas:
5. References
CBD Convention on Biological Diversity (2024). Additional voluntary guidance materials to support case-by-case risk assessments of living modified organisms containing engineered gene drives. https://www.cbd.int/doc/c/175e/90b0/89c0c71660cccc1539adf34f/cp-mop-11-09-en.pdf
Matthew A. Combs, Andrew J. Golnar, Justin M. Overcash, Alun L. Lloyd, Keith R. Hayes, David A. O’Brochta & Kim M. Pepin (2023). Synthesizes eco-evolutionary models to inform quantitative risk predictions for gene drives. https://doi.org/10.1016/j.tig.2023.04.004
John B. Connolly, John D. Mumford, Silke Fuchs, Geoff Turner, Camilla Beech, Ace R. North & Austin Burt (2021) Systematic identification of plausible pathways to potential harm via problem formulation for investigational releases of a population suppression gene drive to control the human malaria vector Anopheles gambiae in West Africa. https://doi.org/10.1186/s12936-021-03674-6
John B. Connolly, John D. Mumford, Debora C. M. Glandorf, Sarah Hartley, Owen T. Lewis, Sam Weiss Evans, Geoff Turner, Camilla Beech, Naima Sykes, Mamadou B. Coulibaly, Jörg Romeis, John L. Teem, Willy Tonui, Brian Lovett, Aditi Mankad, Abraham Mnzava, Silke Fuchs, Talya D. Hackett, Wayne G. Landis, John M. Marshall & Fred Aboagye‑Antwi (2022). Recommendations for environmental risk assessment of gene drive applications for malaria vector control. https://doi.org/10.1186/s12936-022-04183-w
Yann Devos, Wendy Craig, Robert H Devlin, Alessio Ippolito, Rosalind A Leggatt, Jeorg Romeis, Richard Shaw, Claus Svendsen & Christopher J Topping (2019) Using problem formulation for fit-for-purpose pre-market environmental risk assessments of regulated stressors. https://www.doi.org/10.2903/j.efsa.2019.e170708
Yann Devos, Barbara Gallani & Leslie G Firbank (2020). Stakeholder workshop “Problem formulation for the environmental risk assessment of gene drive modified insects”. https://www.doi.org/10.2903/sp.efsa.2020.EN-1819
Yann Devos, Michael B. Bonsall, Leslie G. Firbank, John Mumford, Fabien Nogué & Ernst A. Wimmer (2021) Gene Drive-Modified Organisms: Developing Practical Risk Assessment Guidance. https://doi.org/10.1016/j.tibtech.2020.11.015
Yann Devos, John D. Mumford, Michael B. Bonsall, Debora C.M. Glandorf & Hector D. Quemada (2022a). Risk management recommendations for environmental releases of gene drive modified insects. https://doi.org/10.1016/j.biotechadv.2021.107807
Yann Devos, John D. Mumford, Michael B. Bonsall, Ana M. Camargo, Leslie G. Firbank, Debora C. M. Glandorf, Fabien Nogué, Konstantinos Paraskevopoulos & Ernst A. Wimmer (2022b) Potential use of gene drive modified insects against disease vectors, agricultural pests and invasive species poses new challenges for risk assessment. https://doi.org/10.1080/07388551.2021.1933891
Keith R. Hayes, Geoffrey R. Hosack, Genya V. Dana, Scott D. Foster, Jessica H. Ford, Ron Thresher, Adrien Ickowicz, David Peel, Mark Tizard, Paul De Barro, Tanja Strive & Jeffrey M. Dambacher (2017) Identifying and detecting potentially adverse ecological outcomes associated with the release of gene-drive modified organisms. https://doi.org/10.1080/23299460.2017.1415585
Keith R. Hayes & Geoffrey R. Hosack (2025). Problem formulation for a small-scale field study of non-gene drive, genetically modified, male bias mosquitoes in Burkina Faso. https://doi.org/10.1080/10807039.2025.2484229
Dorington O. Ogoyi, Julia Njagi, Willy Tonui, Brinda Dass, Hector Quemada & Stephanie James (2024). Post-release monitoring pathway for the deployment of gene drive-modified mosquitoes for malaria control in Africa. https://doi.org/10.1186/s12936-024-05179-4
Andrew Roberts, Paulo Paes de Andrade, Fredros Okumu, Hector Quemada, Moussa Savadogo, Jerome Amir Singh & Stephanie James (2017). Results from the Workshop “Problem Formulation for the Use of Gene Drive in Mosquitoes. https://www.doi.org/10.4269/ajtmh.16-0726
Jörg Romeis, Jana Collatz, Debora C.M. Glandorf & Michael B. Bonsall (2020) The value of existing regulatory frameworks for the environmental risk assessment of agricultural pest control using gene drives. https://doi.org/10.1016/j.envsci.2020.02.016
John L. Teem, Aggrey Ambali, Barbara Glover, Jeremy Ouedraogo, Diran Makinde & Andrew Roberts (2019) Problem formulation for gene drive mosquitoes designed to reduce malaria transmission in Africa: results from four regional consultations 2016–2018. https://doi.org/10.1186/s12936-019-2978-5