Cervical disk arthroplasty (CDA) was heralded as a potential solution for adjacent segment disease (ASD) that is commonly seen after ACDF. Initial industry sponsored IDE trials suggested benefits for CDA in terms of patient reported outcomes (PROs) and lower reoperation rates. Consistently lower ASD rates were not observed. Despite many proponents supporting CDA in the literature, the procedure has not been widely adopted. Now that the Level 1 IDE studies have been completed, most of the more recent CDA literature has been observational, which carries a high risk of selection bias and potential confounding. In order to better understand the potential risk of selection bias and its effects on outcomes, Dr. Goz and colleagues from Utah used the National Inpatient Sample to compare baseline characteristics, costs, and complications between patients undergoing ACDF and CDA from 2004 to 2014. They identified nearly 300,000 elective procedures in adults for degenerative conditions, of which 98% were ACDF and 2% were CDA. At baseline, the CDA patients were significantly younger (47 vs. 53), less likely to be black (6.9% vs. 9.2%), more likely to have no comorbidities (79% vs. 69%), more likely to live in a high-income zip code (30% vs. 24%), and less likely to be on Medicare or Medicaid (15% vs. 31%). In their unadjusted model, hospital costs were about $550 less for the CDA group. After adjustment for baseline characteristics, hospital costs were about $575 more for the CDA patients. Similarly, inpatient complications were lower for the CDA group (OR=0.63) in the unadjusted model, but the difference was attenuated and no longer significant after adjustment (OR=0.93).
The authors have done a nice job illustrating the differences in the CDA and ACDF populations in the "real world" of clinical practice that is captured by the National Inpatient Sample. One criticism of IDE trials is that the highly selected patients enrolled in these trials do not represent the overall population, so the results of these trials may not be generalizable. The authors use their findings to urge caution in the interpretation of observational trials, many of which have been retrospective. Given the marked baseline differences between the CDA and ACDF cohorts, comparing outcomes without accounting for these differences puts investigators at risk of making false conclusions due to selection bias and confounding. While the outcomes investigated in this study are probably not the most relevant, they provided a clear demonstration of the effects of selection bias. Major inpatient complications are quite rare after CDA and ACDF, and there is little reason to expect a difference between CDA and ACDF. However, the unadjusted data indicated the CDA patients were 37% less likely to experience a major inpatient complication, and this observed difference almost certainly had nothing to do with which procedure they underwent. Adjusting for age, sex, race, comorbidities, and income eliminated the difference in complication rates, indicating that the baseline differences were driving the differences, not the procedure type (classic confounding). Similarly, a 2.5% difference in cost is not that exciting, but a 2.5% cost savings for CDA in the unadjusted analysis swung to a 2.5% increase in cost vs. ACDF after adjustment. Unfortunately, the National Inpatient Sample does not link subsequent admissions, so the authors could not evaluate reoperations or readmissions. Similarly, the database does not include PROs, probably the most important outcomes to compare between the two procedures. Possibly most telling is that under 2% of patients in this national sample were treated with CDA, indicating that spine surgeons are not enthusiastic about adopting the technique.
Please read Dr. Goz's article on this topic in the July 15 issue. Does this change how you evaluate observational studies comparing CDA and ACDF?
Adam Pearson, MD, MS
Associate Web Editor