General Prediction Theory for Anterior Cruciate Ligament Graft Sizing

Abstract

Predicting hamstring graft size before anterior cruciate ligament (ACL) reconstruction is crucial to avoid subsequent graft failure. Our study aims to (1) develop a generalized algorithm to predict final ACL graft diameter for single- and double-tendon hamstring grafts consisting of any number of folds, (2) evaluate our algorithm with a regression model adjusting for patient and surgical factors, and (3) assess algorithm's specificity, sensitivity, and discriminative ability, defining adequate graft size as ≥9 mm.

We conducted a retrospective review of 105 patients who underwent primary ACL reconstruction with either single- (semitendinosus or gracilis only) or double-tendon (both semitendinosus and gracilis) hamstring grafts from January 2023 to June 2024 at a tertiary institution. Magnetic resonance imaging (MRI) scans were independently measured by two junior members. The average of the measurements was taken. Predicted graft diameter is √(x·AB + y·CD), where A and B are the semitendinosus cross-sectional length and breadth, C and D are the gracilis cross-sectional length and breadth, and x and y are the number of semitendinosus and gracilis folds, respectively.

Pearson correlation shows moderate correlation between predicted and actual graft diameters (R = 0.602, p < 0.01). Univariate and multivariate linear regression, adjusted for age, gender, body mass index (BMI), and graft type, indicate that males, overweight individuals, and those with single-tendon grafts are more likely to have larger actual graft diameters (p < 0.05). Our algorithm has a sensitivity of 95.8%, specificity of 69.7%, with excellent discriminative ability (AUC = 0.889). A high 82.9% agreement rate was achieved, with Cohen's kappa of 0.578 (p < 0.05).

This study has demonstrated a practical generalized algorithm with high sensitivity (95.8%) to predict the final ACL graft diameter for single- and double-tendon hamstring grafts consisting of any number of folds using preoperative MRI measurements.

 

Scroll to top

Keywords

knee - ACL - MRI - hamstring graft

Anterior cruciate ligament (ACL) reconstruction is one of the most commonly performed orthopaedic procedures. While there are a myriad of graft options available, to prevent disruption to the extensor mechanism, many surgeons favor hamstring autografts over bone–patellar tendon–bone (BPTB) grafts.[1] [2] Yet numerous studies demonstrate that hamstring autografts have an inherent variability in graft diameter, which could be detrimental for ACL reconstruction.[3] [4] [5] [6] [7] In biomechanical and clinical trials, smaller hamstring graft sizes have a greater rate of graft failure.[6] [7] [8] Predicting hamstring graft size before ACL reconstruction is, hence, paramount to avoid intraoperative graft failure.

Numerous studies have attempted to predict graft size preoperatively in light of surgeon concerns regarding graft size in ACL surgery, including hamstring autografts.[9] [10] [11] [12] [13] [14] [15] Studies have attempted to discover correlations between the final length and diameter of the semitendinosus tendon and the patients' demographics, including height and weight.[16] [17] [18] As an alternative, preoperative imaging, particularly magnetic resonance imaging (MRI) scans, has been utilized to estimate tendon diameter, with varying degrees of success.[15] [19]

Most published models for predicting hamstring grafts preoperatively using MRI measurements are not as feasible as they require (1) recruiting the help of experienced radiologists or senior orthopaedic surgeons for measurements, which may not always be practical for routine practice in a high-volume institution[9] [20]; (2) the usage of specialized 3D software or freehand region of interest lasso tools that are not widely available on many MRI Picture Archiving and Communication System (PACS); (3) a 3-T MRI to make preoperative measurements, which are costly and uncommon.[21] Furthermore, most, if not all, published models assume that the hamstring graft that the surgeon uses is a four-stranded double-tendon hamstring graft consisting of the semitendinosus tendon folded into two, and the gracilis tendon folded into two.[22] However, we know that surgeons very frequently may utilize five- or six-strand ACL grafts as well, and recent advances in ACL reconstruction demonstrate potential benefits in using single-tendon hamstring grafts.[4] [12]

Hence, our study aims to (1) develop a generalized algorithm to predict final ACL graft diameter, in a continuous function, for single- and double-tendon hamstring grafts consisting of any number of folds, (2) evaluate our algorithm with a regression model adjusting for patient and surgical factors, and (3) assess algorithm's specificity, sensitivity, and discriminative ability, as a discrete function to compare our theory against other methods, defining adequate graft size as ≥9 mm.

