Comparison of intra and post-operative complication rates among surgical approaches in Total Hip Arthroplasty: A systematic review and meta-analysis

Abstract

Background

Various surgical approaches exist for Total Hip Arthroplasty (THA), but approach specific complication rates remain unknown. The purpose of this systematic review and meta-analysis was to compare rates of common complications between surgical approaches.

Methods

Four electronic databases (Medline, Embase, AMED, Ovid Healthstar) were searched from inception to June 2019. Three pairs of reviewers were involved in determining eligibility, rating internal and external validity, and data extraction. Pooled estimates were generated using a random-effects model and relative risk (RR) was calculated for dislocation, intraoperative and early postoperative fracture, early infection, deep vein thrombosis (DVT), wound complication, and failure of implant ingrowth between four approaches (posterior, anterior, direct lateral, and anterolateral).

Results

Sixty-nine studies (n = 283,036) were included with nineteen randomized control trials, fourteen prospective cohort, and thirty-six retrospective cohort studies (included studies ranged from 1987 to 2019). When compared to the posterior approach, the risk for dislocation was significantly lower in the anterior (RR 0.66, 95% CI 0.54–0.77, p < 0.01), anterolateral (RR 0.50, 95% CI 0.32–0.77, p = 0.03) and lateral (RR 0.74, 95% CI 0.58–0.96, p = 0.02). When compared to the posterior approach, we found higher risk of loosening in the anterolateral (RR 1.89, 95% CI 1.59–2.25, p < 0.01) and lateral (RR 1.21, 95% CI 1.02–1.44, p = 0.03). Overall, evidence was deemed very low and low-quality following GRADE assessment.

Conclusion

Our findings reveal that the posterior approach was associated with a higher risk of dislocation (compared to the anterior, lateral, and anterolateral) but lower risk of loosening (compared to the lateral and anterolateral approach). However, the large number of cohorts and imprecision due to low sample size for most pooled comparisons was still insufficient to confidently conclude that one approach is superior to another. Each approach has its own strengths and weaknesses, and surgeons can use the approach they are most comfortable with.

Level of evidence

Level 2.

1. Introduction

The question of which surgical approach is superior during Total Hip Arthroplasty (THA) has recently emerged given the various approaches that now exist. Importantly, how these approaches influence intra and post-operative complications have been analyzed in recent literature through multiple comparison studies. While reported complication rates from THA are low, the most commonly cited complications include wound complication, deep vein thrombosis, neural deficit, dislocation, and fracture.^1,2

Currently used approaches differ in incision, surgical planes and technique utilized, and reported post-operative complication rates.³ Given this variability, debate continues as to which is the most effective. The most common approaches include posterior⁴ (or Moore), direct anterior (or Heuter), direct lateral (or Hardinge), and anterolateral (or Watson-Jones). The posterior approach has been associated with a greater risk of postoperative dislocation due to the disruption of the posterior joint capsule,⁵ while the direct lateral approach has been associated with an increased risk of postoperative abductor weakness, superior gluteal nerve injury, and limping⁶ due to abductor muscle disruption. The anterolateral approach has been hypothesized to cause superior gluteal nerve injury due to prolonged anterior retraction.⁷ Finally, the direct anterior approach has been associated with greater wound complications.^8,9 Less invasive and muscle sparing techniques to facilitate enhanced recovery pathways have increasingly been emphasized, but the risks and complications are not well known.¹⁰

To further understand the risks associated with each approach, we conducted a systematic review and meta-analysis of the literature to compare the incidence of commonly reported complications between different surgical approaches (anterior versus posterior, lateral versus anterior, lateral versus anterolateral, lateral versus posterior, and posterior versus anterolateral).

2. Methods

This systematic review was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)¹¹ and Cochrane guidelines.¹²

2.1. Literature Search

We systematically searched the electronic databases MEDLINE (1946 to June 2019), AMED (1985 to June 2019), OVID Healthstar (1966 to June 2019), and EMBASE (1946 to June 2019) to identify eligible studies comparing two or more surgical approaches for THA. All studies included were published between 1987 and 2019. Our search strategy is available in Appendix A.

2.2. Study Selection

Three sets of two reviewers independently assessed titles and abstracts from the initial search strategy. Eligible studies included those that (1) evaluated the outcomes of patients who had undergone primary THA, (2) compared two or more surgical approaches including anterior (Smith-Peterson or modified Hueter), posterior (Moore or Southern), direct lateral (Hardinge), or anterolateral (Watson-Jones), (3) included at least one reported complication. We excluded studies that included patients undergoing bilateral THA or hip resurfacing, hip fracture patients, studies evaluating a two-incision technique versus one additional approach, or studies combining two approaches in one cohort. Eligible studies underwent full text review, and disagreement between reviewers was discussed and a third reviewer was consulted if necessary. Inter-rater agreement was determined by a Kappa (K) co-efficient for both titles and abstracts and full text stages. The interpretation of K was as follows: k = 0.81–1.00 as almost perfect agreement, k = 0.61–0.80 as substantial agreement, k = 0.41–0.60 as moderate agreement, and k = 0.21–0.40 as fair agreement.¹³

2.3. Quality Assessment

Two individuals independently performed quality assessment using the Cochrane Collaboration's tool for assessing risk of bias (RoB) for randomized control trials (RCTs) and the ROBINS I tool for non-randomized studies.¹⁴ We used the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach to assess the overall quality of the body of evidence as one of four levels: high, moderate, low, or very low quality. Studies were grouped by type and assessed by outcome.¹⁵

2.4. Data Collection

Three sets of two reviewers independently extracted data from eligible studies. Information on study design, sample size, demographic characteristics (age, sex, BMI, diagnosis), study period and follow up duration, length of stay, and intra- and postoperative complications was recorded. For dichotomous outcomes, event rates were extracted. We contacted authors by e-mail for further information when needed for clarity.

2.5. Timing of Incidence and Definitions

Surgical approaches were compared on the following complications (1) dislocation, (2) intraoperative femoral fracture, (3) intraoperative acetabular fracture, (4) intraoperative “other” fracture, which includes studies that used only the words “fracture” or “periprosthetic”, (5) early postoperative femoral fracture, (7) early postoperative acetabular fracture, (7) early postoperative “other” fracture, (8) implant loosening, (9) deep vein thrombosis (DVT), (10) wound complication, and (11) early infection. For the purposes of our study, an early complication was defined as within 90 days of surgery. Despite a study not reporting when complications occurred and authors unavailable when emailed, we elected to report additional later time frames for the complications of dislocation, loosening, DVT, and wound complication by reporting study time frame ranges.

2.6. Statistical Analysis

Using RevMan (Version 5), we generated forest plots using a Mantel-Haenszel random effects model to calculate relative risk (RR) ratios and their 95% confidence intervals (CIs). Heterogeneity was assessed using the I² statistic and Cochrane's X² test. Based on the threshold guidelines available by Cochrane and Higgins et al., I² thresholds were 0%–30% (heterogeneity unlikely to be important), 30%–50% (moderate heterogeneity), 50–90% (substantial heterogeneity) and 90%–100% (significant heterogeneity).^12,16 For the test of overall effect, p < 0.05 was considered statistically significant. Relative risk and I² could not be estimated for studies where the event rate in both groups was zero. Where heterogeneity was greater than 50%, we explored heterogeneity through subgroup analysis using the following a priori hypotheses: (a) differences in study design (hypothesizing that RCTs would yield smaller effect sizes than cohorts) and (b) by RoB (hypothesizing that studies with an overall low RoB may generate a smaller effect size than those with a high RoB).

