Outcome of bone-patellar tendon-bone vs hamstring tendon autograft for anterior cruciate ligament reconstruction: A meta-analysis of randomized controlled trials with a 5-year minimum follow-up.

BACKGROUND: Many systematic reviews have compared the short-term outcomes of anterior cruciate ligment (ACL)reconstruction with hamstring and patellar tendon autograft,but few differences have been observed. The purpose of this meta-analysis was to compare the medium-term outcome of bone-patellar tendon-bone and hamstring tendon autograft for anterior cruciate ligament reconstruction in terms of clinical function, knee stability, postoperativecomplications, and osteoarthritis changes.
METHODS: This meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. The PubMed, Embase, and the Cochrane Library databases were searched from inception to November 2, 2019. This meta-analysis included only randomized controlled trials that compared BPTB and HT autografts for ACL reconstruction with a 5-year minimum follow-up. The Cochrane Collaboration's risk-of-bias tool was used to estimate the risk-of-bias for all included studies. RevMan 5.3 software was used to performed statistical analysis of the outcomes.
RESULTS: Fifteen RCTs, involving 1298 patients (610 patients in the BPTB group and 688 patients in the HT group) were included. In terms of clinical function, no significant difference was found in the objective International Knee Documentation Committee score (OR = 0.94, 95%CI: 0.64-1.37, P = .75), Lysholm knee score (MD = -2.26, 95%CI: -4.56 to 0.05, P = .06), return to preinjury activity level (OR = 1.01, 95%CI: 0.67-1.52, P = .96), and Tegner activity level (OR = 0.03, 95%CI: -0.36 to 0.41, P = .89). There was no statistically significant difference in the Lachman test (OR = 0.86, 95%CI: 0.5-1.32, P = .50), pivot-shift test (OR = 0.68, 95%CI: 0.44-1.06, P = .09), and side-to-side difference (MD = -0.32, 95%CI: -0.81 to 0.16, P = .19). As for postoperative complications and OA changes, there were no statistically significant difference in flexion loss (OR = 1.09, 95%CI: 0.47-2.54, P = .85) and OA changes (OR = 0.76, 95%CI: 0.52-1.10, P = .15), but we found significant differences in favor of the HT group in the domains of kneeling pain (OR = 1.67, 95%CI: 1.04-2.69, P = .03), anterior knee pain (OR = 2.90, 95%CI: 1.46-5.77, P = .002), and extension loss (OR = 1.75, 95%CI: 1.12-2.75, P = .01). There was a significant difference in favor of the BPTB group in the domain of graft failure (OR = 0.59, 95%CI: 0.38-0.91, P = .02).
CONCLUSIONS: Based on the results above, HT autograft is comparable with the BPTB autograft in terms of clinical function, postoperative knee stability, and OA changes, with a medium-term follow-up. The HT autograft for ACL reconstruction carries a lower risk of complications, such as anterior knee pain, kneeling pain, and extension loss, but an increased incidence of graft failure. Patients should be informed of the differences when deciding on graft choice with their physician.

Introduction

Anterior cruciate ligament (ACL) injury is one of the most common orthopedic injuries, with an annual incidence of isolated injury of 68.6 per 100,000 person-years.[ A long-term follow-up study showed that the risk of secondary meniscus tear (hazard ratio 18.0), osteoarthritis (OA, hazard ratio 14.2), and the need for total knee arthroplasty (hazard ratio 5.0) in patients without ligament reconstruction after ACL tear was significantly higher than in those without ACL tear.[ Although there was no consensus on whether to choose conservative treatment or ACLR after ACL injury in many studies,[ numerous young and active individuals still choose ACLR. Typically, bone–patellar tendon–bone (BPTB) and hamstring tendon (HT) autografts are chosen for ACLR.

Historically, the BPTB autograft was the gold standard for ACLR, as it allowed proper bone-to-bone tunnel healing, involved a short fixation distance, and provided excellent biomechanical strength.[ Some studies have shown that BPTB autograft had the advantage of appropriate size, high durability, a low level of laxity, a higher incidence of return to sport activity, and a low rate of revision.[ However, many patients have complained about donor site diseases, such as anterior knee pain, kneeling pain, and extension loss.[ In recent years, HT autografts have been increasingly more frequently used in ACLR[ because of its lower short-term donor site morbidity.[ However, compared to BPTB autografts, HT autografts could lead to delayed graft incorporation, a decrease in knee stability, and significant flexor torque deficiency.[

