1 Introduction

Total shoulder replacements have excellent overall survivorship of 94%–98.3% at 10 years.¹ The most common cause of failure is aseptic loosening of the glenoid component, seen in 24–87% at 10 years.^2–8 The likelihood of aseptic loosening is largely dependent upon seating of the implant, fixation to bone, balancing soft tissues and/or correcting bony deformities to reduce stresses on the implant.¹ In some instances, normalisation of anatomy and implanting a standard ‘off the shelf’ prosthesis cannot be achieved without compromise. There is consensus in the literature that more than 15° retroversion, more than 10° superior/inferior inclination and less than 15 mm of vault depth cannot be corrected with reaming alone and will require either augmentation (bone or metal) or custom implants.⁹ Assessment of glenoid anatomy is contingent on high quality pre-operative imaging.

The type of imaging used in pre-operative planning has evolved from plain radiographs and unformatted non-volumetric (2D) CT^10,11 to volumetric (3D) multiplanar reformation, including fully automated programs, which mitigate distortions in measurement of glenoid version and inclination caused by variable scapula axis.^12–15 Establishing a glenoid plane which distinguishes the paleoglenoid (native plane) and neoglenoid (incorporating degenerative morphology) is done with varying levels of consistency.^16,17 Automated measurements have been assessed previously against manual techniques¹⁸ but inconsistencies have been found between automated programs.^19,20

Given the increasing importance of automated software to aid pre-operative planning, this study aims to evaluate an automated 3D planning system (Blueprint™ v2.1.5, Wright Medical Ltd) against manual multiplanar measurements (MM) of glenoid version and inclination in a clinically derived cohort of arthritic shoulders assessed by several different raters. Secondary objectives were to assess if reliability is affected by glenoid morphology and varying experience of raters. We hypothesized that automated and MM measures would differ and that this would relate to the glenoid version or inclination severity.

2 Materials and methods

A series of consecutive pre-operative planning CT shoulder studies, acquired between 2013 and 2019 (74 shoulders in 70 patients) were collated for analysis. Of the 74 shoulders, 60 had primary osteoarthritis (OA), 14 had OA secondary to rheumatoid arthritis (3), cuff tear arthropathy (7), avascular necrosis (2), renal osteodystrophy (1) and dysplasia (1). CT scans were carried out in a Cannon Aquilion One Vision Scanner. The median patient age was 69 years (range 17–88): 30 male, 40 female. Walch subtypes were 12 A1, 35 A2, 4 B1, 13 B2, 10C.

To generate manual measurements representative of a typical clinical cohort and to verify whether reliability was affected by imaging and/or operative experience, four raters analysed the CT scans: an experienced fellowship-trained consultant shoulder surgeon, an orthopaedic fellow and two consultant musculoskeletal radiologists. Glenoid version and inclination were measured using manual multiplanar reformatting (Sectra AB, Linköping, Sweden) on two separate occasions at least two months apart and in randomised sequence, using a true random number generator (random.org) on each occasion to avoid bias. Prior to analysis, detailed step by step methods to measure glenoid version and inclination were formalised and tested on example cases. These adhered to a modified Friedman analysis for version, and inclination using the Maurier method, both described comprehensively by Boileau et al., in 2018.¹⁸ Briefly, for version the medial most tip of the scapula was aligned with the inferior most tip and glenoid centre to define the scapula plane. Care was taken to align the floor of the supraspinatus fossa and inferior tip of the scapula in all planes then identify the medial most tip of the posterior scapula margin using the convergence of the external cortices of the scapula spine and body, scrolling from superior to inferior, and maintaining alignment between this point, the glenoid centre and inferior tip of the blade in all plains. A line was drawn across the central glenoid viewed in the axial plane to generate an angle for version (expressed as a negative value if retroversion). A line across the glenoid in the coronal plane produced the inclination angle relative to the initial scapula plane. Effort was made to adjust for rather than incorporate osteophytic projections in the measured glenoid margins, more precisely we did not include osteophytic projections that would likely be removed and as such not affect the implant angle. The same cases were then analysed using Blueprint software according to manufacturer instructions. Blueprint stacks slices to create a 3D scapular skeleton and places a best fit plane over this, then uses the intersection of the scapular spine and scapular body to identify the transverse axis. Multiple points on the glenoid surface are used to extrapolate a best fit sphere against the glenoid fossa. Retroversion is identified in the supero-inferior view and inclination in the antero-posterior view, by the angle created from the transverse axis to the radius of the best fit sphere. Blueprint could not produce measures for several studies due to marked scapular dysmorphia, metallic artefact and/or incompatible CT reconstruction algorithms and as such these were excluded from the comparative analysis (n = 15).