3. Results

3.1. Study characteristics

The initial search identified 4107 articles. Following titles and abstract review, 289 full text studies were assessed. Following full-text review, 69 studies were included (Fig. 1). The studies were categorized by design (19 RCTs, 14 prospective cohorts, and 36 retrospective cohorts). Interrater agreement at the titles and abstracts and full text review stage was substantial (k = 0.61 and k = 0.75, respectively). A total of 283,036 THA patients using one of the four surgical approaches: posterior (n = 185,920), direct lateral (n = 57,437), anterolateral (n = 23,603) and direct anterior (n = 16,076). Study characteristics are available in Appendix B.

Fig. 1.

Preferred reporting items for systematic reviews and meta-analysis flow diagram.

3.2. Methodological quality assessment

RoB assessments for each study is available in Appendix C. The RCTs (n = 19) revealed most studies had a low or unclear risk of selection bias in both random sequence generation and concealed treatment allocation. Performance bias was deemed low in the majority of studies (16/19) because surgeons were considered experienced in the approach. Detection bias was primarily low due to sufficient blinding of outcome assessors or unclear when no information was available. Assessment of the cohort studies (n = 50) revealed that the majority of studies were low RoB on intervention classification and deviation from intervention. However, confounding was a concern where almost half of the studies were at a moderate RoB due to different implants used within the same study,¹⁷ differences in post-operative rehabilitation protocol,¹⁸ and differences in follow up period within the same study. Selection bias was also a concern where eight studies were deemed serious risk due to patients being included into a group based on the preference of the surgeon or expertise. Notably, nearly a third of studies were deemed moderate RoB for outcome assessment, where assessors were not blinded, and knowledge of the approach could have influenced the reported complications. The overall quality of evidence using the GRADE approach found that most studies were deemed very low or low, indicating that the true effect might be markedly different from the estimated effect. A common theme was that evidence was rated down for imprecision due to low sample size, and for serious limitations due to high or moderate RoB. In some cases, publication bias was strongly suspected as evident by asymmetrical funnel plots which could be due to selective reporting bias or poor study design. Detailed GRADE evidence profiles are available in Appendix D.

3.3. Anterior versus posterior approach

Twenty-nine studies were included in the analysis between anterior versus posterior approach: 4 were RCTs, 5 were prospective cohorts, and 20 were retrospective cohorts. The results of the meta-analysis on intraoperative femoral fracture, intraoperative “other” fracture, early postoperative femoral fracture, early postoperative “other” fracture, wound complication, loosening, and early infection demonstrated no significant difference between approaches (Table 1). Twenty-five studies reported an overall significantly lower risk of dislocation in the anterior approach (RR 0.66, 95% CI 0.56–0.77, p < 0.001, I² = 0%) with a study time frame range from 1 month to 8 years (Fig. 2).

Table 1.

Summary of findings on complications in anterior versus posterior approach.

Complication	Number of Studies	Overall GRADE	Anterior	Posterior	RR (95% CI)	P-value	I2
Dislocation*†	3 RCTs	⊕⊕⊖⊖	166/22,563	1838/163,582	0.66 (0.56–0.77)	<0.00001	0%
	22 cohorts	⊕⊖⊖⊖
Intraoperative Femoral Fracture	3 RCTs	⊕⊕⊖⊖	19/1821	15/1657	1.31 (0.66–2.61)	0.45	0%
	6 Cohorts	⊕⊖⊖⊖
Intraoperative Acetabular Fracture	2 Cohorts	⊕⊖⊖⊖	1/46	0/46	0.69 (0.04–11.12)	0.79	39%
Intraoperative “Other” Fracture	3 Cohorts	⊕⊕⊖⊖	8/404	3/317	1.16(0.13–10.57)	0.90	47%
Postoperative Femoral Fracture*	1 RCT	⊕⊕⊖⊖	2/101	0/104	3.12 (0.33–29.50)	0.32	0%
	1 Cohort	⊕⊖⊖⊖
Postoperative “Other” Fracture*	3 Cohorts	⊕⊖⊖⊖	6/355	1/273	1.86 (0.07–49.48)	0.52	20%
Wound Complication†	4 RCTs	⊕⊕⊖⊖	33/2747	24/2521	1.15 (0.39–3.42)	0.80	47%
	3 Cohorts	⊕⊖⊖⊖
Infection*	3 Cohorts	⊕⊖⊖⊖	4/177	4/113	0.70 (0.15–3.27)	0.65	0%
DVT†	1 RCT	⊕⊕⊖⊖	49/3352	480/23,882	0.76 (0.57–1.02)	0.06	0%
	3 Cohorts	⊕⊖⊖⊖
Loosening†	5 Cohorts	⊕⊖⊖⊖	118/16,758	677/106,971	1.15 (0.94–1.40)	0.17	0%

*early (within 90 days), †unknown time, “Not estimable”: not enough studies to conduct meta-analysis.

⊕⊖⊖⊖ (Very Low), ⊕⊕⊖⊖ (Low), ⊕⊕⊕⊖ (Moderate), ⊕⊕⊕⊕ (High).

Fig. 2.

Forest plot for dislocation rates between the anterior versus posterior approach.

3.4. Lateral versus anterior approach

Nineteen studies were included in the analysis between lateral versus anterior approach: 6 were RCTs, 4 were prospective cohorts and 9 were retrospective cohorts. The results of the meta-analyses on dislocation, intraoperative femoral fracture, intraoperative “other” fracture, early postoperative femoral fracture, early postoperative “other” fracture, DVT, and early infection demonstrated no significant difference between approaches (Table 2).

Table 2.

Summary of meta-analyses on complications in lateral versus anterior approach.