Numerous RCTs have been conducted to analyze the relative merits of the two mentioned autografts, and a large amount of data published on this topic have been used for several meta-analyses.[ However, most previously published meta-analyses did not investigate the use of BPTB and HT autografts for ACLR after a medium-term follow-up. Xie et al's meta-analysis included 12 RCTs and two prospective cohort studies that used a minimum of 5 years’ follow-up, and finally concluded that ACL reconstruction with BPTB autografts resulted in increased anterior knee pain, kneeling pain, and incidence of OA, as compared with HT autograft.[ Poehling-Monaghan et al[ also reported a systematic review comparing BPTB and HT autografts for ACLR at a minimum of 5 years’ follow-up, but two of the included studies were cohort studies, and finally concluded that the clinical outcomes scores were lower, and greater motion loss and higher rates of OA were seen in the BPTB group. The above meta-analysis included prospective cohort studies, which markedly increased the risk of heterogeneity arising from variations in study design. In addition, some studies observed that the prevalence of OA was higher in the BPTB group during the long-term follow-up, while others found that there was no difference in the prevalence of OA between the BPTB group and the HT group.[ Therefore, this issue remains controversial. In recent years, several RCTs[ have compared BPTB autografts and HT autografts for ACLR after a medium-term follow-up. Therefore, there was a need for a new meta-analysis that included more recent and high-quality RCT studies to provide the latest evidence for clinical decision-making.

Therefore, this meta-analysis aimed to determine whether there are significant differences between BPTB autografts and HT autografts for ACLR in terms of the clinical function, knee stability, postoperative complications, and OA changes, by comparing RCTs with a minimum follow-up of 5 years.

Methods

All analyses were based on previous published studies, thus no ethical approval and patient consent are required. This study is based on the Cochrane review methods.

Search strategy

This meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.[ PubMed, Embase, and Cochrane Library databases were the three major electronic databases used for the literature search from inception to November 2, 2019. Before conducting the search, a systematic review registration was completed on April 28, 2020 using the PROSPERO International prospective register of systematic reviews (registration number CRD42020170821). “anterior cruciate ligament,” “bone patellar tendon,” “hamstring,” were the key words used in different combinations, and the detailed search strategy is (ACL OR anterior cruciate ligament) AND (patella OR patellar OR bone patellar tendon OR bone patella tendon) AND (hamstring OR semitendinosus OR gracilis patellar tendon). Additional studies were identified by reviewing the reference lists in the pertinent articles. Bibliographies of all the relevant retrieved studies were scrutinized to identify additional studies. Two authors (Deng and Lu) independently implemented the literature search, study selection and data extraction, and the reasons for exclusion were recorded and discordance was resolved by discussion or a decision from the corresponding author (Zhao).

Selection criteria

The included studies were randomized controlled trials (RCTs) and that directly comparing BTPB graft and HT graft for primary ACLR. Studies should be required to be published to be English and to have a minimum of 5-year follow-up. Studies with a Level of Evidence I and II (based on the Oxford Centre for Evidence-Based Medicine, used by the American version of the Arthroscopy and Journal of Bone and Joint Surgery[). Studies that reported data on

functional outcome—objective International Knee Documentation Committee score, Lysholm knee score, return to preinjury activity level, Tegner activity level;

stability parameters—Lachman test, pivot test, side-to-side difference;

postoperative complications—anterior knee pain, kneeling pain, failure of graft, loss of extension and flexion; and

radiograph evidence of osteoarthritis were considered to be eligible.

The number of strands of the autografts and the fixation in the proximal and distal part were not limited. Excluded were non-English articles, allograft, in vitro, animal, or cadaveric studies, and systematic reviews and meta-analyses. When multiple studies existed utilizing the same patient population but reporting outcomes at different time points, the study with the longest follow-up was included in our review while the rest were excluded.

Data extraction

The following information was extracted from the enrolled articles: the first author, date of publication, level of evidence, number of patients, number of depletions, mean age, sex, duration of follow-up, strands of the HT, type of graft fixation. The primary outcome was functional outcomes, such as an objective International Knee Documentation Committee (IKDC) score, return to preinjury activity level, Lysholm knee score. Other indicators include stability parameters, such as side-to-side difference, pivot-shift test, Lachman test; postoperative complications, such as anterior knee pain, kneeling pain, extension loss, flexion loss, and graft failure (defined as graft rupture or revision ACLR); and radiographic evidence of osteoarthritis. OA diagnosis was based on radiographic criteria, regarding of the classification or grading scale, OA was defined as the presence of joint-space narrowing of grade 2 and above in the Kellgren and Lawrence classification, grade 1 and above in the Ahlbäck rating system, and grade C and above in the IKDC classification.

Quality assessment

The quality of included studies was appraised by using the Cochrane Collaboration's risk-of-bias tool,[ which contains the following seven items: random sequence generation (selection bias), allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), selective reporting (reporting bias), and other bias. Each of the above items was scored using three options: low risk, unclear risk, and high risk.

