Defining Effect Size Standards in Temporomandibular Joint and Masticatory Muscle Research

Introduction

In scientific research, especially in the context of health and medicine, one of the key issues is the correct interpretation of effect sizes, which allows assessment of the strength and significance of experimental results. The topic of effect sizes in research on the temporomandibular joint (TMJ) and masticatory muscles is becoming increasingly important [1]. The temporomandibular joint is one of the most complex joints in the human body, and its function depends on many factors, ranging from biomechanical to psychological influences [2–6]. Therefore, treating TMJ dysfunctions and issues related to the chewing muscles often requires specialists from various fields, including dentistry, physical therapy, psychology, and speech therapy [3,6–9].

Temporomandibular disorders (TMDs) affect 34% of the global population [3]. TMDs are a group of conditions affecting the temporomandibular joint and masticatory muscles, which can cause pain, limited mobility, and other issues in the face and neck area. Their prevalence varies depending on the continent, with the highest rate (47%) in South America and the lowest (26%) in North America [3]. Joint problems are the largest group of issues associated with TMDs [10].

In the context of masticatory muscles, hyperactivity, such as bruxism (involuntary teeth grinding or clenching), has been observed, which can lead to tooth damage, jaw pain, and other dental issues. The prevalence of sleep bruxism is 21% (43% in polysomnographic studies), while the prevalence of awake bruxism is 23%. There are also geographic differences in prevalence [7]. Interest in this field of research is growing year by year [1].

Use of standardized classification systems is important for diagnosis and treatment of disorders within the TMJ, both in terms of unified disease classification systems, such as TMDs [11], and in establishing clear statistical guidelines [1].

As the number of studies increases, it is essential to improve the quality of the statistics used in these studies. One important part of the analysis is the calculation of effect sizes [1]. To date, no effect size calculations have been performed in the context of TMJ and masticatory muscles. Such calculations have only been conducted in relation to rehabilitation [12] or exercise treatments of tendinopathy [13]. The creation of new effect size intervals follows Cohen’s guidelines [14,15]. In developing his guidelines, Cohen emphasized that the recommended values should only be applied in the absence of detailed information about the distribution of effect sizes in specific fields [14,15]. Based on the increasing number of studies and the global issue related to TMDs and bruxism, the present study was undertaken [1,3,7].

This study aims to propose new guidelines for interpreting group differences in research on the TMJ and masticatory muscles. These guidelines should help determine which values of effect sizes can be considered small, medium, or large in the context of the analyzed data. For this purpose, in the framework of meta-analytic research, popular effect size indicators such as Cohen’s d and Hedges’ g were used to assess how these measures differ in studies examining between-group differences in the area of TMJ and masticatory muscle.

Material and Methods

STATISTICAL ANALYSIS:

Analyses were conducted using the R Statistical language (version 4.3.3; R Core Team,) on Windows 11 pro 64 bit (build 22631), using the packages Matrix (version 1.6.5; [41]), pwr (version 1.3.0; [42]), numDeriv (version 2016.8.1.1; [43]), report (version 0.5.8; [44]), metafor (version 4.6.0; [45]), ggplot2 (version 3.5.0; [46]), readxl (version 1.4.3; [47]), dplyr (version [48]), cowplot (version 1.1.3; [49]) and psych (version 2.4.6.26; [50]).

The study conducted analyses to interpret group differences in research on the TMJ and masticatory muscles. To assess the distribution of effect sizes, percentiles were calculated for Hedges’ g. Particular attention was given to the 25th, 50th, and 75th percentiles, as these align with Cohen’s definitions for small, medium, and large effect sizes [14,15].

PUBLICATION BIAS:

To check for potential inflation bias [51] in the studies, special contour-enhanced funnel plots were created. These plots show the relationship between effect sizes and standard errors, which indicate how accurate the effect size estimates are. Contour areas were added to the plots to help identify whether the results are statistically significant. For example, the orange-shaded area corresponds to the range where p (the value indicating statistical significance) is between 0.10 and 0.05, while the red-shaded area corresponds to the range 0.05> p >0.01.

If most of the studies fall within these areas, it may indicate that the reported effect sizes are inflated, and the results may not be fully reliable. Ideally, the results should be more evenly distributed outside of these areas, indicating solid and trustworthy effect size estimates. Lack of reliability can result from various factors, such as sampling errors, publication bias (the tendency to publish certain results), or manipulating results (commonly referred to as p-hacking) [52,53]. Funnel plots help identify these potential issues and draw attention to possible errors in the analyzed data [17,54,55].

POWER ANALYSIS:

These estimates focused on conducting a series of power analyses aimed at determining the sample size needed for future studies to ensure that the results would be reliable and representative. These analyses were specifically concerned with guidelines for group differences [14,15].

All analyses used a two-tailed significance level of alpha=0.05, meaning that studies were considered significant if the probability of the result was less than 5%. The required sample size was calculated to achieve adequate statistical power (i.e., confidence that the results would be reliable) at levels of 60%, 70%, 80%, and 90% for small, medium, and large effect sizes [17,54,55]. These effect sizes correspond to the percentiles previously defined by Cohen [14,15].

Results

SUMMARY OF KEY FINDINGS:

In studies on TMJ and masticatory muscles, it is recommended to use the following intervals: 0.10, 0.30, and 0.70, corresponding to the 25^th, 50^th, and 75^th percentiles. An analysis of publication bias indicated that the findings are not substantially influenced by estimation errors.

