Inclusion for non-binary athletes in racing events has progressed in the last three years, with five of the six Abbott Major Marathons offering non-binary running categories in 2023. Given that the “main issue for non-binary people is that they cannot compete authentically” without the non-binary category1, non-binary race divisions were created, in part, as inclusion measures for non-binary individuals. As the number of non-binary athletes participating in running is increasing2, the representation and needs of non-binary athletes should be studied using appropriate methods that are in line with current best practices3, particularly in assessing gender and sex measurement and analyses. In their paper, “Performance of non-binary athletes in mass-participation running events,” Armstrong et al. utilize several methodological approaches and analyses that do not appropriately assess the performance of non-binary athletes. As a consequence, the conclusions of the paper introduce numerous biases with results that do not accurately reflect the reality of non-binary athletes.

In designing their study, the authors attempt to determine the sex assigned at birth (labeled natal sex in their model) of non-binary athletes based on the presumed sex assigned at birth of their names. This approach is methodologically unreliable and flawed4. Specifically, the authors fail to distinguish between sex assigned at birth, legal sex, and any sex-related medical intervention that the athletes may have had, conflating them. This technique is not novel, but rather a misapplication of methods. As a result, the authors are analyzing assumed sex assigned at birth and race times, but assume that it accounts for gender. The model does not analyze gender. As a member of an organization whose athletes are a part of this analysis, we can confirm that errors are present.

Principles of research methods dictate that the design of a study must align with the study's objectives and the context of the phenomenon. However, in Amstrong’s case, the authors’ particular use of language describing natal sex shows their use of a pathologizing and stigmatizing framework in viewing the phenomenon, which is misaligned with current best practices in trans health research5. Confirmation bias is present in the way the authors attempt to guess sex assigned at birth and design the independent variable labels. Treating gender identity as “natal-female nonbinary” or “natal-male nonbinary” is still miscategorizing non-binary athletes. Finally, the authors introduce non-neutrality bias as they favor sex assigned at birth as a concept over gender identity by saying “sex has considerable explanatory advantages over gender identity” without providing evidence to support this.

These poorly chosen methods, apparent biases, and lack of reasonable evidence lead to inaccurate conclusions. The authors state that "...if one wishes to understand the needs of gender non-conforming individuals, it is vital to control for sex as it is likely to play a significant role in any analysis." Because the analysis only supports a conclusion for the relationship between sex and race time, as a result of the poorly constructed and likely miscategorized sex variable, we know nothing about the needs of gender non-confirming individuals at all.

The methods of this paper are misapplied and not robust. Importantly, the non-binary category in running helps to validate the identities of non-binary individuals1. The author's conclusion that sex and gender data should both be collected does not disprove any concepts regarding the importance of gender identity and relating to social determinants is an inappropriate conclusion. As a result of these misapplications, the findings cannot and should not have any implication in law or policy. Future research to better understand the needs of this population should include non-binary runners themselves, as non-binary runners are likely experts in their own needs6,7.

References

1. Erikainen, S., Vincent, B., & Hopkins, A.. Specific Detriment: Barriers and Opportunities for Non-Binary Inclusive Sports in Scotland. Journal of Sport and Social Issues, 2022; 46(1), 75-102.

2. Bernhard, J. What Does It Take to Design an Inclusive Running Race? - Uncommon Path – An REI Co-op Publication. REI. 2023. https://www.rei.com/blog/run/inclusion-in-running (accessed 11 January, 2024)

3. Reisner SL, Deutsch MB, Bhasin S, et al. Advancing methods for US transgender health research. Curr Opin Endocrinol Diabetes Obes. 2016 Apr;23(2):198-207.

4. Blevins, C., & Mullen, L. DHQ: Digital Humanities Quarterly: Jane, John … Leslie? A Historical Method for Algorithmic Gender Prediction. Alliance of Digital Humanities Organizations. 2015

5. Ruane, J. M. Essentials of research methods: A guide to social science research. Malden, MA: Blackwell Publishing. 2005

6. Solle JC, Steinberg A, Marathe P, Gray TF, Emmert A, Abel GA. Patients as experts: characterizing the most relevant patient-reported outcomes after hematopoietic cell transplantation. Bone Marrow Transplant. 2020 Jan;55(1):242-244.

