PART I — THE SHORT VERSION
Nine papers on the reverse total shoulder arthroplasty came across the desk in the past few weeks. Eight of them measured something other than what the patient felt about their result from the surgery. Only one asked the patient something, but only tested whether the question was reproducible.
Here is what each paper reported, and what it did not.
Study | Endpoint reported | Endpoint missing |
Reverse in patients ≤65 — Obana (n = 103) [1] | Range of motion; revision | Any patient-reported outcome |
Implant selection and positioning — Parsons (n = 49) [2] | Simulated impingement-free motion | Any patient-reported outcome |
3-D preoperative planning — Zampetakis (635 shoulders) [3] | Version, inclination, screw placement accuracy | Evidence that 3-D planning leads to superior clinical outcomes |
Scapular stress fracture risk factors — Bengart (n = 662 screened) [4] | Acromion-to-lateral-humerus distance | Any patient-reported outcome |
Predicting fracture by machine learning — Schneller (n = 2,256) [5] | Fracture incidence; model discrimination | Whether the model prediction changed a decision or a result |
Managing acromial and scapular spine fractures — Abukar (376 fractures) [6] | Radiographic union, by treatment | Statistical comparison of the outcome scores |
Short versus standard-length stems — Hollo (n = 114) [9] | Coronal alignment; distal filling ratio | Any patient-reported outcome |
Late central screw breakage — Romem (n = 3) [7] | Mode and timing of construct failure | How many similar cases in the surgeon’s practice did not fail |
Reliability of the Forgotten Joint Score — Feik (n = 60) [8] | Reliability of the FJS-12 | Validity, responsiveness, reverse-specific behavior |
Every endpoint in the middle column got sharper. Version error is now measured in degrees rather than described in words. Fracture risk carries a probability. Stem alignment carries a threshold.
The column on the right is mostly empty. The largest single-center series of young patients having RSA collected no patient-reported outcomes. The most faithfully executed plan has not been shown to change what patients report about the outcome of their surgery. Fixing an acromial fracture heals the bone more often than non operative treatment, but does not make the patient significantly better. The one instrument built to hear from the patient about their outcomes has been tested in only eleven reverse shoulders.
When the accessible endpoints improve every year, yet the patient’s perception of their outcome is not collected, we can try to convince ourselves that we’re making progress, but we can do more meaningful patient outcomes research.
PART II — THE PAPERS, ONE AT A TIME
The younger patient
Obana and colleagues at Columbia [1] reported on 103 patients aged 65 or younger finding improved range of motion after RSA. While 187 patients met the age criterion, only 30 (16%) contributed the two-year motion data. Patient-reported outcome measures were not collected routinely in clinic; none are presented. Yet the abstract concludes that the surgery “can reliably improve clinical outcomes.” Degrees of elevation and a patient’s account of their shoulder are not the same. Neel and colleagues found lower satisfaction and function in younger patients despite comparable postoperative motion. Eight patients (7.8%) required revision at a mean of 25.8 months, baseplate failure was the most common indication.
Optimizing what we can simulate
Parsons and colleagues at Duke [2] ran 49 preoperative plans through 19 implant configurations each and 12 motions, using a model that includes scapulothoracic as well as glenohumeral motion. The model indicated that larger, eccentric, lateralized glenospheres produced more impingement-free motion. The authors state that the surgeries were too recent to correlate with any postoperative result, that the model captures bone-on-bone and implant-on-bone contact but not deltoid or cuff tension, that osteophytes were left in place, and that the motion patterns simulated are those of healthy subjects rather than reverse patients.
Zampetakis and colleagues [3] conducted a review covering 11 studies and 635 reverse arthroplasties. Three-dimensional planning, with or without patient-specific instrumentation or navigation, improved the accuracy of glenoid component and peripheral screw placement with respect to a preoperative plan, with version error falling from 6.7 ± 5.4 to 1.5 ± 1.1 degrees in the largest comparative study. However, they found no evidence that patient comfort and function were improved by the use of 3D planning. Clinical outcomes were assessed in only 3 of the 11 studies. In two of them the Constant, ASES, and DASH scores were comparable to cases in which 3D planning was not used. The third study had no conventional-technique arm. So, the glenoid component placement became more accurate without evidence that the patient’s outcome improved.
