Introduction

Total hip arthroplasty (THA) continues to be one of the most effective and commonly performed orthopaedic procedures, with over 500,000 performed each year in North America and a projected rise to almost 1.5 million by 2040 [1,2,3]. With the growing demand for minimally invasive surgical methods, there has been a recent resurgence of the direct anterior approach (DAA) to the hip in the context of THA [4]. Though originally described by Hueter, the DAA is more commonly known by the report published by Smith-Petersen in 1917 [5, 6]. This approach uses a true intermuscular, internervous plane to approach the hip, which, along with the size of the skin incision, contributes to its reputation as a minimally invasive technique.

Due to its muscle-sparing nature, proponents of the DAA believe it leads to less post-operative pain and faster recovery times [7,8,9,10,11,12]; however, it is thought that the functional advantages offered by the DAA are equivalent to other approaches by as early as two weeks post-operatively [7]. Additionally, it is suggested that the DAA leads to decreased post-operative dislocation rates, as the posterior capsule and soft tissues are preserved with this approach [13,14,15]. Lastly, with shorter average hospital stays due to enhanced early recovery, the use of the DAA has opened the door for outpatient THA [16,17,18]. Despite its many potential advantages, the use of the DAA is technically challenging and overall complication rates may be higher in DAA THA [19, 20], especially among surgeons new to the approach [21, 22]. This, along with the large number of surgeons transitioning to the DAA, has led to discussion regarding the procedure’s learning curve.

The concept of a learning curve for surgical procedures is not a new one. The relationship between operative procedures performed by a surgeon and lower mortality rates was reported in 1979 [23]. This concept of a learning curve in a surgical context was further described as having four stages: (1) a rapid ascent during the early stages of training; (2) a zone of decreasing improvement, where additional experience yields only marginal improvement; (3) a plateau, in which further experience has no effect on the measured outcome; and (4) possible age-related decline in the measured outcome [24]. Due to its increasing popularity and technical challenges, the learning curve of the DAA THA has garnered considerable attention [25,26,27,28,29]. Identifying this learning curve has considerable implications for patient safety, surgical training, and cost-effectiveness as it relates to operative time. The purpose of this systematic review is to identify the reported learning curves associated with the DAA to THA, primarily by analysis of operative time and complications. Additionally, the purpose was to determine, based on the best available evidence, a point on the learning curve after which the surgeon can be considered proficient.

Methods

This systematic review was conducted in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines for conducting and reporting systematic reviews [30]. The study protocol was registered prospectively on The International Prospective Register of Systematic Reviews (PROSPERO) (ID: CRD42020195680).

Search strategy

Three online databases (MEDLINE, Embase, and Web of Science) were searched from database inception to June 25, 2020, for literature addressing the learning curve associated with the use of the DAA for THA. Search terms used to identify eligible studies included “direct anterior approach,” “total hip arthroplasty,” “learning curve,” “clinical competence,” “outcome assessment,” and “complication” (Appendix).

Study screening

Studies identified during the comprehensive literature search were screened at the title/abstract as well as full-text stages by two reviewers independently and in duplicate using the online software Rayyan (2010, Qatar Computing Research Institute, Doha, Qatar). Any discrepancies at the title/abstract stage were resolved with automatic inclusion into the next stage of screening for more in-depth review. At the full-text stage, discrepancies were discussed and resolved by consensus between the reviewers, and a more senior author was consulted for any remaining discrepancies. In addition, the references of relevant studies were screened manually to identify any eligible studies potentially missed by the database search.

Assessment of study eligibility

The research question and study eligibility criteria were established a priori. The inclusion criteria were as follows: (1) all levels of evidence, (2) studies performed on human patients, (3) operative studies using the DAA for THA, and (4) formal discussion or analysis of the learning curve based on the results of the study. Exclusion criteria were (1) review articles, opinion pieces, editorials, or basic science studies, and (2) multiple studies reporting on the same group of patients (only the most recent study is to be included).

Data abstraction

Three reviewers independently extracted data from included studies into a Google Sheet (Google, CA, USA) online collaborative spreadsheet, designed a priori, and piloted prior to its use. Collected data included study characteristics, patient demographics, data on the learning curve, and complications both intra-operative and post-operative. Learning curve data was extracted from figures using WebPlotDigitizer (Version 4.3, Pacifica, CA, USA) to estimate values for individual data points.

Quality assessment

The methodological quality of non-randomized studies was evaluated using the Methodological Index for Non-Randomized Studies (MINORS) criteria [31]. Using the MINORS checklist, non-comparative studies can achieve a maximum score of 16, while comparative studies can achieve a maximum score of 24. No literature currently exists for categorizing MINORS scores; however, we categorized the quality of evidence a priori based on a previous systematic review by our group: <5 indicated very low quality evidence, 6–9 low quality, 10–14 fair quality, and >14 good quality [32].