7. Streed CG Jr, Perlson JE, Abrams MP, Lett E. On, With, By-Advancing Transgender Health Research and Clinical Practice. Health Equity. 2023 Mar 3;7(1):161-165.

The systematic review by Paultre et al. supports the use of turmeric or curcumin extract for knee osteoarthritis pain.

They did not perform a formal meta-analysis but summarize the results of individual studies by calculating effect sizes based on the data in the original papers. Unfortunately there are two problems with these, one major and the other more modest.

The major issue is with the last study reported in table 3, Srivastava (2016). Paultre et al. report very large effect sizes for this study, such as 8.6, 9.5, and even 11 for a visual analogue scale. These effect sizes are the usual "d" value, that is the mean difference divided by the standard deviation. Effect sizes of such high magnitudes should raise a red flag that something is wrong, as they are rarely attained in clinical studies.

The authors' impressive effect sizes for Srivastava are errors due to using a standard error of the mean (SE) as if it were a standard deviation (SD). Srivastava et al. define the statistic used in the statistical methods: "The results are presented as mean ± SE." The values shown are also impossibly small to be standard deviations, which is what caught my attention. Both at 60 days and 120 days, the "standard deviations" shown for a 10-point VAS scale are around 0.1. This suggests a range of responses of about 0.5, which is not plausible.

The SEM is the SD divided by the square root of the sample size and represents the accuracy of the sample mean, not the distribution of individual subject values. Effect sizes incorrectly calculated using the SEM will thus be inflated by a factor of the square root of the sample size. Correcting for this leaves several still large effects (near 1) but other effect sizes drop to below 0.5. None are over 2.

The modest issue is with the first study in the table, Panda. I don't see any actual errors, but the standard deviations on the VAS are smaller than is typical for a 100 point scale. I believe this is because an inclusion criterion for that study was, "VAS score during the most painful knee movement between 40-70mm." Restricting to a max of 70 for "most painful" is a significant restriction and results in rather small standard deviations for the VAS.

On a 100-point scale a mean difference of 10 is not a large effect clinically, but Paultre et al. show differences of 5 and 7 as "large" for the Panda study because of the small standard deviations. Not until day 60 is there a mean difference much bigger than 10 on the VAS.

I would be interested to see if the authors would modify their conclusions after addressing these issues.

====
References (both are from the reviewed paper)

Srivastava S, Saksena AK, Khattri S, et al. Curcuma longa extract reduces inflammatory and oxidative stress biomarkers in osteoarthritis of knee: a four-month, double-blind, randomized, placebo-controlled trial. Inflammopharmacology 2016;24:377–88

Panda SK, Nirvanashetty S, Parachur VA, et al. A randomized, double blind, placebo controlled, parallel-group study to evaluate the safety and efficacy of Curene® versus placebo in reducing symptoms of knee oa. Biomed Res Int 2018;2018:1–8.

Dear editor/ dear authors,

We read with interest your editorial ‘Sport and exercise medicine around the world: global challenges for a unique healthcare discipline’ [1] in BMJ Open Sport and Exercise Medicine. We would like to congratulate the authors for bringing the challenges of our speciality back into the spotlight again.

While Sports and Exercise Medicine (SEM) may be a modern and more inclusive terminology than sports medicine, many of the challenges of our speciality have remained the same over the years. Societies such as the European College of Sports and Exercise Physicians (ECOSEP) have championed for years for the advancement of sports medicine/ SEM speciality across Europe by providing education, publishing research, organising congresses, collaborating with other organisations and serving as a source of information to the public [2–4]. ECOSEP has been promoting exercise for prevention and treatment to policy holders, creating post-graduate programmes and seminars to provide further training for physicians and bringing practitioners together, not only with biannual congress but also through promoting professional dialogue and standards [2–4]. Other societies, like the European Federation of Sports Medicine Associations (EFSMA) have been champing for a common sports medicine speciality within Europe for over 20 years, providing a detailed curriculum for sports medicine practitioners [5]. Even back then they recognised that sports medicine is a multidisciplinary speciality dealing with health promotion for the general population by stimulating a physically active lifestyle as well as providing diagnoses, retreatment, prevention programmes and rehabilitation following injuries or illnesses sustained by participation in physical activities, exercises and sport at all levels [5]. So, while the terminology SEM may be new, the goals and challenges to the speciality have remained the same.