Statistical analysis

All data were analyzed using RevMan 5.3 (Cochrane Collaboration, Oxford, UK) software. The mean difference (MD) was used to compare continuous variables, such as the Lysholm knee score and side-to-side difference. The odds ratio (O) was used for some continuous variables, such as return to preinjury activity level, objective IKDC score with grades C and D, pivot-shift test, Lachman test, anterior knee pain, kneeling pain, extension loss, flexion loss, and graft failure. All results were reported with 95% confidence intervals (95%CIs), and P values < .05 were considered to indicate statistical significance. I2 statistics and chi-square tests were used to evaluate heterogeneity. We chose the random-effects model if I2 > 50% and P < .1, because such studies had substantial heterogeneity. Alternatively, we chose the fixed-effects model if I2 < 50% and P > .1. Funnel plots were used to screen for potential publication bias.

Results

Search

We initially searched 890 citations in the three databases and three citations were identified from other sources, of which 354 were excluded as duplicate articles. Screening the titles and abstracts of the remaining 539 citations, we excluded 514 citations, which were irrelevant to the topic of interest, or were not RCT or clinical studies, or were studies with follow-up of shorter than 5 years. The full text of the remaining 25 citations was analyzed by a senior author, and finally, 15 articles[ were included in the meta-analysis (Fig. 1). All 15 studies that compared clinical outcomes between BPTB and HT autografts for ACLR were RCTs. The characteristics of the 15 included studies are summarized in Table 1. Overall, 1298 patients (n = 610 in BPTB, n = 688 in HT groups) were enrolled. The mean follow-up ranged from 60 months to 204 months.

Figure 1

Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) flowchart of the literature search.

Table 1

The baseline data and characteristics of the studies.

			NO. of follow up patients		Sex (male/female)		Mean age (year)					BPTB fixation		HT fixation
Study, year	Country	Level of evidence	BPTB	HT	BPTB	HT	BPTB	HT	Follow up (month)	NO. of depletions	Strands of HT	Femoral	Tibial	Femoral	Tibial
Gifstad[35] 2013	Norway	II	41	36	NA	NA	27	27	84	12	4	ISC	ISC	Sc + WL	Sc + WL
Wipfler[40] 2011	Germany	II	28	25	19/12	18/13	29.8	34.2	105	8	4	Press-fit	Su	Knot	Su
Ibrahim[37] 2005	Kuwait	I	40	45	40/0	45/0	22.3	22.3	81	25	4	EB	ISc	EB	Sc + W/Pl + Sc + St
O’Neill[39] 2001	America	I	75	150	2/1	2/1	Na	Na	102	1	4	Na	Na	Na	Na
Lidén[38] 2007	Sweden	I	32	36	23/11	26/11	28	29	86	3	3-4	ISC	ISC	ISC	ISC
Harilainen[33] 2006	Finland	I	40	39	Na	NA	Na	NA	60	16	4	ISC	ISC	EB	Sc+W
Holm[36] 2010	Norway	I	28	29	18/10	15/14	25	27	120	15	4	ISC	ISC	EB	ISC
Konrads[27] 2016	Germany	II	24	23	22/9	23/8	29.8	29.8	120	14	3–4	ISC	ISC	EB	Su
Zaffagnini[41] 2006	Italy	I	25	25	16/9	15/10	30.5	31.3	60	0	4	ISC	ISC	EB	SC
Ahlden[23] 2009	Sweden	I	22	25	14/8	18/7	26	28	86	24	3–4	ISC	ISC	ISC	ISC
Barenius[24] 2014	Sweden	I	69	66	35/34	44/21	39.2	41.6	169	29	4	ISC	ISC	EB	Su
Mohtadi[29] 2019	Canada	I	103	105	60/43	58/47	33.8	33.7	60	12	4	EB	ISC	EB	ISC
Sajovic[28] 2018	Slovenia	II	24	24	15/9	13/11	45.7	42.5	204	16	4	ISC	ISC	ISC	ISC
Matsumoto[34] 2006	Japan	I	37	35	21/16	15/20	23.7	24.4	87	8	5	ISC	ISC	ISC	ISC
Webster[39] 2016	Australia	I	22	25	16/6	20/5	26.6	26.1	184	18	4	EB	ISC	EB	SC

Quality

The details of the quality assessment are shown in Figure 2. RCTs were assessed by determining the risk-of-bias. For 15 (100%) studies, allocation sequence generation (Criterion 1) was scored as “Low risk.” Allocation concealment (Criterion 2) was scored as “Low risk” in 10 (66.7%) studies and “Unclear risk” in 5 (33.3%) studies. Criterion 3 was scored as “Low risk” in 2 (13.3%) and as “High risk” in 13 (86.7%) studies. Criterion 4 was scored as “Low risk” in 11 (73.3%) and as “Unclear risk” in 4 (26.7%) studies. Criterion 5 was scored as “Low risk” in 8 (53.3%) and as “High risk” in 7 (46.7%) studies. Criteria 6 and 7 were scored as “Low risk” in all (100%) studies.