Methods

A retrospective review of 105 patients from three senior surgeons subspecializing in sports surgery who underwent primary ACL reconstruction with either single-tendon (i.e., either semitendinosus or gracilis only) hamstring grafts or double-tendon (both semitendinosus and gracilis) hamstring grafts from January 2023 to June 2024 at a tertiary institution was conducted. Ethics approval was obtained from the National Healthcare Group Domain Specific Review Board (DSRB number 2021/00555) before the initiation of the study. The inclusion criteria consisted of skeletally mature patients who had undergone primary ACL reconstruction using either single- or double-tendon hamstring grafts, regardless of the number of folds, with preoperative MRI images showing intact hamstrings. Exclusion criteria included patients whose MRI was performed at private institutions, making the images inaccessible, those with prior surgical interventions on the imaged knee, as well as patients who underwent ACL reconstruction with quadriceps or BPTB autografts, allografts, or synthetic grafts ([Fig. 1]).

Fig. 1 Flow diagram showing the inclusion and exclusion criteria of the study. ACL, anterior cruciate ligament; MRI, magnetic resonance imaging.

Semitendinosus and Gracilis Tendon Measurements

Senior radiographers at the authors' tertiary institution used a 1.5-T Siemens MAGNETOM Aera scanner (Erlangen, Germany) to provide MRI images of the afflicted knees using sequences such as axial medic, coronal PD/T2-weighted spin echo, and sagittal PD/T2-weighted spin echo. Two independent evaluators—a fourth-year medical student and a junior medical officer (postgraduate year 1), both without prior radiology posting experience, separately measured the cross-sectional lengths and breadths of the semitendinosus and gracilis tendons using preoperative MRI scans for all included patients. The two evaluators were separately trained by LZQG, the senior author of the study.

The average of the measurements was taken. The actual graft sizes were measured intraoperatively using graft sizing tubes and sizing cylinders in increments of 0.5 mm, and the evaluators were blinded to them. To standardize the measuring procedure, an instructional document detailing the measurement protocol and technique was created for all reviewers to utilize. As explained in the following paragraph, both evaluators underwent a similar 1-hour training session and were blinded to each other's findings.

The previously described measurement protocol by Liau et al (2024) is as follows[22]: The standard PACS (General Electric, NY) was used to measure the semitendinosus and gracilis grafts on MRI knees. The axis of the pes tendons, which are often located close to the subcutaneous level of the medial tibial plateau, was first identified using sagittal T2-weighted images of the injured knee ([Fig. 2A]). Individual differences exist in the pes tendon axis ([Fig. 2B]).

Fig. 2 T2-weighted MRI displaying sagittal knee slices from two different patients (A, B), illustrating variations in the pes tendon axes used as reference points for subsequent measurements. MRI, magnetic resonance imaging.

Following the same sagittal sequences, sagittal cuts were made laterally from the sagittal view from which the axis of the pes tendons was taken, usually ranging from 12 to 21 mm (four to seven sagittal cuts, each measuring 3 mm). This was done until a distinct separation between the cartilage and the subchondral bone of the medial tibial plateau could be seen. This reveals the previously described Liau subchondral bony ridge.[22] It could appear on one to three consecutive 3-mm sagittal cuts ([Fig. 3A]). Line A, an annotation line was positioned such that it originated from the ridge on the posterior edge of the subchondral surface of the tibial plateau (immediately anterior to the downward-curving part of the tibial plateau) and extended 30 mm proximally parallel to the pes tendon axis ([Fig. 3A]). The nearest axial MRI cut is taken from the proximal aspect of this 30-mm line. Cross-sectional lengths and breadths of the semitendinosus and gracilis tendons were measured from this axial cut using enlarged axial slices ([Fig. 3B], [C]) using at least ten times magnification, that is, a 4-mm length would appear as 4 cm on the screen. Measurements were made at the midpoint of the zone between the clear edges of the tendon and the subcutaneous tissues in instances where the zone of transition grays.

Fig. 3 T2-weighted MRI displaying (A) sagittal view and (B) axial views of the affected knee. MRI, magnetic resonance imaging.

The axis along the pes tendons was selected for line A. The Liau subchondral bony ridge was selected as a reproducible bony landmark because it approximates the level of the tibial tunnel's intra-articular opening. With the exception of a few millimeters because of the pes tendons' oblique tibial insertion, the length of the pes tendons from the level of its tibial insertion to the level of the Liau ridge would therefore almost match the length of the ACL graft in the tibial tunnel. End-to-end average intraarticular ACL graft lengths were reported to be 30.75 mm.[23] Thus, 30 mm proximally from the Liau ridge would approximately correspond to the level of the pes tendons that, when folded on itself at that level, would be inserted into the femoral tunnel. This level was then used to measure the semitendinosus and gracilis dimensions. Since this is a retrospective review, the surgeons were all blinded to the aforementioned measurements.

 

Scroll to top

Mathematical Proof for General Prediction Theory

The semitendinosus and/or gracilis tendons' cross-sectional areas were used to calculate the diameter of the predicted graft size ([Fig. 4]), assuming an ovoid cross-sectional profile of the tendons, and depending on whether the graft is a single- or double-tendon graft.

Fig. 4 Proposition of a circular profile of predicted ACL graft tendon, based on the ovoid cross-sectional profiles of the harvested semitendinosus and gracilis grafts. ACL, anterior cruciate ligament.