Complication	Number of Studies	Overall GRADE	Lateral	Anterior	RR (95% CI)	P-value	I2
Dislocation*†	12 Cohorts	⊕⊖⊖⊖	191/37,965	87/17,770	1.15 (0.88–1.49)	0.31	0%
	3 RCTs	⊕⊕⊕⊖
Intraoperative Femoral Fracture	2 RCTs	⊕⊕⊕⊖	7/582	12/664	0.59 (0.22–1.61)	0.30	0%
	3 Cohorts	⊕⊖⊖⊖
Intraoperative Acetabular Fracture	1 Cohort	⊕⊖⊖⊖	0/258	1/372	Not estimable
Intraoperative “Other” Fracture	1 RCT	⊕⊕⊖⊖	10/766	8/737	1.10 (0.44–2.74)	0.84	0%
	5 Cohorts	⊕⊖⊖⊖
Postoperative Femoral Fracture*	2 Cohorts	⊕⊖⊖⊖	1/392	5/278	0.25 (0.04–1.66)	0.15	0%
Postoperative Acetabular Fracture*	1 Cohort	⊕⊖⊖⊖	0/372	4/258	Not estimable
Postoperative “Other” Fracture*	1 Cohort	⊕⊖⊖⊖	0/40	1/40	Not estimable
Infection*	3 RCTs	⊕⊕⊕⊖	4/610	13/907	0.57 (0.23–1.52)	0.26	0%
	2 Cohorts	⊕⊖⊖⊖
DVT†	2 RCT	⊕⊕⊕⊖	0/771	6/615	0.26 (0.06–1.15)	0.08	0%
	4 Cohorts	⊕⊖⊖⊖
Loosening†	4 Cohorts	⊕⊕⊖⊖	314/36,687	118/15,414	0.80 (0.40–1.58)	0.52	17%
	2 RCT	⊕⊕⊖⊖

*early (within 90 days), †unknown time, “Not estimable”: not enough studies to conduct meta-analysis.

3.5. Lateral versus anterolateral approach

Six studies were included in the analysis between lateral versus anterolateral approach: 5 were RCTs, and one was a retrospective cohort. The results of the meta-analysis on dislocation, intraoperative femoral fracture, intraoperative “other” fracture, wound complication, early infection, DVT and loosening demonstrated no significant difference between approaches (Table 3).

Table 3.

Summary of meta-analyses complications in lateral versus anterolateral approach.

Complication	Number of Studies	Overall GRADE	Lateral	Anterolateral	RR (95% CI)		P-value	I2
Dislocation*†	3 RCTs	⊕⊕⊖⊖	173/37,015	68/17,475	1.07 (0.80–1.43)		0.64	0%
	4 Cohorts	⊕⊖⊖⊖
Intraoperative Femoral Fracture	2 RCTs	⊕⊕⊕⊖	4/530	3/163	0.66 (0.16–2.76)		0.57	0%
	2 Cohorts	⊕⊕⊖⊖
Intraoperative “Other” Fracture	1 Cohort	⊕⊖⊖⊖	126/35,830	43/12,744	Not estimable
Wound Complication†	2 RCTs	⊕⊕⊖⊖	2/47	2/43	0.88 (0.13–5.72)		0.89	0%
Infection*	2 RCTs	⊕⊕⊖⊖	1/93	1/92	0.99 (0.10–9.35)		0.99	0%
DVT†	1 RCT	⊕⊕⊖⊖	0/42	1/41	Not estimable
Loosening†	2 Cohorts	⊕⊖⊖⊖	313/36,223	159/12,774	0.36 (0.05–2.65)	0.32		60%

*Early (within 90 days), †Unknown Time, “Not estimable”: not enough studies to conduct meta-analysis.

3.6. Lateral versus posterior approach

Twelve studies were included in the analysis between lateral versus posterior approach: 2 were RCTs, 4 were prospective cohorts, and 6 were retrospective cohorts. The results of the meta-analysis on intraoperative femoral fracture, early infection, and DVT demonstrated no significant difference between approaches (Table 4). When pooled, twelve studies found a significantly lower risk in dislocation in the lateral approach (RR 0.74, 95% CI 0.58–0.96, p = 0.02, I² = 24%) with a study time frame of 90 days to 17 years (Fig. 3). On the other hand, four cohort studies found a lower risk of implant loosening in the posterior approach (RR 1.21, 95% CI 1.02–1.44, p = 0.03, I² = 48%) with a study time frame range of 5 months–17 years (Fig. 4). Importantly, the lower boundary of the 95% CIs suggests that the benefit could be as low as 2% in favour of the posterior approach.

Table 4.

Summary of meta-analyses complications in lateral versus posterior approach.

Complication	Number of Studies	Overall GRADE	Lateral	Posterior	RR (95% CI)	P-value	I2
Dislocation*†	2 RCTs	⊕⊕⊖⊖	314/60,705	1145/151,943	0.74 (0.58–0.96)	0.02	24%
	10 Cohorts	⊕⊖⊖⊖
Intraoperative Femoral Fracture	2 RCT	⊕⊕⊕⊖	5/257	2/231	1.57(0.36, 6.95)	0.55	0%
	2 Cohorts	⊕⊖⊖⊖
Postoperative “Other” Fracture*	1 Cohort	⊕⊕⊖⊖	0/40	1/40	Not estimable
Infection*	1 RCT	⊕⊕⊕⊖	2/30	0/30	Not estimable
DVT†	1 RCT	⊕⊕⊕⊖	2/109	2/127	1.48 (0.18–12.24)	0.71	28%
	2 Cohorts	⊕⊖⊖⊖
Loosening†	4 Cohorts	⊕⊖⊖⊖	896/52,674	877/107,549	1.21 (1.02–1.44)	0.03	48%

*Early (within 90 days), †Unknown Time, “Not estimable”: not enough studies to conduct meta-analysis.

Fig. 3.

Forest plot for dislocation rates between lateral versus posterior approach.

Fig. 4.

Forest plot for implant loosening rates between lateral versus posterior approach.

3.7. Anterolateral versus posterior approach

Nine studies were included in the analysis between anterolateral versus posterior approach: 2 were RCTs, 3 were prospective cohorts, and 4 were retrospective cohorts. The results of the meta-analysis on intraoperative femoral fracture, early infection, and loosening demonstrated no significant difference between approaches (Table 5). Eleven studies found a lower dislocation risk in favour of the anterolateral approach (RR 0.50, 95% CI 0.32–0.77, p = 0.03) with a study time frame range from 90 days to 10 years (Fig. 5). We detected considerable heterogeneity (I² = 71%) but subgroup analyses based on a priori hypotheses were unable to resolve heterogeneity. Two studies report an overall lower risk of loosening in favour of the posterior approach (RR 1.89, 95% CI 1.59–2.25, p < 0.001, I² = 0%) with a study time frame range from 6 months to 6 years (Fig. 6). However, one large cohort study was contributing nearly all of the weighted estimated (99.7%) and the limitations and findings of the pooled analysis were not different from the single study.

Table 5.

Summary of complications in anterolateral versus posterior approach.

Complication	Studies (N)	Overall GRADE	Anterolateral	Posterior	RR (95% CI)	P-value	I2
Dislocation*†	10 Cohorts	⊕⊖⊖⊖	175/23,298	1228/139,704	0.50 (0.32, 0.77)	0.002	71%
	1 RCT	⊕⊕⊕⊖
Intraoperative Femoral Fracture	1 RCT	⊕⊕⊖⊖	1/73	6/65	0.20 (0.04–1.16)	0.07	0%
	1 Cohort	⊕⊖⊖⊖
Wound Complication†	1 Cohort	⊕⊖⊖⊖	0/35	1/43	Not estimable
Infection*	2 RCTs	⊕⊕⊕⊖	2/45	1/63	2.13 (0.35–12.83)	0.42	0%
	1 Cohort	⊕⊖⊖⊖
DVT†	1 Cohort	⊕⊖⊖⊖	4/37	0/37	Not estimable
Loosening†	1 RCT	⊕⊕⊖⊖	160/12,786	663/100,874	1.89 (1.59–2.25)	0.0001	0%
	2 Cohorts	⊕⊖⊖⊖

*Early (within 90 days), †Unknown Time, “Not estimable”: not enough studies to conduct meta-analysis.