Figure 2

Assessment of risk-of-bias.

Functional outcomes

Objective IKDC score

Overall, 863 patients were included in nine studies[ in which patients with IKDC grades C and D were found in 71 of 469 patients in the BPTB group and 69 of 394 patients in the HT group. There was no heterogeneity among the studies (P = .76, I2 = 0%), and the difference was not significant between the two groups (OR = 0.94, 95%CI [0.64, 1.37], Z = 0.32, P = .75). These results are shown in Figure 3.

Figure 3

Forest plot of comparison of the postoperative objective International Knee Documentation Committee (IKDC) score with grades C and D between the bone–patellar tendon–bone (BPTB) vs hamstring tendon (HT) autograft groups for anterior cruciate ligament reconstruction.

Lysholm knee score

Eight studies[ were analyzed for the Lysholm knee score, involving a total of 484 patients (238 BPTB, 246 HT). The analysis showed that there was significant heterogeneity among the studies (P = .004, I2 = 66%), and the difference was not significant between the two groups (MD = −2.26, 95%CI [−4.56, 0.05], Z = 1.92, P = .06). These results are shown in Figure 4.

Figure 4

Forest plot of comparison of Lysholm knee score between the bone–patellar tendon–bone (BPTB) vs hamstring tendon (HT) autograft groups for anterior cruciate ligament reconstruction.

Return to preinjury activity level

We reviewed five studies[ for the return to preinjury activity level, which included 396 patients (203 BPTB, 193 HT). There was no heterogeneity among the studies (P = .33, I2 = 12%), and the difference was not significant between the two groups (OR = 1.01, 95%CI [0.67, 1.52], Z = 0.05, P = .96). The results are shown in Figure 5.

Figure 5

Forest plot of comparison of return to preinjury activity level between the bone–patellar tendon–bone (BPTB) vs hamstring tendon (HT) autograft groups for anterior cruciate ligament reconstruction.

Tegner activity level

We reviewed nine studies,[ which included 620 patients (308 BPTB, 312 HT) for the postoperative Tegner activity level. There was significant heterogeneity among the studies (P = .0005, I2 = 72%), but the difference was not statistically significant between the two groups (OR = 0.03, 95%CI [−0.36, 0.41], Z = 0.13, P = .89). The results are shown in Figure 6.

Figure 6

Forest plot of comparison of postoperative Tegner activity level between the bone–patellar tendon–bone (BPTB) vs hamstring tendon (HT) autograft groups for anterior cruciate ligament reconstruction.

Stability

Side-to-side difference

Nine studies,[ including 548 patients (274 BPTB, 274 HT) reported data regarding side-to-side difference. There was significant heterogeneity among the studies (P = .005, I2 = 64%), and the difference was not significant between the two groups (MD = −0.32, 95%CI [−0.81,0.16], Z = 1.31, P = .19). The results are shown in Figure 7.

Figure 7

Forest plot of comparison of postoperative side-to-side difference between the bone–patellar tendon–bone (BPTB) vs hamstring tendon (HT) autograft groups for anterior cruciate ligament reconstruction.

Lachman test

Eight studies[ were analyzed for the Lachman test, involving 501 patients (248 BPTB, 253 HT). There was no heterogeneity among the studies (P = .73, I2 = 0%), and the difference was not significant between the two groups (OR = 0.86, 95%CI [0.56, 1.32], Z = 0.67, P = .50). The results are shown in Figure 8.

Figure 8

Forest plot of comparison of the postoperative Lachman test results between the bone–patellar tendon–bone (BPTB) vs hamstring tendon (HT) autograft groups for anterior cruciate ligament reconstruction.

Pivot-shift test

Five studies,[ involving 455 patients (225 BPTB, 230 HT) reported the outcomes of the pivot-shift test. The analysis showed that there was no heterogeneity among the studies (P = .56, I2 = 0%), and the difference was not significant between the two groups (OR = 0.68, 95%CI [0.44, 1.06], Z = 1.70, P = .09). The results are shown in Figure 9.

Figure 9

Forest plot of comparison of the postoperative pivot-shift test results between the bone–patellar tendon–bone (BPTB) vs hamstring tendon (HT) autograft groups for anterior cruciate ligament reconstruction.

