Analysis of risk factors and construction of nomogram model for arthroscopic single-row rivet repair

Abstract

Background

The factors influencing the clinical outcome of arthroscopic rotator cuff repair are not fully understood.

Purpose

To explore the factors related to the postoperative outcome of arthroscopic single-row rivet rotator cuff repair in patients with rotator cuff injury and to construct the related nomogram risk prediction model.

Methods

207 patients with rotator cuff injury who underwent arthroscopic single-row rivet rotator cuff repair were reviewed. The differences of preoperative and postoperative Visual Analogue Score (VAS) scores and University of California, Los Angeles (UCLA) scores were analyzed and compared. The postoperative UCLA score of 29 points was taken as the critical point, and the patients were divided into good recovery group and poor recovery group, and binary logstic regression analysis was performed. According to the results of multivariate logistic regression analysis, the correlation nomogram model was constructed, and the calibration chart was used, AUC, C-index. The accuracy, discrimination and clinical value of the prediction model were evaluated by decision curve analysis. Finally, internal validation is performed using self-random sampling.

Results

The mean follow-up time was 29.92 ± 17.20 months. There were significant differences in VAS score and UCLA score between preoperative and final follow-up (p < 0.05); multivariate regression analysis showed: Combined frozen shoulder (OR = 3.890, 95% CI: 1.544 ∼ 9.800), massive rotator cuff tear (OR = 3.809, 95%CI: 1.218 ∼ 11.908), More rivets number (OR = 2.118, 95%CI: 1.386 ∼ 3.237), lower preoperative UCLA score (OR = 0.831, 95%CI: 0.704–0.981) were adverse factors for the postoperative effect of arthroscopic rotator cuff repair. Use these factors to build a nomogram. The nomogram showed good discriminant and predictive power, with AUC of 0.849 and C-index of 0.900 (95% CI: 0.845 ∼ 0.955), and the corrected C index was as high as 0.836 in internal validation. Decision curve analysis also showed that the nomogram could be used clinically when intervention was performed at a threshold of 2%∼91%.

Conclusion

Combined frozen shoulders, massive rotator cuff tears, and increased number of rivets during surgery were all factors associated with poor outcome after arthroscopic rotator cuff repair, while higher preoperative UCLA scores were factors associated with good outcome after arthroscopic rotator cuff repair. This study provides clinicians with a new and relatively accurate nomogram model.

Introduction

Rotator cuff (RC) is mainly composed of supraspinatus, infraspinatus, teres minor and subscapularis tendons [1]. Rotator cuff injury or tear is a tear or separation of one or more tendons from the humerus [2]. Rotator cuff injury is the most common cause of shoulder pain. The onset population is usually over 40 years old, with an average age of 55 years [3]. The clinical features of rotator cuff injury are often as follows Shoulder pain, especially at night. And shoulder joint dysfunction, which has caused great problems to patients ‘mental and life in the long run. The current treatment of rotator cuff injury mainly includes conservative treatment and surgical treatment. For patients who have long-term conservative treatment but ultimately ineffective, surgery Treatment has become the only treatment for rotator cuff injury, and Arthroscopic rotator cuff repair (ARCR) is currently the most widely used surgical treatment for patients with rotator cuff injury [4]. Compared to traditional open surgery, Patients suffer less pain, recover faster, and have fewer postoperative complications.