3 Results

3.1

3.1 Version

Intra-observer reliability of MM for glenoid version for the group of four observers was excellent (mean ICC 0.92). Inter-observer reliability between all four readers was good (ICC 0.89). Specifically, concordance between the mean readings for radiologists and surgeons was excellent (Lin's CCC 0.92, 95%CI 0.88–0.95). Mean MM version values ranged between 5° and −39°, 10/60 cases were greater than 15° retroversion. Blueprint values ranged between 4° and −47°, with 20/60 cases measuring greater than 15° retroversion (Fig. 1). Bland-Altman analysis of Blueprint v MM version measurements (Fig. 2) demonstrated a mean pair difference of −5.77 (95%CI -7.25 – 4.29) with an upper limit of agreement of 5.73 (95%CI 3.15–8.30), a lower limit of agreement −17.27 (95%CI -19.84 to −14.69) and there was no relationship between differences and mean retroversion to indicate any proportional bias. Consistent with the mean pair difference, concordance was poor (Lin's CCC 0.71, 95%CI 0.61–0.78).

Close

Fig. 1 Summary of version analysis MM values (x = mean, box and whisper plot for median, IQR and range) are plotted in order of severity, with corresponding Blueprint values (blue circles). The superimposed red line shows the 15° cut-off for use of augmentation/implant. There are 20 Blueprint measurements >15° vs 10 MM.

Export to PPT

Close

Fig. 2a Bland Altman analysis of glenoid version measured by MM vs Blueprint.

Export to PPT

Close

Fig. 2b Bland Altman analysis of inclination measured by MM vs Blueprint.

Export to PPT

4 Discussion

CT based planning is now routine and has evolved for shoulder arthroplasty. In recent years automated planning software packages have become readily available, are more consistent, and may be more accurate than manual measurements since the data is calculated using multiple points on the face of the glenoid, instead of the mid transverse plain of the glenoid. Using automated software can help the surgeon prepare for the case, assess need for custom implants or the use of augmentation (bone graft of wedged baseplates) and allows for templating of the implant.

Commercially available programs have been assessed in the literature. Glenosys (Imascap) has been validated against 2D manual measures of arthritic and cadaveric shoulders without adjustment using multiplanar reformatting.²² Neyton et al. assessed a cohort of 51 glenoids (B2 only) for version and subluxation using 2D manual versus 3D automated measurement (Blueprint) finding no significant difference for version and a low correlation between version and subluxation.²³ Similar studies however have shown a consistently greater degree of retroversion for 3D automated.^22,24 Werner et al. found a difference >5° for almost half of the automated version values relative to 2D, but no significant difference for inclination, which is consistent with our findings. Dagget et al. also found a minimal difference in inclination −1° (SD 0.5) between 3D MM and Blueprint in 51 arthritic shoulders.²⁵

Our study design aimed for accuracy and while concordance between manual and Blueprint measures exceeded a similar analysis for inclination,¹⁸ our values for version showed a tendency for Blueprint toward a greater degree of retroversion across a range of mean with more than expected variability but not proportional to severity. The relevance of this is a clinical judgement, in the context of intra-operative anatomical and technical variation and margin of error in each step, but if relying on this planning program, more shoulders would pass the arbitrary 15° cut-off mark for retroversion (20 Blueprint vs 10 MM in our study), which could drive the surgeon to change their implant choice. We postulate this might be due to Blueprint incorporating osteophytes in its retroversion measurements which we aimed to adjust for (Fig. 3). Since these osteophytes mostly form anteriorly, this should not affect inclination measurements as much. Care should be taken when using the planning software to interpret with caution and manually measure when close to the threshold for augmentation or custom implants. Preoperative planning also does not obviate intra-operative variations in surgical technique, and surgical implantation cannot replicate planning when it is done free hand. In future, technology incorporating real-time feedback would be helpful in this regard.