The acromion
Bengart and colleagues [4] screened 662 primary reverse arthroplasties and found 14 acromial or scapular spine stress fractures, a rate of 2.1%, matched 3:1 to 42 controls. The fracture group had a postoperative acromion-to-lateral-humerus distance of 13.1 mm versus 8.5 mm (P = 0.034), and that distance increased by a median 2.3 mm from preoperative values while the control group’s decreased by 3.0 mm (P = 0.024). Receiver operating characteristic analysis identified 9.78 mm as the most discriminant cutoff, with an area under the curve of 0.690. Their practice is now to avoid lateralizing beyond 10 mm. Their data supporting this practice is thin. Two measurements are in play: (1) the absolute distance on the postoperative film, which is what the 10 mm rule is based on, and (2) the change in that distance from before surgery to after. The authors built a separate model for each. The model using the absolute distance did not reach significance (P = 0.064). The model that did reach significance used the change, and its coefficient runs the other way: β = −0.132 per millimeter. Each additional millimeter of increase in acromion-to-lateral-humerus distance was associated with a 12% lower odds of fracture, which is opposite to the group comparison the paper draws its recommendation from.
Schneller and colleagues [5] used 2,256 registry patients to build a logistic regression model. Acromial and scapular spine fracture incidence was 4.1%, of which 85% were Levy type 2 or 3. On the held-out test set the model achieved sensitivity 0.71, specificity 0.61, and an area under the curve of 0.71. Of the 235 shoulders in which the model predicted a fracture, only 17 actually had one and 218 did not; of the 352 shoulders predicted not to fracture, 7 had one. Implant design did not differ significantly between the fracture and no-fracture groups on straightforward comparison (P = .348). The authors did not indicate whether the use of the model changed choices made in their practice.
Abukar and colleagues [6] reviewed fourteen studies, 376 fractures in 374 patients, mean age 72.9 years, 78% women, 70% atraumatic, mean time to diagnosis 11.3 months. Union was 68% overall: 78% after fixation and 59.8% after nonoperative care, a difference that reached statistical significance (P = 0.004). The abstract and conclusion both state that functional outcomes are comparable between the two approaches. However, that claim was not tested. Four scores were reported. On three of them the operated patients did slightly better: VAS 1.7 against 2.4, ASES 66.9 against 60.2, and Subjective Shoulder Value 63.5 against 60.7. On the Constant score they did worse, 37.9 against 46.0, but only two of the fourteen studies reported a Constant score after surgery, and those two contributed only 12 patients between them. The authors state that none of these scores could be compared statistically because patient-level data were inconsistently reported.
The stem and the screw
Hollo and colleagues [9] compared coronal alignment between short and standard-length humeral stems of the same design in 114 consecutive primary reverses at two Swiss centers, 57 stems in each group, on true anteroposterior radiographs at a minimum of six months. Median varus-valgus deviation was 3.50 degrees (IQR 1.90 to 5.10) for short stems and 1.90 degrees (IQR 1.00 to 3.90) for standard-length stems (P = 0.002). Deviation beyond 5 degrees occurred in 26.3% of short stems and 12.3% of standard-length stems. Standard-length stems held alignment above an observed distal filling ratio of 48%, short stems above 59%, with an absolute risk reduction of 30.4%. The authors call this threshold analysis exploratory, note that it was derived from the observed data without receiver operating characteristic analysis, and say it has not been validated in an independent cohort. The difference between groups was a median of 1.2 degrees. The primary regression model explains 6.7% of the variance in alignment; most of what decides where a stem sit was not measured: for example, broaching trajectory, insertion force, humeral deformity, and canal flare. The clinical importance of stem alignment was not tested; no patient-reported outcome, range of motion, or pain data were collected
Romem and colleagues [7] report three patients, 61 to 81 years old, in whom a monoblock central screw broke between 5.5 months and 8 years after reverse arthroplasty performed with bone graft behind the baseplate. Two had abrupt pain and functional decline after uneventful recoveries, one of them while catching a falling door; the third developed activity-related pain three and a half years after surgery. The authors did not report the number of patients in the practice that did not have central screw failure after a reverse total shoulder with a monoblock central screw performed with bone graft behind the baseplate. The series does establish a pattern worth recognizing: when the baseplate sits largely on graft rather than native bone, incomplete incorporation may permit cyclic toggling that loads the central screw to fatigue. Plain radiographs were the best diagnostic test; computed tomography rarely added decisive information, An abrupt decline in the outcome after a good early result deserves an x-ray.