A numerically continuous prediction for the graft diameter is √(x·AB + y·CD) where A and B represent the semitendinosus cross-sectional length and breadth, respectively, C and D represent the gracilis cross-sectional length and breadth, respectively, and x and y represent the number of semitendinosus and gracilis folds, respectively ([Fig. 5]). If only the semitendinosus was used, y would be 0, and if only the gracilis was used, x would be 0. The total area of the predicted graft (πr² ) was determined using its circular profile, by adding together (1) the multiplication of semitendinosus tendon's area by x number of times based on the number of folds, and (2) the multiplication of gracilis tendon's area by y number of times based on the number of folds ([Fig. 5]). The radius (r) and subsequently the diameter of the predicted graft size (2r) are then calculated based on the formula. The diameter of the predicted graft size was calculated using measurements of both hamstring tendons using an automated calculator in Microsoft Excel ([Supplementary Data] [available in the online version only]). Since intraoperative graft size is 0.5 mm, measurements are made in increments of 0.5 mm, and an accurate predicted graft size is defined as ±0.5 mm of the actual graft size.

Fig. 5 Mathematical proof of our proposed generalized algorithm.

 

Scroll to top

Statistical Analysis

All statistical analyses were conducted in R (R 4.1.0). The means and 95% confidence intervals (CIs) were computed for all continuous variables. Pearson's correlation coefficient was computed. A confusion matrix was used to calculate specificity and sensitivity. Receiver operating characteristic (ROC) curves and logistic regression were calculated accordingly. Interobserver reliability between the evaluators was calculated using percentage agreement and Cohen's kappa. The study did not encounter any missing data.

 

Scroll to top

 

Scroll to top

Results

Predicted Anterior Cruciate Ligament Hamstring Graft Sizes

About 105 patients ([Table 1]) who underwent primary ACL reconstruction with ST-G hamstring grafts were recruited for this study.

Table 1 Demographic profile of patients

Total number	105
Mean age (95% CI)	27.06 (25.71–28.41)
Gender
Male	72
Female	33
Body mass index[a]
Overweight	68
Not overweight	36
No height/weight documented	1
Graft type
Single-tendon[b]	20
ST only, quadrupled	19
GT only, quadrupled	1
Double-tendon[c]	85
ST doubled, GT doubled	40
ST tripled, GT doubled	45
Accuracy
Within 0.5 mm[d]	89

Abbreviations: CI, confidence interval; GT, gracilis tendon; ST, semitendinosus tendon.

^a Overweight is defined using the Asia-Pacific body mass index cutoff values of ≥23 kg/m².

^b Single-tendon graft type is defined as the harvesting of one tendon only, that is, either a gracilis or semitendinosus tendon.

^c Double-tendon graft type is defined as the harvesting of both the semitendinosus and gracilis tendons.

^d Within 0.5 mm is defined as the number of patients where the predicted graft diameter was within 0.5 mm of the actual graft diameter.

[Table 2] presents the cross-sectional MRI measurements of the semitendinosus and gracilis tendons, the predicted hamstring graft diameter as calculated using the proposed generalized algorithm, and the actual hamstring graft diameter measured intraoperatively.

Table 2 MRI measurements of tendons, with the predicted and actual graft diameters (n = 105)

	Cross-sectional measurements done on MRI (mm)				Diameter (mm)
	SemiT, larger diameter	SemiT, smaller diameter	Gracilis, larger diameter	Gracilis, smaller diameter	Predicted graft diameter	Actual graft diameter
Fourth-year medical student	5.28 (5.12, 5.44)	3.74 (3.63, 3.85)	4.6 (4.43, 4.78)	2.47 (2.38, 2.55)	8.6 (8.48, 8.72)	8.6 (8.5, 8.71)
Junior medical officer	5.32 (5.16, 5.48)	3.86 (3.75, 3.97)	4.53 (4.35, 4.71)	2.61 (2.52, 2.71)	8.8 (8.64, 8.95)	8.6 (8.5, 8.71)
Average	5.3 (5.15, 5.45)	3.8 (3.69, 3.91)	4.57 (4.4, 4.73)	2.54 (2.46, 2.62)	8.7 (8.57, 8.83)	8.6 (8.5, 8.71)

Abbreviation: MRI, magnetic resonance imaging.

The values are indicated as mean (95% CI).

 

Scroll to top

Correlation between the Predicted and Actual Graft Sizes

Pearson's correlation coefficient between the predicted and actual graft diameter was 0.602 (p < 0.01), which shows a moderate positive correlation in accordance with Schober et al's interpretation[24] ([Fig. 6]).

Fig. 6 Predicted graft diameter versus actual graft diameter showing a moderate positive correlation. The line of best fit follows the equation y = 0.9438x + 0.5787, while the gray zone represents the 95% confidence interval.