Statistical analysis

Inter-class correlation (ICC) was calculated to determine agreement between reviewers on MINORS assessments and studies were assigned the mean score in cases when there was disagreement between reviewers. Descriptive statistics including means, standard deviations, and ranges are presented where applicable. Due to the heterogeneity of existing literature and inconsistency in reported outcomes, meta-analysis was unable to be performed.

Results

Study characteristics

The initial search yielded 2083 studies, 21 of which met the inclusion criteria for this review (Fig. 1) [21, 25, 27,28,29, 33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48]. These studies, which were published between 2008 and 2020, included five retrospective cohort studies (level 3 evidence) and 13 case series (level 4 evidence) (Table 1). The included studies were completed in North America (10 studies), Europe (8 studies), Asia (2 studies), and Australia (1 study). These studies analyzed the learning curves of 63 surgeons. Main methodology and findings of included studies are described in Table 2.

Fig. 1
figure 1

PRISMA flowchart for the systematic search strategy utilized

Full size image
Table 1 Characteristics of included studies
Full size table
Table 2 Summary of included studies main methodology and findings
Full size table

Study quality

There was excellent overall agreement between assessors for quality assessment scores using the MINORS criteria (ICC=0.945, 95% CI: 0.868–0.977). The included studies had an average MINORS score of 14 ± 3 which indicates a fair quality of evidence of included studies.

Patient demographics

There were 9908 patients included, 60% of which were women. Included patients had an average age of 63.5 years (range: 13–94) and body mass index of 27.1 kg/m2 (range: 16.79–58.94). The average length of follow-up was 19 months (range: 1.5–100). Most studies used exclusively cementless stems, but four studies also used cemented stems for some cases (range: 1–5% of cases) [29, 33, 39, 40]. Twelve studies used cementless acetabular components [21, 25, 27, 28, 33, 34, 37,38,39, 41, 43, 48], two used cemented [29, 40], and seven studies did not report the type of acetabular component used [25, 35, 36, 42, 44, 46, 47]. Seven studies used intra-operative fluoroscopy to assist with component positioning [25, 36, 38, 39, 44,45,46]. Osteoarthritis was the most common indication for THA in most studies, with the exception of two in which femoral neck fractures [37] and osteonecrosis [38] were the most common indications for THA.

Learning curve groups and surgeon experience

All studies investigated the learning curve by grouping consecutive cases and comparing outcomes between the groups. Each study chose slightly different grouping variations, but most frequently studies compared the first number cases to subsequent cases or divided their cases into two equal groups. Groups ranged from 15 cases to 100 cases. Pooled analysis of early and late groups was performed when appropriate. Surgeon experience also varied between studies. Most studies investigated the learning curve in experienced surgeons who had performed hundreds of THAs via alternative approaches. One study compared the learning curve of a senior surgeon to junior surgeons, showing that the learning curve of junior surgeons joined that of the senior surgeon after ten cases [25].

Operative time learning curve

Only six studies reported individual case data to depict a true learning curve, with five reporting on operative time [34, 39, 40, 42, 43] and individual studies reporting on blood loss [43], setup time [40], and component placement [48]. Although five studies reported a learning curve for operative time, each study reported these results slightly differently. Individual studies reported on the first 30 cases [39], the first 50 cases [43], the first 100 cases [34], the first 210 cases [40], and the first 1000 patients [42]. Mean operative time for the first case was 156.59 ± 41.71 min but decreased to 93.18 ± 14.68 minutes by case 30 and 80.45 ± 12.28 min by case 100 (Fig. 2). Although mean operative time did not reach a true plateau, it remained less than 90 minutes for all cases after the 100th case.

Fig. 2
figure 2

Learning curve of DAA: operative time by case

Full size image

All studies that evaluated operative time across the learning curve revealed a downward trend as surgeons became more experienced with the procedure. Most studies reported a statistically significant reduction in operative time in the late group compared to the early group [27, 29, 36,37,38,39,40, 45,46,47], though two did not reach significance [21, 42]. Mean operative time for all early and late groups was pooled (see Table 3). The grouping of early DAAs had a mean operating time of 109.5 ± 20.7 min (range: 64.7–135.9), whereas it was 82.6 ± 17.3 min (range: 47.4–113.9) in the late group.

