Design and prioritization of the French surveillance program of health effects in the context of exposure to endocrine disruptors using a Delphi study

Fifty-nine health effects were identified through a narrative review of the literature and contributions from participants. These effects were classified into eleven categories corresponding to endocrine glands or biological functions affected by EDs (Table  2 ). These categories encompass a wide range of health effects attributed to EDs, reflecting the systemic impact these chemicals can have on the human body. The categorization serves as a foundation for prioritizing which health effects should be monitored and addressed within the context of ED exposure. Table 2 Categories of health effects suspected to be related to exposure to endocrine disruptors according to the literature review CATEGORY 1 : Impaired reproductive health in women Endometriosis Primary ovarian insufficiency Fibroids Irregular menstrual cycles Polycystic ovary syndrome (PCOS) Premature or early menopause Gestational diabete CATEGORY 2 : Impaired reproductive health in men Cryptorchidism Hypospadias Erectile dysfunction* Impaired sperm quality Testicular cancer Benign prostatic hyperplasia (BPH)* CATEGORY 3 : Impaired reproductive health without sex distinction Altered sex ratio at birth Precocious puberty Adverse pregnancy outcomes (miscarriage, pre-eclampsia, premature delivery, low birth weight, fetal death) Reduced fertility/under-fertility Infertility Intersex Variation (IV)* CATEGORY 4 : Thyroid disorders Hyperthyroidism Congenital hypothyroidism Hypothyroidism or subclinical hyperthyroidism CATEGORY 5 : Paediatric neurodevelopmental disorders Behavioural disorders : emotional disorders, relational disorders and cognitive disorders Neurodegenerative diseases in adults : Alzheimer and Parkinson* Intellectual deficit -Decrease in intelligence quotient points (IQ) Attention deficit hyperactivity disorder (ADHD) Cerebral palsy Autism spectrum disorder (ASD) CATEGORY 6 : Hormone-dependent cancers Breast cancer Thyroid cancer Prostate cancer Colon cancer (non-hormone dependent)* Endometrial cancer Lung cancer (non-hormone-dependent)* Ovarian cancer CATEGORY 7 : Adrenal disorders Cushing’s disease (chronic hypercortisolism) Addison’s disease (Insufficiency hyposecretion of adrenal hormones) CATEGORY 8 : Bone disorders Bone fractures Osteoporosis Impaired skeletal development Enamel development anomalies, hypomineralization, dental fluorosis, hyperdontia, hypodontia CATEGORY 9 : Metabolic disorders Being overweight and obesity Type 2 diabete Type 1 diabete Metabolic syndrome Cardiovascular diseases Non-alcoholic fatty liver disease* CATEGORY 10 : Immune function disorders Allergies Lymphoma and leukaemia in children Autoimmune thyroid disease (e.g. Basedow’s disease) Disorders of haematopoiesis and malignancies Asthma CATEGORY 11 : Skin and eye disorders Chloracne Ageing of the skin Skin pigmentation disorders Ocular surface impairment Atopic dermatitis Retinal disorders Skin cancer ects added because proposed by Delphi panel Categories of health effects suspected to be related to exposure to endocrine disruptors according to the literature review Cryptorchidism Hypospadias Erectile dysfunction* Impaired sperm quality Testicular cancer Benign prostatic hyperplasia (BPH)* Hyperthyroidism Congenital hypothyroidism Behavioural disorders : emotional disorders, relational disorders and cognitive disorders Neurodegenerative diseases in adults : Alzheimer and Parkinson* Intellectual deficit -Decrease in intelligence quotient points (IQ) Attention deficit hyperactivity disorder (ADHD) Cerebral palsy Autism spectrum disorder (ASD) Bone fractures Osteoporosis Impaired skeletal development Enamel development anomalies, hypomineralization, dental fluorosis, hyperdontia, hypodontia Being overweight and obesity Type 2 diabete Type 1 diabete Metabolic syndrome Cardiovascular diseases Non-alcoholic fatty liver disease* ects added because proposed by Delphi panel The Delphi consultation process involved two groups: the scientific group and the societal group. The scientific group was composed of 29% academic researchers, 19% clinical physicians, 35% researchers, 12% individuals in other roles, and 6% university professors and hospital practitioners. Out of the 91 participants who registered for the scientific questionnaire, 52 completed the first round (57% response rate), 39 completed the second round (75% response rate), and 34 completed the third round (an 87% response rate). The societal group consisted of 33% associations, 17% learned societies, 11% elected officials and community representatives, and 39% individuals in other roles, such as industrial actors involved in the ED theme. Among the 35 participants registered for the societal questionnaire, 18 completed the first round (a 51% response rate), 12 completed the second round (a 66% response rate), and 9 completed the third round (a 75% response rate). In the scientific group, the largest number of respondents were in the categories of metabolic disorders (24 responses), male reproductive health (21 responses), female reproductive health (18 responses), and gender-neutral (18 responses). Three categories of health effects could not be presented in round 2 due to a lack of participants: adrenal disorders, bone disorders, and skin and eye disorders. In the societal group, the highest number of respondents were in the categories of metabolic disorders (12 responses), thyroid disorders (12 responses), and neurodevelopmental disorders (12 responses) (see Appendix A). Out of the 59 proposed health effects, 43 could be assessed in the consultation process. The results are summarized in Table  3 , with detailed statistical analyses available in Appendix B. The health effects were categorized based on their prioritization. Twenty-one health effects were identified as priorities to be surveilled for their links with EDs: 13 with a high priority and 8 with a moderate priority. Among these, Hurgent [ 9 ] had already prioritized several reproductive health effects in 2015. Some are currently being surveilled by SpFrance (endometriosis, cryptorchidism, precocious puberty, alteration of sperm quality, testicular cancer), while others are not yet surveilled (ovarian and endometrial cancers). Effects beyond the realm of reproductive health were also assessed as high priorities and are not currently surveilled: metabolic disorders (overweight and obesity, cardiovascular diseases, type 2 diabetes, metabolic syndrome), neurodevelopmental disorders in children (behavioral disorders, intellectual deficits, Attention Deficit/Hyperactivity Disorder), and lymphoma/leukemia in children. Twenty-three health effects were classified as low or non-priorities. Sixteen effects could not be prioritized by the end of the consultation, either due to a lack of participants (skin and eye disorders, adrenal disorders, bone disorders) or due to a lack of consensus (erectile dysfunction, benign prostatic hyperplasia, allergies). Table 3 Prioritization of 43 endocrine disruptor-related health effects based on the weight of evidence and interest in developing surveillance estimated from the Delphi method Priority category Prioritization criterion 1: Weight of evidence Strong Moderate Low Undocumented Priortization criterion 2 : Interest in setting up a surveillance (severity , evolution of incidence , societal concern) Strong High priority High priority Low priority Non priority - Breast cancer - Prostate cancer - Endometriosis - Cardiovascular diseases - Endometrial cancer - Ovarian cancer - Lymphoma and leukaemia in children - Autism spectrum disorder (ASD) - Neurodegenerative diseases in adults : Alzheimer and Parkinson - Thyroid cancer* - Colon cancer (non-hormone dependent) - Lung cancer (non-hormone-dependent) - Disorders of haematopoiesis and malignancies Moderate High priority Moderate priority Non priority Non priority - Impaired sperm quality - Precocious puberty - Infertility - Being overweight and obesity - Cryptorchidism - Reduced fertility/under-fertility - Testicular cancer - Attention deficit hyperactivity disorder (ADHD) - Type 2 diabete - Metabolic syndrome - Behavioural disorders - Asthma - Intellectual deficit - Decrease in intelligence quotient points (IQ) - Type 1 diabete - Hyperthyroidism - Cerebral palsy Low Moderate priority Low priority Non priority Non priority - Hypospadias - Adverse pregnancy outcomes - Polycystic ovary syndrome (PCOS) - Irregular menstrual cycle - Altered sex ratio at birth - Hypothyroidism and subclinical hyperthyroidism - Non-alcoholic fatty liver disease - Primary ovarian insufficiency - Gestational diabete - Premature or early menopause - Hypothyroïdie - Intersex Variation (IV) - Uterine fibroid - Autoimmune thyroid disease *This health effect was the subject of a strong divergence of opinion regarding the level of WoE (“strong” VS “undocumented”) and the prioritization result does not reflect the overall opinions of the participant panel. Prioritization of 43 endocrine disruptor-related health effects based on the weight of evidence and interest in developing surveillance estimated from the Delphi method Priortization criterion 2 : Interest in setting up a surveillance (severity , evolution of incidence , societal concern) - Breast cancer - Prostate cancer - Endometriosis - Cardiovascular diseases - Endometrial cancer - Ovarian cancer - Lymphoma and leukaemia in children - Autism spectrum disorder (ASD) - Neurodegenerative diseases in adults : Alzheimer and Parkinson - Thyroid cancer* - Colon cancer (non-hormone dependent) - Lung cancer (non-hormone-dependent) - Disorders of haematopoiesis and malignancies - Impaired sperm quality - Precocious puberty - Infertility - Being overweight and obesity - Cryptorchidism - Reduced fertility/under-fertility - Testicular cancer - Attention deficit hyperactivity disorder (ADHD) - Type 2 diabete - Metabolic syndrome - Behavioural disorders - Asthma - Intellectual deficit - Decrease in intelligence quotient points (IQ) - Type 1 diabete - Hyperthyroidism - Adverse pregnancy outcomes - Polycystic ovary syndrome (PCOS) - Irregular menstrual cycle - Altered sex ratio at birth - Hypothyroidism and subclinical hyperthyroidism - Non-alcoholic fatty liver disease - Primary ovarian insufficiency - Gestational diabete - Premature or early menopause - Hypothyroïdie - Intersex Variation (IV) - Uterine fibroid - Autoimmune thyroid disease *This health effect was the subject of a strong divergence of opinion regarding the level of WoE (“strong” VS “undocumented”) and the prioritization result does not reflect the overall opinions of the participant panel. Despite reaching a consensus on many aspects, the consultation revealed divergent opinions among participants, particularly regarding the WoE and the trends in the incidence of health effects. These differences were identified through two main methods: the analysis of response dispersion (standard deviation) and qualitative analysis of participant justifications. The question of the WoE generated the most disagreement among experts, reflected in a wide dispersion of responses and conflicting opinions. Notably, the highest standard deviation was observed for thyroid cancer, where experts were evenly split, with four rating the WoE as ‘strong,’ and four ‘undocumented.’ This disparity indicates significant disagreement within the panel. In addition to the standard deviation, qualitative analysis helped shed light on these differing opinions. Experts who indicated ‘strong’ WoE argued that there were numerous studies on the subject and provided literature, including a 2020 systematic review demonstrating links between organochlorines and some PCBs and the risk of thyroid cancer [ 30 ]. However, experts who indicated ‘undocumented’ did not provide any reasoning. Disagreement was also evident regarding the WoE for immune system disorders, though no dissenting opinions were explained or justified. This could be attributed to the small number of experts who responded on this topic (only three experts) and the limited availability of studies on the subject. On the scientific side, several participants had divergent opinions on whether the incidence of health effects was truly increasing or if it was due to improved diagnosis. Specific health effects where this challenge was noted include : ADHD and ASD: Some participants argued that better diagnostic methods might explain the perceived increase in cases. Endometriosis: Similar concerns were raised about the distinction between actual incidence and improved detection. Neurodegenerative Diseases, Prostate Cancer, Testicular Cancer, and PCOS: Some opinions suggested that the rise in incidence might be more closely related to an aging population rather than increased exposure to endocrine disruptors. ADHD and ASD: Some participants argued that better diagnostic methods