Influence of surgical approach on complication risk in primary total hip arthroplasty

Background and purpose Systematic comparisons of anterior approach (A) versus posterior approach (P) in primary total hip arthroplasty (THA) have largely focused on perioperative outcomes. In this systematic review with meta-analysis, we compared complication risk of A versus P in studies of primary THA with at least 1-year mean follow-up. Patients and methods We performed a systematic review of prospective and retrospective studies with at least 1-year mean follow-up that reported complications of A and P primary THA. Complications included infection, dislocation, reoperation, thromboembolic event, heterotopic ossification, wound complication, fracture, and nerve injury. Random effects meta-analysis was used for all outcomes. Complication risk was reported as rate ratio (RR) to account for differential follow-up durations; values >1 indicated higher complication risk with A and values <1 indicated lower risk with A. Results 19 studies were included; 15 single-center comparative studies with 6,620 patients (2,278 A; 4,342 P) and 4 multicenter registries with 157,687 patients (18,735 A; 138,952 P). Median follow-up was 16 (12–64) months) with A and 18 (12–110) months with P. Anterior approach was associated with lower rate of infection (RR =0.55, p = 0.002), dislocation (RR =0.65, p = 0.03), and reoperation (RR =0.84, p < 0.001). No statistically significant differences were observed in rate of thromboembolic event (RR =0.59, p = 0.5), heterotopic ossification (RR =0.63, p = 0.1), wound complication (RR =0.93, p = 0.8), or fracture (RR =1.0, p = 0.9). There was a higher rate of patient-reported nerve injury with A (RR =2.3, p = 0.01). Interpretation Comparing A with P in primary THA, A was associated with lower risk of reoperation, dislocation, and infection, but higher risk of patient-reported nerve injury.


■
The durability of total hip arthroplasty (THA) is excellent with 10-year survivorship exceeding 90% (Hailer et al. 2015, Makela et al. 2014. All standard approaches to the hip have been shown to be safe and effective, with certain advantages and disadvantages of each approach (Mjaaland et al. 2017). While the anterior approach (A) has been increasingly used in the United States, little is known about the safety of the A relative to other common surgical approaches. Several groups (Higgins et al. 2015, Meermans et al. 2017, Putananon et al. 2018) have performed systematic reviews comparing the A with the posterior approach (P) in primary THA. However, follow-up durations of the included studies varied widely, with most studies having less than 1-year follow-up. Comparative safety evaluation of these surgical techniques over a longer period is warranted. The purpose of this systematic review with meta-analysis was to compare the complication risk of A versus P in studies with at least 1-year mean follow-up.

Literature search and data extraction
In accordance with the PRISMA guidelines, we searched MEDLINE and EMBASE for comparative studies of primary THA performed using the A or P. Therapeutic search terms consisting of THA and total hip arthroplasty were combined with the following surgical approach-specifi c search terms: anterior, direct, posterior, posterolateral, and Smith-Petersen. We also manually searched the Directory of Open Access Journals (DOAJ), Google Scholar, and the reference lists of included papers and relevant systematic reviews. No language or date restrictions were applied to the searches. The fi nal search was conducted on June 30, 2017.
When data were reported at multiple intervals during followup, the fi nal value was extracted for analysis. Complications included infection, dislocation, reoperation (for any reason), thromboembolic event, heterotopic ossifi cation, wound complication, fracture, and nerve injury. To account for differential follow-up durations, complication data were extracted by determining the number of events and then calculating the number of person-years in each group to determine incidence rates. Risk of bias in each study was assessed with the Cochrane Collaboration tool, which included evaluations of sequence generation, allocation concealment, blinding, incomplete outcome data, selective outcome reporting, and other sources of bias (Higgins et al. 2011).

Data analysis
We assumed heterogeneous effects among studies a priori and conservatively applied a random effects model for all outcomes. Denominators were adjusted to include the number of patients or hips, as appropriate. The rate ratio (RR) was the effect size statistic of interest, which indicates the ratio of incidence rates (events per person-year) between A and P. A RR value > 1 indicates higher complication incidence rate with A and a value < 1 indicates lower complication incidence rate with A. For each complication, the RR and 95% confi dence interval (CI) were calculated in each study and pooled among all studies. Inconsistency in complication risk among studies was quantifi ed with the I 2 statistic; values of ≤ 25%, 50%, and ≥ 75% represented low, moderate, and high inconsistency, respectively (Higgins et al. 2003). Publication bias was visually assessed with funnel plots (not shown) and quantitatively assessed using Egger's regression test. Post hoc random effects meta-regression using the Knapp-Hartung method (Knapp and Hartung 2003) was performed to assess the possible infl uence of study design, median surgery year, inclusion of learning cases, and follow-up duration on complication risk. P-values were 2-sided with a signifi cance level < 0.05. Analyses were performed using Comprehensive Metaanalysis (version 3.3, Biostat, Englewood, NJ, USA).