Table 3 Summary of pooled outcome data
Full size table

Intra-operative outcomes

Seven studies reported on estimated blood loss across the learning curve [21, 27, 36, 37, 39, 41, 47][21, 27, 36, 37, 39, 40, 47]. Three of these studies demonstrated significant reductions in blood loss from early to late groups [27, 36, 37][27, 36, 37], three showed trends toward blood loss reduction that were not statistically significant [21, 39, 47], and one did not change [40]. When early and late groups were pooled, estimated blood loss decreased from 642.5 ± 219.0 mL (range: 400–1071) to 468.6 ± 115.9 mL (range: 347.5–643). Three studies reported on fluoroscopic time during the procedure [36, 38, 39]. Two studies showed a significant reduction in fluoroscopic time between early and late groups [38, 39] and one showed only a downward trend [36]. Mean fluoroscopic time for the early groups was 19.5 ± 13.7 s (range: 4.8–32.1) and decreased to 11.5 ± 7.5 s (range: 2.86–17.0).

Post-operative outcomes

Five studies reported on duration of hospitalization following DAA [21, 27, 38, 47, 48]. Experience performing DAA did not have a significant impact on length of stay in any of the studies. The mean length of stay for the early group of cases was 4.4 ± 1.7 days (range: 2.5–6.6) compared to 4.0 ± 1.3 days (range: 2.5–5.4) for the late group. Three studies reported on acetabular component placement between early and late procedures, though there was no substantial differential between groups [36, 39, 40]. Mean acetabular inclination of early DAAs was 46.8 ± 4.3° (range: 41.9–50) and 47.5 ± 2.9° (range: 44.4–50.1) for late DAAs. Mean acetabular anteversion was 14.3 ± 1.8° (range: 13–15.6) for early DAAs and 12.9 ± 1.6° (range: 11.7–14) for late DAAs. Six studies reported on leg length discrepancy (LLD) [33, 36, 38,39,40, 44]. The early groups of DAAs had a mean LLD of 3.2 ± 1.4 mm (range: 2–5.04) and late groups of DAAs had a mean LLD of 2.0 ± 1.2 mm (range: 1.07–3.74). Two studies only reported on the number of unacceptable LLDs. One showed a decrease in unacceptable LLDs in the late group from seven to two cases [33] while the other showed no difference between groups [44].

Functional outcomes

Two studies evaluated subjective pain using a visual analogue scale from 0–100 post-operatively, revealing a downward trend of post-operative pain with surgeon experience performing the DAA [29, 48]. Mean pain VAS scores decreased from 21.3 ± 0.5 (range: 20.9–21.6) in early DAAs to 14.2 ± 0.2 (range: 14.0–14.3) in late DAAs. Two studies also compared post-operative Harris Hip Scores (HHS) between early and late groups [36, 38]. Mean HHS did not change significantly, increasing slightly from 89.4 ± 8.8 (range: 83.2–95.7) in early groups to 90.3 ± 8.5 (range: 84.2–96.3) in late groups.

Complications

Eight studies reported increased rates of complications in early groups of DAA procedures compared to late groups [25, 27, 33, 35, 38, 40, 44, 48], while one study did not reveal any difference between groups [46]. Mean complication rate in early groups was 20.8 ± 12.7% (range: 7–44) whereas in late groups it decreased to 7.6 ± 7.1% (range: 0–20) (Fig. 3). Seven studies compared surgical revision rates between early and late procedures [11, 28, 29, 33, 35, 44, 48]. Across early groups, the mean revision rate was 7.1 ± 5.0% (range: 2–15) compared to 1.1 ± 0.9% (range: 0–3) for late groups. Revision rate decreased from early to late groups across all studies. Lateral femoral cutaneous nerve injury was more frequent in early procedures than late, decreasing from 15.0 ± 11.0% (range: 4–25) to 5.0 ± 4.8% (range: 2–12) in late groups [27, 33, 37, 47]. The most common complications during the DAA learning curve were fracture, dislocation, component malposition, and infection [25, 27, 33, 34, 38, 47, 48]. Five-year implant survival was evaluated by three studies, showing a rising trend in implant survival with DAA experience from 85.9 ± 8.7% (range: 78.9–95.6) in early groups to 96.4 ± 0.6% (range: 95.7–96.8) in late groups [34, 35, 41].

Fig. 3
figure 3

Comparison of mean complication and revision rates for early and late DAA groupings. Error bars represent standard deviation

Full size image

Discussion

This systematic review revealed a steep learning curve for the DAA to THA over the first 30 cases and a relative plateau after approximately 100 cases. Mean operative time decreased by more than 50 minutes over the first 30 cases, showing a significant improvement in surgeon skill and comfort with the procedure. This is consistent with previous literature for other orthopaedic procedures, which have often defined the learning curve as the first 30 cases [32, 49, 50]. Although mean operative time was plotted for more than 200 cases, the learning curve never reached a true plateau or inflection point. This suggests that even after performing hundreds of DAA procedures, surgical technique may continue to improve. It should be noted, however, that the learning curve determined by this systematic review only involved the results of two surgeons past 50 cases.

