Challenges of considering both extremities of the weight status spectrum to better understand obesity: insights from the NUTRILEAN project in constitutionally thin individuals

While the physiology of obesity has been so extensively investigated to date, only an extremely small number of studies (less than 50) have focused on the other extremity of the weight spectrum: constitutional thinness. Yet, this important state of underweight in the absence of any eating disorders provides a mirror model of obesity that might be particularly insightful in understanding obesity. Nevertheless, important methodological and recruitment-related issues appear when it comes to this complex constitutionally thin phenotype, as experienced by our research group with the realization of the ongoing NUTRILEAN clinical trial. To face this challenge, the present paper aims at identifying, analyzing, and discussing the quality of such recruitment processes in publications about constitutional thinness. In this order, a group of experts collectively created a new grading system to assess the level of rigour and quality achieved by each study based on different criteria. The main results were that (i) metabolic-related biasing criteria were poorly observed despite being crucial, (ii) recruitment processes were not detailed enough and with sufficient explicitness, and (iii) recruiting among already identified patients would be associated with both higher sample sizes and better scores of quality. The present work encourages investigators to adopt a high level of rigour despite the complexity and duration of recruitment processes for this specific population, and readers to pay close attention to the quality of recruitment when interpreting the data. To better understand obesity and its physiological adaptations, it seems essential not only to compare it to normal-weight conditions, but also to the other extremity of the weight status spectrum represented by constitutional thinness.

When it comes to weight-related disorders, hundreds of thousands of scientific articles have been published on obesity, focusing on many aspects of this condition including recruitment issues [1,2].While the studies conducted so far exclusively involve normal-weight people as controls when questioning the physiopathology of obesity, future studies would benefit from including participants from the other end of the weight status spectrum to better mirror obesity and improve our understanding of this chronic disease.Indeed, studying obesity by considering not only the profile of what we classify as normal weight, but also considering underweight profiles [3], might help better identify new insights and potential targets when it comes to the understating of the physiological adaptations and impairments that occur in people with overweight and/or obesity.Although the integration of such participants in obesity research is of interest, their identification faces the challenge of properly diagnosing and recruiting them, carefully avoiding associated eating and hormonal disorders, and targeting then individuals with constitutional thinness (CT) only.
CT is a rare condition, characterized by an untypical but healthy thin profile, which has only recently begun to receive widespread attention.Despite a growing interest for CT, only 6 studies [4][5][6][7][8][9] were published before 2000, 14 [10][11][12][13][14][15][16][17][18][19][20][21][22][23] between 2000 and 2010, and 23 [3, since 2010 (last PubMed search run in February 2023).According to the World Health Organisation (WHO), the last estimation of the prevalence of underweight in Europe (from all aetiologies) was 2.3% for women and 0.8% for men [46].While CT represents a particular subsection of the underweight population, its specific prevalence remains unclear [47].Interestingly, data derived from the Estonian Genome Centre of the University of Tartu (EGCUT) cohort [25], with CT identified in 881 out of 47 102 participants, suggest a global prevalence of 1.9%.Despite the absence of robust data, the prevalence of CT in the general population is recognized to be extremely low, making it difficult to identify and recruit individuals with CT to studies.Furthermore, the lack of consensus in defining and diagnosing CT exacerbates heterogeneity and hampers comparability of research findings.Based on a recent systematic review of the literature conducted by our research team, a decision tree was proposed to harmonize and support the diagnosis of CT [48].This review identified the following criteria: absence of undernourishment/ malnutrition, eating disorders, associated diseases, overexercising, presence of regular menstruations, weight gain resistance, and stable body weight throughout life [48].Although the phenotype in CT may be very similar to anorexia [22,23,49,50], many objective parameters (such as biochemical analyses) clearly distinguish these two conditions [49].However, people with CT often experience the same weight-related stigma as people suffering from anorexia [24].