Arthroscopic single-row rivet repair as a rotator cuff in situ suture form of ARCR, compared with double-row rivet repair, the cost of patient surgery is lower, the audience is wider, with the orthopedic surgeon’s understanding of rotator cuff injury continues to deepen, the diagnosis of rotator cuff injury patients is also increasing, the need for arthroscopic surgery treatment of rotator cuff injury patients is also increasing, clinically inevitable will appear some ARCR postoperative shoulder function recovery is not ideal patients. However, the factors affecting the clinical outcome of arthroscopic single-row rivet rotator cuff repair are not fully understood, and most studies have always focused on the effects of age, body mass index, diabetes mellitus, fatty infiltration of subscapularis and infraspinatus muscles, duration of symptoms, bone mineral density, tear length, tear width, tear area, retraction, critical shoulder angle, scapulohumeral septum, distance from tendon junction to glenoid, operation time, and long head of biceps tendon (LHBT) surgery on postoperative rotator cuff retear and healing [5, 6]. However, rotator cuff retear and rotator cuff healing are not always equivalent to patient outcomes such as pain and function. Ball CM et al. [7] used magnetic resonance arthrography (MRA) evaluation of patients at least six months after arthroscopic rotator cuff repair found that even though interstitial division and/or delamination occurred in 65.6% of patients, significant partial thickness tears were observed in 14.6% of patients (> 50%), which did not affect the outcome (ASES score and patient satisfaction). Oh JH et al. [8] was also found that patients with poor rotator cuff integrity after ARCR had better shoulder function improvement. Therefore, it seems to be of great significance to study the factors that influence the clinical outcome of ARCR in patients with rotator cuff injury. At the same time, there is no accurate prediction in clinic Patients with rotator cuff injury. The evaluation and prediction model of ARCR postoperative effect may bring great benefits to clinicians and patients。.

This article mainly reviews the diagnosis of rotator cuff injury in our department from January 2016 to August 2021. Patients undergo ARCR with single-row rivet repair technique. Clinical data of patients were collected and analyzed. (1) To explore the factors related to the postoperative outcome of arthroscopic single-row rivet rotator cuff repair in patients with rotator cuff injury. (2) To observe the postoperative outcome of arthroscopic single-row rivet rotator cuff repair in elderly patients over 65 years old. (3) To develop and validate the prediction model of postoperative adverse risk in patients with rotator cuff injury treated with ARCR.

Materials and methods

Inclusion criteria

• 1) All patients who underwent arthroscopic single-row rivet rotator cuff repair in our hospital from January 2016 to August 2021;

• 2) all patients diagnosed with rotator cuff tear by arthroscopy;

• 3) Patients aged 18 years and older.

Exclusion criteria

• 1) combined shoulder fractures included glenoid fractures, humeral fractures, clavicle fractures;

• 2) shoulder with glenoid labrum injury, severe osteoarthritis, shoulder joint tumor;

• 3) the shoulder has a history of surgery.

Data acquisition

All the patients were given DR scan of shoulder joint and supraspinatus muscle, and MR scan of shoulder at the same time. The number of rivets in operation was recorded, whether it was full-thickness tear or partial tear, whether it was a huge tear of rotator cuff, and the specific ways of treatment (no treatment, simple cutting, cutting and fixing, cutting and fixing). The blood lipid of patients with fasting blood test, TC ≥ 6.19 mmol/L or TG ≥ 2.27 mmol/L was defined as dyslipidemia. A traumatic rotator cuff tear is defined when the onset of shoulder pain can be attributed to a specific event considered sufficient to cause a rotator cuff tear [9];Rotator cuff tear defects greater than 5 cm or involving two or more rotator cuff tendons were defined as massive rotator cuff tears [10]. According to the score of UCLA at the last follow-up, the patients were divided into two groups, ≥ 29 as good recovery group, < 29 as poor recovery group.

Surgical method

All the patients were placed in lateral decubitus position, and then disinfected and draped routinely after general anesthesia. The affected limb was abducted about 60°, and the traction force was equal to the gravity of 2 bags of 3 L saline solution. The rotator cuff was fixed by single-row repair technique with thread rivet through the rotator cuff. The LHBT was cut off from the insertion of the superior glenoid tuberosity or cut off and fixed at the intermuscular sulcus. For individual huge rotator cuff injuries that were difficult to repair with rivets alone, the LHBT was cut off at the intermuscular sulcus and the proximal end was transferred to the footprint area to reconstruct part of the joint capsule and bridge the rotator cuff.

Postoperative treatment

All patients were fixed in neutral position with shoulder joint abduction 45° by brace after operation. Passive functional rehabilitation exercises including flexion, abduction, internal rotation and external rotation were performed 2 weeks later. Active functional exercises and active wall climbing exercises were performed 6 weeks later.