might explain the perceived increase in cases. Endometriosis: Similar concerns were raised about the distinction between actual incidence and improved detection. Neurodegenerative Diseases, Prostate Cancer, Testicular Cancer, and PCOS: Some opinions suggested that the rise in incidence might be more closely related to an aging population rather than increased exposure to endocrine disruptors. Concerning the societal component, the statement “The French population is concerned about the occurrence of this health effect in the general population” generated the most disagreement among stakeholders. This was particularly true for conditions like hypospadias and neurodegenerative diseases in adults. However, no justifications were provided for these dissenting opinions. A preliminary hierarchy of health effects based on the WoE was also conducted using the method developed by Hurgent (when data were available) [ 9 ]. This hierarchy aligns with the consultation results, confirming the ‘strong’ WoE or several health effects previously categorized as having ‘sufficient’ evidence in the literature (e.g., breast cancer, prostate cancer, endometriosis, cryptorchidism). However, limitations such as the lack of a universal methodology for evaluating WoE, incomplete and sometimes contradictory scientific data, and the need for more up-to-date research highlight the prospective nature of this ranking. Two health effects (uterine fibroid and autoimmune thyroid diseases) were classified as ‘undocumented’ by the consultation despite being categorized as having ‘sufficient’ evidence in previous literature reviews. This discrepancy underscores the ongoing need for robust methodologies and updated data to accurately evaluate the WoE for health effects linked to endocrine disruptors.

To select and prioritize relevant indicators for the surveillance of effects linked to endocrine disruptors (EDs) in France, considering literature data and the views of multiple EDs stakeholders and experts, we designed a four-stages methodology (see Fig.  1 ). The first stage aimed to establish an exhaustive list of health effects linked to EDs exposure for prioritization. The second stage involved developing a method to estimate the prioritization criteria. The third stage was a three-round Web-Delphi process with a panel of international experts and national stakeholders to reach consensus on the prioritization criteria. The fourth stage combined the prioritization criteria to prioritize the health effects, followed by critical analyses of the results. Fig. 1 Overview of the 4-stage approach used to prioritize key health effects to track in the context of EDs exposure based on literature data and a consensus method Overview of the 4-stage approach used to prioritize key health effects to track in the context of EDs exposure based on literature data and a consensus method The first stage aimed to identify, as exhaustively as possible, all the health effects suspected of being connected to exposure to EDs. This identification was based on weight of evidence (WoE) for linking health effect to an endocrine mode of action from a mechanistic perspective, thereby focusing on biological plausibility. This list served as the foundation for continuously updating and validating suspected pathologies with new knowledge thanks to a comprehensive review of the existing literature. The WHO and European Commission reports in 2012 provided an initial list of suspected pathologies linked to EDs [ 2 , 12 ]. This list was then expanded and validated with new knowledge gained since 2012 through a narrative literature review. Health effects were also proposed by participants in the Delphi consultation [ 18 , 29 ]. During the first round of consultations, we also asked participants if they were aware of any health effects suspected to be related to EDs that were not addressed by the Delphi process. Relevant suggestions were incorporated in the second round for consideration in the prioritization process. Initially, we focused solely on health effects for the purpose of prioritization and planning. However, during the consultation, we explored the possibility of identifying these effects through biological parameters corresponding to the mechanisms of action of EDs (such as epigenetic mechanisms and inflammatory processes) and investigated the potential for biomarkers to characterize exposure to EDs. We prioritized health effects based on two criteria: (1) the WoE supporting the association between the occurrence of the health effect and exposure to EDs and (2) a composite criterion for the epidemiological and societal value of surveilling this effect. This composite criterion combines the severity of the health effect, the changes in the incidence rate of the health effect (known or suspected), and the level of concern about the health effect in French society. WoE involves structuring the synthesis of lines of evidence, which may vary in quality, to determine the extent of support for hypotheses [ 12 ]. Our hypothesis is that there is a relationship between exposure to EDs and adverse health effects. The panel evaluated the WoE by considering all the knowledge within their field of expertise and sources of data they were aware of (including epidemiological and toxicological studies, biochemistry, cell biology, studies of mechanisms, and effects on laboratory animals, expology, etc.). The participants categorized WoE into four levels: Strong : Specific effects of exposure to EDs, Moderate : Relatively specific effects of exposure to EDs, Low : Less specific effects of exposure to EDs, where the proportion attributable to EDs is not yet known, Undocumented : Effects not currently worth surveillance for their relationship with EDs due to insufficient scientific data to justify this link. Severity. Strong : Specific effects of exposure to EDs, Moderate : Relatively specific effects of exposure to EDs, Low : Less specific effects of exposure to EDs, where the proportion attributable to EDs is not yet known, Undocumented : Effects not currently worth surveillance for their relationship with EDs due to insufficient scientific data to justify this link. Severity. To assess the severity of a health effect, we used the method developed by the Centre for Research, Studies, and Documentation in Health Economics, which classifies diseases based on the degree of severity using two morbidity indicators: risk to life and physical disability [ 30 ]. We