Performance of Our General Prediction Theory When Defining an Adequate Actual Graft Size as ≥9 mm

Defining 9 mm as an adequate graft size, a confusion matrix ([Table 3]) reveals that the generalized algorithm achieved a 95.8% sensitivity, correctly identifying 69 out of 72 patients with inadequate actual graft diameters (<9 mm). Augmentation in the form of a lateral extra-articular tenodesis (LET) was performed for all of the abovementioned patients to improve rotational stability and decrease re-tear rates. At the same time, the algorithm also achieved a specificity of 69.7%, correctly identifying 23 out of 33 patients with adequate actual graft diameters (≥9 mm).

Table 3 Confusion matrix showing the number of patients correctly classified as having inadequate (<9 mm) or adequate (≥9 mm) actual graft diameter

	Actual graft diameter (mm)
Predicted graft diameter (mm)	<9	≥9
<9	69 (95.8%)	10 (30.3%)
≥9	3 (4.2%)	23 (69.7%)

Univariate and multivariate linear regressions between predicted and actual graft diameter were performed ([Table 4]) while adjusting for age, gender, BMI, and type of graft (single- vs. double-tendon).

Table 4 Univariate and multivariate linear regressions (adjusted R ² = 0.498, p < 0.0001)

	Adjusted odds ratio	p-Value
Predicted graft diameter	1.41 (1.28–1.56)	<0.0001
Age	0.99 (0.98–1)	0.112
Gender (male)	1.31 (1.11–1.56)	0.00176
BMI (overweight)	1.22 (1.04–1.44)	0.0140
Double-tendon graft	0.72 (0.6–0.88)	0.000115

Abbreviation: BMI, body mass index.

Univariate linear regression showed that the larger the predicted graft diameter, the larger the actual graft diameter (adjusted R ² = 0.356). The generalized algorithm is significant (p < 0.01). Multivariate linear regression showed that the odds of having a bigger actual graft diameter are higher if you are male (p = 0.00160), if you are overweight (p = 0.0130), and if a single-tendon graft is used (p = 0.000799). Only age does not influence the actual graft diameter (p = 0.0850). For the regression models, an area under the ROC curve was generated ([Fig. 7]), indicating good discrimination (AUC = 0.889).

Fig. 7 Area under receiver operating characteristic plotted with the respective logistic regression models showing good discrimination. Adequate actual graft diameter is defined as ≥9 mm.

 

Scroll to top

 

Scroll to top

Interobserver Reliability of Our General Prediction Theory

Finally, an assessment of the interobserver reliability of the measurements of the semitendinosus and gracilis tendons between the two separate, blinded assessors ([Table 2]) was conducted. These assessors were trained junior members of the surgical team. More precisely, the extent of agreement in terms of the number of patients who were accurately categorized as having an actual graft diameter that was either adequate (≥9 mm) or inadequate (<9 mm) was assessed. In accordance with Landis and Koch's interpretation,[25] our findings indicate a “moderate” agreement between the two evaluators, with a high percentage agreement of 82.9% and a Cohen's kappa of 0.578.

 

Scroll to top

 

Scroll to top

Discussion

General Prediction Theory for Anterior Cruciate Ligament Graft Sizing

To the authors' knowledge, this is the first generalized algorithm that enables predicted single- or double-tendon ACL hamstring graft sizes consisting of any number of folds to be computed in a continuous numerical manner rather than in a binary fashion that produces dichotomous results. As previously shown in our novel mathematical proposition of a circular profile of predicted ACL graft tendon based on the ovoid cross-sectional profiles of the harvested semitendinosus and gracilis grafts, our method enables users to ascertain the likely exact value of the axial graft diameter. Unlike previous studies that assume the hamstring graft is going to be quadrupled, our study is the first that allows the exact value of the graft diameter to be calculated from five-stranded grafts, six-stranded grafts, single-stranded grafts, or grafts with any number of folds. Other techniques only use a cutoff surface area based on MRI measurements to produce a dichotomized result of whether a graft is sufficient or not, based on a predefined target size.[9] [15]

Our approach offers the surgical team significantly more detailed information regarding the extent of the discrepancy between the actual and ideal hamstring graft size, enabling more informed decisions when selecting a graft method. Additionally, it may assist in preoperative planning by providing insight/s into the appropriate femoral and tibial tunnel sizes required for the procedure. Furthermore, we have developed an Excel calculator ([Supplementary Data] [available in the online version only]) with our inputted generalized algorithm that would automatically calculate the predicted graft size based on the four parameters in our General Prediction Theory (semitendinosus cross-sectional length and breadth, gracilis cross-sectional length and breadth).