PROPOSAL OF A NEW GRADING SYSTEM TO ASSESS THE RECRUITMENT QUALITY IN STUDIES INVESTIGATING CT
Whilst diagnostic tools exist in clinical practice [24,48,49,[51][52][53][54][55][56], the inclusion of participants with CT in scientific research requires more extensive inclusion and exclusion criteria.As previously mentioned, the clinical diagnosis of CT mainly relies on morphological (state of thinness, weight history), hormonal (regular cycling in women, normal levels of some markers of undernutrition), and psychological (absence of eating disorders) criteria.However, additional issues about energetic considerations should be considered within scientific research, as CT is underpinned by metabolic issues.This means that any extraneous factor influencing energy metabolism should be an exclusion criterion within research studies.For instance, as tobacco influences energy and substrate metabolism [57][58][59][60], smokers cannot be recruited in scientific research on CT even if, in absolute terms, an individual can present a CT and be a smoker.Consequently, to facilitate recruitment in scientific studies, we proposed here to divide the clinical criteria for CT into two categories entitled 'Main clinical criteria' for a CT diagnosis and 'Metabolic-related biasing criteria'.According to these criteria, a scale that reflects a level of quality in the recruitment processes has been collectively established (score 0-lower grade to score 10-upper grade) (Table 1).The development of such a scale has been a complex issue that required extensive and accurate considerations.To address this question appropriately, we constituted a group of experts from our research team and made different rounds of discussions based on the Delphi method until a consensus was reached and decisions made.Given the heterogeneity of the contexts and levels of detail found in the publications, some criteria could be split into many degrees of achievement while others did not have this potential.For more consistency in the scoring system, the threshold of '8 out of 10' was finally set as the minimum level for each criterion to be considered as 'appropriate' recruitment.
The 'Main clinical criteria' and 'metabolic-related biasing criteria' scores are expressed in percentage (0%: lower quality, 100%: maximum quality score) after adding up the scores obtained for each criterion that respectively belongs to each of these two main categories (Fig. 1).A 'Global Quality' score is expressed in percentage (0%: lower quality, 100%: maximum quality score) after adding up all the scores (from both 'main clinical' and 'metabolic-related biasing' criteria).A 'Quality score per each individual criterion' is also expressed in percentage (0%: lower quality, 100%: maximum quality score) after adding up all the scores of each individual recruitment process.

QUALITY ASSESSMENT OF RECRUITMENT STRATEGIES IN PREVIOUS STUDIES ENROLLING PARTICIPANTS WITH CT
The studies presented in the Fig. 1 have been identified by an update (last bibliographic research run in February 2023) of the systematic reviews of the literature in CT completed in 2020 and 2021 [48,49] (for further details in the research strategies, see these references 48,49).From these previous works, 4 new studies have been identified [26,[42][43][44].In the present work, we refer to 'recruitment processes' instead of 'studies' as a single recruitment process sometimes led to several studies [13,14,18,19,39,40,43,44] (Fig. 1).
When calculating a global quality score per each individual criterion rather than counting the number of recruitment processes that reached a minimal grade, data once more shows that 'metabolic-related biasing criteria' were less considered by studies than the 'main clinical criteria' (last line Fig. 1).Graphically illustrated by Fig. 2A, these results highlight that four criteria did not reach a global quality score of 50%: 'tobacco' (score: 22%), 'physical activity level' (score: 30%), 'medication' (score: 46%), and 'eating disorders' (score: 46%).While 'tobacco' and 'eating disorders' criteria have already been discussed in this article, these results also highlight the necessity to more rigorously evaluate screening criteria for medication use and physical activity within CT studies.

NUTRILEAN EXPERIMENT-AN EXAMPLE OF DIFFICULT RECRUITMENT
The realization of the ongoing NUTRILEAN clinical trial conducted by our research group raised all the methodological and recruitment-related issues discussed in the present paper and led our group to deepen the question and conduct the present analysis.The NUTRILEAN study specifically questioned the potential functional, metabolic, hormonal, nutritional, and energetic specificities that might help better understand and characterize CT.The aims of the study obviously impose on us a strict and careful recruitment process.As highlighted by Fig. 2B, we strive to follow the previously discussed criteria to the best of our ability.Yet, we would like to catch scientific attention that striving for an important level of rigour in the inclusion process of participants with CT really represents a tremendous challenge.From our 2-year-long experience of active recruitment, multiplying advertisement strategies (solicitations of general practitioners, specialized physicians, nutritionists, dieticians, nutrition centres, and naturopaths, articles in local newspapers, online publications, mailing to large lists of students and employees and recruitment brochures in shops/street/university campus), only 11 participants with CT finally ended the protocol while the study information has been spread to about 30,000 people (crude estimation).A rate of recruitment relative to the total number of canvassed people has been calculated as 0.03% and relative to participants with a presumed CT before the inclusion process 7.0% (Fig. 2B).