Fig. 5.

Forest plot for dislocation rates between anterolateral versus posterior approach.

Fig. 6.

Forest plot for implant loosening rates between anterolateral versus posterior approach.

4. Discussion

Our findings reveal that few differences exist in common intra- and post-operative complications between frequently performed surgical approaches for THA. When comparing other approaches (anterior, lateral, and anterolateral) to the posterior group, the risk of dislocation was 50%–75% lower. This is consistent with the literature that suggests that the anterior and lateral approaches may have lower rates of dislocations.^10,19 In a meta-analysis evaluating the association between soft-tissue repair and dislocation in the posterior approach, an eight times greater risk of dislocation was observed when soft tissue was not repaired, demonstrating the importance of the soft tissues involved.20, 21, 22

The posterior approach was also associated with a lower risk of implant loosening compared to the anterolateral (based on three cohort studies, but one large study primarily contributed to the overall estimate) and lateral approach (based on four cohort studies). However, it is important to note that we did not look at whether studies included cemented or non-cemented implants or differences in prosthesis design which is known to significantly impact loosening and fracture rates, where cemented has been shown to have fewer rates of loosening long term.^23,24 Another confounder for the findings of loosening is study time frame. A study with a longer follow up is more likely to find late loosening rates, regardless of surgical approach. Since the range for the study period for our meta-analysis was from six weeks to 17 years, this has implications on our findings since the farther we are from the time of surgery, the more challenging it becomes to attribute to surgical approach alone. In the included studies that were significant for loosening, three admitted the learning curve may influence results,25, 26, 27 while one attempted to control for it by excluding the first 150 cases by each hospital.¹⁷ The effect of learning curve has been well demonstrated, with Laffosse et al. reporting a higher rate of intra- and peri-operative complications during the learning curve of the anterolateral approach.²⁸ Our study did not control for the learning curve, but studies were deemed at risk for selection bias due to expertise differences.

The overall quality of evidence was deemed low or very low since the majority of studies were imprecise due to few events. Our meta-analysis was constructed primarily of underpowered RCTs and cohort studies, and large database studies. In particular, Zjilstra et al. used the Dutch Arthroplasty Registry and reported on complications, where the posterior approach group had nearly three times as many patients.¹⁷The ability to discern case details is challenging in registry data given the difficulty in obtaining the timing of the complication, as well as the preponderance of posterior approach cases and the recent introduction of new approaches. Further, Zjilstra et al. only reported on dislocations that were treated with a surgical intervention and did not capture non-surgical management of dislocations, which likely means dislocations were under reported. The nature of registry studies also means that selection and detection bias are unclear and complication rates may be influenced by capture rates, and procedure expertise. There is a need for quality registries with uniform reporting techniques and consistency in complication recording in order to better understand differences in approaches.

Limitations include that our conclusions are only as strong as the evidence available, which was primarily level 2 and 3 cohort studies. We did not stratify anterior approach studies for those that evaluated the use of a specialized table for direct anterior cases, which has been associated with increased intra-operative trochanteric fractures,²⁹ and implant type, which has been reported to influence complication rates.²⁵ We also did not capture the sheer breadth of complications that may occur postoperatively, such as limb-length discrepancy, bleeding, limping, and nerve damage. Further, more studies were available for some approaches (posterior, lateral) while fewer were available for less frequently used approaches (anterolateral, anterior) and the effect of these studies may be larger as they are still overcoming the learning curve and random sampling error compared to older procedures. Another important limitation is the variability amongst length of follow up between studies included in our analysis. While some studies gave indications on the timing of complication, others included only final follow up times, or study periods, which ranged from six weeks to 17 years. The changing time periods of these studies as well as the experience of the surgeons may have a major influence on the complications and may explain the extreme variability of the data collected.

The strengths of this study are that it is the first to compare complications rates between multiple approaches, with the inclusion of both RCTs and cohort studies, allowing for a broad scope of the literature. Little heterogeneity allowed us to aggregate all data into forest plots and the inclusion of a range of follow up times in certain complications allowed for studies to be included that otherwise would not have been.

In conclusion, our findings reveal that the posterior approach has higher risk of dislocation (when compared to the anterior, lateral, and anterolateral) but lower risk of loosening (when compared to the lateral and anterolateral approach). However, we believe that the low-quality evidence was insufficient for us to confidently conclude that one approach is superior to another. Given that each approach has its own strengths and weaknesses, surgeons can use the approach they are most comfortable with in context of the patient.

Source of funding

None.

Declaration of competing interest

Dr. Lanting reports grants, personal fees and other from Smith and Nephew, grants, personal fees and other from Stryker, grants, personal fees and other from DePuy, personal fees from IntelliJoint, other from Zimmer, outside the submitted work; Dr. Somerville reports other from Smith and Nephew, Zimmer, Stryker, Depuy, outside the submitted work.

Appendix A.

Search Strategy in OVID Healthstar, AMED, MEDLINE, EMBASE

• 1.Arthroplasty, replacement, hip/
• 2.Hip Prosthesis/
• 3.Or/1-2
• 4.Arthroplasty/or arthroplasty, replacement/
• 5.Joint Prosthesis/
• 6.“Prostheses and Implants"/
• 7.(arthoplasty or replacement or prosthes#s).tw.
• 8.Or/4-7
• 9.Hip/or hip joint/or hip.tw.
• 10.8 and 9
• 11.3 or 10
• 12.Kocher-langenbeck.tw.
• 13.Posterior.tw.
• 14.12 or 13
• 15.3 and 14
• 16.Hardinge.tw.
• 17.Lateral.tw.
• 18.16 or 17
• 19.3 and 18
• 20.3 and 18
• 21.(anterior or direct anterior).tw.
• 22.3 and 21
• 23.Watson-jones.tw.
• 24.3 and 23
• 25.(smith-petersen or heuter).tw.
• 26.3 and 25
• 27.Minimally invasive.tw.
• 28.3 and 27
• 29.Two incision.tw.
• 30.3 and 29
• 31.(mt or su).fs.
• 32.11 and 31
• 33.15 or 19 or 22 or 24 or 26 or 28 or 30
• 34.32 and 33

Appendix B.

Table 1.

Study demographics for anterior versus posterior approach.

