Acute Versus Delayed Reverse Shoulder Arthroplasty for the Primary Treatment of Proximal Humeral Fractures

Henry D. Seidel, BS; Sarah Bhattacharjee, BS; Jason L. Koh, MD; Jason A. Strelzow, MD; Lewis L. Shi, MD


J Am Acad Orthop Surg. 2021;29(19):832-839. 

In This Article


Making up 4% to 6% of all fractures in adults, fractures of the proximal humerus pose a profound burden for elderly patients.[1] The management of these fractures can be particularly challenging, and a thorough understanding of the treatment methods is important for surgical decision making and counseling patients on reasonable postoperative expectations. Although the success of rTSA treatment has been previously reported in the literature, the effect that timing of primary rTSA has on revision and complication rates is less defined. Here, we sought to compare the rates of revision and complication for patients treated with acute or delayed primary rTSA.

For elderly patients undergoing management of a proximal humerus fracture, the rate of future revision is of particular interest. Any additional surgery is especially burdensome in this cohort. Our study found that the delayed cohort was more than twice as likely to require a revision procedure, a result that remained notable even after accounting for demographic and comorbidity risk factors through multivariate analysis. Several previous studies have reported on revision after acute primary rTSA for proximal humeral fracture. However, sample sizes in these studies were small, making the findings difficult to interpret. Seidl et al[34] reported that none of their 15 acute primary rTSA patients went on to revision. Cicak et al[36] had a similar finding; none of their seven acute primary rTSA patients required revision. Although both Seidl et al and Cicak et al had acute groups that received primary rTSA treatment, their delayed groups included patients who underwent rTSA after operative treatment. Kuhlmann et al directly compared acute primary rTSA with delayed primary rTSA after initial nonsurgical treatment. The authors reported one revision because of component failure in the acute primary rTSA group and another revision because of periprosthetic fracture in their delayed primary rTSA group. This resulted in a 1.0% revision rate in the acute group and a 2.6% rate in the delayed group. These rates were not markedly different after univariate analysis. Our revision rate finding of 1.7% for the acute group was similar to the rate reported by Kuhlmann et al, but our delayed group's rate of 4.5% was higher. Contrasting Kuhlmann et al, we found these rates to be markedly different, even after multivariate analysis.

In addition to revision, we also investigated a broad range of postoperative complications. Although most of the surgical complications trended higher in the delayed cohort, after multivariate analysis, timing of primary rTSA treatment was shown to be independently associated only with dislocation risk. We hypothesize that the increased rate of dislocation may be because of partial malunion, callus and adhesion formation complicating exposure, and necessitating additional releases during surgery. Kuhlmann et al[32] primarily focused on functional outcome measures and active range of motion in their analysis; however, they did report overall complication rates of 6.9% and 7.9% for the acute and delayed groups, respectively. The pooled surgical complication rate for our study was comparatively higher at 8.7% and 13.6%. Of note, although we found a strong trend toward increased mechanical complications in the delayed primary rTSA group, multivariate analysis failed to detect a significant difference (P = 0.051). This was unexpected because our revision rate in the delayed primary cohort was markedly higher than the acute primary rTSA cohort. Aseptic loosening, periprosthetic fracture, periprosthetic osteolysis, and other mechanical complications are leading indications for revision. We speculate that this result may be because of secondary indications for revision, such as infection and dislocation, contributing to the elevated revision rate in the delayed primary rTSA cohort. Specifically, the breakdown of complications among patients who required revision demonstrated that the increased rate of revision in the delayed group may have been coupled to the higher rate of dislocation; 55.0% of the revisions in the delayed group were associated with dislocation compared with 44.6% in the acute group. Thus, the higher instability rate in the delayed group (6.1% versus 2.8%) may be the driver for higher revision rates. With instability rates higher than the overall observed revision rate, however, some of the dislocations may have been managed with closed reduction, not resulting in a coded revision procedure.

The acute treatment group had a higher day of surgery cost compared with the delayed treatment group. This was unexpected because we predicted that the delayed group might have a longer time of surgery and thus an increased cost. Because of the lack of recorded surgical times in the database, we could not evaluate how length of surgery affected cost; however, the authors suspect various additional factors may have affected the cost of the acute group. It is possible that in the acute setting, some rTSA surgeries started with ORIF, which was then changed to reverse shoulder replacement and may have extended the length and increased the cost of the surgery. In addition, in the acute setting, tuberosity repair is part of the rTSA procedure, whereas in the delayed setting the tuberosity may have malunited and no tuberosity repair was needed. Additional investigation into how the setting of rTSA affects surgical length and procedural variability is needed for additional conclusions regarding cost.