PRACTICAL CONSIDERATIONS WHEN RECRUITING PARTICIPANTS WITH CT
Such rates of recruitment raise an important question: why is it so difficult to recruit participants with CT? Beyond the low prevalence of CT and the need to consider metabolic-related biasing criteria, we thought that the sample size would be strongly explained by the level of recruitment quality reached by the studies.We hypothesized that the higher the recruitment quality, the lower the sample size.For instance, we indeed had to exclude dozens of people because of smoking according to our experience in the NUTRILEAN Study.Contrary to our expectations, the data did not provide a clear relationship between recruitment quality and sample size.This led to the consideration of other factors, more particularly the type of recruitment (Fig. 3).As presented in Fig. 1, recruitment processes were divided into three main categories: (C) existing cohorts or databases without physical contact with patients, (F) medical files of already known patients or participants, and (N) new recruitment through advertising campaigns.After setting aside large existing cohorts or data without physical contact with patients (C), Fig. 3 reveals that the type of recruitment (F vs. N) appears as a strong discriminating factor.To recruit among already known patients/ participants (F) (yellow bars) would be related to both a higher sample size and a higher score of quality the inclusion process compared with the recruitment of new patients/participants through advertising campaigns (N) (grey bars) (Fig. 3).To have access to a pool of medical cases of patients/participants already identified seems therefore to be associated with recruitment strategies of higher quality but also with larger numbers of participants with CT.To conclude, while this seems of high interest for both a better consideration of CT and also when it comes to obesity research, the present commentary highlights and discusses the considerable challenge to recruit individuals presenting a rare condition such as CT, especially when recruiting new patients/ participants.Through the example of CT in the lower extremity of the body weight spectrum, present data showed that metabolicrelated biasing criteria (such as 'medication', 'tobacco', and 'physical activity level') remain poorly observed while the main clinical criteria were found quite well-considered in the inclusion processes.Metabolic-related biasing criteria however seem crucial given that metabolic issues are inherent to the atypical physiology of CT.Another important point emphasized here was that many methodological sections of the included studies contain implicit information that would benefit from being more explicitly detailed.Finally, the present work invites readers to always be cautious with the interpretation of the data provided in the literature and to consider them in light of the observance of all the aforementioned inclusion criteria.Indeed, only a few studies were found to rigorously observe all these criteria, which should encourage investigators to be as rigorous as possible in the diagnosis and inclusion of participants with CT in clinical trials, even if it would constrain the timelines of the studies.Although challenging, we believe that future researchers conducted in the field of overweight and obesity should not only consider normal-weight people as natural controls but should enlarge their analysis to individuals from the other extremity of the weight status spectrum, for a better understanding of obesity itself and of its associated complications.

Fig. 1
Fig.1Heatmap of the evaluation of quality scores in the inclusion process of participants with constitutional thinness in clinical trials.CT: constitutional thinness; C: among an existing cohort or database without physical contact with patients; F: medical files of already known patients or participants; N: new recruitment through advertising campaign; * Single recruitment process re-used for: Ling et al.[39,40], Bailly et al.[41,42] Cominetti et al.[44], Gabriel et al.[43].The 'Main clinical criteria' and 'metabolic-related biasing criteria' scores are expressed in percentage (0%: lower quality, 100%: maximum quality score) after adding up the scores obtained for each criterion that respectively belongs to each of these two main categories.A 'Global Quality' score is expressed in percentage (0%: lower quality, 100%: maximum quality score) after adding up all the scores (from both 'main clinical' and 'metabolic-related biasing' criteria).A 'Quality score per each individual criterion' is also expressed in percentage (0%: lower quality, 100%: maximum quality score) after adding up all the scores of each individual recruitment process.

Fig. 2 A
Fig. 2 A real challenge to recruit with constitutional thinness: insights from the NUTRILEAN study.A Quality of the recruitment in the NUTRILEAN Study.B Flowchart of the recruitment process in the NUTRILEAN Study.BMI body mass index, CT constitutional thinness.a: crude estimation.

Table 1 .
Rating scale to estimate the level of quality in the recruitments processes.BMI body mass index, DSM diagnostic and statistical manual of mental disorders, FRS figure rating scale.