 

Scroll to top

Performance of Our General Prediction Theory When Defining an Adequate Actual Graft Size as ≥9 mm

Most studies in current literature deem adequate graft size as ≥8 mm.[9] [20] A study conducted by Magnussen et al (2012) shows that the revision rate of grafts measuring >8.5 mm was 1.7% as compared with 6.5% for grafts measuring 8 to 8.5 mm.[26] Moreover, increments in graft sizes up to 10 mm can be beneficial for patients.[27] Given the incremental benefits of using grafts of ≥8.5 mm up to 10 mm, a more stringent adequate actual graft size of 9 mm was defined in this study.

Studies have shown that when it comes to determining the size of hamstring autografts utilized in ACL restoration, preoperative MRI images are more accurate than anthropometric factors.[21] [28] It has been demonstrated that preoperative MRI measurements of the hamstring tendons can help more accurately estimate the intraoperative ACL graft size. Given that semitendinosus, gracilis, and total CSA showed a significant positive connection with the actual graft diameter (p < 0.001), Thwin et al (2020) concurred with our findings. When it came to identifying adequate graft size (defined as ≥7 mm), MRI's sensitivity and specificity were 84.1% and 100%, respectively.[28] Erquicia et al (2013) reported similar findings as well, obtaining a 96.2% sensitivity and 100% specificity for tendons with CSA ≥8 mm.[11] Similarly, Wernecke et al reported that semitendinosus CSA, gracilis CSA, and combined CSA were among the significant positive correlations between preoperative MRI and intraoperative findings (p = 0.0006, 0.001, and 0.001, respectively). They only provided a cutoff value for the semitendinosus and gracilis CSAs, not the total CSA, to forecast the 7-mm graft diameter.[12]

The General Prediction Theory based on a generalized algorithm yields a high sensitivity of 95.8% when defining an adequate actual graft size as ≥9 mm, which is comparable to contemporaneous studies conducted in terms of sensitivity and specificity. Our approach does not require any specific software or programs. This algorithm offers a much more convenient method for predicting the size of ACL hamstring grafts by merely measuring the larger and smaller diameters of semitendinosus and gracilis tendons.

 

Scroll to top

Correlation between the Predicted and Actual Graft Sizes

Although previously published studies report a positive correlation between the predicted CSA on MRI with the actual intraoperative graft diameters, most of these correlations were low to moderate, with Pearson's correlation coefficient of 0.419,[10] 0.495,[14] and 0.536.[11] Our Pearson's correlation coefficient between the predicted and actual graft diameter was 0.602 (p < 0.01), which shows a moderate positive correlation ([Fig. 6]). CSAs of the tendons were assessed at the widest point of the medial femoral condyle,[11] the musculotendinous junction,[10] or at the physis/physeal scar of the femur in other investigations that found poorer correlations.[14] The tendons in our study, however, were measured 3 cm proximal to the Liau ridge, which is roughly where the ACL graft would be inserted into the femoral tunnel. This anatomical consideration has not been considered in any previous investigations, which may have led to the superior correlation reported.

 

Scroll to top

Factors Resulting in Larger Actual Graft Diameters

Anthropometric measures and graft size have been the subject of numerous research studies. Our study found that the odds of having a larger actual graft diameter are higher if the patient is (1) male, (2) overweight, or (3) using a single-tendon graft.

It is well-known that graft diameter varies by gender, with males often having larger grafts than females.[29] [30] [31] Our study reports that the odds of having a larger actual graft diameter are indeed higher if the patient is male, which we hypothesize is due to males having larger muscle mass and tendon size compared with females. Pinheiro et al (2011) reported that the average graft diameter was significantly smaller in women than in men (p < 0.001).[30] Treme et al (2008) analyzed the hamstring graft diameters of 50 patients and found that, on average, women, particularly those with smaller stature and lower weight, eventually presented with smaller graft diameters.[31]

Our study also found that the odds of having a larger graft diameter are increased in overweight individuals, likely due to an increased body mass that leads to increased tendon sizes and consequently graft diameters harvested. This is in concordance with Papastergiou et al (2012), who discovered a moderate correlation between patient weight and the diameter of the semitendinosus and gracilis graft, as well as a correlation between BMI and graft diameter.[32] However, not all studies found a similar trend. Thomas et al (2013) measured these parameters in 121 patients and found no statistically significant relationship between BMI and actual graft size.[33] Atbaşi et al. (2017) found a statistically significant relationship between patient weight and graft diameter (p < 0.0001), yet there was no statistically significant relationship between BMI and graft diameter (p > 0.05).[1] According to Tuman et al (2007), the actual hamstring graft diameter is associated with the patient's height, weight, age, and sex, but BMI and actual graft diameter did not correspond.[34]

We also found that the odds of a larger actual graft diameter were higher in patients using a single-tendon graft. We postulate that, due to the nature of graft preparation—where a single tendon is folded a total of three times to make a quadrupled tendon, as opposed to two tendons being folded only twice—the single-tendon grafts have a higher likelihood of achieving a larger graft diameter, with the folds themselves potentially contributing to the graft's thickness. This is further supported by the fact that single-tendon grafts typically involve harvesting thicker grafts compared with double-tendon grafts. There is currently no literature that reports the relationship between actual graft diameter and single- versus double-tendon grafts.