The average operative time for the late groups in this review was 82.6 ± 17.3 minutes. A recent systematic review including 630 675 THA procedures showed that the average operative time for THAs of all approaches was between 90–99 minutes [51]. This review suggests that surgeons who are new to the DAA can reach the average operative time of more traditional THA approaches after approximately 50 cases.

While operative time has been used to evaluate learning curve in many procedures, there is some debate as to the utility of the measure [52, 53]. It is important to consider that there are many factors which may influence operative time, aside from surgeon proficiency. For example, as surgeons become more comfortable performing a procedure, they may take on more technically complex cases which take longer to complete [49]. While operative time is an important outcome with respect to cost-effectiveness of a procedure, small reductions in operative time do not directly lead to patient benefit [52, 54]. Recent literature has shown that each 20 minutes increase in operative time increases the rate of periprosthetic joint infection following total joint arthroplasty by 25% [55]. This suggests that the large reductions in operative time seen early in the learning curve may substantially impact patient outcomes and the small reductions seen past the first 50 cases are less significant. Increased operative time in surgeons who are learning a procedure is likely to represent lack of comfort from the entire surgical team and attention to detail, and thus does not necessarily correlate to patient-centered outcomes such as complication rates [32]. Learning curves may be better evaluated using multiple outcomes rather than operative time alone.

Another key finding from this review was that complication and revision rates showed considerable reductions from early to late groups. Although the specific number of cases needed to significantly reduce complication rates was unable to be determined, the decline in complications and revisions between early and late groups suggests that patient-centered outcomes improve with surgical experience. The average rate of complications in early groups was 20.8 ± 12.7% but ranged from 7 to 44% between individual studies. These complications include fracture, infection, and dislocation, which can require prolonged hospitalization or revision surgery. This trend is also seen in revision rates, five year implant survival, and leg length discrepancy, with improved outcomes as surgeons became more experienced with the procedure. Average rate of revision in the late DAA groups was 1.1 ± 0.9%, which is quite comparable to the rates demonstrated in the literature in alternative approaches beyond the learning curve [19]. The reductions in leg length discrepancy from 3.2 ± 1.4 mm (range: 2–5.04) to 2.0 ± 1.2 mm (range: 1.07–3.74) are especially important, as this is a common reason for litigation and patient dissatisfaction following THA [56, 57]. These increased risks during the learning curve are important for surgeons to discuss with patients when receiving informed consent and consideration should be made to select technically favourable cases as surgeons gain proficiency with the procedure. Female patients with low BMI have been considered technically favourable cases to maximize patient safety during initial procedures [35].

Proper component positioning in THA is important for minimizing the risk of component wear, instability, and impingement, leading to dislocation and revision [58, 59]. Acetabular anteversion is a key aspect of component positioning, with literature showing that ideal anteversion is 15° [60]. Excessive acetabular anteversion is a common concern with the DAA due to the limited view of the anterior acetabular wall [61, 62]. This review showed that mean acetabular anteversion was 14.3 ± 1.8° (range: 13–15.6) for early DAAs and 12.9 ± 1.6° (range: 11.7–14) for late DAAs, suggesting that early DAAs are not at risk of excessive anteversion and that with experience anteversion tends to decrease. The use of intra-operative fluoroscopy in seven of the included studies may help explain the precise component positioning seen in this review.

This systematic review was strengthened by its rigorous methodology. This includes a comprehensive search strategy that involved three major databases and criteria designed to be as inclusive as possible. Reviewer bias was minimized by independently completing each stage of the process in duplicate and automatically including any conflicts. This allowed for the inclusion of 21 studies and 63 surgeons.

The findings of this review are limited by the overall low quality of evidence of included studies. Despite the broad search strategy utilized, the highest level of evidence was level 3 and most included studies were case series. The lack of consecutive case data presented in the included studies was another challenge for this review. Most included studies divided patients into distinct groups of patients with significant heterogeneity between groups. This heterogeneity made it difficult to draw conclusions about the slope of the learning curve beyond early versus late procedures. This limited our ability to analyze the early learning curve for outcomes other than operative time and prevented us from identifying the point at which surgeons reach proficiency.

Future studies should continue to investigate the learning curve beyond the first 50 cases to better characterize the point at which a plateau is reached, suggesting that mastery of the procedure has been achieved. Furthermore, including patient-reported outcomes in these learning curve studies, such as pain and functional scores, would clarify the impact of the DAA learning curve on patient benefit. By reporting continuous case data, future studies could improve the understanding of the learning curve and allow for the integration of risk mitigation strategies as surgeons transition to the procedure.

The DAA is a minimally invasive approach to THA that optimizes post-operative outcomes but is technically complex and has a significant learning curve. This learning curve means that there are increased risks for patients undergoing the procedure by surgeons who are new to the procedure. Operative time reached a relative plateau after approximately 100 cases, suggesting that it takes 100 cases for surgeons to develop proficiency in the DAA to THA.