Early or delayed anterior cruciate ligament reconstruction: Is one superior? A systematic review and meta-analysis

Since the previous systematic review and meta-analysis in 2010, several studies have been published comparing outcomes of early versus delayed ACL reconstruction [6, 26‐30]. The purpose of the present study was to reassess the available literature from the last review to the present day, to determine whether newer evidence is available to support any benefit of early (< 3 weeks) or delayed (≥ 3 weeks) ACL reconstruction in skeletally mature humans. The definitions of early and delayed surgery have been based on the previous systematic review [16] with an amendment in definition for the delayed surgical group to include sub-acute patients in the 3 to 6-week post-injury window; a period where many patients undergo surgery and whose outcomes also warrant inclusion.

Given that early surgery provides faster tibiofemoral stabilisation after injury, reduces joint laxity and the potential for subsequent degenerative changes [18, 19], the alternative hypothesis in our review is that in a population of young healthy adults, early ACL reconstruction is superior to delayed surgery. As per the previous review, the key outcome measures were meniscal/chondral lesions noted at time of reconstruction, PROMs, Lysholm, Tegner, IKDC and VAS scores, and objective clinical examinations of ROM and stability.

This systematic review was performed along preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines [31].

Eligibility criteria was specified as English language articles of randomised and non-randomised control trials and comparative cohort studies of ACL-deficient skeletally mature humans published between January 2009 and January 2018. A decision on the timeframe was made to ensure inclusion of all published research since the time of the previous systematic review in 2010. Studies were eligible for inclusion irrespective of open or arthroscopic reconstruction, ACL tear grade, graft type, rehabilitation protocol or gender. Skeletally immature and non-human models were excluded (Fig. 1).

A scoping search was conducted in early April 2017 with a subsequent decision to expand the retrieval. An electronic database search was conducted in Embase and Medline via the Ovid interface by a senior research librarian at the University of Oxford Cairns Library in May 2017 and updated in January 2018. The strategy consisted of MeSH terms and free text terms (“Appendix 1”). 496 Medline and 358 Embase articles were returned. The only limitation was the year of publication from 2009 to current as this review was subsequent to and built upon the previous systematic review and meta-analysis of Smith 2010 [16]. The following inclusion criteria were applied to all retrieved articles after the removal of duplications: (1) all studies were required to have patients undergoing isolated ACL reconstruction or in combination with a secondary procedure, e.g., meniscectomy; (2) a minimum of one early (surgery < 3 weeks) and one delayed (surgery ≥ 3 weeks) group per study; (3) a minimum of one reported subjective or objective outcome post-intervention; (4) a minimum period of 6-month follow-up. Of the 658 articles, 6 were selected for cross-referencing in PubMed and the Web of Science to identify related articles and explore citation histories. No new articles were identified. A further search on OpenGrey was conducted with no time limit ("Appendix 1"). Review paper reference lists were scrutinised for any further publications not identified in the electronic search and when necessary corresponding authors were contacted for clarification of published work. Editorials, comments, guidelines, case reports, review papers and articles on ACL repair were reviewed but excluded from the analysis. One investigator selected articles meeting inclusion criteria extracting data onto a standardised proforma, and was not blinded to journal names, author names or source.

The selected studies comprised a range of randomised control trials, non-randomised trials and observational studies with outcomes across varying time frames ("Appendix 2"). The Downs and Black checklist [32] has previously demonstrated reliability and validity in the assessment of a variety of study designs including observational studies and non-randomised intervention trials [33]. The checklist attributes significant weighting to study design, methodology and statistical power. A modified version of the original checklist was employed with a maximum score of 32. Assessment and scoring was performed by one investigator. The purpose of this was to provide a descriptive summary of sources of bias within the included studies with none excluded on the basis of this assessment.

Four meta-analyses were performed using RevMan V.5.3.5 (The Cochrane Collaboration, Copenhagen, Denmark): for Lysholm score and for Tegner activity scale at final follow-up, and for intra-operative incidence of meniscal and chondral lesions. Continuous variables were extracted and analysed as a mean ± SD. When SD was not reported, the corresponding author was contacted and asked to provide the value. In the case of no response, the study was excluded from the meta-analysis. The mean difference and 95% CI were calculated for continuous variables. Relative risk (RR) and 95% CI were calculated for dichotomous variables. Heterogeneity was measured with X² and I² statistical tests. Data were pooled using a random-effects model if statistical heterogeneity was ≥ 10% (I² test); a fixed-effects model was used if heterogeneity was below 10%. A probability of p < 0.05 was deemed statistically significant.

A statistically significant difference of + 0.39 points on the Tegner activity scale was found in favour of earlier surgery. In this calculation the two arms of the Herbst 2017 [6] study (isolated ACL reconstruction^ and combined ACL reconstruction plus meniscal repair*) were analysed separately to permit a greater sensitivity in analysis. No differences were found between the groups for the patient-reported Lysholm scores. There was no statistically significant difference between intra-operative findings of meniscal lesions. A trend towards significance was observed for the incidence of chondral lesions in the early surgery group (Table 2).