It is interesting to note that in our study, only age does not influence the actual graft diameter. We postulate this is due to tendons maintaining their dimensions with age. This is in concordance with Moghamis et al (2019), who reported a positive correlation between age and the final graft diameter.[35] However, we also recognize that other studies found that the final quadrupled hamstring graft diameter specifically was negatively correlated with age.[31] [34] [36]

 

Scroll to top

Interobserver Reliability of Our Generalized Prediction Theory and Its Clinical Translation

Our study demonstrates good interobserver reliability, as evidenced by a high percentage agreement of 82.9% and a Cohen's kappa of 0.578, which indicates “moderate” agreement between the two evaluators, namely the fourth-year medical student and junior medical officer. These results are particularly noteworthy because the evaluators had no prior radiological experience, underscoring the robustness of our measurement protocol.

Clinically, this moderate interobserver reliability translates into greater confidence for the surgical team, as it shows that even less experienced trainees can accurately predict hamstring graft sizes using our generalized algorithm. This reliability ensures consistent preoperative planning and decision-making, reducing variability and enhancing the team's ability to anticipate appropriate graft selection and tunnel sizes. By offering a standardized, reproducible approach to measuring hamstring grafts, our method supports surgical efficiency and potentially improves patient outcomes through better preparation and precision in ACL reconstruction procedures.

In cases where our model predicts a smaller graft size, we recommend that surgeons consider additional strategies to reinforce the hamstring graft or opt for an alternative graft to ensure optimal outcomes. Our predictive method allows the surgical team to be well-prepared in advance, offering flexibility in decision-making during the procedure. If the predicted graft size is smaller than ideal, the surgeon may choose to augment the graft by tripling or quadrupling the hamstring strands to achieve the desired thickness and strength. Alternatively, the surgeon may decide—at the preoperative stage—to switch to a different autograft, such as a BPTB graft or a quadriceps tendon graft, which may provide more structural integrity by allowing a potentially larger harvested graft thickness. We would also suggest additional augmentation, such as extra-articular stabilization with the use of an LET or an anterolateral ligament reconstruction, to help control rotational instability, as is routinely practiced within our institution. It may even be appropriate for additional intra-articular stabilization with suture tape augmentation to assist with load-sharing in grafts that are in the 8- to 9-mm range.

Conversely, if the predicted graft size is larger than expected, the surgeon can proceed with greater confidence in the adequacy of the hamstring graft, reducing concerns about graft failure and ensuring the patient has a strong, durable ACL reconstruction. This approach significantly enhances preoperative planning and allows for informed, real-time decisions, ultimately improving surgical outcomes by tailoring the graft choice to the specific anatomical and procedural needs of each patient.

 

Scroll to top

 

Scroll to top

Limitations and Future Work

The General Prediction Theory has its limitations. First, we did not account for the variable degrees of knee flexion during MRI. However, this is not reported in the literature and does not appear to affect our accuracy. Second, all MRI scans were performed with a single 1.5-T Siemens MAGNETOM Aera Scanner (Erlangen, Germany), whereas 3-T MRI may be used clinically. Nonetheless, we postulate that the use of 3-T MRI will likely further improve our algorithm's precision and prediction accuracy. Third, our model recorded a specificity of 69.7%. However, overestimating graft size has minimal impact and implications on surgical management and preoperative planning. Fourth, there was a loss to follow-up as imaging was unable to be reviewed for 45 patients because they were done privately and were unavailable in the institution's PACS to be reviewed. Fifth, the measurements were performed by junior members of the surgical team, which may have introduced variability due to limited experience. However, this was an intentional aspect of the study design, as the algorithm was developed to be simple and reproducible even by less experienced users, ensuring broader clinical utility. We expect the sensitivity and specificity to be even higher when senior surgeons conduct the measurements. Lastly, we did not include intraobserver reliability in this study. Inclusion of repeated measurements by the same observer at different time points would further strengthen future studies to further validate measurement consistency.

Future work can be done with a focus on refining the prediction model with larger datasets, exploring the use of 3-T MRI, and using machine learning to improve the accuracy of our generalized algorithm. Real-time testing in surgical planning will also validate our theory's clinical utility.

 

Scroll to top

Conclusion

This study has demonstrated a practical generalized algorithm with high sensitivity (95.8%) to predict the final ACL graft diameter for single- and double-tendon hamstring grafts consisting of any number of folds using preoperative MRI measurements. The described general prediction theory requires no specialized software and can be reliably performed by junior members of the surgical team, providing valuable information for graft selection and for facilitation of preoperative planning.

 

Scroll to top

 

Scroll to top

Conflict of Interest

The authors declare that they have no conflict of interest.

