The accuracy and reliability of preoperative digital 2D templating in prosthesis size prediction in uncemented versus cemented total hip arthroplasty: a systematic review and meta-analysis

Abstract

The purpose of this study was to compare the accuracy and the inter- and intra-observer reliability of preoperative digital 2D templating in prosthesis size prediction for the planning of cemented or uncemented THA. This study was registered in the NIHR PROSPERO database (ID: CRD42020216649) and conducted according to the PRISMA guidelines. A search of electronic databases in March 2021 found 29 papers overall. The quality of evidence was assessed using the IHE Quality Appraisal of Case Series Studies Checklist and the CASP Randomised Controlled Trials Checklist. A meta-analysis was conducted, and the accuracy was presented as proportions and the inter- and intra-observer reliability were measured using intraclass correlation coefficients (ICC). Accuracy within one prosthesis size (±1) for cemented stems was 0.89 (95% confidence interval (CI) 0.83–0.95), cemented cups 0.78 (95% CI 0.67–0.89), uncemented stems 0.74 (95% CI 0.66–0.82) and uncemented cups 0.73 (95% CI 0.67–0.79) (test of group differences: p = 0.010). Inter-observer reliability (ICC) for uncemented cups was 0.88 (95% CI 0.85–0.91), uncemented stems 0.86 (95% CI 0.81–0.91), cemented stems 0.69 (95% CI 0.54–0.84) and cemented cups 0.68 (95% CI 0.55–0.81) (test of group differences: p = 0.004). Due to lack of data, intra-observer reliability (ICC) could only be calculated for uncemented prostheses, which for the stems was 0.90 (95% CI 0.88–0.92) and for the cups was 0.87 (95% CI 0.83–0.90) (test of group differences: p = 0.124). The accuracy of preoperative digital templating is greater for cemented prostheses, but the inter-observer reliability is greater for uncemented prostheses. The intra-observer reliability showed a high level of agreement for uncemented prostheses. Cite this article: EFORT Open Rev 2021;6:1020-1039. DOI: 10.1302/2058-5241.6.210048

Deep Learning and Computer Vision Techniques for Automated Total Hip Arthroplasty Planning on 2-D Radiographs

“…The accuracy was higher for the preoperative templating results using the deep learning-based prototype software developed by the authors than with previously used 2-D acetate methods. The results were consistent with those reported in recently published meta-analysis studies [58], [59]. The accuracy of the proposed method is acceptable considering its automation and completion within 30 s; the methods used in the meta-analysis are manual and require more time.…”

“…± is acceptable at 78.9% for the AC and 70.9% for the FC. This result is reasonable compared with the meta-analysis of existing templating methods [58], [59]. As a result of the proposed method for all prostheses, ± was 36.1%, which is higher than 35% (min=25%; max=36%) of 2D acetate and lower than 48% (min=43%; max=61%) of 2D digital.…”

“…± for each of the AC and FC using uncemented prostheses, comparable to 73% and 74% of meta-analyses, respectively [58].…”

“…Most published studies directly related to preoperative templating of THA have evaluated the accuracy and reliability of 2-D acetate and 2-D/3-D digital templating by comparing preoperative templating results with actual surgical results, prospectively [49]- [51] and retrospectively [52]- [57]. Systematic reviews and meta-analyses of these studies have also been published [4], [58], [59]. In summary, 1) 2-D and 3-D digital templating are reliable techniques for preoperative THA planning; 2) templating results are accurate in order of 3-D digital, 2-D digital, and 2-D acetate; and 3) templating is less accurate with uncemented prostheses than with cemented prostheses.…”

“…± is the exact accuracy that considers only predictions identical to the ground truth as correct. The accuracy within one size error, ± , is commonly used for evaluating preoperative THA templating accuracy [58], [59].…”