Postoperative complications

Kneeling pain

Six studies,[ including 481 patients (238 BPTB, 243 HT), reported data regarding postoperative kneeling pain. There was no heterogeneity among these studies (P = .25, I2 = 25%), and there was a significant difference between the two groups (OR = 1.67, 95%CI [1.04, 2.69], Z = 2.12, P = .03). The results are shown in Figure 10. Three studies[ reported on the degree of kneeling pain; moderate or severe pain occurred in 21/167 (12.6%) in the BPTB group and in 6/168 (3.6%) in the HT group.

Figure 10

Forest plot of comparison of the postoperative kneeling pain between the bone–patellar tendon–bone (BPTB) vs hamstring tendon (HT) autograft groups for anterior cruciate ligament reconstruction.

Anterior knee pain

Five studies,[ involving 331 patients (165 BPTB, 166 HT), reported postoperative anterior knee pain. There was no heterogeneity among these studies (P = .73, I2 = 0%), and there was a significant difference between the two groups (OR = 2.90, 95%CI [1.46, 5.77], Z = 3.04, P = .002). The results are shown in Figure 11. Only one study[ reported the degree of anterior knee pain; moderate or severe pain occurred in 2/37 (5.4%) in the BPTB group, that in 1/34 (2.9%) of the HT group.

Figure 11

Forest plot of comparison of the postoperative anterior knee pain between the bone–patellar tendon–bone (BPTB) vs hamstring tendon (HT) autograft groups for anterior cruciate ligament reconstruction.

Extension loss

Nine studies,[ involving 749 patients, reported postoperative extension loss. Extension loss occurred postoperatively in 61 of 377 patients in the BPTB group and in 38 of 372 patients in HT group. There was no heterogeneity among the studies (P = .12, I2 = 38%), and the difference was significant between the two groups (OR = 1.75, 95%CI [1.12, 2.75], Z = 2.46, P = .01). The results are shown in Figure 12. The specific conditions of extension loss between the two groups are shown in Table 2.

Figure 12

Forest plot of comparison of postoperative extension loss between the bone–patellar tendon–bone (BPTB) vs hamstring tendon (HT) autograft groups for anterior cruciate ligament reconstruction.

Table 2

The specific conditions of extension loss and flexion loss in each study.

Study	Measuring method	BPTB: extension loss	HT: extension loss	BTBP: flexion loss	HT: flexion loss
Ibrahim[37]	Physical examination	≥10°:2 (5%),≤5°:12 (30%)	≤5°:8 (17.7%)	≥15°:5 (12.5%)	≥15°: 1 (3%)
Gifstad[35]	Physical examination	≥5°:2 (5%)	≥5°:3 (8%)	≥5°:4 (10%)	≥5°:6 (17%)
Ahlden[23]	Unknown	3 (14%)	3 (12%)	6 (27%)	12 (48%)
Barenius[24]	Unknown	3–5°:26 (39%)	3–5°:8 (13%)	6–15°:13 (19%)	6–15°:16 (27%)
Matsumoto[34]	Unknown	5°:3 (8.1%),10°:1 (2.7%)	5°:2 (5.7%)	10°:2 (5.4%)	10°:1 (2.8%)
Konrads[27]	Unknown	5°:1 (4.2%)	Normal	Normal	Normal
Mohtadi[29]	Long-arm goniometer	>3°:0 (0%)	>3°:0 (0%)	Passive flexion: 142.5°±7.3°	Passive flexion: 142.2°±7.8°
Zaffagnini3[41]	Physical examination	≤2°:21 (84%),3–5°:3 (12%),6–10°:0 (0%),>10°:2 (8%)	≤ 2°:18 (72%),3–5°:5 (20%),6–10°:1 (4%),>10°:1 (4%)	≤5°:16 (64%),6–15°:6 (24%), 16–25°: 2 (8%),>25°:1 (4%)	≤5°:22 (88%),6–15°:2 (8%),16–25°:1 (4%),>25°:0 (0%)
Sajovic[28]	Unknown	2 (8.3%)	1 (4.2%)	0 (0%)	0 (0%)

Flexion loss

Seven studies,[ including 505 patients (257 BPTB, 248 HT), reported data regarding postoperative flexion loss. There was heterogeneity among these studies (P = .06, I2 = 52%), and the difference was not significant between the two groups (OR = 1.09, 95%CI [0.47, 2.54], Z = 0.19, P = .85). The results are shown in Figure 13. The specific conditions of flexion loss between the two groups are shown in Table 2.