• References

• 1 Atbaşi Z, Erçin E, Erdem Y, Emre TY, Atilla HA, Parlak A. Correlation between body mass index and quadrupled hamstring tendon autograft size in ACL reconstruction. Joints 2017; 4 (04) 198-201
• 2 Bade D, Malayko G, Johnson L. et al. Single versus double hamstring tendon graft in anterior cruciate ligament reconstruction in the paediatric patient: a single-blind randomised controlled trial study protocol. BMJ Open 2022; 12 (08) e057465
• 3 Banovetz MT, Kennedy NI, LaPrade RF, Engebretsen L, Moatshe G. Biomechanical considerations for graft choice in anterior cruciate ligament reconstruction. Ann Jt 2023; 8: 17
• 4 Brzezinski A, Nasra M, Pfaff W. et al. Five-strand hamstring grafts are biomechanically comparable to four-strand grafts and offer greater diameter for anterior cruciate ligament reconstruction. Arthrosc Sports Med Rehabil 2022; 4 (05) e1731-e1738
• 5 Calvo R, Figueroa D, Figueroa F. et al. Five-strand hamstring autograft versus quadruple hamstring autograft with graft diameters 8.0 millimeters or more in anterior cruciate ligament reconstruction: Clinical outcomes with a minimum 2-year follow-up. Arthroscopy 2017; 33 (05) 1007-1013
• 6 Tang SPK, Wan KHM, Lee RHL, Wong KKH, Wong KK. Influence of hamstring autograft diameter on graft failure rate in Chinese population after anterior cruciate ligament reconstruction. Asia Pac J Sports Med Arthrosc Rehabil Technol 2020; 22: 45-48
• 7 Gupta RK, Poudel N, Dudeja V, Bansal H, Kapoor A. Variability in quadrupled hamstring tendon autograft diameter in North Indian population and its impact on anterior cruciate ligament (ACL) reconstruction outcomes. J Arthrosc Surg Sports Med 2024; 4 (40) 40-42
• 8 Maeda A, Shino K, Horibe S, Nakata K, Buccafusca G. Anterior cruciate ligament reconstruction with multistranded autogenous semitendinosus tendon. Am J Sports Med 1996; 24 (04) 504-509
• 9 Ashford WB, Kelly TH, Chapin RW, Xerogeanes JW, Slone HS. Predicted quadriceps vs. quadrupled hamstring tendon graft size using 3-dimensional MRI. Knee 2018; 25 (06) 1100-1106
• 10 Beyzadeoglu T, Akgun U, Tasdelen N, Karahan M. Prediction of semitendinosus and gracilis autograft sizes for ACL reconstruction. Knee Surg Sports Traumatol Arthrosc 2012; 20 (07) 1293-1297
• 11 Erquicia JI, Gelber PE, Doreste JL, Pelfort X, Abat F, Monllau JC. How to improve the prediction of quadrupled semitendinosus and gracilis autograft sizes with magnetic resonance imaging and ultrasonography. Am J Sports Med 2013; 41 (08) 1857-1863
• 12 Wernecke G, Harris IA, Houang MTW, Seeto BG, Chen DB, MacDessi SJ. Using magnetic resonance imaging to predict adequate graft diameters for autologous hamstring double-bundle anterior cruciate ligament reconstruction. Arthroscopy 2011; 27 (08) 1055-1059
• 13 Hamada M, Shino K, Mitsuoka T, Abe N, Horibe S. Cross-sectional area measurement of the semitendinosus tendon for anterior cruciate ligament reconstruction. Arthroscopy 1998; 14 (07) 696-701
• 14 Leiter J, Elkurbo M, McRae S, Chiu J, Froese W, MacDonald P. Using pre-operative MRI to predict intraoperative hamstring graft size for anterior cruciate ligament reconstruction. Knee Surg Sports Traumatol Arthrosc 2017; 25 (01) 229-235
• 15 Bickel BA, Fowler TT, Mowbray JG, Adler B, Klingele K, Phillips G. Preoperative magnetic resonance imaging cross-sectional area for the measurement of hamstring autograft diameter for reconstruction of the adolescent anterior cruciate ligament. Arthroscopy 2008; 24 (12) 1336-1341
• 16 Sundararajan SR, Rajagopalakrishnan R, Rajasekaran S. Is height the best predictor for adequacy of semitendinosus-alone anterior cruciate ligament reconstruction? A study of hamstring graft dimensions and anthropometric measurements. Int Orthop 2016; 40 (05) 1025-1031
• 17 Calvo R, Meleán P, Figueroa D, Vaisman A, Scheu M, Figueroa F. ¿Existe correlación entre el peso y la talla del paciente con el largo y diámetro del injerto semitendinoso?. Rev Esp Cir Ortop Traumatol 2011; 55 (01) 2-8
