Risk factors for nonunion following surgically managed, traumatic, diaphyseal fractures: a systematic review and meta-analysis

Abstract

Background: There are several studies on nonunion, but there are no systematic overviews of the current evidence of risk factors for nonunion. The aim of this study was to systematically review risk factors for nonunion following surgically managed, traumatic, diaphyseal fractures.

Methods: Medline, Embase, Scopus, and Cochrane were searched using a search string developed with aid from a scientific librarian. The studies were screened independently by two authors using Covidence. We solely included studies with at least ten nonunions. Eligible study data were extracted, and the studies were critically appraised. We performed random-effects meta-analyses for those risk factors included in five or more studies. PROSPERO registration number: CRD42021235213.

Results: Of 11,738 records screened, 30 were eligible, and these included 38,465 patients. Twenty-five studies were eligible for meta-analyses. Nonunion was associated with smoking (odds ratio (OR): 1.7, 95% CI: 1.2-2.4), open fractures (OR: 2.6, 95% CI: 1.8-3.9), diabetes (OR: 1.6, 95% CI: 1.3-2.0), infection (OR: 7.0, 95% CI: 3.2-15.0), obesity (OR: 1.5, 95% CI: 1.1-1.9), increasing Gustilo classification (OR: 2.2, 95% CI: 1.4-3.7), and AO classification (OR: 2.4, 95% CI: 1.5-3.7). The studies were generally assessed to be of poor quality, mainly because of the possible risk of bias due to confounding, unclear outcome measurements, and missing data.

Conclusion: Establishing compelling evidence is challenging because the current studies are observational and at risk of bias. We conclude that several risk factors are associated with nonunion following surgically managed, traumatic, diaphyseal fractures and should be included as confounders in future studies.

Keywords: healing; nonunion; pseudoarthrosis; risk factors; trauma.

Figure 1

PRISMA 2020 flow diagram for new systematic reviews (8). *Wrong setting includes eight conservative fracture treatment, six periprosthetic fractures, five pediatric, two gunshots, one fusion study, one pathological fractures, one osteotomy, six pooling of data from conservative and operative treatments, thirty-six other wrong setting. **Other: contact to authors, and duplicates found when full-text were retrieved. ***Language includes one Persian, one Turkish, one Japanese, two Chinese, four Russian, one Spanish, one Hebrew, and two Czech.

Figure 2

Risk of bias assessment in the cohort studies. Domains were selection Q1, exposure Q2–Q3, confounding Q4–Q5, outcome Q6–Q8, missing data Q9–Q10, and reported results Q11. Green (✓) indicates the best possible answer, yellow (?) is ‘unclear’, red (✕) is ‘no’, and white (0) is ‘non-applicable’.

Figure 3

Risk of bias assessment in the case–control studies. Domains were selection Q1–Q3, exposure Q4–Q5+Q9, confounding Q6–Q7, outcome Q8, and reported results Q10. Green (✓) indicates the best possible answer, yellow (?) is ‘unclear’, and red (✕) is ‘no’.

Figure 4

Number of each risk factor occurrences (black bar) and number of significant risk factor occurrences (gray bar). Data stem from the univariate analyses, unless only data from the multivariable analysis were reported.