Study	Study Type	Participants	Male (%)	Female (%)	Mean Age (SD)	Mean BMI (SD)	Study Time Frame
Bergin et al., 2011	P.C.						4 weeks
Anterior		29	34	66	69 (9)	26 (5)
Posterior		28	50	50	65 (11)	29 (5)
Hananouchi et al., 2009	P.C.						1 year
Anterior		20	10	90	55 (6)	22 (3)
Posterior		20	10	90	57 (8)	21 (3)
Panichkul et al., 2016	R.C.						3 years
Anterior		594	42	58	62.3 (11.1)	28.4 (5.6)
Posterior		88	22.7	77.3	72.6 (11)	28.4(7.2)
Lateral		421	46.1	53.9	62.2(12.4)	29.1(6.3)
Nakata et al., 2009	R.C.						6 months
Anterior		99	16	84	63 (1)	23 (1)
Posterior		96	14	87	65 (11)	23 (1)
Spaans et al., 2012	R.C.						12 months
Anterior		46	52	48	69 (10)	25 (3)
Posterior		46	30	70	68 (11)	29 (4)
Sugano et al., 2009	R. C.						24 months
Anterior		39	8	92	57 (12)	23 (4)
Posterior		33	12	88	56 (13)	23 (4)
Barrett et al., 2013	R.C.T.						12 months
Anterior		43	67	33	61 (9)	31 (5)
Posterior		44	43	57	63 (8)	29 (4)
Rodriguez et al., 2014	P.C.						12 months
Anterior		67	47	53	60 (10)	27 (4)
Posterior		65	43	57	59 (6)	28 (4)
Taunton et al., 2014	R.C.T.
Anterior		27	44	56	62	28
Posterior		27	48	52	66	29
Sheth et al., 2015	R.C.						10 years
Anterior		1851	44	56	65 (11)	28 (5)
Posterior		31,747	43	58	66 (12)	29 (6)
Lateral		667	42	58	65(11)	30(6)
Anterolateral		4226	42	58	67 (11)	29 (6)
Hamilton et al., 2015	R.C.						Early (90 days) Late (up to 2.5 years)
Anterior		100	36	64	63 (15)	29 (7)
Posterior		100	39	61	61 (13)	29 (6)
Kobayashi et al., 2016	R.C.						4 years
Anterior		75	11	89	63 (13)	24 (4)
Posterior		77	17	83	64 (14)	24 (4)
Tripuraneni et al., 2016	R.C.						3 years (mean)
Anterior		66	39	61	60.2 (N/A)	27.6 (N/A)
Posterior		66	39	61	60.2 (N/A)	27.8 (N/A)
Petis et al., 2016	P.C.						3 months
Anterior		40	37.5	62.5	66.9(9.5)	27.9 (4.3)
Posterior		38	35	65	66.7(9.2)	28.2(5.3)
Lateral		40	35	65	65.5(10.4)	29.1(5.6)
Cheng et al., 2016	R.C.T.						3 months
Anterior		35	43	57	59 (4)	28 (1)
Posterior		38	47	53	63 (4)	28 (2.5)
Fransen et al., 2016	R.C.						12 months
Anterior		45	34	66	64 (9)	28 (3)
Posterior		38	37	63	63 (9)	25 (3)
Tsukada et al., 2015	R.C.						14 years
Anterior		139	10	90	67 (10)	23 (3)
Posterior		177	17	83	62 10)	24 (4)
Zjilstra et al., 2017	R.C.						6 years
Anterior		14,446	32.2	67.8	N/A	N/A
Posterior		100,823	32.3	67.7	N/A	N/A
Anterolateral		12,744	32.9	67.1	N/A	N/A
Lateral		35,830	31.8	68.2	N/A	N/A
Poehling-Monaghan et al., 2015	P.C.						8 weeks
Anterior		126	46	54	65 (12)	30 (6)
Posterior		96	45	55	64 (13)	30 (6)
Purcell et al., 2018	R. C.						12 months
Anterior		2424	41.5	58.5	62.6(N/A)	27.7 (N/A)
Posterior		2227	47	53	62.7 (N/A)	28.2 (N/A)
Leucht et al., 2014	R.C.						3 years (min)
Anterior		100	52	48	59 (14)	28 (5)
Posterior		100	57	43	60 (13)	29 (7)
Zawadasky et al., 2013	R.C.						6 weeks
Anterior		50	44	56	60.8 (11.8)	28.6(6.2)
Posterior		50	28	72	56. (11.4)	27.9(6.2)
Balasubramaniam et al., 2016	R. C.						12 months
Anterior		50	50	50	62.5(9.01)	31.3(5.2)
Posterior		42	33.3	66.7	57 (12.84)	29.9(6.7)
L'Hommedieu et al., 2016	R.C.						3 months
Anterior		3120	N/A	N/A	N/A	N/A
Posterior		23,653	N/A	N/A	N/A	N/A
Malek et al., 2016	R.C.						18.1 months (mean)
Anterior		265	44.2	55.8	70.8 (N/A)	28.5(N/A)
Posterior		183	47	53	70 (N/A)	29 (N/A)
Taunton et al., 2018	R.C.T.						12 months
Anterior		52	51	49	65 (10)	29 (22)
Posterior		49	51	49	64 (11)	30 (4)
Angerame et al., 2018	R.C.
Anterior		2431	N/A	N/A	N/A	N/A	7 years
Posterior		4463	N/A	N/A	N/A	N/A
Ponzio et al., 2018	R.C.						3.2 years
Anterior		289	44.7	55.3	64.7 (11.2)	65.1 (9.8)
Posterior		4249	42.2	57.8	28.1 (5.7)	28.4 (5.5)
Aggrawal et al., 2019	R.C.						2 years
Anterior		1329	43.0	57	63.3	27.7
Posterior		1657	45.5	54.5	62.5	30.1
Lateral		393	44.3	56.7	61	29.9
Anterolateral		30	53.3	47.7	63.9	27.7

R.C.T.: Randomized Control Trial, R.C.: Retrospective Cohort, P.C.: Prospective Cohort.

Table 2.

Study demographics for anterior vs. lateral approach.