modified the physical disability indicator to a global indicator of “quality of life,” which is more suitable for the chronic effects of EDs. Severity was assessed considering: Risk to life: A fatal prognosis, similar to a severe condition that could lead to death in the short-term or theoretically life-threatening in the longer term. Health-related quality of life: The WHO distinguishes six areas that impact quality of life [ 31 ]. For this project, to facilitate participants’ assessment of the effect of a disease related to EDs on quality of life in a simple and easily understandable way, while maintaining robustness, three dimensions of health-related quality of life were chosen: physical, psychological, and social functioning. Risk to life: A fatal prognosis, similar to a severe condition that could lead to death in the short-term or theoretically life-threatening in the longer term. Health-related quality of life: The WHO distinguishes six areas that impact quality of life [ 31 ]. For this project, to facilitate participants’ assessment of the effect of a disease related to EDs on quality of life in a simple and easily understandable way, while maintaining robustness, three dimensions of health-related quality of life were chosen: physical, psychological, and social functioning. Participants had to judge the two previous criteria and combine them to classify severity into three levels: very severe, severe, and mild (see Supplementary Material). Some effects may have different severity levels depending on circumstances, such as age, sex, medical history, etc. For this project, participants were required to assess the severity of the health effect linked with EDs by considering the worst possible case. The incidence of a disease is the number of new cases of that disease observed over a given period. Participants classified the incidence rates over the last 20 years into four levels: increasing, stable, decreasing, and no opinion. Societal concern was estimated using a scale of agreement toward a series of proposals related to the perception of the level of concern in the general population, the level of information and prevention, the level of scientific data available, and the value of surveilling each stakeholder interviewed (see Supplementary Material). We assigned points for severity, incidence rate, and societal concern and combined them to classify the final estimation of criterion 2 into three categories: high, moderate, low (see Supplementary Material). Health effects are categorized into priority groups by combining the two criteria listed in Table  1 . These criteria are : Weight of Evidence (WoE): This criterion evaluates the level of scientific support for the association between exposure to endocrine disruptors (EDs) and specific health effects. Composite Criterion for Epidemiological and Societal Value: This criterion combines the severity of the health effect, changes in the incidence rate of the health effect, and the level of societal concern about the health effect. Weight of Evidence (WoE): This criterion evaluates the level of scientific support for the association between exposure to endocrine disruptors (EDs) and specific health effects. Composite Criterion for Epidemiological and Societal Value: This criterion combines the severity of the health effect, changes in the incidence rate of the health effect, and the level of societal concern about the health effect. Table 1 Priority classification based on the assessment of the weight of evidence and value of surveilling Prioritization criterion 1: Weight of evidence Strong Moderate Low Undocumented Priortization criterion 2 : Interest in setting up a surveillance (severity , evolution of incidence , societal concern) Strong High priority High priority Low Priority Not a priority Moderate High priority Moderate priority Not a priority Not a priority Low Moderate priority Low Priority Not a priority Not a priority Priority classification based on the assessment of the weight of evidence and value of surveilling Priortization criterion 2 : Interest in setting up a surveillance (severity , evolution of incidence , societal concern) By integrating these two criteria, health effects are classified into priority groups. This categorization helps in identifying which health effects should be prioritized for surveillance and further research, ensuring that both scientific evidence and societal concerns are considered in the decision-making process. The objective of the Web-Delphi method was to establish a consensus on the prioritization criteria for each of the health effects related to EDs identified in the first stage. We conducted a three-round Web-Delphi process. Since the criteria required different areas of expertise, we proposed two distinct questionnaires: a scientific questionnaire to assess the weight of evidence, the severity of the effects, and the evolution of the incidence rate; and a societal questionnaire to evaluate societal concerns in France. The list of invited participants was compiled from bibliographic searches (e.g. PubMed), expert directories, national and international institutions, and stakeholder organizations (e.g. NGOs, learned societies, local authorities), ensuring broad representation across disciplines, areas of expertise, and societal sectors. The target number of around 500 invitations was determined to ensure sufficient coverage across the 11 health effect categories considered, while maintaining representativeness within each group. We acknowledge that participants had varying levels of expertise or institutional status, which reflects the diversity of actors involved in endocrine disruptor–related issues. This heterogeneity was considered a strength in capturing a wide range of knowledge, but it also introduces variability in participation and judgment, which was accounted for through stratified analysis and transparency in response rates. This panel consisted of two groups: International scientific experts : This group included diverse expertise across all eleven categories of health effects studied in stage 1, such as epidemiologists, toxicologists, academics, hospital practitioners, physicians, and others. Their role was to address the scientific aspect. French stakeholders : This group included representatives from associations, non-governmental organizations, elected officials, communities, learned societies, and others involved in the field of EDs. Each institution