Figure 13

Forest plot of comparison of postoperative flexion loss between the bone–patellar tendon–bone (BPTB) vs hamstring tendon (HT) autograft groups for anterior cruciate ligament reconstruction.

Graft failure

Twelve studies,[ involving 1280 patients (678 BPTB, 602 HT), reported postoperative graft failure. There was no heterogeneity among these studies (P = .78, I2 = 0%), and there was a significant difference between the two groups (OR = 0.59, 95%CI [0.38, 0.91], Z = 2.38, P = .02). The results are shown in Figure 14. The proportion of graft failure was 5.5% (37/678) in the BPTB group and 9.0% (54/602) in the HT group.

Figure 14

Forest plot of comparison of postoperative graft failure between the bone–patellar tendon–bone (BPTB) vs hamstring tendon (HT) autograft groups for anterior cruciate ligament reconstruction.

Radiographic evidence of osteoarthritis

Nine of the 15 studies[ presented information about degenerative OA of the knee (Table 3). Three studies[ reported OA using the IKDC score, according to radiograph evidence, two[ used the Kellgren and Lawrence classification, and one[ used the Ahlbäck rating system. Zaffagnini et al[ measured the tibiofemoral space on Rosenberg X-rays preoperatively and at follow-up, and diagnosed OA when there was a difference of ≥2 mm between the two X-ray images. In another two studies,[ the methods used to evaluate change related to OA were not clear. The following cases with tibiofemoral joint or medial tibiofemoral compartment OA were combined in our meta-analysis: grade 2 and above in the Kellgren and Lawrence classification, grade 1 and above in the Ahlbäck rating system, and grade C and above in the IKDC classification. Pooled outcomes showed that there was no heterogeneity among the studies (P = .72, I2 = 0%), and the difference was not significant between the two groups (OR = 0.76, 95%CI [0.52, 1.10], Z = 1.44, P = .15) (Fig. 15).

Table 3

Radiographic outcome measures.

Study	Time (months) from injury to operation (BPTB/HT)	Radiographic grading scale	BPTB: radiographic evidence of OA	HT: radiographic evidence of OA
Holm[36]	41.3 ± 41.0/40.5 ± 41.6	Kellgren and Lawrence classification	(grade 0-1-2-3-4):2-8-11-7-0	(grade 0-1-2-3-4):4-9-14-2-0
Barenius[24]	13.9 ± 23/16.3 ± 23	Kellgren and Lawrence classification	(grade 0-1/2-4):31/38	(grade 0-1/2-4):20/46
Webster[42]	NA	Kellgren and Lawrence classification	(grade 0-1/2-3):14/5	(grade 0-1/2-3):13/6
Ahlden[23]	11 (2-252)/17 (3-240)	Ahlbäck rating system	(grade 0-1-2-3-4-5):19-2-0-0-0-0	(grade 0-1-2-3-4-5):20-2-0-0-0-1
Sajovic[28]	22.1 ± 28.8/27.5 ± 43.5	IKDC	(grade A-B-C-D):0-16-5-3	(grade A-B-C-D):7-12-3-2
Matsumoto[34]	11.4 ± 9.9/8.9 ± 9.9	IKDC	3	3
Ibrahim[37]	9.7 (4-13)/9.7 (4-13)	NA	PFJ:8,MTFC:6	PFJ:3,MTFC:8
O’Neill[39]	NA	NA	TFJ:14/150	TFJ:12/75
Zaffagnini[41]	8 (3-13)/9 (2-12)	NA	MTFC:0/25	MTFC:1/25

Figure 15

Forest plot of comparison of osteoarthritis (OA) between the bone–patellar tendon–bone (BPTB) vs hamstring tendon (HT) autograft groups for anterior cruciate ligament reconstruction.

The results of all the above forest plot are summarized in Table 4.

Table 4

Summary of the meta-analyses.

Analyses item	Number of studies	Heterogeneity and P	Analysis model	MD/OR (95%CI)	P
Objective IKDC score	9	0% (.66)	Fixed-effects	0.94 (0.64,1.37)	.75
Lysholm knee score	8	66% (.004)	Random-effects	−2.26 (−4.56,0.05)	.06
Return to preinjury activity level	5	12% (.33)	Fixed-effects	1.01 (0.67,1.52)	.96
Tegner activity level	9	72% (.0005)	Random-effects	0.03 (−0.36,0.41)	.89
Side-to-side difference	9	0% (.72)	Fixed-effects	0.76 (0.52,1.10)	.15
Lachman test	8	0% (.73)	Fixed-effects	0.86 (0.56,1.32)	.50
Pivot shift test	5	0% (.56)	Fixed-effects	0.68 (0.44,1.06)	.09
Kneeling pain	6	25% (.25)	Fixed-effects	1.67 (1.04,2.69)	.03
Anterior knee pain	5	0% (.73)	Fixed-effects	2.90 (1.46,5.77)	.002
Extension loss	9	38% (.12)	Fixed-effects	1.75 (1.12,2.75)	.01
Flexion loss	7	52% (.06)	Random-effects	1.09 (0.47,2.54)	.85
Graft failure	12	0% (.78)	Fixed-effects	0.59 (0.38,0.91)	.02
Radiograph evidence of osteoarthritis	9	0% (.72)	Fixed-effects	0.76 (0.52,1.10)	.15