• 18 Janssen RPA, van der Velden MJF, van den Besselaar M, Reijman M. Prediction of length and diameter of hamstring tendon autografts for knee ligament surgery in Caucasians. Knee Surg Sports Traumatol Arthrosc 2017; 25 (04) 1199-1204
• 19 Zakko P, van Eck CF, Guenther D, Irrgang JJ, Fu FH. Can we predict the size of frequently used autografts in ACL reconstruction?. Knee Surg Sports Traumatol Arthrosc 2017; 25 (12) 3704-3710
• 20 Pérez-Mozas M, Payo-Ollero J, Montiel-Terrón V, Valentí-Nin JR, Valentí-Azcárate A. Preoperative prediction of autologous hamstring graft diameter in anterior cruciate ligament reconstruction [In Spanish]. Rev Esp Cir Ortop Traumatol (Engl Ed) 2020; 64 (05) 310-317
• 21 Bhamare DS, Sirasala S, Jivrajani P, Nair A, Taori S. Preoperative MRI assessment of hamstring tendons to predict the quadruple hamstring graft diameter in anterior cruciate ligament reconstruction. Cureus 2022; 14 (01) e21753
• 22 Liau ZQG, Ng MSP, Low SSE, Chin BZ, Hui JHP, Kagda FHY. A novel practical method to predict anterior cruciate ligament hamstring graft size using preoperative MRI. Knee Surg Relat Res 2024; 36 (01) 17
• 23 Matsumoto A, Yamaguchi M, Sasaki K, Kanto R. Prediction of graft length by body height in anatomic double-bundle anterior cruciate ligament reconstruction. Asia Pac J Sports Med Arthrosc Rehabil Technol 2018; 12: 17-21
• 24 Schober P, Boer C, Schwarte LA. Correlation coefficients: Appropriate use and interpretation. Anesth Analg 2018; 126 (05) 1763-1768
• 25 Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977; 33 (01) 159-174
• 26 Magnussen RA, Lawrence JTR, West RL, Toth AP, Taylor DC, Garrett WE. Graft size and patient age are predictors of early revision after anterior cruciate ligament reconstruction with hamstring autograft. Arthroscopy 2012; 28 (04) 526-531
• 27 Figueroa F, Figueroa D, Espregueira-Mendes J. Hamstring autograft size importance in anterior cruciate ligament repair surgery. EFORT Open Rev 2018; 3 (03) 93-97
• 28 Thwin L, Ho SW, Tan TJL, Lim WY, Lee KT. Pre-operative MRI measurements versus anthropometric data: Which is more accurate in predicting 4-stranded hamstring graft size in anterior cruciate ligament reconstruction?. Asia Pac J Sports Med Arthrosc Rehabil Technol 2020; 22: 5-9
• 29 Movahedinia M, Movahedinia S, Hosseini S. et al. Prediction of hamstring tendon autograft diameter using preoperative measurements with different cut-offs between genders. J Exp Orthop 2023; 10 (01) 4
• 30 Pinheiro Jr LFB, de Andrade MAP, Teixeira LEM. et al. Intra-operative four-stranded hamstring tendon graft diameter evaluation. Knee Surg Sports Traumatol Arthrosc 2011; 19 (05) 811-815
• 31 Treme G, Diduch DR, Billante MJ, Miller MD, Hart JM. Hamstring graft size prediction: a prospective clinical evaluation. Am J Sports Med 2008; 36 (11) 2204-2209
• 32 Papastergiou SG, Konstantinidis GA, Natsis K, Papathanasiou E, Koukoulias N, Papadopoulos AG. Adequacy of semitendinosus tendon alone for anterior cruciate ligament reconstruction graft and prediction of hamstring graft size by evaluating simple anthropometric parameters. Anat Res Int 2012; 2012: 424158
• 33 Thomas S, Bhattacharya R, Saltikov JB, Kramer DJ. Influence of anthropometric features on graft diameter in ACL reconstruction. Arch Orthop Trauma Surg 2013; 133 (02) 215-218
• 34 Tuman JM, Diduch DR, Rubino LJ, Baumfeld JA, Nguyen HS, Hart JM. Predictors for hamstring graft diameter in anterior cruciate ligament reconstruction. Am J Sports Med 2007; 35 (11) 1945-1949
• 35 Moghamis I, Abuodeh Y, Darwiche A, Ibrahim T, Al Ateeq Al Dosari M, Ahmed G. Anthropometric correlation with hamstring graft size in anterior cruciate ligament reconstruction among males. Int Orthop 2020; 44 (03) 577-584
• 36 Asif N, Ranjan R, Ahmed S, Sabir AB, Jilani LZ, Qureshi OA. Prediction of quadruple hamstring graft diameter for anterior cruciate ligament reconstruction by anthropometric measurements. Indian J Orthop 2016; 50 (01) 49-54

Publikationsverlauf

Eingereicht: 10. April 2025

Angenommen: 30. August 2025

Accepted Manuscript online:
04. September 2025

Artikel online veröffentlicht:
16. September 2025