Reply: Assessing Risks in Intramuscular Fat Grafting: A Systematic Review and Meta-Analysis of Complications

We are grateful to Bohler et al.1 for their careful reading of our systematic review and meta-analysis on the risks associated with intramuscular fat grafting2 and for the opportunity to clarify several methodological points. Our goal has been to transparently synthesize heterogeneous evidence and clearly communicate uncertainty for clinicians and trainees. Regarding inclusion criteria and potential overlap, our quantitative synthesis included the article by Condé-Green et al.3 This followed our a priori eligibility, which allowed large aggregates or surveillance sources when event counts and denominators were reported. However, including an aggregate report alongside primary series raises concerns about design heterogeneity and possible cohort overlap. Following guidance on sensitivity analyses and meta-analyses of proportions,4–7 we repeated the analysis using 2 alternative study sets to isolate the influence of these sources. In both analyses, we maintained the same statistical framework: a random-effects DerSimonian–Laird model applied to logit-transformed proportions with a 0.5 continuity correction and back-transformation for interpretability. As several studies reported very few events, study-level confidence intervals (CIs) in forest plots are the Clopper–Pearson exact intervals, a conservative choice for small binomial counts.8–10 Analyses were conducted in R with the meta and metafor packages for reproducibility.11,12 We recognize the paradox noted regarding Mofid et al.13 In random-effects models, weight is driven by variance rather than sample size, which explains its modest contribution despite a large denominator. When the study by Condé-Green et al.3 was excluded, the pooled proportion was 0.060 (95% CI, 0.03 to 0.13), with N = 204,587, I2 = 99.1%, and τ2 = 3.2422. (See Figure, Supplemental Digital Content 1, which shows a forest plot of major complication proportion, excluding Condé-Green et al.3 [random-effects DerSimonian–Laird model on the logit scale with 0.5 continuity correction; pooled estimate back-transformed to the proportion scale; study-level 95% CIs by Clopper–Pearson exact method]. Random-effects weights are shown [right]. The pooled proportion was 0.06 [95% CI, 0.03 to 0.13], with N = 204,587, I2 = 99.1%, and τ2 = 3.2422. For heterogeneity, I2 was the percentage of total variation attributable to between-study heterogeneity, and τ2 was the between-study variance [model fitted on the logit scale], https://links.lww.com/PRS/I595.) Compared with the primary pooled estimate of 0.062 (95% CI, 0.031 to 0.121),2 this estimate is virtually the same as the primary pooled value, with a CI of similar breadth, indicating that the central interpretation is unchanged. When both Condé-Green et al.3 and Mofid et al.13 were excluded, the pooled proportion was 0.08 (95% CI, 0.04 to 0.15), with N = 5730, I2 = 93.2%, and τ2 = 2.2240. (See Figure, Supplemental Digital Content 2, which shows a forest plot, excluding Condé-Green et al.3 and Mofid et al.13 The model and display conventions are the same as those used in Figure, Supplemental Digital Content 1. The pooled proportion was 0.08 [95% CI, 0.04 to 0.15], with N = 5730, I2 = 93.2%, and τ2 = 2.2240. For heterogeneity, I2 was the percentage of total variation attributable to between-study heterogeneity, and τ2 was the between-study variance [logit scale], https://links.lww.com/PRS/I596.) Here, the proportion increased from 0.062 (95% CI, 0.031 to 0.121)2 to 0.080 (95% CI, 0.04 to 0.15), and the CI widened modestly, as expected, with a substantially smaller denominator after removal of the 2 very large aggregate/survey sources. CIs overlapped across all analyses, reinforcing the conclusion that complication risks are nontrivial and consistently higher than those observed in many other aesthetic procedures.6,7 The sensitivity estimates across scenarios bracketed primary results and supported the same clinical interpretation. Large denominators with few events can dilute pooled prevalence, while random-effects models limit single-study influence when heterogeneity is substantial.6–8 We agree that the I2 value is high; this is common in proportion syntheses across populations, techniques, and perioperative safeguards, cautioning against overprecision rather than invalidating synthesis.5,14 In gluteal fat grafting safety, large surveillance efforts, such as those conducted by the Aesthetic Surgery Education and Research Foundation task force,13 are invaluable yet methodologically distinct from adjudicated series; examining their influence separately is a prudent robustness check rather than critiquing any dataset. We appreciate the engagement of Bohler et al. in our study. In summary, our sensitivity analyses showed that alternative study sets did not materially alter the pooled estimates. The direction and magnitude of risk remain consistent, reinforcing the clinical implication that while intramuscular fat grafting offers aesthetic benefits, it carries complication risks that justify caution and demand rigorous prospective research. DISCLOSURE The authors have no competing interests or conflicts of interest to declare. No external funding was received for this study.