CORR Insights®: Are Higher Global Alignment and Proportion Scores Associated With Increased Risks of Mechanical Complications After Adult Spinal Deformity Surgery? An External Validation

Where Are We Now?

Adult spinal deformity correction is challenging because of the high likelihood of complications that can unintentionally push a carefully planned surgical process off course either during or after surgery. Taking good care of patients who undergo these resource-intensive procedures typically calls for dedicated facilities and multidisciplinary care teams. Postoperatively, patients may have wound healing complications including but not limited to serious infections. But the potential for trouble does not end right after surgery. Even if the patient recovers well from surgery in the short-term, a number of intermediate and long-term complications may arise as a function of over- or undercorrection adjacent segment degeneration, and/or proximal junctional complications—any of which may result in reoperation. The ability to estimate the likelihood of problems and results in advance of surgery would be a great boon for patients, as well as surgeons and their hospital groups.

A 2017 study [10] examined a prediction model called the global alignment and proportion (GAP) score, which attempts to predict mechanical complications after adult spinal deformity surgery. There is still an ongoing debate about the clinical value of GAP scores. In this month’s Clinical Orthopaedics and Related Research®, Kwan et al. [7] add to the debate by investigating the external validation of the GAP score.

In their study, the authors involved a secondary analysis of data from a prospective trial, which helped show that the GAP score may have some important holes. Specifically, higher GAP scores—a measurement of moderately to severely disproportioned spinopelvic balance after corrective surgery—were not associated with a higher risk of revision surgery at a minimum follow-up of 2 years. Additionally, these higher GAP scores were also not associated with an increased risk of mechanical revision surgery. However, patients with disproportioned GAP scores were burdened by lower SF-36 physical and mental component summary scores, as well as lower Scoliosis Research Society Outcomes Questionnaire (SRS-22) and Oswestry Disability Index (ODI) total scores postoperatively.

Where Do We Need To Go?

The authors theorize, and I agree, that future revisions to the GAP score (or development of other potentially predictive tools) should consider nonmechanical parameters for estimating the risk of mechanical complications and revision surgery. I believe, based on findings from other studies [1-3, 6, 9], that medical comorbidities such as smoking, diabetes, and cardiopulmonary disease, as well as variations in surgical technique and postoperative care, should be considered as we develop future tools. In the meantime, surgeons should not solely rely on the GAP tool to try to estimate the risk of revision surgery after spinal deformity correction. The increase in outcomes score differences observed in the secondary analysis by Kwan et al. [7] is likely too small to make a difference to surgeons or patients in clinical practice.

As I suggested just above, nonmechanical parameters—things unrelated to deformity correction and spinopelvic balance—may have important implications in anticipating the likelihood of mechanical complications following adult spinal deformity surgery. Future predictive analytic algorithms need to consider these factors. Before utilizing artificial intelligence (AI), researchers relied on linear and logistical regression analyses, and these methods were and remain difficult to use when large numbers of variables are present, particularly if there are outliers. Fortunately, predictive analytics with computational techniques, utilizing AI through deep machine learning, can process incomprehensibly large amounts of data [5, 8]. Investing in and advancing our abilities to utilize AI and deep machine learning algorithms in close collaboration with our computer engineering colleagues can lead to robust predictive analytics models for spinal deformity correction surgery.

How Do We Get There?

Any future research should aim to accomplish two main goals: (1) provide surgeons with a dashboard to leverage far more accurate and individualized predictive tools that guide both the surgeon and better inform each individual patient about the ideal surgical technique, and (2) achieve the best anticipated mechanical correction and improvements in SF-36, SRS-22, ODI, and EQ-5D outcomes scores and anticipated time to return to work or play.

Each patient should be notified of their personalized potential postoperative complications associated with adult spinal deformity correction surgery. These complications include but are not limited to rod breakage, pseudoarthrosis, spinal column failure adjacent to the instrumentation, reoperations, and readmissions. Preventative measures (think “prehab”) and modifiable risk factors (such as elevated HbA1C, smoking status, and osteoporosis) should be discussed with the patient as a way to improve his or her chance for a successful outcome.

The ideal predictive system of the future should offer automatic entry of all radiographic parameters, including bone quantity and bone quality data as well as the relative rigidity of the individual spinal motion segments. This can be obtained from upright bending films, spatial matching, and allowing the upright restacking of preoperative three-dimensional (3-D) supine imaging for surgical planning.

High-resolution, CT-like bone detail can be directly extracted from MRI imaging [4] to minimize ionizing radiation while improving data acquisition from one scan modality. This 3-D system gets transferred into a robotic system with multiple active end-effectors to more predictably and more accurately plan and perform a highly individualized adult spinal deformity correction, while also reducing interoperator variability. An AI review of the CT-like data and direct comparison with the MRI source is one way of validating this approach. Another option is placing the scan over the real anatomy via augmented reality to identify the true original bony landmarks upon which the original MRI (and then the subsequent CT extraction) was based.

The best predictive analytics could benefit immensely from minimizing unwanted in-OR variations that currently exist between community hospitals, academic institutions, and surgeons themselves. This is particularly true regarding the approaches to, and execution of, adult spinal deformity surgery. One example of this is the difference in results between open spine surgery and the minimally invasive approach (likely an anterior column realignment and fusion with percutaneous tension band recreation posteriorly only), which offers a quicker recovery, fewer complications, and lower readmission and reoperation rates. The minimally invasive approach may also lead to less muscle damage, which could reduce the rate of adjacent segment degeneration due to loss of the muscular tension band that typically just turns into scar after open surgery. Finally, the use of automation and robotics could standardize this approach, further minimizing intraoperative variation between facilities and surgeons, benefitting a more predictable postoperative outcome and longevity of these resource-intense surgical approaches.