selected one representative to respond to the social questionnaires. International scientific experts : This group included diverse expertise across all eleven categories of health effects studied in stage 1, such as epidemiologists, toxicologists, academics, hospital practitioners, physicians, and others. Their role was to address the scientific aspect. French stakeholders : This group included representatives from associations, non-governmental organizations, elected officials, communities, learned societies, and others involved in the field of EDs. Each institution selected one representative to respond to the social questionnaires. We pre-defined a maximum of three rounds for the Delphi method, as research suggests that a higher number of rounds may result in a lower response rate [ 32 ]. We established four consensus categories based on response percentages: Strong consensus : ≥80% identical responses, Moderate consensus : 71%−80% identical responses, Low consensus : 50%−70% identical responses, No consensus : <50% identical responses. Strong consensus : ≥80% identical responses, Moderate consensus : 71%−80% identical responses, Low consensus : 50%−70% identical responses, No consensus : <50% identical responses. We conducted an online Delphi procedure using LimeSurvey version 2.05+ ( www.limesurvey.org ). Initially, a registration proposal was sent by email to all identified participants in either French or English, along with a methodological summary of the study and an explanation of the issues. Subsequently, the questionnaires (additional file 1) were distributed to all registered participants. In each round, panel members received a unique link (secured by a token) to access the questionnaire. We conducted three rounds between November 2021 and June 2022, with each round lasting 3 weeks, followed by an optional 2-week period for follow-ups. Email reminders were sent at various stages of the consultation, including during the registration invitation and throughout the different rounds. A maximum of three reminders were sent for each round of the questionnaire per participant. Participants who did not complete the previous round within the deadline were not invited to the next round. Both questionnaires began with an introduction outlining the consultation’s objectives and the Delphi method, along with practical instructions on how to complete and submit the questionnaire. Round 1: Participants were asked to provide their profile information to validate several pieces of information (such as family name, email address, and institution). The eleven categories of health effects available for consultation were presented to the participants, allowing them to select the one(s) they wished to provide input on, based on their knowledge, expertise, professional experience, or personal knowledge (see Table  1 ). Participants were only given access to questions related to the categories of health effects they had chosen in advance. Rounds 2 and 3: Participants had access to anonymous responses from all other participants, along with a reminder of their own responses from the previous round. They were encouraged to maintain or revise their answers if they deemed it necessary. Only categories for which at least two participants had provided responses were retained for further consultation and analysis. During these rounds, participants also had the opportunity to justify their answers, especially if their responses differed from the majority. Additionally, they could attach supporting documents to substantiate their answers. We assessed consensus based on response percentages. When responses fell between 50% and 70% identical, we considered the achieved consensus to be low. In such instances, we opted to use the median response as the measure of central tendency, as it was deemed to provide a more objective representation of the collective opinion of the panel. For example, if 60% of the participants believed the weight of evidence was “strong” and 40% felt it was “low,” we would assign the “moderate” category. If the median response fell between two categories, we would arbitrarily select the weaker category. For instance, if 50% of participants rated the weight of evidence as “moderate” and 50% as “low,” we would choose the “low” category. We employed descriptive statistics to summarize response distributions, including mean, median, median class, standard deviation, quartiles, and the 95th percentile. To facilitate descriptive statistics, we assigned quantitative values to qualitative categories, such as “0” for “not very serious” severity, “1” for “serious” severity, and “2” for “very serious” severity. This allowed for comparisons of standard deviation magnitudes among different health effects. The median, as a measure of central tendency, represents the collective opinion of the panel, while the standard deviation, as a measure of dispersion, indicates the degree of disagreement within the panel [ 32 ]. In addition, we conducted qualitative analyses. Participants were asked several open-ended questions, providing them with the opportunity to justify their answers for each round. They could provide justifications through text (in the scientific and societal questionnaires) and also by sending documents (a designated option was available to upload various document formats). These responses were collected, organized by relevance, and used for assessing the feasibility of surveillance. An analysis of divergent responses was also conducted to identify various controversies and interpret the dispersion of opinions. We analyzed differences in opinion using two methods. Firstly, we considered standard deviation, as a measure of dispersion, to represent the degree of disagreement within the panel. We considered questions with consensus but with a standard deviation greater than the 95th percentile value of the distribution as questions with disagreement. Secondly, starting from the second round onward, participants were given the opportunity to justify their decision to maintain or change their initial answers after reviewing the group’s responses. A qualitative analysis of divergent answers was conducted to uncover different controversies, identify key messages, and interpret the dispersion of opinions. Additionally, we conducted a preliminary ranking of health effects based on the Weight of Evidence (WoE) when data were available [ 33 ], utilizing a method developed by Hurgent [ 9 ].