Ultimately, the results of our study suggest that timing of rTSA treatment may make a difference in risk for revision and dislocation in the short-term postoperatively. Although several other studies have investigated the effect of acute rTSA versus salvage rTSA after failed operative treatment,[33,34,36,37] only the aforementioned study by Kuhlmann et al and a study by Roberson et al[35] examined rTSA timing where the delayed cohort was defined by patients initially managed nonoperatively. The main focus of Roberson et al's study compared rTSA treatment with nonsurgical treatment; however, they included a subanalysis of 20 patients looking at surgical timing in the rTSA group. Their analysis focused on range of motion and patient-reported outcomes rather than rates of revision and complication. They found no substantial differences between the acute and delayed treatment groups and concluded that rTSA outcomes were not compromised by delayed treatment. Kuhlmann et al arrived at a similar conclusion, finding that the timing of primary rTSA had little effect on mid-term outcomes. They ultimately suggested that a trial of nonsurgical treatment should be considered.[32] However, Kuhlmann et al recognized that their small cohort size may have limited detection of notable outcome differences. With the largest study to date on this topic, our findings suggest that a trial of nonsurgical management in elderly patients with proximal humeral fracture may not be as prudent as previously suggested.

There exist limitations to this study. Although this study demonstrates increased risk of complication with delayed surgery, it does not address patient-reported outcomes or functional outcome scores because these are not available in the database. In addition, as a database study, our results rely on the individual coding practices of providers and hospitals. Because the ICD-9 does not separate proximal humeral fractures by the Neer Classification, we were unable to control for the severity of fracture, which may affect patient outcomes and surgical decision making regarding the timing of treatment. We were also unable to control for the status of greater tuberosity healing or type of implant used. Overall, nonsurgical treatment produces excellent results in most cases.[38] The chance at avoiding any surgery in these patients may outweigh the risk of possibly worse outcomes with delayed primary rTSA treatment. Thus, this study represents a small proportion of patients that may require additional study. In our analysis, 319,034 patients were managed nonoperatively and did not require delayed rTSA. The delayed primary rTSA cohort (n = 892) accounted for only 0.3% of all nonoperatively treated patients. Additional understanding of the risk factors for failed conventional treatment strategies is desperately needed. Furthermore, although our rates of complication and revision were comparable with other small comparative studies that investigated rTSA in proximal humeral fracture patients, our rates of revision were notably lower than revision rates from studies published on rTSA used for any indication. A 2010 review of 21 studies by Zumstein et al[39] identified a 10.1% revision rate after rTSA surgery. In addition to database limitations, the lack of long-term follow-up in our study may limit the conclusions about implant failure and potential differences between the acute and delayed primary rTSA patients. Finally, significant demographic and comorbidity differences existed between the acute and delayed treatment cohorts. Although the age reporting of the Medicare data set is limited to certain ranges, and the reported average age of the two groups ended up in the same range (75 to 79 years), a breakdown of the age groups in Table 1 demonstrates that the acutely treated patients may have been older and also tended to have fewer comorbidities. This finding matches the results of other studies, suggesting that age and comorbidity may affect the decision to attempt a trial of nonsurgical treatment.[32,34,40] Older, healthier patients seem to be more suitable candidates for acute rTSA operations after proximal humeral fracture. These demographic differences between the acute and delayed cohorts, including higher comorbidities and obesity, may be the driver of the differences we observed. Although we attempted to control for this with multivariate analysis, there may be other unaccounted for differences that are responsible for the differences we found.

Our study has several strengths. By using a national database, we avoided regional or institutional biases, improving the generalizability of our findings. In addition, our sample size is notably larger than previous studies, where multiple authors recognized their cohort sizes as a potential limitation to detecting differences between the groups. Finally, our study attempted to control for demographic and comorbidity risk factors through the use of logistic regression multivariate analysis, which was not done in previous studies on this topic. Although these results should be interpreted cautiously with the understanding that we were still unable to control for fracture patterns and surgical decision making, our findings suggest that surgical timing may be a risk factor for increased short-term rates of revision and dislocation. Patients undergoing delayed primary rTSA treatment may require closer follow-up and additional counseling regarding reasonable postoperative expectations. If nonsurgical treatment fails or delayed rTSA is required for another reason, more complications and a higher chance of revision can be expected. Additional study in this area is required.