Clinical efficacy evaluation

Visual analog scale (VAS) pain score, ranging from 0 to 10 points, in this range, patients choose the score according to the subjective feeling of pain. UCLA score, a total of 35 points, divided into excellent, good, poor three grades, excellent between 34 points and 35 points; Good between 29 and 33 points; difference less than 29 points [11].

Statistical method

The data were analyzed by SPSS 22.0 software. The quantitative data were described by mean ± standard deviation, and the counting data were represented by the number of cases or percentage. The data of numerical variables were evaluated by shapiro-wilke test, and the paired sample t test was used to compare the data before and after the normal distribution Wilcoxon rank sum test was used when the normal distribution was not met. The sex, age, course of disease, trauma, frozen shoulder, hypertension, diabetes, dyslipidemia, the number of rivets used in operation, full-thickness tear or partial tear, massive rotator cuff tear, the specific treatment of LHBT, preoperative VAS score, preoperative UCLA score were analyzed by univariate binary logistic regression analysis. In this paper, the cutting and fixing and the cutting and fixing in the LHBT processing mode are combined into one cutting and fixing mode, that is, the four processing modes of LHBT are combined into three types (untreated, simple cutting, cutting and fixing). When p < 0.05, it is considered to have statistical difference. R software (version 4.2.1) is used to construct a risk nomogram model for postoperative poor effect of patients with rotator cuff injury treated with ARCR, and the calibration curve, ROC curve, decision analysis curve. Calibration curve represents the calibration between the nomogram prediction and the actual result. In a perfectly calibrated nomogram, the calibration curve will fall on the 45-degree diagonal of the calibration curve [12]. Calculate the area under the curve (AUC) of the ROC and the calculation model C-index, for predicting performance, C-index similar to AUC, but more suitable for reviewing data [13], higher C-index values mean better prognostic models, The nomograms were validated by internal self-random sampling (1,000 bootstrap samples) to calculate the relatively-adjusted C-index. Decision curve analysis was performed to determine the clinical utility of the nomograms for the risk of poor outcome after ARCR in patients with rotator cuff injuries by quantifying the net benefit at different threshold probabilities in a cohort of patients with rotator cuff injuries treated with ARCR [14].

Results

Patient profile

A total of 519 patients underwent arthroscopic rivet rotator cuff repair in our hospital, of which 280 met the inclusion and exclusion criteria, 73 were excluded from follow-up, and 207 patients (42 patients ≥ 65 years old) were followed up, including 79 males and 128 females, with an average age of 57.35 ± 8.42 (27 ∼ 76) years, with a total of 207 shoulders (153 right shoulders and 54 left shoulders); The mean course of disease was 12.89 ± 30.67 (0.03–240) months. 96 cases were traumatic tears and 111 cases were non-traumatic tears 38 cases had hypertension, 16 cases had diabetes, 44 cases had dyslipidemia, 21 cases had frozen shoulder. The preoperative VAS score was 6.08 ± 1.26, and the preoperative UCLA score was 12.83 ± 2.74 (the preoperative VAS score was 5.88 ± 1.25, and the preoperative UCLA score was 12.86 ± 3.11 in patients ≥ 65 years old). There were 32 cases of massive rotator cuff tear and 175 cases of non-massive rotator cuff tear 131 cases were untreated and 76 cases were treated (35 cases were cut off, 37 cases were cut off and fixed, 4 cases were cut off and fixed); The average number of rivets was 2.26 ± 1.32 (72 cases with 1 rivet, 63 cases with 2 rivets, 41 cases with 3 rivets, 18 cases with 4 rivets, 7 cases with 5 rivets, 3 cases with 6 rivets and 3 cases with 7 rivets) (see Table 1).

Table 1 Correlated variables (n = 207)

Clinical outcomes

The mean follow-up time was 27.93 ± 16.18 (6–72) months. The mean postoperative VAS score was 1.00 ± 1.09 and the mean postoperative UCLA score was 30.00 ± 4.88 in all patients (0.83 ± 1.08 and 29.40 ± 6.21 in patients ≥ 65 years). There was a significant difference between the preoperative VAS score and the final follow-up VAS score in all 207 patients and 42 elderly patients ≥ 65 years (P < 0.05). There was a significant difference between the UCLA score before surgery and the UCLA score at the last follow-up (P < 0.001) (see Tables 2 and 3).