The one paper that asks the patient to provide the outcome
Feik and colleagues [8] took the twelve-item Forgotten Joint Score into shoulder arthroplasty and tested its reproducibility in 60 patients with glenohumeral osteoarthritis. Intraclass correlation coefficients were 0.96 to 0.97. Only 11 of the 60 patients had a reverse for osteoarthritis. The minimal detectable change at 90% confidence was 13.5 points at six months and 14.2 at one year, and 15.1 for the cohort as a whole, on a 0 to 100 scale. That is a wide band for an instrument meant to separate good results from very good ones.
Where that leaves us
That is the last few months in nine papers. The surrogate measures are getting better. The effects on the patient’s comfort and function are mostly missing.
Keeping our focus on the patient
Great horned owl
San Antonio
Follow on twitter/X: https://x.com/RickMatsen
Follow on facebook: https://www.facebook.com/shoulder.arthritis
Follow on LinkedIn: https://www.linkedin.com/in/rick-matsen-88b1a8133//
References
1. Obana KK, Chen JY, Weiss DL, Luzzi AJ, Knudsen ML, Jobin CM, Levine WN. Outcomes of reverse total shoulder arthroplasty in patients ≤65 years old. J Shoulder Elbow Arthroplast. 2026;10:100041. doi:10.1016/j.jsea.2026.100041
2. Parsons KE, Shenoy DA, Lorentz SG, Hurley ET, Navacchia A, Moverman M, Levin JM, Klifto CS. Optimization of implant selection and positioning for reverse total shoulder arthroplasty using three-dimensional computed tomography–guided simulation software with scapulothoracic motion. J Shoulder Elbow Arthroplast. 2026;10:100049. doi:10.1016/j.jsea.2026.100049
3. Zampetakis K, Sakellaridis G, Lepetsos P, Tsiotsias A, Leonidou A. Three-dimensional preoperative planning in reverse shoulder arthroplasty: a systematic review of implant positioning accuracy and clinical outcomes. Cureus. 2026;18(7):e112080. doi:10.7759/cureus.112080
4. Bengart JJ, Kohut KT, Haider MN, Feng L, Duquin TR. Radiographic risk factors for scapular stress fractures after reverse total shoulder arthroplasty: a case-control study. J Am Acad Orthop Surg. 2026;34(16):e2233–e2240. doi:10.5435/JAAOS-D-25-00976
5. Schneller T, Cina A, Maggini E, Klimov A, Braun M, Pfender A, Lazaridou A, Scheibel M. Prediction of acromial and scapular spine fractures after reverse total shoulder arthroplasty using machine learning: a retrospective cohort study. J Shoulder Elbow Surg. 2026 [in press]. doi:10.1016/j.jse.2026.06.023
6. Abukar A, Case C, Sheth U, Henry P, Nam D. Outcomes of operative and non-operative management of acromial and scapular spine fractures after reverse shoulder arthroplasty: a systematic review. J Shoulder Elbow Arthroplast. 2026 [in press]. doi:10.1016/j.jsea.2026.100081
7. Romem R, Kalva SR, Fucich D, Perry AJ, Yao JJ, Kwon YW. Late central screw breakage of a monoblock glenoid baseplate placed with bone graft: a report of 3 cases. JBJS Case Connect. 2026;16(3):e25.00538. doi:10.2106/JBJS.CC.25.00538
8. Feik ML, Smith AZ, Chen KK, Gehring ZA, Myers NL, Gregory JM. The reliability of the forgotten joint score for shoulder arthroplasty. J Shoulder Elbow Arthroplast. 2026;10:100059. doi:10.1016/j.jsea.2026.100059
9. Hollo D, Soproni I, Toft F, Ateschrang A, Shirinskiy I, Blakeney WG, Bauer S. Standard-length stems require lower distal filling ratios than short stems to achieve neutral alignment in reverse total shoulder arthroplasty. JSES Int. 2026 [in press]; article 101765.
Secondary citations referenced in the text — Neel GB et al. (J Shoulder Elbow Surg. 2022;31:1803–1809), Gauci MO et al. (J Shoulder Elbow Surg. 2024;33:1771–1780), Levy JC, Anderson C, Samson A (J Bone Joint Surg Am. 2013;95:e104), and Schoch BS et al. (J Shoulder Elbow Surg. 2022;31:1647–1657) — are cited as they appear within the nine primary sources above and have not been independently retrieved.