Another challenge is the common recognition of the speciality across countries; however, this has not even been achieved within Europe, even after years of negotiations with governments and stakeholder. In an open market such as the European Union it makes it challenging for health care professionals, working and traveling with sports teams across Europe as well as being disadvantaged in seeking job opportunitiesin the common market.

We believe adequate employment opportunities for SEM/ sports physicians are another important challenge to be addressed. Historically sports doctors did not receive payment for attending match days or overseeing competitions/ training camps and although this has changed in some countries, this type of work remains voluntary or poorly paid in other countries. Many big international sporting events such as Olympic Games, Athletic European or World Championship and others rely on SEMs as volunteers and while this may be interesting for some SEM to gain experience or for other reasons, we believe it is a practice that should also be addressed. Many SEMs have a portfolio career, e.g., combining working with a sports team, health prevention programmes, private practice, university etc., however employers/ institutions often do not recognise this working type of pattern and are reluctant or unwilling to adapt to modern ways of working.

We agree with your statements that ‘SEM should be more visible in the health-based undergraduate curricula’ [1], and we believe it should be taught by appropriately qualified staff with the necessary qualifications. In some countries, however it is still possible to hold a university chair/ professorship in SEM/ sports medicine without even having obtained the necessary country specific sports medicine qualification or specialised training. What message does this send for the promotion of our speciality?

While we understand that an editorial cannot address all the challenges of our speciality, we hope that some of our challenges maybe also be addressed in future consultations, and we wish the current ‘generation’ of SEMs more success in tackling those challenges compared to previous generations of sports physicians.

References

1 Carrard J, Morais Azevedo A, Gojanovic B, et al. Sport and exercise medicine around the world: global challenges for a unique healthcare discipline. BMJ Open Sport Exerc Med 2023;9:e001603. doi:10.1136/bmjsem-2023-001603
2 ECOSEP. 2023.https://ecosep.eu (accessed 4 May 2023).
3 Heron N, Malliaropoulos NG. International differences in sport medicine access and clinical management. Muscles Ligaments Tendons J 2012;2:248–52.
4 West L, Malliaropoulos N, De Jonge S. ECOSEP: bringing the European SEM family together. Br J Sports Med 2014;48:1585–1585. doi:10.1136/bjsports-2014-094167
5 Ergen E, Pigozzi F, Bachl N, et al. Sport medicine specialty training core curriculum for European countries. J Sports Med Phys Fitness 2008;48:419–33.
 

We would like to thank Bache-Matiesen et al.(1) for their thoughtful article on non-linear modelling in sport medicine. Our own study on the non-linear relationship between acute: chronic workload ratio (ACWR) and injury risk in children was published as a preprint (2) and recently accepted by the American Journal of Epidemiology.(3) Below, we highlight some additional underlying principles in non-linear modelling that readers should understand.

GENERAL CONCEPTS
Models are based on information, which includes both data and assumptions. Simple linear models are more prone to bias because they assume a data generating process that is likely incorrect. The flexibility of non-linear models leads to less risk of bias, but also less precision. The optimal choice between bias and uncertainty depends on the context.(4)

Bache-Matiesen describe three non-linear modelling options: quadratic modelling, fractional polynomials (FP), and restricted cubic splines (RCS, where knots are determined by either data driven or a priori methods). These all fall under generalized additive models (GAMs), or generalized additive mixed models (GAMMs; if one uses “random effects” to adjust for repeated measures on participants).

FP methods use a single polynomial function over the entire range of exposures to predict the outcome. Quadratic models are special cases of FP (with exponents of 0, 1 and 2) and are too restrictive to be generally recommended. RCS separate data into sections by “knots”, determine which polynomials best predict the observed data within the sections defined by the knots, and apply a smoothing function to join the polynomials. The placement of the knots is important. In simulations, we know the true data generating process. In real-world observational data, we do not. Although we believe subjective a priori knot placement is sometimes better than data driven methods, researchers should be aware that gross errors may occur if the assumptions are incorrect.