Study	Study Type	Participants	Male (%)	Female (%)	Mean Age (SD)	Mean BMI (SD)	Study Time Frame
Berend et al 2009	R.C.						5 months
Anterior		258	N/A	N/A	63 (N/A)	28.9 (N/A)
Lateral		372	NA	NA	63 (N/A)	30.4(N/A)
Pogliacomi & DeFilippo et al., 2012	R.C.						12 months
Anterior		35	54	46	65 (8)	27 (2)
Lateral		35	51	49	65 (8)	27 (2)
Sendtner et al., 2011	P.C.						12 months
Anterior		74	68	32	68 (8)	29 (5)
Lateral		60	25	75	68 (9)	29 (5)
Dienstknecht et al., 2014	R.C.T.						3 months
Anterior		55	40	60	62 (12)	28 (6)
Lateral		88	47	53	61 (12)	30 (6)
Pogliacomi & Paraskevopoulos al. 2012	R.C.						12 months
Anterior		30	50	50	68 (N/A)	27 (N/A)
Lateral		30	47	53	69 (N/A)	27 (N/A)
Reichert et al., 2015	R.C.
Anterior		85	43	57	68 (N/A)	28 (N/A)	3.3 years (mean)
Lateral		86	56	44	64 (N/A)	29 (N/A)	5.4 years (mean)
Chen et al., 2016	R.C.						309.7 days (mean)
Anterior		186	52	48	68 (10)	30 (5)
Lateral		186	49	51	68 (11)	30 (5)
Ilchmann et al., 2013	P.C.						6 weeks
Anterior		113	53	47	70 (13)	27 (5)
Lateral		142
De Anta-Diaz et al., 2016	R.C.T.						6 months
Anterior		50	52	48	64.8(10.1)	26.6(3.9)
Lateral		49	53	47	63.5 (12.5)	26.1 (3.1)
Alecci et al., 2011	R.C.						NR
Anterior		221	45.2	54.8	70.7 (8.2)	N/A
Lateral		198	37.9	62.1	70.15(9.6)	N/A
Restrepo et al., 2010	R.C.T.						2 years
Anterior		50	34	66	62.02 (N/A)	25.18(N/A)
Lateral		50	44	54	59.91 (N/A)	25.17 (N/A)
Ilchmann et al., 2016	P.C.						2 years
Anterior		700	53	47	71 (median) (10.6)	26.6 (4.2)
Lateral		404	49	51	71 (median) (10.6)	27.2 (5.2)
Reichert et al., 2018	R.C.T.						12 months
Anterior		77	58	42	63.2 (8.2)	28.1 (3.7)
Lateral		71	55	45	61.9 (7.8)	28.3 (3.4)
Brismar et al., 2018	R.C.T.
Anterior		50	36	64	66 (58–74)	27 (24–29)	5 years
Lateral		50	34	66	67 (60–76)	27 (24–30)
D'Arrigo et al., 2009	R.C.T.
Anterior		20	60	40	64 (8)	23.1 (1.5)	6 weeks
Lateral		20	70	30	66.3 (10.4)	37.6 (3)
Anterolateral		20	55	45	66 (7.5)	23.1 (1.5)
Wayne & Reinhard et al., 2009	R.C.						NR
Anterior		100	29	71	68 (35–90)	26.6 (16–38)
Lateral		100	34	66	68 (32–90)	27.0 (18–42)

R.C.T.: Randomized Control Trial, R.C.: Retrospective Cohort, P.C.: Prospective Cohort, NR: Not reported.

Table 3.

Study demographics in the posterior vs. lateral approach.

Study	Study Type	Participants	Male (%)	Female (%)	Mean Age (SD)	Mean BMI (SD)	Study Time Frame
Downing et al., 2001	P.C.						12 months
Lateral		49	41	59	65(6)	(N/A)
Posterior		51	49	51	67 (N/A)	(N/A)
Schleicher et al., 2011	P.C.						6 months
Lateral		64	31	69	69 (9)	29 (4)
Posterior		64	25	75	68 (10)	27 (4)
Witzleb et al., 2009	R.C.T.						3 months
Lateral		30	47	53	55 (4)	29 (5)
Posterior		30	50	50	58 (5)	27 (5)
Vincente et al., 2008	P.C.						6 months
Lateral		42	62	38	57 (11)	27 (4)
Posterior		34	62	38	50 (13)	27 (4)
Jameson et al., 2014	R.C.						12 months
Lateral		816	35.9	64.1	73.2(7.2)	28.9 (5.1)
Posterior		1121	31.8	68.2	72.6(8.1)	28.6 (4.9)
Vicente et al., 2014	R.C.T.						7.2 years (mean)
Lateral		121	57	43	56 (N/A)	27 (N/A)
Posterior		103	53	47	56 (N/A)	27 (N/A)
Arthursson et al., 2007	R.C.						17 years
Lateral		16,381	28	72	71 (median)	(N/A)
Posterior		6604	30.4	69.5	73 (median) (N/A)	(N/A)
Amlie et al., 2014	R.C.						3 years
Lateral		431	36	64	66 (7.3)	N/A
Posterior		421	36	64	66 (7.1)	N/A
Anterior		421	31	69	67 (7.1)	N/A
Chomiak et al., 2015	R.C.						9 months
Anterolateral		22	N/A	N/A	60.7 (N/A)	N/A
Posterior		33	N/A	N/A	62 (N/A)	N/A
Lateral		15	N/A	N/A	66.2(N/A)	N/A

R.C.T.: Randomized Control Trial, R.C.: Retrospective Cohort, P.C.: Prospective Cohort.

Table 4.

Study demographics for the posterior vs. anterolateral approach.

Study	Study Type	Participants	Male (%)	Female (%)	Mean Age (SD)	Mean BMI (SD)	Study Time Frame
Goosen et al. 2009	R.C.T.						12 months
Anterolateral		27	53	47	62 (7)	26 (3)
Posterior		29	43	57	62 (6)	27 (3)
Carlson et al., 1987	P.C.						12 months
Anterolateral		37	92	8	67 (7)	N/A
Posterior		37	84	16	64 (13)	N/A
Abdel et al., 2016	R.C.						27 months (mean) 133 months (last)
Anterolateral		3384	N/A	N/A	N/A	N/A
Posterior		5765	N/A	N/A	N/A	N/A
Laffosse et al., 2007	P.C.						6 months
Anterolateral		33	60	40	57 (13)	26 (4)
Posterior		43	65	35	56 (14)	25 (3)
Meneghini et al., 2008	R.C.T.						6 weeks
Anterolateral		7	N/A	N/A	54 (9)	26 (2)
Posterior		8	N/A	N/A
Takao et al., 2016	P.C.						12 months
Anterolateral		32	9	91	60 (12)	23 (N/A)
Posterior		57	18	82	63 (11)	24 (N/A)
Smith et al., 2012	R.C.						3 years
Anterolateral		246	N/A	N/A	N/A	N/A
Posterior		665	N/A	N/A	N/A	(N/A)
Ritter et al., 2001	R.C.						12 months
Anterolateral		122	43	57	68	N/A
Posterior		184	45.3	54.7	67	(N/A)
Edmunds et al., 2011	R.C.						1 year (dislocation only)
Anterolateral		2471	42.3	57.7	68.9	N/A
Posterior		362	41.2	58.8	65.7	(N/A)

R.C.T.: Randomized Control Trial, R.C.: Retrospective Cohort, P.C.: Prospective Cohort.

Table 5.

Study demographics in the anterolateral vs. lateral approach.

Study	Study Type	Participants	Male (%)	Female (%)	Mean Age (SD)	Mean BMI (SD)	Study Time Frame
Bernasek et al., 2010	R.C.						12 months (min)
Anterolateral		47	N/A	N/A	N/A	27 (3)
Direct Lateral		45	N/A	N/A	N/A	31 (4)
Inaba et al., 2011	R.C.T.						12 months
Anterolateral		50	24	76	64 (11)	23 (4)
Direct Lateral		52	25	75	65 (11)	24 (5)
Martin et al., 2011	R.C.T.						12 months
Anterolateral		42	29	71	67 (10)	31 (6)
Direct Lateral		41	34	66	63 (10)	29 (6)
Muller et al., 2012	R.C.T.						3 months
Anterolateral		15	40	60	64.3 (7)	26.9 (3.3)
Direct Lateral		15	33.3	66.7	66.2 (8)	27 (3.1)
Landgraeber et al., 2013	R.C.T.						3.5 years
Anterolateral		28	33.3	66.7	70.26 (4.05)	27.03(2.82)
Direct Lateral		32	35	65	71.03 (5.38)	26.76 (3.83)
Pospischill et al., 2010	R.C.T.						12 weeks
Anterolateral		20	40	60	61.9 (N/A)	25.7(N/A)
Direct lateral		20	40	60	60.6 (N/A)	25.7 (N/A/)

R.C.T: Randomized Control Trial, R.C.: Retrospective Cohort, P.C.: Prospective Cohort.