The primary aim of this study was to compile a list of health effects identified as priorities for surveillance by SpFrance due to their association with EDs, while considering both scientific and societal criteria. This serves as a decision-making tool for shaping the Agency’s surveillance program on the topic of EDs. Additionally, this approach allows to capture a snapshot of the current shared knowledge between health effects and EDs. The methodology combined a comprehensive literature review and a Delphi process to foster consensus on health effects to prioritize surveillance. The optimal sample size for a Delphi is not precisely defined, but it is generally accepted that each group should include 12 to 18 participants [ 26 ]. While our initial group sizes met this recommendation, the societal group experienced a decline in participation from the third round onwards, which may have affected the stability of consensus. Nonetheless, the participation rates in this study align with, and in some cases surpass, those documented in the literature [ 34 ]. There is no universally accepted definition of consensus in the Delphi method. Various thresholds have been suggested. For instance, McKenna’s work suggests that consensus is achieved with 51% agreement among respondents [ 32 ], Sumsion recommends 70% (1998) [ 35 ], while Green et al. advocate for 80% (1999) [ 36 ]. A 2014 systematic literature review on Delphi studies found that the most common definition of consensus was a threshold of percent agreement, with an average threshold of 75%, followed by ranges of percent agreement among participants [ 37 ]. Given the complex nature of issues related to EDs and the conflicting evidence in the literature, it was anticipated that the study might elicit widely divergent opinions. Setting a too high consensus threshold might have resulted in a low number of responses reaching consensus, thus hindering our ability to achieve the agency’s surveillance program objectives. To avoid being overly restrictive and to gain insight into the level of agreement among participants, we defined four categories of consensus. Strong consensus rates (≥ 80% identical responses) were observed at 73% for the scientific questionnaire and 63% for the societal questionnaire. These results demonstrate that our Delphi study effectively identified consensus on complex issues. The prioritization process highlighted several reproductive health outcomes as priorities for current surveillance, aligning with the previous prioritization conducted by Hurgent. Some of these effects are already under surveillance in France [ 38 , 39 ]. The provisional hierarchy of health effects, based on the WoE according to the literature, corroborated the Delphi results. However, the prioritization of three health outcomes raised questions. Firstly, thyroid cancer triggered a significant difference of opinion among experts regarding the strength of evidence, with some experts classifying it as “strong” evidence while others deemed it “undocumented” WoE. Consequently, it received a low priority ranking. It’s important to note that this result doesn’t reflect the consensus of all participants. As a result, we will consider this health effect during the next feasibility evaluation, acknowledging its relevance for further investigation, and recognizing that its prioritization as “low” is not shared unanimously by the panel. Furthermore, the WoE for two health effects were categorized as “undocumented” by the consultation process, even though our previous literature review had classified them as “sufficient.” Consequently, an update will be necessary in the coming years to either confirm or revise the WoE ranking for uterine fibroids and autoimmune thyroid diseases. It is essential to emphasize that expert opinion cannot replace solid epidemiological evidence regarding the connections between EDs and diseases. Furthermore, the existence of a consensus does not necessarily indicate that the correct answer has been definitively determined. Nevertheless, uncertainty is a reality in various aspects of decision-making within the field of EDs and public policy. The method presented herein identified the areas that experts and stakeholders deemed significant at a specific point in time concerning this topic. Despite the inherent limitations of the Delphi method and the absence of a universally accepted method or regulation, the proposed approach provided the agency with initial insights to determine which health effects should be prioritized for surveillance in connection with EDs. The Delphi method facilitated the engagement of a diverse and geographically dispersed group of participants, both in France and internationally, with varied skills, backgrounds, and expertise. This diversity was essential for addressing the complex topic of EDs health impact. The method also allowed for the inclusion of difficult-to-access knowledge, such as unpublished findings, while ensuring participants’ anonymity to minimize potential biases associated with factors like notoriety, charisma, or leadership. Furthermore, the approach developed for assessing the level of confidence allowed us to validate the prioritization results by considering the level of consensus achieved the number of participants, and their justifications. Beyond the prioritization outcomes, this method enabled the collection of valuable information about existing or potential monitoring systems and specific biomarkers related to the EPs of interest. This information will be instrumental in the feasibility assessment phase and the development of a surveillance strategy for priority diseases. However, the Delphi consensus method used for prioritization has several limitations. Some of these limitations are inherent to the Delphi method itself. It can be challenging to prevent group fatigue when conducting multiple rounds of questionnaires, potentially leading to a false consensus on complex or controversial issues. To mitigate this, we initially set a limit of three questionnaire rounds. While the Delphi technique is gaining popularity in health and social research, there are no universally accepted guidelines, which can result in methodological challenges and difficulties in result interpretation. Nonetheless, this study adhered to several recommendations from the scientific literature to ensure rigor [ 32 , 40 ], including defining consensus criteria in advance, maintaining panel expertise and homogeneity, providing reminders of the method and objectives between rounds, and more. Moreover, the questionnaires were only available in French and English, which may have limited participation of non-fluent experts in other languages. Despite our efforts to involve as many stakeholders in the field of EDs as possible, the number of participants in the societal questionnaire was relatively low compared to recommendations from the literature, limiting the strength of consensus in this part. However, these results were made available to all stakeholders of the French national strategies for EDs before the official publication of the surveillance program, allowing for additional input that may not have been captured during the consultation. The requirement for continuous participant engagement across multiple rounds, which was achieved