Four studies, all with similar designs, methodology and surgical technique [6, 26, 27, 29], assessed Tegner activity scale at the end of the follow-up period (26–36 months), reporting no differences between the groups on individual assessments. One further study [28] recorded Tegner activity scale at a much earlier final follow-up of 6 months with a trend approaching significance in favour of early 4.15 ± 1.45 versus delayed surgery 3.72 ± 1.34 (p = 0.06). Similarly, the four studies also assessed Lysholm scores at the same (26–36 months) follow-up with no differences found between the groups when assessed individually. One paper considered the International Knee Documentation Committee subjective rating score (IKDC) [35] with no difference between the groups, but with a trend towards significance in favour of early surgery when followed-up at 6 months (p = 0.08) [28]. Visual analogue pain scores (VAS) were reported in one study [6] with no differences found between the groups at any of the follow-ups up until the final reviews at 24 months.

There was significant heterogeneity in both the follow-up times and the reporting methods for the recovery in range of motion (ROM). No significant differences were found between the groups when considering the mean time taken (14 weeks) to recover full ROM [26] or when ROM between groups was assessed at 6 and 24 months [27‐29]. The only significant difference was found in one study at the midterm (12 months) follow-up in patients having combined ACL reconstruction and meniscal repair [6], where more patients in the delayed group had a lack of IKDC extension grade B (p = 0.04). Notably, however, this lack of extension was not present at either the 6 or 24-month follow-up in the same group, or at 6, 12 or 24 months in the parallel arm when patients underwent isolated ACL reconstruction.

Clinical examination by Lachman, pivot shift, and anterior draw testing showed no differences between the groups at any point up to the completion of follow-up [6, 26, 29]. On KT-1000 arthrometric evaluation of laxity at 24 months Li et al. [27] found evidence of greater stability with early surgery, but this was inconsistent with Raviraj et al. [26], who found no differences between the groups at a mean 32-month follow-up. Meniscal repair failures rates were assessed in 2 studies with divergent findings; in Karuppiah et al. [30] 23.0% failed after delayed and 4.8% following early surgery (p = 0.048), whilst Herbst et al. [6] found no differences in failure rates between the two groups. Objective IKDC [35] was not different between groups in two studies at final 24-month follow-up [6, 27].

Appraisal findings using the Downs and Black checklist [32] is graphically represented in Fig. 2. Two studies attempted randomisation; one via a computer-generated sequence [26], the second on basis of odd/even hospital numbers [28]. Whilst two studies attempted uniformity of follow-up using a single observer [26, 27], only one study was blinded to the timing of surgery [26]. Power calculations were performed in two studies, one a priori [26] and one post hoc [6]. Given the relatively low number of studies in any of the analyses, an assessment of publication bias via a funnel plot was considered to yield little value and was thus not performed.

When assessed individually, there were no differences found in Tegner activity scales between the groups in any of the five papers. Mananadhar et al. [28] observed a trend towards significance in favour of early surgery at only the 6-month follow-up, whilst all the other similar studies reported no differences in outcomes from 12 to 24 months [26, 27, 29, 36]. Our meta-analysis found a 0.39 point greater Tegner activity scale score in the early surgical group (p = 0.008) compared with the delayed group, suggesting a potentially beneficial role for early surgery in the young active cohort under review. This finding contrasts with Smith’s 2010 meta-analysis [16] where no difference was recorded between the groups; however, the cohorts in that analysis had a maximum of 70 patients and none of the studies were adequately powered. With no objective value for what constitutes a minimum clinically important difference (MCID) in Tegner activity scale [37, 38], the observed effect size of 0.39 must be taken in context. In a review of measures of knee function, Collins 2011 [37] reported that post-ACL reconstruction score differences of 0.74 (were classed as moderate) and 1.0 (as large differences) at 6 and 12 months, respectively; thus, the observed effect of 0.39 is likely to equate to a negligible or small clinical difference and should not be the basis for any change in clinical practice.

Four papers reported Lysholm scores and found no differences between the groups when assessed separately or in the meta-analysis. As a validated and reliable instrument [39] and the most widely used subjective assessment of knee function worldwide [40], the score is a key tool in the evaluation of outcomes within and between studies. With only short-term (2-year) follow-up available, a cautious interpretation is mandated, and medium (5-year) and long-term (10-year) [41] data might help to better inform decisions on the timing of ACL reconstructions.

In papers reviewing objective [6, 27], and subjective [28] IDKC, no differences were found between the groups. In a systematic review, Wera 2014 [40] evaluated the use of both IKDC forms and recommended its interpretation in combination with the Tegner activity scale [40]. In our study, the IKDC findings reflected the Tegner scores, and no differences were found between the groups. One author assessing pain scores with VAS found no differences [6]. The VAS is a unidimensional measure with excellent test–retest reliability for chronic painful musculoskeletal conditions [42], but may be less valid in ligamentous knee injuries [43, 44].