Funding and potential confl icts of interest
This work was supported by DePuy Synthes (Raynham, MA, USA). LM received a research grant from DePuy Synthes for data analysis. JW and SB are employees of DePuy Synthes. JG, AK, and FB declare no confl ict of interest in this work.

Study selection
After screening 340 records for eligibility, 19 studies were included in this review, including 15 single-center comparative studies with 6,620 patients (2,278 A; 4,342 P) and 4 multicenter registries with 157,687 patients (18,735 A; 138,952 P). Primary reasons for study exclusion included mean follow-up less than 1 year (27 studies), complications not reported (25 studies), and no comparison of A with P (20 studies) (Figure).

Study and patient characteristics
This review included 4 randomized controlled trials, 1 prospective nonrandomized study, 10 retrospective studies, and 4 multicenter registries. Surgeries in each group occurred during the same period in 11 studies. In 7 studies, learning curve cases comprised some or all of the A group. Median follow-up duration was 16 months (range: 12-64 months) with A and 18 months (range: 12-110 months) with P. Comparing patients treated with A versus P, baseline patient characteristics were well matched for age (median 63 years per group), female sex (median 60% versus 58%), and BMI (median 28 per group) Study eligibility was determined by 2 independent researchers (LM, DF). Disagreements were resolved by discussion. Main inclusion criteria included comparison of A versus P in primary THA, predominant diagnosis of osteoarthritis, mean follow-up duration at least 1 year, and extractable complication data. Titles and abstracts were initially screened to exclude review articles, commentaries, letters, case reports, and obviously irrelevant studies. Full-texts of the remaining articles were retrieved and reviewed. Studies were excluded if patients received revision or bilateral THA. When multiple studies included overlapping series of patients, only the study with the largest sample size was included. Data were independently extracted from eligible peer-reviewed articles by the same 2 researchers. Data discrepancies were resolved by discussion.

Defi nitions and outcomes
( Table 1). The primary risks of bias were attributable to inclusion of retrospective nonrandomized studies (Table 2).

Complications
The A was associated with lower rates of infection (RR = 0.55, p = 0.002 from 7 studies), dislocation (RR = 0.65, p = 0.03 from 11 studies), and reoperation (RR = 0.84, p < 0.001 from 16 studies). In a subgroup analysis of infection, the rate of superfi cial (RR = 0.47, p = 0.5) and deep infection (RR = 0.23, p = 0.1) remained low with A, but neither was statistically signifi cant. When explicitly reported, the most common reasons for reoperation were aseptic loosening, dislocation, fracture, and infection in the A group and dislocation, aseptic loosening, infection, and fracture in the P group. No statistically signifi cant differences were observed in the rate of thromboembolic event (RR = 0.59, p = 0.5 from 4 studies), heterotopic ossifi cation (RR = 0.63, p = 0.1 from 4 studies), wound complication (RR = 0.93, p = 0.8 from 5 studies), or fracture (RR = 1.0, p = 0.9 from 10 studies). Most fracture reports were of intraoperative periprosthetic fractures; however, type and time to fracture was not consistently reported. There was a higher rate of patient-reported nerve injury with A vs. P (RR = 2.3, p = 0.01 from 2 studies). Nerve injuries were described as patient-reported sensory defi cit (Luo et al. 2016) or patient-reported nerve injury with no distinction between   (Amlie et al. 2014). For each complication, heterogeneity among studies was low and publication bias was not evident (Table 3).

Post hoc meta-regression
Post hoc meta-regression was performed to assess the possible infl uence of study design, median surgery year, inclusion of learning cases, and follow-up duration on complication risk. No covariate was statistically signifi cantly associated with the risk of any complication. In comparative studies, there was no statistically signifi cant difference between A vs. P in the rate of any complication. In registries, the rate of patient-reported nerve injury was higher with A while the rates of infection and reoperation were lower with A (Table 4).