Appendix C. Risk-of-bias assessment of included articles

Table 1.

Methodological and Quality Assessment of Randomized Control Trials (N = 19)

Study (Name, Year)	Selection Bias: Random Sequence Generation	Selection Bias: Concealed Treatment Allocation	Performance Bias: Blinded participants and personnel	Detection Bias: Blinded Outcome Assessment	Attrition Bias: Incomplete Outcome Data	Reporting Bias	Other
Barrett 2013	Unclear	Unclear	Low	High	Unclear	Low	None
Cheng 2016	Low	Low	Low	Unclear	Low	Low	None
Taunton 2017	Unclear	Unclear	Low	Unclear	Unclear	Low	None
Retrespo 2010	Low	Low	Low	Low	Low	Low	None
Dienstknecht 2014	Unclear	Unclear	Low	Unclear	Unclear	Low	None
De Anta-Diaz 2016	Low	Low	Unclear	Low	Low	High	None
Inaba 2011	Unclear	High	Unclear	Unclear	Low	Low	None
Martin 2011	Unclear	Unclear	Low	Low	Unclear	Low	None
Muller 2012	Low	Unclear	Low	Low	Low	Low	None
Pospichill 2010	Low	Low	Low	High	Low	Low	None
Landgraeber 2013	Low	Low	Low	Low	Unclear	Low	Study sponsored by Stryker
Meneghini 2008	Low	Unclear	Low	Unclear	Low	Low	None
Witzleb 2009	Low	Low	Low	Low	Unclear	Low	None
Vicente 2014	Low	Low	Unclear	Unclear	Unclear	Low	None
Goosen 2011	Unclear	Unclear	Low	Low	Low	Low	None
Taunton 2018	Unclear	High	High	High	Low	Low	None
Reichert 2018	Low	Unclear	High	High	Low	Low	None
Brismar 2018	Low	Low	High	High	Low	Low	Sponsored by Stryker
D'Arrigo 2009	Unclear	Unclear	High	High	Unclear	Low	None

Table 2.

Methodological and Quality Assessment of Comparative Cohort Studies (N = 50)

Study (Name, Year)	Confounding	Study Participant Selection	Classification of Interventions	Deviation from interventions	Missing Data	Outcome Measurement	Reporting Bias
Rodriguez 2014	Low	Low	Low	Low	Low	Moderate	Low
Zawadasky 2014	Low	Low	Low	Low	Low	NI	Low
Bergin 2011	Moderate	Low	Low	Low	Low	Moderate	Low
Hananouchi 2009	Moderate	Low	Low	Low	Low	NI	Low
Sugano 2009	Low	Moderate	Low	Low	Low	NI	Low
Nakata 2009	Moderate	Low	Low	Low	Low	Moderate	Low
Spaans 2012	Moderate	Low	Low	Low	Low	NI	Low
Fransen 2016	Moderate	Low	Low	Low	Low	Low	Low
Tsukada 2015	Moderate	Low	Moderate	Moderate	Low	NI	Low
Hamilton 2015	Low	Low	Low	Moderate	Low	NI	Low
Tripuraneni 2016	Low	Serious	Moderate	Low	Low	NI	Low
Balasubramaniam 2016	Low	Low	Low	Low	Low	NI	Low
Leutch 2015	Low	Low	Low	Low	Low	NI	Low
L'Hommedieu 2016	Low	Low	Moderate	Low	Low	NI	Low
Malek 2016	Low	Low	Low	Low	Serious	Moderate	Low
Poehling-Monaghan 2015	Low	Low	Low	Low	Low	Moderate	Low
Panichkul 2016	Moderate	Low	Low	Moderate	Low	Low	Low
Sheth 2015	Low	Low	Low	Low	Low	Low	Low
Chomiak 2015	Moderate	Low	Low	Low	Low	NI	Low
Purcell 2018	Low	Low	Low	Low	Low	Low	Moderate
Zjilstra 2017	Low	Low	Low	Low	Low	Low	Low
Petis 2016	Low	Low	Low	Moderate	Low	NI	Low
Pogliacomi & De Filippo 2012	Low	Low	Low	Low	Unclear	NI	Low
Pogliacomi & Paraskevopoulos 2012	Moderate	Low	Low	Low	Low	NI	Low
Schliecher 2011	Low	Low	Low	Low	NI	Moderate	Low
Alecci 2011	Moderate	Low	Low	Moderate	Low	NI	Low
Ilchmann 2013	Serious	Serious	Moderate	Low	Low	NI	Low
Berend 2009	Low	Serious	Low	Low	Low	Low	Low
Reichert 2015	Moderate	Low	Low	Low	Moderate	NI	Low
Chen 2016	Low	Low	Low	Low	Low	Low	Low
Ilchmann 2016	Moderate	Serious	Low	Moderate	Low	Moderate	Low
Bernasek 2010	Moderate	Low	Low	Low	Low	Low	Low
Downing 2001	Moderate	Serious	Low	Low	Serious	Moderate	Low
Vicente 2008	Moderate	Serious	Moderate	Low	Serious	Moderate	Serious
Jameson 2014	Low	Low	Low	Low	Low	Moderate	Low
Arthursson 2007	Low	Low	Low	Low	Low	Low	Low
Sendtner 2011	Low	Low	Low	Low	Unclear	Moderate	Low
Amlie 2014	Moderate	Low	Low	Low	Low	Moderate	Low
Smith 2012	Moderate	Low	Low	Low	Moderate	Moderate	Low
Carlson 1987	Low	Low	Low	Low	NI	NI	Low
Ritter 2001	NI	Low	Low	Low	Low	NI	Low
Laffosse 2007	Moderate	Low	Low	Low	Unclear	Moderate	Low
Edmunds 2011	Serious	Moderate	Low	NI	Serious	NI	Serious
Abdel 2016	Moderate	Serious	Low	Moderate	Low	Low	Low
Takao 2016	Moderate	Low	Low	Low	Low	Moderate	Low
Jelsma 2016	Low	Low	Moderate	Low	Moderate	NI	Low
Wayne 2009	Low	Serious	Low	Low	Low	NI	Serious
Ponzio 2018	Low	Moderate	Low	Low	Low	Low	Low
Aggrawal 2019	Low	Low	Low	Low	Low	NI	Low

NI: Not indicated.

APPENDIX D. GRADE ASSESSMENT FOR INCLUDED STUDIES

Table 1.

Summary of Findings and Evidence Profile Table for Anterior versus Posterior Approach Studies.