The examination of large samples in terms of statistical power and effect size is a crucial aspect of designing future research on TMJ and masticatory muscles. In increase in statistical power leads to a larger sample size, regardless of effect size. A challenge in future studies may be determining the appropriate sample size for small effect sizes, where the required sample size ranges from 1,000 to over 2,000 participants.

Discussion

The objective of this study was to develop new guidelines for analyzing group differences in research on the TMJ and masticatory muscles. The number of studies related to TMJ has been increasing annually [1], and the assessment of effect size in research TMDs remains a key topic in scientific discussions [56,57]. Notably, no guidelines have been established for effect size estimation in TMJ and masticatory muscle research to date, underscoring the significance of this publication. Only standardized diagnostic and statistical criteria can aid in accurately diagnosing patients [1,11,57,58].

The results obtained in this study differ from Cohen’s guidelines (Table 2). In the context of TMJ and masticatory muscle research, Cohen’s effect size benchmarks appear to be overestimated [14,15]. A comparison with other studies, such as the recommendations in audiology and speech-language pathology by Gaeta and Brydges, which present values corresponding to the 25th, 50th, and 75th percentiles (0.25, 0.55, and 0.95, respectively), highlights these discrepancies [19]. Given that the structures of the stomatognathic system participate in sound articulation, comparing effect size estimates with those currently available in related fields appears justified. However, similar to Cohen’s benchmarks, the effect sizes proposed by Gaeta and Brydges exceed the values obtained in this study [19]. This discrepancy may suggest that effect sizes in speech and hearing sciences are generally larger due to different physiological mechanisms, measurement methodologies, or sample characteristics. Gaeta and Brydges’ findings emphasize the importance of context-specific benchmarks, which may not directly translate to TMJ and masticatory muscle research. The observed differences could indicate that research in this field exhibits inherently smaller effect sizes, necessitating adjusted interpretative frameworks tailored to the unique biomechanical and functional properties of the masticatory system. Future studies should further investigate these distinctions to refine effect size benchmarks specific to TMJ and masticatory muscle studies.

When comparing the results of the present study to effect size guidelines for exercise therapy in tendinopathy treatment, as proposed by Swinton et al., which specify effect size values for objective physical function measures (small=0.2, medium=0.4, large=0.7), the small and medium effect sizes obtained in this study are comparable, and the large effect size is identical [13]. This similarity suggests that research dynamics concerning physical function and masticatory muscle function share common characteristics.

In addition to defining effect size thresholds for group differences, this study also proposed required sample sizes to achieve various levels of statistical power. Notably, large sample sizes were required across all power levels for small effect sizes. A similar issue regarding sample size requirements for small effects was observed in Brydges’ research in gerontology [17].

The study’s limitations must also be acknowledged. The first limitation concerns the use of a contour-enhanced funnel plot. While the analysis revealed that 64.79% of studies fell within the white area, suggesting result stability (Table 3), the presence of 20.76% of studies in the gray area raises concerns about potential biases, including sampling errors and publication bias. Although these findings align with previous research, such as Brydges’ work [17], further verification is necessary to confirm their reliability.

Another limitation stems from the assumption that all included studies adhered to appropriate statistical methodologies. This assumption, while common in meta-analyses, introduces a risk of bias if some studies employed suboptimal statistical techniques. For instance, both Cohen’s and Hedges’ effect size measures may be inflated in cases of small samples or unequal variances, potentially affecting the accuracy of effect size benchmarks. However, the methodology applied in this study has been consistently used in prior research [17,19,55].

Additionally, in accordance with the registered protocol, the study aimed to develop guidelines for Pearson’s r [16]. However, due to the nature of the collected data, it was not possible to formulate specific recommendations. This limitation is not uncommon in effect size research – Gignac and Szodorai, for example, provided guidelines only for individual differences, omitting group comparisons [22].

Finally, research on effect size estimation in dental studies remains an evolving field [18]. The lack of formally published guidelines underscores the need for continued investigation. Given the multifactorial etiology of TMJ and masticatory muscle disorders, which necessitates interdisciplinary collaboration among dentists, physiotherapists, psychologists, and speech-language pathologists [3,6–9], the development of dedicated, field-specific effect size benchmarks remains a critical objective.

Practical Implications: The suggested effect size thresholds (0.10, 0.30, and 0.70) offer a more context-specific framework for interpreting research findings on TMJ and masticatory muscles, potentially enhancing clinical decision-making. Larger and moderate effect sizes require smaller sample sizes, whereas detecting small effects necessitates significantly larger cohorts, often exceeding 1,000 participants. By acknowledging that effect sizes in this field tend to be smaller, clinicians and researchers can adjust their expectations regarding treatment efficacy and study design. The findings also emphasize the need for larger sample sizes to achieve sufficient statistical power, particularly when investigating small effect sizes. These insights contribute to more precise methodological guidelines, ultimately improving the reliability of evidence in TMJ research.

Conclusions

Researchers studying the TMJ and masticatory muscles are encouraged to adopt the following thresholds: 0.1 for small effects, 0.3 for medium effects, and 0.7 for large effects, for both Cohen’s d and Hedges’ g. These guidelines provide a standardized approach to effect size interpretation, enhancing the reliability and relevance of findings in TMJ research.