Discussion
We conducted a systematic review and meta-analysis of comparative studies of A versus P primary THA with at least 1-year mean follow-up. An anterior approach was associated with a lower risk of reoperation, dislocation, and infection, but higher risk of patient-reported nerve injury. No difference was seen in the rate of thromboembolic event, heterotopic ossification, wound complication, or fracture. While heterogeneity or publication bias was not evident for any outcome, the possibility of such infl uences cannot be ruled out given the small number of studies reporting each complication. A criticism of the A in primary THA is the presence of a learning curve, during which complication rates may be elevated. In an analysis of over 5,000 THA procedures, 50 or more A procedures were required to overcome the learning Notes: A = anterior approach; P = posterior approach. a Rate ratio >1 indicates higher complication incidence rate with anterior approach; rate ratio < 1 indicates lower complication incidence rate with anterior approach. b Inadequate number of studies to calculate value. a Rate ratio > 1 indicates higher complication incidence rate with anterior approach; RR < 1 indicates lower complication incidence rate with anterior approach. b Comparison of rate ratio in comparative studies versus registries, derived from Knapp-Hartung random effects meta-regression model.
curve (de Steiger et al. 2015). In a single-surgeon experience with the fi rst 500 A cases, the most dramatic reduction in complication rates occurred after the fi rst 100 cases (Hartford and Bellino 2017). We identifi ed no substantial infl uence of learning case inclusion on complication rates in meta-regression although this analysis was limited since it was not possible to determine the percentage of the entire sample comprising learning cases. We identifi ed a higher rate of patient-reported nerve injury with A. In the study of Amlie et al. (2014), nerve injury was self-reported in 5.9% of A patients at 24 months follow-up and 3.3% of P patients at 30 months follow-up; however, there was no distinction between sensory or motor involvement. In another comparative study (Luo et al. 2016), sensory defi cit was 3.8% with A and 0% with P at 14 months' follow-up. While comparative nerve injury data were limited to these 2 studies, a high incidence of sensory defi cit with A has been reported in other studies (Bhargava et al. 2010, Goulding et al. 2010. This is primarily attributable to likely iatrogenic injury of the lateral cutaneous femoral nerve. Despite the higher patient-reported nerve injury rate with A, long-term functional limitations or higher reoperation rates are unlikely with these events based on the fi ndings from other studies (Bhargava et al. 2010, Goulding et al. 2010. In a meta-analysis comparing A and P (Higgins et al. 2015), there were no group differences in risk of intraoperative fracture and lower risk of dislocation with A. More recently, a systematic review compared anterior, posterior, and lateral approaches in primary THA (Meermans et al. 2017). In that review, complications were not systematically evaluated although the authors concluded that there were similar rates of complications between surgical approaches. In a network meta-analysis of randomized controlled trials (Putananon et al. 2018), complication risk was reported to be lower with P vs. A (1.0% vs. 1.4%); however, specifi c complications were not described. Among these reviews, follow-up duration varied considerably and was generally less than 1 year. Key differences in our meta-analysis are inclusion of only those studies with mean follow-up of at least 1 year, reporting of multiple specifi c complications, and statistical adjustment to account for differential follow-up periods among studies.
Several aspects of our meta-analysis are novel including the longest duration follow-up of any A versus P review and a comprehensive assessment of complication rates. There are also several limitations. First, despite the longest mean follow-up of any review on this topic, it must be acknowledged that data derived from 16 (A) to 18 (P) months median follow-up must be considered preliminary. Further, while the RR statistic allows for group comparison of event rates on a common scale (per person-year), event rates that are nonconstant with respect to time may complicate interpretation of these results. Second, while osteoarthritis was the predominant diagnosis in each study, reporting of THA indications was inconsistent and may have confounded outcomes.
Third, due to the small number of studies reporting certain complications, some complication estimates reported in this review may change with the addition of data by future studies. Further, the infl uence of study design on complication rates should be interpreted cautiously given the small number of studies for subgroup comparisons. Fourth, complication reporting was generally inconsistent among studies. Adherence to standardized complication reporting guidelines would greatly improve data transparency and consistency in the THA literature. Fifth, no conclusions regarding complication risk with anterolateral or lateral approaches in THA may be derived from this review. Finally, 14 of 19 included studies were retrospective in nature, which are inherently prone to bias.
In summary, comparing A with P in primary THA, A was associated with a lower rate of reoperation, dislocation, and infection, but a higher rate of patient-reported nerve injury.