Close

Fig. 3 Blueprint vs MM version measurement example case. In this example, Blueprint (blue line) produced glenoid version of −12°; MM mean representation superimposed (green line) −0.1.

Export to PPT

Boileau compared modified Friedman and Maurier measurement versus those generated by Glenosys in 60 arthritic patients and found a mean difference in version of 1.5° ± 4.5° (CCC = 0.94) and 0.2° ± 4.7° (CCC = 0.78) between the inclination measurements.¹⁸ Boileau used a manual measurement of volumetric studies adjusting for the scapular axis. This is employed to improve accuracy by utilizing three dimensions for setting the scapular plane, but it could be less consistent than 2D measures, particularly if the measured 2D slice is stipulated a priori,²³ as there are a greater number of occasions for adjustments to diverge. Reliability, factoring both consistency and accuracy, may be increased due to consistency in spite of systematic errors in either manual or automated methods. Waltz et al. used manual analysis of volumetric studies against Blueprint and VIP (Arthrex, Naples, FL, USA) and demonstrated variations in the region the glenoid is measured by Blueprint and VIP for version and inclination and automated results did not correlate with manual measurements of the central regions.²⁰ While Levy et al. compared Blueprint and Matchpoint Surgicase and found no significant differences in version or inclination,²⁶ a more recent study noted discrepancies between a number of automated programs (Blueprint, VIP, Treu-Sight, ExactechGPS) in which Blueprint retroversion measures were consistently greater than both manual and other automated values. They were least concordant with manual measures (CCC 0.702; 95% CI 0.5875–0.789), a result very similar to our own (0.71, 95%CI 0.61–0.78).²⁷

Limitations of our study include cohort heterogeneity with respect to age and Walch types. A greater proportion of A2 glenoids were observed compared with previous studies as we chose not to adjust the clinically derived cohort of consecutive patients. While the overall sample size was satisfactory for our analysis, the smaller number of other Walch types were insufficient for reliable subgroup analysis, in particular type C glenoids. Consequently, meaningful conclusions cannot be drawn with regards to this group, an issue common with several previous studies.^19,27 It should also be noted that the high number of outliers with severe version/inclination might affect our results. These outliers could be in-part justified by a more varied and representative population dataset. Regardless, we found a tendency for Blueprint to overpredict retroversion in our cohort.

5 Conclusion

Pre-operative CT has become the mainstay in the planning of total shoulder replacement, more recently automated planning programs have been developed. Our study compared manual multiplanar measurements against Blueprint. We showed no clinically important difference in inclination between manual and Blueprint measurements. However, our study identified a systematic difference between manual measurements and blueprint, with a tendency for blueprint to overpredict retroversion in comparison. This could potentially result in unnecessary augmentation in the glenoid components. Whilst 3D planning software is a useful tool, the absolute values measured by the software should be interpreted with caution. Nevertheless, it is still a useful tool as long as surgeons are aware of this limitation.

Funding/sponsorship

This research did not receive any specific grant from funding agencies in the public, commercial or not-for-profit sectors.

Informed consent

NA.

Institutional ethical committee approval

No Institutional Review Board approval was required for this study.

CRediT authorship contribution statement

Tim Hall: Visualization, Formal analysis, Investigation, Data curation, Writing – original draft, Writing – review & editing. Joshua Rui Yen Wong: Formal analysis, Writing – original draft, Writing – review & editing, Project administration. Margo Dirckx: Formal analysis, Investigation, Data curation, Writing – original draft, Writing – review & editing. Kannan Rajesparan: Formal analysis, Methodology, Resources, Writing – review & editing, Supervision. Abbas Rashid: Visualization, Formal analysis, Conceptualization, Methodology, Resources, Writing – review & editing, Supervision, Project administration.