Avascular necrosis in proximal humeral fractures in patients treated with operative fixation: a meta-analysis

We used established search strategies and retrieved literature in a systematic way from various databases, including PubMed, MEDLINE, Springer, Elsevier Science Direct, Cochrane Library, Google Scholar, CNKI (China National Knowledge Infrastructure, in Chinese), and Wanfang (in Chinese) up to December 2013. The key words ‘avascular necrosis’, ‘operative fixation’, and ‘proximal humeral fractures’ were used for the searches. References from retrieved papers were checked for additional studies. We collected data only from fully published papers—not from meeting or conference abstracts. The publication date was not restricted in our research.

Studies were included in the meta-analysis if they met the following criteria. The study had to contain (1) the occurrence rates of AVN in patients with proximal humeral fractures (prospective, retrospective, and cross-sectional studies), (2) operative fixation versus conservative treatment data, and (3) the effect size reported as odds ratios (ORs) and 95% confidence intervals (CIs) or it could be calculated. The sample size and age range were not limited. Studies were excluded if the study (1) described only conservative treatment data in a review or report, (2) was a reduplicated study, or (3) did not compare operative fixation with conservative treatment.

Articles were reviewed and filtered out independently by two investigators according to our criteria. The data were then extracted independently in duplicate using a standardized form to assess eligibility for inclusion. Data items included study details (e.g., the first author's name, research year(s) of the study, year of the study's publication, location of participants, design of the study, follow-up time) and characteristics of the participants (e.g., age, sex, sample size). Discrepancies were resolved by consensus.

Evaluation of quality mainly included the sample size and recruitment of the studies. The papers were first selected by reading the document titles and abstracts. Then, we read the full text of each paper to determine whether the study conformed to our inclusion criteria. Two investigators independently completed this task. Assuming that differences would occur, agreements were to be reached by discussion.

The meta-analysis was performed in fixed or random effect models as appropriate. The effect sizes of the ORs with 95% CIs were pooled to assess the AVN incidence associated with operative fixation in proximal humeral fractures compared with conservative treatment. Heterogeneity among studies was evaluated by Cochran's Q-statistic[12] and I² parameter testing[13]. Also, p < 0.05 or I² > 50% was considered to indicate a heterogeneous nature. When substantial heterogeneity was detected, we calculated the summary ORs and their 95% CIs with the DerSimonian and Laird method in the random effect model[14]. If heterogeneity was not detected, the pooled estimate was presented based on the Mantel-Haenszel method in the fixed effect model[15]. To test the reliability of the results, we undertook a sensitivity analysis by repeating the meta-analysis after removing one study each time.

We further conducted subgroup analysis according to the study design (retrospective/prospective), sample size (≤40/>40), ethnicity (European/Asian), and specific operative approach (plate fixation/tension band wiring fixation/others) to investigate the impact of study characteristics on our outcomes. Publication bias was also assessed by Egger's regression asymmetry test[16].

Analyses were performed using the software Review Manager 5.1 (Cochrane Collaboration: http://​ims.​cochrane.​org/​revman) and the STATA software package version 11.0 (Stata, College Station, TX, USA). All p values were two-sided. A value of p < 0.05 was considered to indicate statistical significance.

The details of the literature search are presented in a flow diagram (Figure 1). We identified 1,120 papers potentially relevant to the search terms (PubMed, 245; MEDLINE, 147; Springer, 276; Elsevier Science Direct, 132; Cochrane Library, 19; Google Scholar, 165; Wanfang, 57; CNKI, 79). There were 117 studies after removing duplicates. During the screening phase based on titles and abstracts, 123 articles were excluded (28 were review articles; 34 did not report AVN; 32 did not report conservative treatment; 29 were case reports). Among the remaining 54 studies for full publication review, only 7 met the inclusion criteria after full publication review.

The characteristics of the seven included studies[5, 9, 11, 17‐20] are presented in Table 1. The studies were published between 1988 and 2012. A total of 291 patients (142 operative fixation cases and 149 conservatively treated cases) with proximal humeral fractures were considered in this meta-analysis. The studies' sample sizes ranged between 12 and 93, and patients' ages ranged between 65.2 and 75.0 years. Various operative approaches were conducted. Plate fixation was performed in two of the seven studies[11, 20], tension band wiring fixation in two studies[9, 18], and Steinmann pin fixation in one study[17]. Also, AO angular blade plate, screw fixation and cerclage, antegrade screw nails, and retrograde pin fixation were used in one study each[5]. Data for single operative approaches were not provided. In the remaining article[19], no details were given for the open reduction with internal fixation approach.

In an overall analysis of the seven selected studies, the heterogeneity test showed that there were heterogeneities (Q² = 16.68, I² = 64%, p = 0.01) among studies. Thus, the random effect model was applied. The results showed that there was a relatively higher risk of AVN in patients with proximal humeral fractures who underwent operative fixation than in those treated conservatively (Figure 2) (total OR = 1.42, 95% CI 0.33–6.11, p > 0.05), but the difference was not statistically significant. The result of Egger's linear regression test (Table 2) indicated that there was no publication bias in this study (t = 1.52, p > 0.05).

The outcomes of the subgroup analysis stratified by study design, sample size, ethnicity, and operative approach are shown in Table 2. The results indicated that a nonsignificant increase of AVN in patients with proximal humeral fractures was consistent in the subgroup analyses when stratified by sample size. In the study design analysis, patients in retrospective studies showed that operative fixation had a nonsignificantly decreased risk of AVN (OR 0.73, 95% CI 0.15–3.55, p > 0.05). In the stratified analysis by ethnicity, Asian patients (OR 0.74, 95% CI 0.04–12.82) showed less risk of developing AVN than European patients (OR 1.60, 95% CI 0.30–8.51) when treated with operative fixation. The result of the subgroup analysis by operative approach demonstrated that plate fixation had a significantly increased risk of AVN compared with conservative treatment (OR 0.20, 95% CI 0.05–0.76, p = 0.019), whereas tension band wiring fixation showed a risk similar to that with conservative treatment (OR 1.77, 95% CI 0.41–7.70, p > 0.05).

We performed the sensitivity analysis by removing one study each time and rerunning the model to determine the effect on each overall estimate. By omitting the Fjalestad et al. study[5], for instance, the pooled OR was changed to 0.64 (95% CI 0.29–1.44, p > 0.05), and the heterogeneity test showed that there were heterogeneities (I² = 38%, p = 0.15). Also, the estimates changed little, which implied that our results were statistically reliable.