The Fragility of Statistical Findings in Cervical Disc Arthroplasty: a Systematic Review of Randomized Controlled Trials

• Published in: Archives of orthopaedic and trauma surgery Vol. 144; no. 6; pp. 2609 - 2617
• Main Authors: Megafu, Emmanuel C., Megafu, Michael N., Nguyen, Janet T., Du Jour, Elisabeth Point, Bronson, Wesley H., Lin, James D., Hecht, Andrew C., Parisien, Robert L.
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.06.2024; Springer Nature B.V
• Subjects: Arthroplasty - methods; Arthroplasty - statistics & numerical data; Cervical Vertebrae - surgery; Clinical trials; Data Interpretation, Statistical; Humans; Intervertebral Disc Degeneration - surgery; Joint surgery; Medicine; Medicine & Public Health; Orthopedics; Randomized Controlled Trials as Topic; Total Disc Replacement - methods; Trauma Surgery; Statistical fragility; Fragility quotient; Cervical disc arthroplasty; Fragility index; Randomized controlled trials
• ISSN: 1434-3916; 0936-8051; 1434-3916
• DOI: 10.1007/s00402-024-05353-y

Purpose This study employs both the fragility index (FI) and fragility quotient (FQ) to assess the level of robustness in the cervical disc arthroplasty (CDA) literature. We hypothesize that dichotomous outcomes involving CDA would exhibit statistical vulnerability. Methods A PubMed search was conducted to evaluate dichotomous data for randomized controlled trials (RCTs) in CDA literature from 2000 to 2023. The FI of each outcome was calculated through the reversal of a single outcome event until significance was reversed. The FQ was calculated by dividing each fragility index by the study sample size. The interquartile range (IQR) was also calculated for the FI and FQ. Results Of the 1561 articles screened, 111 met the search criteria, with 35 RCTs evaluating CDA included for analysis. Six hundred and ninety-three outcome events with 130 significant ( P  < 0.05) outcomes and 563 nonsignificant ( P   ≥  0.05) outcomes were identified. The overall FI and FQ for all 693 outcomes were 5 (IQR 3–7) and 0.019 (IQR 0.011–0.043). Fragility analysis of statistically significant outcomes and nonsignificant outcomes both revealed an FI of 5. All of the studies reported loss to follow-up (LTF) data where 65.7% (23) did not report or reported an LTF greater or equal to 5. Conclusions The literature regarding CDA RCTs lacks statistical robustness and may misrepresent the conclusions with the sole use of the P value. By implementing the FI and FQ along with the P value, we believe the interpretation and contextualization of the clinical data surrounding CDA will be better understood.