Five papers assessed post-operative ROM, but due to considerable heterogeneity in measurement methods, a meta-analysis was not indicated and a narrative review was preferred. In the 1990s, rehabilitation protocols typically focused on restricting ROM and surgical intervention utilised non-anatomical ACL reconstruction—factors which may account for reduced ROM noted in studies during this period [46, 47]. Through Shelbourne’s 1990 findings that patients who were non-compliant with their rehabilitation programme and who ambulated on their reconstructed knee without splints had fewer ROM problems with better strength gains [45], various accelerated rehabilitation programmes were developed and became the norm after support in multiple subsequent systematic reviews [46‐49]. The five papers in this study measuring post-operative ROM utilised an accelerated rehabilitation protocol which may explain why no differences were found between the groups at final follow-up in any study (ranging 14 weeks to 24 months); this is in keeping with findings from similar studies with accelerated protocols [36, 50‐53].

Two studies objectively assessed stability scores on a KT-1000 arthrometer with contrasting results. In the smaller study [27] which failed to score for external validity, selection bias or power on the checklist, the authors reported significantly greater laxity in the delayed surgery group. However, the authors of the paper with the highest methodological score in this review conducted randomisation, assessor blinding, and an adequate power calculation, and actually reported no differences between the groups (p = 0.9) [26]. This is in keeping with the wider literature where no difference was evident in a meta-analysis [16], suggesting the findings of Li et al. [27] should be interpreted in context of methodological compromise.

Proponents of early reconstruction cite the odds of a knee cartilage lesion increasing by almost 1% per month in ACL-deficient knees between time of injury and surgery [19], as a driver to early surgery; cartilage lesions and untreated meniscal tears are recognised predictors for OA [20, 21, 54]. When meta-analysing the incidence of chondral lesions, there was a borderline significant (p = 0.06) effect favouring early surgery, whilst there was no apparent effect on the incidence of meniscal lesions.

Our definition of delayed surgery might be considered ‘early’ by some [24, 55‐60]. The longest delay from injury to surgery in all studies in this review was 74 weeks [29], and subsequently, the early versus delayed heterogeneity may be too small to permit any significant findings. This is reflected in the literature where four authors with similar definitions of early (< 3 weeks) and delayed (≥ 3 weeks) interventions failed to find any differences between groups [26, 36, 50, 53]. These results are contrasted in papers utilising a later definition of ‘early’ to include periods up to 5 months post-injury, and ‘delayed’ ranging from 3 months to 14 years. Across studies utilising these later/longer timeframes all five authors reported a significantly higher incidence of meniscal/chondral defects in the delayed groups. Thus, whilst no statistically significant findings are present between the groups in our review, on the balance of current evidence, a difference might have existed if surgery had been delayed by months or years.

Although lesions noted in this study may remain clinically silent for decades in many patients, their presence is a cause for concern given the increased risk of OA with its associated physical, psychological and financial impact [61‐63]. In a methodologically robust study, Frobell et al. [24] found no differences between early ACL reconstruction and non-surgical treatment with the option for a later reconstruction, initially indicating that a ‘watch and wait’ approach might be permissible. However, there were markedly increasing proportions of reconstructions in the delayed groups of 7% at 6 months, 20% at 12 months and 51% at midterm (5-year) follow-up, with more frequent meniscal signs, symptoms and knee instability in the ‘optional delayed’ reconstruction group (p < 0.001) [25]. Thus, whilst there were no differences in observed meniscal/chondral lesions between early or delayed surgery groups seen in our review, on the basis of the wider current literature, early intervention would appear to confer a decreased risk of developing later meniscal/chondral lesions and potentially might then lower the subsequent risk of OA.

Assessment of meniscal repair failures rates yielded contrasting findings in two papers. Karuppiah et al. [30] retrospectively analysed the notes of patients undergoing combined ACL reconstruction plus meniscal repair and noted a higher failure rate in delayed versus early repairs (p < 0.05). Herbst et al. [6] found no difference between early (mean < 1 day) and delayed (mean 49.2 days) surgery. The size and locations of the tears was only reported in the retrospective analysis, thus a direct comparison is difficult. One should bear in mind that the discrepancy in tear rates is likely to relate to the small study sizes (of approximately 100 patients), which were not adequately powered to detect a difference. Additionally, there is no single accepted definition of what constitutes a failure in meniscal repair, how one identifies when those patients represent and how one actually makes a correct diagnosis.  In addition, a maximum of 2-years follow-up is probably too short a time to make a true judgement. Given that the inclusion/exclusion criteria, patient demographics, surgical technique and instrumentation are actually similar, and with the recognised increased risks of OA with meniscal lesions, an urgent well-designed randomised control trial is required to determine whether timing of intervention actually affects the rates of meniscal repair failure.

There are substantial limitations to this paper. A solely electronic search of three databases (Medline, Embase, OpenGrey) was performed. The reviewing author was not blinded to the journal titles, the authors of the studies, or sources. Arbitrary time frames of early and delayed reconstructions were applied based on previous meta-analysis, with a minor modification. The degree of ACL tear and presence or absence of meniscal lesions were inconsistently reported across the studies with no controls applied. The follow-up periods were limited to a maximum of 2 years.