Complication	Number of Studies	GRADE Assessment					Overall Quality (GRADE)
		Limitations	Inconsistency	Indirectness	Imprecision	Publication Bias
Dislocation*†	3 RCTs	Serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊕⊖⊖
	22 Cohorts	Serious limitations	No serious inconsistency	No serious indirectness	No serious imprecision	Strongly suspected	⊕⊖⊖⊖
Intraoperative Femoral Fracture	3 RCTs	No serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊕⊖⊖
	6 Cohorts	No serious limitations	Serious inconsistency	No serious indirectness	Very serious imprecision	Strongly suspected	⊕⊖⊖⊖
Intraoperative Acetabular Fracture	2 Cohorts	No serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊖⊖⊖
Intraoperative “Other” Fracture	3 Cohorts	Serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊖⊖⊖
Postoperative Femoral Fracture*	1 RCT	Serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊕⊖⊖
	1 Cohort	No serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊖⊖⊖
Postoperative “Other” Fracture*	3 Cohorts	No serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊖⊖⊖
Wound Complication†	4 RCT	Serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊕⊖⊖
	3 Cohorts	Serious limitations	Very serious inconsistency	No serious indirectness	Serious imprecision	Strongly suspected	⊕⊖⊖⊖
Infection*	3 Cohorts	No serious limitations	Serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊖⊖⊖
DVT†	1 RCT	Serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊕⊖⊖
	3 Cohorts	No serious limitations	No serious inconsistency	No serious indirectness	No serious imprecision	Undetected	⊕⊖⊖⊖
Loosening†	5 Cohorts	Serious limitations	No serious inconsistency	No serious indirectness	No serious imprecision	Strongly suspected	⊕⊖⊖⊖
*Early (within 90 days) †Unknown Time

Table 2.

Summary of meta-analyses on complications in Lateral versus Anterior Approach Studies.

Complication	Number of Studies	GRADE Assessment					Overall Quality (GRADE)
		Limitations	Inconsistency	Indirectness	Imprecision	Publication Bias
Dislocation*†	12 Cohorts	Serious limitations	Serious inconsistency	No serious indirectness	No serious imprecision	Undetected	⊕⊖⊖⊖
	3 RCTs	Serious limitations	No serious inconsistency	No serious indirectness	Serious imprecision	Undetected	⊕⊕⊕⊖
Intraoperative Femoral Fracture	2 RCTs	No serious limitations	No serious inconsistency	No serious indirectness	Serious imprecision	Undetected	⊕⊕⊕⊖
	3 Cohorts	Serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊖⊖⊖
Intraoperative Acetabular Fracture	1 Cohort	Serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊖⊖⊖
Intraoperative “Other” Fracture	1 RCT	No serious limitations	No serious inconsistency	No serious indirectness	Serious imprecision	Undetected	⊕⊕⊖⊖
	5 Cohorts	No serious limitations	Serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊖⊖⊖
Postoperative Femoral Fracture*	2 Cohort	Serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊖⊖⊖
Postoperative Acetabular Fracture*	1 Cohort	Serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊖⊖⊖
Postoperative “Other” Fracture*	1 Cohort	Serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊖⊖⊖
Infection*	3 RCT	No serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊕⊕⊖
	2 Cohorts	No serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊖⊖⊖
DVT†	2 RCT	No serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊕⊕⊖
	4 Cohorts	Serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Strongly suspected	⊕⊖⊖⊖
Loosening†	4 Cohorts	No serious limitations	Serious inconsistency	No serious indirectness	Very serious imprecision	Strongly suspected	⊕⊖⊖⊖
	2 RCT	No serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊕⊖⊖

*Early (within 90 days), †Unknown Time.

Table 3.

Summary of Findings and Evidence Profile for Complications in Lateral versus Anterolateral Approach Studies.

Complication	Number of Studies	GRADE Assessment					Overall GRADE
		Limitations	Inconsistency	Indirectness	Imprecision	Publication Bias
Dislocation*†	3 RCTs	Serious limitation	No serious inconsistency	No serious indirectness	Very serious imprecision	Strongly suspected	⊕⊕⊖⊖
	4 Cohort	Serious limitations	No serious inconsistency	No serious indirectness	Serious imprecision	Undetected	⊕⊖⊖⊖
Intraoperative Femoral Fracture	2 RCTs	No serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊕⊕⊖
	2 Cohort	No serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊕⊖⊖
Intraoperative “Other” Fracture	1 Cohort	No serious limitations	No serious inconsistency	No serious indirectness	Serious imprecision	Undetected	⊕⊖⊖⊖
Wound Complication†	2 RCTs	No serious limitations	Serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊕⊖⊖
Infection*	2 RCTs	No serious limitations	Serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊕⊖⊖
DVT†	1 RCT	Serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊕⊖⊖
Loosening	2 Cohorts	No serious limitations	No serious inconsistency	No serious indirectness	No serious imprecision	Undetected	⊕⊖⊖⊖

*early, †unknown time.

Table 4.

Summary of Findings and Evidence Profile for Complications in Lateral versus Posterior Approach Studies.

Complication	Number of Studies	GRADE Assessment					Overall GRADE
		Limitations	Inconsistency	Indirectness	Imprecision	Publication Bias
Dislocation*†	2 RCTs	No serious limitations	Serious inconsistency	No serious indirectness	No serious imprecision	Undetected	⊕⊕⊖⊖
	10 Cohorts	Serious limitations	Serious inconsistency	No serious indirectness	Very serious imprecision	Strongly suspected	⊕⊖⊖⊖
Intraoperative Femoral Fracture	2 RCTs	No serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊕⊕⊖
	2 Cohorts	Serious limitations	Serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊖⊖⊖
Postoperative “Other” Fracture*	1 Cohort	Serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊕⊖⊖
Infection*	1 RCT	No serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊕⊕⊖
DVT†	1 RCT	No serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊕⊕⊖
	2 Cohorts	No serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊖⊖⊖
Loosening†	4 Cohorts	Serious limitations	No serious inconsistency	No serious indirectness	No serious imprecision	Undetected	⊕⊖⊖⊖
*Early (within 90 days) †Unknown Time

Table 5.

Summary of Findings and Evidence Profile for Complications in Anterolateral versus Posterior Approach Studies.

Complication	Number of Studies	GRADE Assessment					Overall GRADE
		Limitations	Inconsistency	Indirectness	Imprecision	Publication Bias
Dislocation*†	10 Cohorts	Serious limitations	Serious inconsistency	No serious indirectness	No serious imprecision	Strongly suspected	⊕⊖⊖⊖
	1 RCT	No serious limitations	Serious inconsistency	No serious indirectness	No serious imprecision	Undetected	⊕⊕⊕⊖
Intraoperative Femoral Fracture	1 RCT	No serious limitations	Serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊕⊖⊖
	1 Cohort	Serious limitations	Serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊖⊖⊖
Wound Complication†	1 Cohort	No serious limitations	Serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊖⊖⊖
Infection*	2 RCTs	No serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Strongly suspected	⊕⊕⊕⊖
	1 Cohort	Serious limitation	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊖⊖⊖
DVT†	1 Cohort	No serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊖⊖⊖
Loosening†	1 RCT	No serious limitations	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	⊕⊕⊖⊖
	2 cohort	No serious limitations	No serious inconsistency	No serious indirectness	No serious imprecision	Undetected	⊕⊖⊖⊖

*Early (within 90 days), †Unknown Time.