Differential Exclusion of Low‐SES Women May Bias Sex‐Specific Mediation Findings in Happiness Research

Dear Editor,

Beltran Patino and colleagues [1] deserve commendation for their cross‐sectional study examining the mediating role of group exercise participation (GEP) in the association between socioeconomic status (SES) and happiness among older Japanese adults. The use of principal component analysis to construct a multidimensional SES indicator and the sex‐stratified mediation analysis represents notable methodological strengths. We also appreciate the authors' transparency in reporting the characteristics of participants excluded from the final analysis. However, we believe that the pattern of systematic exclusion revealed in this comparison raises important questions about the validity of the sex‐specific mediation findings. Additional sensitivity analyses would substantially strengthen the interpretation.

The authors report that excluded participants (n = 307) differed systematically from those included in the final analysis. Those excluded were significantly older, more likely to be female, and had lower educational attainment. They also had lower employment rates and lower subjective economic status [1]. This exclusion pattern directly affects the population for which the authors claim their key finding: “SES facilitates GEP in women, whereas in men, both factors independently contribute to their happiness.” Women were disproportionately excluded, particularly those with lower SES. The mediation pathway in women is the study's most novel finding. The exclusion of the very subgroup in which this pathway is hypothesized to operate therefore warrants careful scrutiny.

Selection bias can affect mediation estimates in ways that are not predictable by simple intuition [2, 3]. The bias can arise in either the exposure‐to‐mediator path (SES → GEP) or the mediator‐to‐outcome path (GEP → happiness), or both. If excluded lower‐SES women have different GEP participation patterns than included women, the estimated association between SES and GEP (the a‐path) may be distorted. If the relationship between GEP and happiness differs between included and excluded women, the b‐path may also be biased. The direction and magnitude of this bias cannot be known without empirical assessment. The authors acknowledge in their limitations that “the exclusion of this population restricted our analysis to a relatively healthier cohort.” They also state that findings “may have underestimated the true strength of the associations.” Yet no quantitative assessment of this potential bias is provided. Without sensitivity analyses, readers cannot determine whether the sex‐specific mediation finding would persist if the excluded participants were represented. Inverse probability weighting could account for differential exclusion [4]. Multiple imputation could assess the impact of missing data [5]. These methods are feasible and would provide valuable information about the robustness of the conclusions.

The low survey response rate (22.9%) further compounds this concern. Non‐respondents are not characterized in the study. However, if they resemble the missing‐data excluded group, the analytical sample may be even less representative than the comparison suggests. Low response and differential exclusion by key sociodemographic factors together mean that the generalizability of the sex‐specific mediation findings remains uncertain. This is especially true for lower‐SES women, who are underrepresented in the sample [6]. In Japan, widening socioeconomic disparities and barriers to GEP among low‐SES older women make this limitation particularly relevant [7].

We suggest three refinements for future work. First, conduct inverse probability weighting or multiple imputation to assess the potential impact of exclusion bias on the mediation estimates. This is particularly important for the female subsample, where the exclusion rate was higher. Second, report the proportion of the total SES‐happiness association mediated by GEP, along with its confidence interval. This would help readers evaluate the practical importance of the indirect pathway [8]. Third, frame the sex‐specific findings as hypothesis‐generating observations requiring replication in more representative samples. They should not be presented as established sex differences in causal pathways. Our comments are intended to enhance the methodological transparency of this otherwise valuable contribution, not to diminish its overall significance.

Author Contributions

Q.W. conceived the letter concept, performed the literature analysis, and drafted the manuscript. H.D. and Q.L. critically revised the manuscript for important intellectual content and provided supervision. All authors approved the final version and agree to be accountable for all aspects of the work.

Funding

The authors have nothing to report.

Ethics Statement

The authors have nothing to report.

Consent

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

Data sharing not applicable to this article as no datasets were generated or analyzed during the current study.