Publication bias

A funnel plot was created to assess if there was bias in this study. The funnel plot appeared mild asymmetrical, suggesting publication bias in the data (Fig. 16).

Figure 16

Funnel plot with 95% CL is showing mild publication bias for comparison of graft failure between BPTB and HT autograft.

Discussion

The results of this meta-analysis indicated that, compared with BPTB autografts, HT autografts were associated with a decreased incidence of anterior knee pain, kneeling pain, and extension loss, but increased incidence of graft failure after ACLR in medium-term follow-up. Differences in other measures were not statistically significant between the two types of autografts used for ACLR.

Although ACL reconstruction is a mature technology, there is still considerable controversy about which autograft provides the best results for ACLR. BPTB autografts raise concerns about donor site morbidity, including anterior knee pain, kneeling pain, patella fracture, and patellar tendon rupture.[ In their review, Kartus et al[ pointed out that 40% to 60% of patients reported various donor site problems after harvesting of the BPTB autograft. It has been reported that patella fracture and patellar tendon rupture after ACLR are quite rare (with an incidence of 0.4% to 1.3% and 0.18% to 0.25%, respectively), and are related to irregular surgical procedures, such as deep and irregular saw cutting, poor centralization during the harvest leading to the lateral or medial side being too thin.[ However, anterior knee pain and kneeling pain were quite common after ACLR with BPTB autograft, and it has been reported that these pains are related to injury to the infrapatellar branches of the saphenous nerve during the operation;[ In BPTB autografts, their incidences were 52% and 65% at the 2-year follow-up, respectively, which was higher than the 17% and 35% incidences, respectively, of HT autograft.[ It has been reported that the transplantation of β-tricalcium phosphate or autologous bone to the donor-site bone defects at the tibial and patellar tubercle can effectively reduce the occurrence of anterior knee pain and kneeling pain.[ Additionally, the double-incision technique has been used to avoid injury of the saphenous nerve infrapatellar branches and to preserve the paratenon and the patellar tendon, which can also reduce the donor site morbidity associated with BPTB.[ Gaudot et al[ used the double-incision technique to reduce the frequency of anterior knee pain to 18% (as compared to 58% for the single-incision technique), but there was no significant difference in the degree of anterior knee pain between the two techniques. The results of this meta-analysis showed that the HT group had apparent advantages in anterior knee pain and kneeling pain during medium-term follow-up. The incidence of moderate or severe kneeling pain was significantly higher in the BPTB group than in the HT group. Clinically, the technique of double-incision and bone grafting could be used to reduce donor site morbidity. BPTB may thus not be recommended for patients who need to kneel frequently during their work. The HT autograft also has some disadvantages. The gracilis and semitendinosus muscle mainly function as internal tibial rotators and knee flexors, so that HT autograft harvests may lead to flexion strength deficits. Gifstad et al[ reported that the peak flexion torque of the operated knee decreased significantly after 1 year in the HT group. However, there was no difference after 2 years and after 7 years.[ The results of this meta-analysis showed that the HT group had apparent advantages in terms of anterior knee pain and kneeling pain during the medium-term follow-up. This finding was consistent with the results of previous published meta-analyses.[