RCS methods are more flexible than FP because they allow the functions between knots to be different from each other. We highlight two technical points. First, the “restriction” in the RCS method used by Bache-Matiesen appears restricted to using linear functions before the first knot and after the last knot. RCS can include other types of restrictions as well. Second, FP and RCS are just two forms of GAM(M)s. Two other popular forms are penalized regression spline and thin plate regression spline (e.g. default in mgcv package in the statistical program R (5)). The major difference between RCS and the penalized/thin plate regression splines is the shape of the polynomial that is used. The optimal choice depends partly on the research question and partly on the observed data.

SPECIFIC COMMENTS ON RECOMMENDATIONS BY BACHE-MATIESEN
1. Based on their simulations, Bache-Matiesen suggests RCS performs better than FP for predictive modeling. We believe the conclusion is too strong. RCS can be used in situations when there is a more complex underlying structure than FP. However, fitting more complex structure requires more data. When there is limited data and less complex structure, FP could outperform RCS.
2. Bache-Matiesen claim that RCS allowed the authors to model the relationship at high training loads where there were few data points. We caution against this due to the limited information. In our own study on children,(3) we restricted our conclusions about the relationship between acute:chronic workload ratio (ACWR) and injury risk to ACWR <3 where there were enough data. We show the full range of the relationship in the supplementary material to be fully transparent and to help generate hypotheses for future studies but did not feel it appropriate to make inferences.
3. We disagree with Bache-Matiesen that FP is more interpretable than RCS for causal effects, and that RCS results “can only be interpreted in the form of p values and visualisation”. First, we hope that most sport medicine researchers have moved beyond making inferences on p-values because of its severe limitations.(6, 7) Second, for causal questions, we generally estimate the difference in outcome when exposure is set to two different levels. When causal inference assumptions are reasonable, one estimates the causal effect with g-computation using predicted data from the model;(8) this is applicable for both linear and non-linear models. In brief, the magnitude of the causal effect in a non-linear model necessarily depends on the two chosen exposure levels. Although FP has only a single function over its entire range, the causal effect between the chosen exposure levels still requires using g-computation and the predicted values from the FP function. The causal effect using other GAM(M) methods is obtained with the same process; one uses the predicted values provided by the statistical software over the chosen exposure range.
4. The authors discuss the need to add a small constant to training load when it can equal 0, in order to allow for analyses on the log scale. The choice should have theoretical justification. For example, if activity (i.e. training load) is a proxy for time at risk (e.g. game injuries and a game was not played), adding a constant is inconsistent with the research question. However, in the ACWR, activity is a proxy for fatigue in the numerator and a proxy for fitness in the denominator. Both fatigue and fitness are affected by activities of daily living (e.g. occupation, transportation) outside of regular sports. In our analysis on children, we added a constant of 0.1 to activity to reflect our belief that activities of daily living might contribute 10% to each of fatigue and fitness. In supplementary analyses, we explored the effects if the contribution were 25% or 50%.
5. Finally, if we believe tripling the activity will triple the injury risk (e.g. 3 games vs 1 game), then activity (or ACWR) should be plotted on the log scale.(3)
 
REFERENCES
1. Bache-Mathiesen LK, Andersen TE, Dalen-Lorentsen T, et al. Not straightforward: modelling non-linearity in training load and injury research. BMJ Open Sport Exerc Med. 2021;7(3):e001119.
2. Wang C, Stokes T, Vargas JT, et al. Injury risk increases minimally over a large range of changes in activity level in children. arXiv. 2021;2010.02952v2 [q-bio.QM].
3. Wang C, Stokes T, Vargas JT, et al. Injury risk increases minimally over a large range of changes in activity level in children (In Press). Am J Epidemiol. 2021.
4. Kaufman JS. Commentary: Why are we biased against bias? Int J Epidemiol. 2008;37(3):624-626.
5. Wood S. mgcv. In: R: A language and environment for statistical computing R Foundation for Statistical Computing Vienna, Austria; 2021.
6. Amrhein V, Greenland S, McShane B. Scientists rise up against statistical significance. Nature. 2019;567(7748):305-307.
7. Wasserstein RL, Lazar NA. The ASA statement on p-values: context, process, and purpose. Am Stat. 2019;70(2):129-133.
8. Hernán MA, Robins JM. Causal inference: What if. Boca Raton: Chapman & Hall/CRC, 2020.