Unmeasured Confounding and Classification Issues in Rod Fracture Risk Analysis

We read with great interest the study by Salmanian et al ¹ examining predictors of rod fracture in adult spinal deformity surgery. The multivariate analysis identifying anterior column manipulation as an independent risk factor (OR 4.69, P = .004) provides valuable insights for surgical planning. However, we wish to engage in dialogue regarding the exclusion of pseudarthrosis from the multivariate model, as this may affect the interpretation of the causal relationships reported. The authors transparently acknowledge that 46% of the rod fracture group had confirmed pseudarthrosis, but could not include this variable due to a lack of systematic CT imaging in asymptomatic patients. We respectfully suggest this exclusion warrants careful consideration, as established literature demonstrates that pseudarthrosis is the dominant precursor to rod failure, with nonunion at ≥ 1-year conferring extraordinary risk (OR 28.9, P < .001). ² When biological fusion fails, instrumentation designed for temporary stabilization must assume indefinite load-bearing, inevitably leading to material fatigue. With nearly half of rod fracture patients having confirmed pseudarthrosis, we wonder whether the reported anterior lumbar interbody fusions (ALIF)/anterior column realignment (ACR) association reflects direct biomechanical stress or is mediated through increased pseudarthrosis risk; a distinction with important clinical implications. This interpretation is further complicated by unmeasured patient factors affecting fusion biology. Smoking increases pseudarthrosis risk (OR 1.38), while diabetes substantially elevates revision risk for pseudarthrosis (RR 2.80, P = .033).^3,4 We respectfully ask whether these factors were distributed equally between surgical groups, as differential prevalence could represent confounding by indication. Additionally, we seek clarification regarding fracture location. Literature demonstrates that rod fractures cluster at the thoracolumbar junction and around connectors where stress concentration is maximal. ⁵ In Figure 1E in Salmanian et al ¹ study, the authors did not report whether fractures occurred at the connector sites where the configurations transitioned. The apparent protection of multi-rod constructs may represent load redistribution rather than true risk reduction.

We also note a terminological clarification: the Results section reports “extended dependency on opioids,” though the measurement was duration of use (days), not clinical dependency. A qualitative diagnosis requiring specific assessment of physiological adaptation and withdrawal symptoms. ⁶ While prolonged opioid consumption after a rod fracture is clinically meaningful, the terminology should reflect the quantitative measure rather than suggest assessment of opioid dependence. We suggest future investigations might benefit from systematic fusion assessment, enabling inclusion of pseudarthrosis in multivariate models, documentation of smoking and comorbidity status, and analysis of fracture location relative to construct geometry. These refinements could clarify whether anterior column manipulation independently causes rod fracture or whether the relationship is mediated through patient selection and fusion biology. We commend the authors for their comprehensive analysis and actionable findings regarding construct reinforcement strategies, which advance our understanding of mechanical complications in this challenging patient population.