during this study, spanning from November 2021 to June 2022. Various strategies, such as clear explanations of questionnaires and objectives, protocol provision, contact email availability for queries, time estimation for questionnaire completion, and personalized reminders, helped maintain satisfactory participation rates despite some dropouts. Certain health effects could not be prioritized due to insufficient responses (e.g., skin and eye disorders, adrenal disorders, bone disorders) in the scientific component. Although invitations were extended to a wide array of internationally recognized experts, it’s noteworthy that these are categories of health effects that have only recently gained attention in relation to EDs. The lack of responses can thus reflect both a lack of interest in these topics and a dearth of relevant knowledge and literature. Regarding the assessment of societal concern by stakeholders in the field of EDs, it was observed that respondents tended to provide unqualified answers, indicating predominantly strong interest in surveillance and recognizing very serious health consequences. While this questionnaire allowed them to express their deep concern, it may have complicated the process of prioritizing the societal aspect. It is crucial to acknowledge that the Delphi method has faced criticism in the literature [ 41 , 42 ] for potentially fostering consensus and for not facilitating participant discussions. Therefore, it is important to emphasize that the existence of consensus and our prioritization results do not imply that the definitive answer has been found. Rather, they signify the areas that experts and stakeholders considered important at a particular time regarding the topic. Despite the inherent limitations of the Delphi method and the absence of universally accepted standards or regulations, the proposed method has provided initial insights to assist the agency in identifying health effects that warrant priority surveillance in relation to EDs. The next steps will involve making definitive selections of the health effects classified as “high” and “moderate” priorities. Addressing feasibility issues will be crucial, including determining measurement and collecting methods, exploiting existing indicator, adaptating existing databases, or initiating de novo data collection. Comprehensive knowledge of the existing databases in France and fostering multidisciplinary collaborations will be essential. This assessment will also evaluate the suitability of these indicators for surveillance in the context of EDs and their effectiveness in driving actionable interventions. For example, integrating diverse methodologies and tools within a comprehensive approach holds promise for identifying effective prevention strategies and actions, particularly in a context marked by uncertainty and knowledge gaps. The notion of endocrine, in conjunction with the exposome concept, has fostered the development of a more integrated analytical framework combining epidemiology, exposomics, and toxicology—a transdisciplinary approach that has been particularly promoted within European research programs. Applying the exposome concept to surveillance involves aggregating available data beyond health data and employing diverse methodological approaches. The emergence of new environmental and toxicological databases provides opportunities to better characterize population multi-exposure patterns. However, this requires prior consideration of the technical prerequisites for harnessing extensive data, constructing composite exposure indicators, and achieving interoperability among tools and data from various sources and types. Additionally, methodologies for assessing the environmental burden of disease enable the combined use of incidence data, biomonitoring data, and epidemiological relationships to estimate attributable cases for risk factor and translate them into economic cost to help anticipating and managing the ensued public health problems. Throughout the consultation process, information on biological parameters and biomarkers of effects was also gathered, which will be integrated into the feasibility assessment phase. Numerous proposals were put forth, such as those related to surveillance cryptorchidism and hypospadias (including the establishment of congenital malformation registries at the national level, implementing stringent recording methodology and measuring ano-genital distance), thyroid disorders (mandatory iodemia screening during pregnancy), or neurodevelopmental disorders in children (systematic school-based detection at a specific age). This information may prove particularly valuable for the implementation of surveillance within the framework of the French national biomonitoring program lead by SpFrance.

SpFrance has undertaken a comprehensive prioritization of health effects suspected of being linked to exposure to EDs, aiming to establish a robust framework for its surveillance program. In the absence of a universal methodology and incomplete, contradictory, or low level evidence, SpFrance incorporated both scientific and societal criteria trough participative consensus methods to guide priorization. The prioritization method was developed based on a thorough literature review and the engagement of experts and stakeholders through the Delphi consensus method. In total, 43 health effects were prioritized using two key criteria: (1) the strength of available evidence and (2) the epidemiological and societal relevance for establishing health effect surveillance. Expert opinion cannot replace solid epidemiological evidence regarding the connections between EDs and diseases. Furthermore, the existence of a consensus does not necessarily indicate that the correct answer has been definitively determined. Nevertheless, uncertainty is a reality in various aspects of decision-making within the field of EDs and public policy. The Delphi method proved crucial in structuring the information gathered from expert input and societal concerns, effectively identifying the health effects that are currently deemed important to monitor by both experts and stakeholders. The results of this consultation provide the Agency with initial decision-making elements to identify which health effects should be prioritized for surveillance in relation to EDs. This prioritization approach is replicable and adaptable over time, allowing for selection of health indicators for study based on evolving knowledge, data availability, and societal changes. The work and initiatives are intended to public policy and are integrated into national plans related to EDs. By identifying and prioritizing health effects linked to EDs, SpFrance aims to create a focused and effective surveillance program that can inform public health interventions, support further research, and ultimately contribute to the mitigation of health risks associated with ED exposure. A detailed presentation of the operational surveillance strategy, including indicator selection and feasibility assessments, will be addressed in a dedicated article. Additionally, proposals related to biomarkers of exposure and effect collected during the Delphi process are being integrated within the national biomonitoring program coordinated by Sante publique France agency.

