Long‐term outcomes of lateral sinus floor elevation: A machine‐learning analysis, systematic review, and meta‐analysis of predictive factors

Abstract Objectives To quantify long‐term (≥ 5 years) implant survival after lateral sinus floor elevation (LSFE) and to identify clinical predictors of long‐term survival using conventional and machine‐learning meta‐analytic techniques. Methods A systematic search was conducted through MEDLINE, Embase and Scopus. Randomized trials, prospective or retrospective studies with ≥ 10 patients and ≥ 5‐year follow‐up were eligible. Risk of bias was assessed with RoB‐2 (RCTs) and the Newcastle–Ottawa Scale (observational studies); certainty of evidence was graded with GRADE approach. A multilevel random‐effects meta‐analysis (logit‐transformed proportions) estimated pooled survival while accounting for clustering of multiple implants per patient. Moderator effects were explored by mixed‐effects meta‐regression. A MetaForest machine‐learning model examined non‐linear interactions among predictive factors. Results Thirty‐two studies (48 cohorts) involving 7,902 implants and ≈ 2,800 patients met the criteria (3 RCTs, 7 prospective non‐randomized, and 22 retrospective studies; follow‐up 5–13 years). Pooled long‐term survival was 95.8 % (95 % CI: 94.5–96.8 %); Heterogeneity was substantial (I² = 82.8 %) but fell to 53.2 % in MetaForest residuals. Meta‐regression identified lower survival with Graftless procedures versus allografts(β = 1.11, p = 0.016), autografts(β =2.40, p = 0.005), Xenografts(β =1.53, p = 0.04), Xenografts+Allografts(β =1.55, p = 0.04) and Xenografts+Autografts(β =1.82, p = 0.03) with no significant difference versus Alloplasts. Residual bone height (RBH) < 4 mm (β = –0.53, p = 0.039) as well as smoking prevalence (1.96 % per 10 % increase, p = 0.009) reduced survival. MetaForest ranked smoking, age, RBH, follow‐up duration and membrane use as the principal predictors; barrier membranes mitigated the negative effect in smokers and in low‐RBH sites. Time‐to‐failure analysis of 205 failed implants showed 59 % of losses within the first year and 96 % by year 5. Overall certainty of evidence was low. Conclusions Implants placed after LSFE exhibit high 5–13‐year survival (96 %). Use of bone graft materials and their mixtures yield more favorable outcomes compared to Graftless protocol, whereas RBH < 4 mm and smoking significantly impair implant survival. Barrier‐membrane coverage is advisable for smokers and short‐RBH sites. Despite the long‐term nature of this review (≥5 years of follow‐up), most failures occur within the first three years of function, underscoring the need for intensified monitoring during this early phase.