Another focus of the debate about the two autografts was graft failure. The leading causes of postoperative graft failure include graft rupture and graft laxity after ACLR, which might be affected by the biomechanical and histologic properties of the graft. One of the advantages of BPTB autografts is its excellent biomechanical properties, which was similar to that of the ACL. Woo et al[ proved that, when the ACL was tested in the anatomical direction, for young patients (aged 22–35 years), the maximum load to failure of the ACL was 2160N, while those of the BPTB autograft and quadrupled HT graft were 800N and 4590N, respectively.[ These studies mentioned above suggested that the HT autograft may have better mechanical properties than the BPTB autograft. However, the present meta-analysis suggested that, at medium-term follow-up, the HT group had a higher incidence of graft failure than the BPTB group; these results were mainly consistent with Samuelsen et al[ and Chee et al's meta-analysis.[ In terms of postoperative knee stability, the results of the present study showed that there was no significant difference in side-to-side difference, Lachman test, and pivot-shift test results between the two groups after medium-term follow-up.[ This was mainly consistent with the result of Samuelsen et al[ and Xie et al's meta-analysis.[ The above results show that some indexes of knee stability, including the side-to-side difference, Lachman test, and pivot-shift test, have limitations in that they are carried out on an open kinetic chain with a relaxed lower limb and cannot reliably predict the functional stability of the knee in sports activities.[ In this study, the graft failure rate was 5.5% in the BPTB group and 9.0% in the HT group. This finding may indicate that it was not only the strength of the graft itself that affected the structural strength of the graft structure, but also the fixation points on both the femoral and tibial side, which seem to be the weakest points.[ The proper bone-to-bone tunnel healing of the BPTB autograft could provide firm fixation, while the HT autograft involved a slow tendon to bone healing process and a lack of rigid fixation.[ A previous study[ has shown that soft tissue could frequently take up to 6 weeks longer to incorporate into the host bone than the BPTB autograft. Therefore, the increased failure and laxity of the graft may be a clinical manifestation of the slower ligamentization process and different biomechanical characteristics of the HT autograft.[

The occurrence of OA after ACLR was also a concern. In the published meta-analysis, there was not much discussion about the occurrence of OA after ACLR, but the problem was universal. However, it was not clear which type of autograft would increase the risk of OA progression. Xie et al's meta-analysis showed that the incidence of OA in the BPTB group might be increased, as compared with that in the HT group.[ However, the above meta-analysis did not limit the included studies to RCTs. The results of present meta-analysis showed that there was no significant difference in the incidence of OA between the two autografts after medium-term follow-up, which was consistent with the conclusion of Belk et al's systematic review.[ ACL injury is often accompanied by cartilage injury of the knee joint, which persists after ACLR. After injury, some inflammatory responses in the knee could ultimately lead to OA progression.[ Therefore, according to our knowledge of the mechanism of post-traumatic OA[ and the results of this meta-analysis, the degree of a knee injury after trauma, rather than the types of autograft, might be the key factor affecting the progression of OA after ACLR. However, these results should be cautiously interpreted. First, the occurrence of OA after ACLR may be related to meniscectomy. We were unable to confirm whether the results were affected by meniscectomy due to the lack of information in the included study. Second, different radiological classification systems for OA (such as Kellgren and Lawrence, Ahlback, IKDC) were used in this meta-analysis, and a conversion methods for various OA score need to be developed to allow subsequent statistical analysis. In addition, due to the lack of information provided in the included studies, we did not further explore whether the progression of OA was mainly based on the patellofemoral joint or the tibiofemoral joint. If it was mainly the progression of patellofemoral osteoarthritis, it may represent technique of graft harvest and suggest that extreme care be taken during harvest to avoid chondral injury. If it was mainly the progression of tibiofemoral osteoarthritis, it may be related to meniscectomy and the increased rate of stiffness after utilizing BPTB autograft reported by some authors. Regardless of the cause of the effect, it was recommended that both BPTB and HT autograft could be selected in patients who already exhibit early signs of osteoarthritis.

Several potential limitations of this study should be noted. First, the final clinical results may be affected by the differences in population heterogeneity, follow-up duration, follow-up loss, method of graft fixation, and postoperative rehabilitation regimen, but it is not possible to control these variables individually. Second, although the flexion torque after HT harvest is important in the discussion between choosing BPTB and HT autografts for ACLR, there have been few reports about this in the included studies, and the reported indicators are different, making it impossible to carry out a meta-analysis of flexion torque. Third, not all the studies use the same scale to grade radiographic signs of OA and there was an apparent lack of reported data combining both clinical and radiological aspects of OA. More high-quality RCTs with long-term follow-up are needed to come to a definite conclusion in future.

Conclusions

Based on the results above, HT autograft is comparable with the BPTB autograft in terms of clinical function, postoperative knee stability, and OA changes, with a medium-term follow-up. The HT autograft for ACL reconstruction carries a lower risk of complications, such as anterior knee pain, kneeling pain, and extension loss, but an increased incidence of graft failure. Patients should be informed of the differences when deciding on graft choice with their physician.

Author contributions

LL Z and MF L have made substantial contributions to conception and design of the study, written the manuscript; MC D, LL H, CB W, and JS X unsearched literature, extracted data from the collected literature and analyzed the data; LL Z revised the manuscript; all authors have read and approved the final version of the paper for submission.

Data curation: Jisi Xing.

Formal analysis: Lilei He.

Resources: Mingcong Deng, Changbing Wang.

Writing – original draft: Mingfeng Lu.

Writing – review & editing: Lilian Zhao.