The World Health Organization (WHO) first highlighted the rising incidence of human diseases with endocrine origins in 2002 [ 1 ], and reiterated this concern in 2013 [ 2 ], noting the increasing prevalence of endocrine-related disorders such as reproductive dysfunctions, hormone-dependent cancers (e.g., breast, prostate), thyroid disorders, metabolic syndromes, and neurodevelopmental impairments. Numerous studies have explored potential links between environmental exposure and these pathologies [ 3 ]. Of particular concern are endocrine disruptors (EDs), which are believed to significantly contribute to the deterioration of reproductive health [ 4 , 5 ]. Biomonitoring studies of the French population have shown widespread exposure to several endocrine disruptors (phthalates, bisphenols, pesticides, PCBs, etc.), with particularly high levels observed in children. These findings indicate that such exposures are frequent and occur at measurable levels in biological samples (e.g., urine, blood), even in the general population [ 6 , 7 ]. In France, the impact of EDs on both the environment and public health has become a growing concern. To enhance the understanding of these EDs, and to better protect our citizens and ecosystems from their harmful effects, France has initiated the French Strategy on Endocrine Disruptors (SNPE). This comprehensive strategy involves all stakeholders in defining the government’s vision and action regarding EDs. Despite mounting evidence suggesting the health risks associated with ED exposure, there remains an urgent need for comprehensive epidemiological surveillance data. Santé publique France (SpFrance) has developed a dedicated surveillance program for ED-related diseases, which not only helps anticipate and manage public health issues but also constructs epidemiological indicators using national databases. These indicators quantify disease incidence for public health purposes and explore potential ED exposure [ 8 – 11 ]. This program initially focused on reproductive health since the potential health effects of EDs were first highlighted by reproductive dysfunction. However, these effects now extend to many other organs and functions of the body, including immune system alteration, thyroid disorders, and hormone-dependent cancers [ 2 , 12 , 13 ]. Understanding the mechanisms through which EDs act and their long-term effects on both animals and humans, especially during development, has become a major focus of scientific research [ 14 – 17 ]. The scientific literature continues to report new findings, including neurodevelopmental disorders [ 13 ], metabolic disorders [ 18 ], immune system dysregulation [ 15 ], and thyroid disorders [ 19 , 20 ]. Consequently, the surveillance strategy for ED-related effects in France needs to be reconsidered to encompass all suspected effects linked to EDs. This strategic shift is essential for quantifying the disease burden attributable to EDs and other factors, guiding public health interventions, and informing future assessments. However, the allocation of resources for this surveillance strategy necessitates a rational prioritization of dimensions and indicators. Methods are available for prioritizing chemicals with potential endocrine-disrupting properties [ 21 ], but no universal method exists for prioritizing health effects associated with risk factors. One of the central challenges in studying EDs and their impact on human health is establishing definitive causal links or assessing the “weight of evidence” (WoE). Unlike many other environmental factors, the effects of EDs are often subtle, pervasive, and may manifest over extended periods. This complexity makes it challenging to apply traditional epidemiological methods. In the health and environmental fields, especially for EDs, scientific conclusions are rarely made based on a single evidentiary modality, necessitating WoE approaches to draw evidence from various modalities, converging on a coherent and consistent conclusion [ 22 , 23 ]. One such initiative is the international network Hurgent (HUman Reproductive health and Global ENvironment Network) [ 9 ], which played a crucial role in the prioritizing key indicators for reproductive health within the French ED exposure surveillance program. The prioritization was established based on a combination of WoE evaluations from two reports: the World Health Organization (WHO)/United Nations Environment Program (UNEP) report [ 2 ], which provided qualitative appraisals of WoE for the causal relationship with ED exposure using expert panels [ 23 ], and the European Commission (EC) report, which quantitatively documented WoE using eight criteria derived from a 2002 International Program of Chemical Safety (IPCS) report [ 12 ]. However, the characteristics of ED mechanisms, such as non-dose-response effects, cocktail effects, critical exposure windows, and transgenerational effects, have hindered the development of universally accepted WoE assessment methods tailored for EDs. Additionally, relying solely on WoE evaluations from previous reports is insufficient, as scientific knowledge has evolved considerably since 2012. Furthermore, these reports solely utilized the Weight of Evidence (WoE) approach—based primarily on mechanistic and toxicological data—as a prioritization criterion, without integrating epidemiological evidence (e.g., human studies, population-level burden) or societal dimensions such as public concern, vulnerability, or policy relevance. Given the incomplete, contradictory, or low-evidence nature of available scientific data, participatory processes play a crucial role in integrating both scientific and societal prioritization criteria [ 24 , 25 ]. Consensus methods, such as the Delphi method, offer systematic approaches to gather and synthesize opinions, expertise, and knowledge from diverse groups of experts and stakeholders. They are particularly valuable in environmental health studies, where multidisciplinary insights and societal perspectives are integral to shaping policies and interventions. The Delphi method, specifically, is a consensus-building approach widely used in the health sciences when knowledge is incomplete or uncertain, and empirical evidence cannot provide definitive answers [ 26 , 27 ]. In this method, experts are surveyed in successive rounds, with controlled feedback between rounds, allowing them to reassess, alter, or develop their opinions based on group input. It’s an iterative process that refines and converges consensus over time, making it ideal for integrating diverse perspectives and expertise in environmental health studies [ 26 , 27 ]. The Web-Delphi can also include a geographically dispersed group of international experts through remote interviews [ 28 ]. Given the far-reaching implications of EDs on human health, the Delphi method enables the harnessing of collective wisdom from experts and stakeholders. This approach ensures a holistic and robust prioritization strategy, reflecting both scientific and societal considerations. Recognizing the multifaceted nature of the issue, SpFrance adopted an innovative approach, gathering a diverse panel of experts and stakeholders and employing the Delphi consensus method. This comprehensive consultation integrates scientific insights with societal considerations, shedding light on the complexity of the issue. In this article, we present the outcomes of this consultation process, including the prioritized health effects. This study aims to prioritize health effects potentially linked to ED exposure, using a consensus-based approach that integrates scientific and societal perspectives, in order to guide future surveillance efforts. Our findings provide critical insights into ongoing efforts to safeguard public health in the presence of EDs. This work underscores the importance of evidence-based decision-making in environmental epidemiology and emphasizes the necessity of consensus methods that bridge the scientific and societal dimensions of this complex issue.

Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2.