Table 2 Comparison of preoperative and postoperative pain and function in all patients

Table 3 Comparison of preoperative and postoperative pain and function in elderly patients ≥ 65 years old

Univariate and multivariate logistics regression analysis

According to the UCLA score at the last follow-up, the postoperative efficacy of patients was divided into two groups, ≥ 29 was divided into good postoperative recovery group, < 29 was divided into poor postoperative recovery group, and finally good recovery group (n = 155) and poor recovery group (n = 52) were obtained. The corresponding variables of the two groups are shown in Table 4 below. Univariate analysis was performed on age, gender, course of disease, traumatic rotator cuff tear, history of hypertension, diabetes mellitus, dyslipidemia, frozen shoulder, full-thickness tear or partial tear, number of rivets used during operation, massive rotator cuff tear, treatment method of LHBT (untreated, simple cut off, cut off fixation), preoperative VAS score, preoperative UCLA score and other influencing factors by using binary logistics regression analysis, showing frozen shoulder (OR = 4.574, 95%CI: 1.331 ∼ 15.721), massive rotator cuff tear (OR = 15.750, 95%CI: 6.426 ∼ 38.600), LHBT cut-off (OR = 2.702, 95%CI: 1.139 ∼ 6.408) and transection fixation (OR = 6.825, 95%CI: 3.119 ∼ 14.936), rivet number (OR = 3.028, 95%CI: 2.134 ∼ 4.297), preoperative UCLA score (OR = 0.753, 95%CI: 0.652 ∼ 0.869) were the influencing factors of good recovery in patients with rotator cuff injury treated with ARCR (P < 0.05), the univariate factors with significant difference were screened out and the multivariate logistic regression analysis was performed again. Multivariate analysis showed that the combined frozen shoulder (OR = 4.574, 95% CI: 1.331 ∼ 15.721), massive rotator cuff tear (OR = 3.809, 95%CI: 1.218 ∼ 11.908), rivet number (OR = 2.107, 95%CI: 1.379 ∼ 3.219), preoperative UCLA score (OR = 0.831, 95%CI: 0.704 ∼ 0.981) were the influencing factors of the recovery of postoperative curative effect in patients with rotator cuff injury treated by ARCR (P < 0.05), in which the combination of frozen shoulder, huge rotator cuff tear, and the increase in the number of intraoperative rivets are the influencing factors for poor recovery of curative effect after arthroscopic rotator cuff repair, while the higher preoperative UCLA score is the influencing factor for good recovery of curative effect after arthroscopic rotator cuff repair (see Table 5).

Table 4 Differences in demographic and clinical characteristics between good and poor recovery groups

Table 5 Univariate and multivariate analysis of influencing factors

Nomogram of the risk of poor postoperative outcomes in patients with rotator cuff injuries for whom ARCR treatment was developed

Based on multiple factors logistics Regression analysis results, whether or not Combined frozen shoulder, whether it is a huge rotator cuff tear, the number of rivets used during surgery, and the preoperative UCLA score are included in the prediction model of poor postoperative effect of patients with rotator cuff injury treated with ARCR, and are represented by a nomogram (as shown in Fig. 1). The nomogram simply adds the scores above the nomogram corresponding to each influencing factor to obtain the total score, and then corresponds the total score to the probability of adverse risk below, so that the probability of adverse risk after rotator cuff repair can be quickly estimated.

Fig. 1

Nomogram of risk of poor postoperative outcome in patients with rotator cuff injuries treated with the developed ARCR

Prediction performance and validation of nomogram model

The calibration curve of the nomogram used to predict the risk of poor postoperative outcome in patients with rotator cuff injuries treated with ARCR basically fits the ideal model curve in this cohort (as shown in Fig. 2), indicating that the model has good prediction accuracy. At the same time, the ROC curve of the nomogram of the postoperative adverse risk of rotator cuff injury patients treated with ARCR was drawn (as shown in Fig. 3), and the corresponding area under the curve (AUC) value was calculated. By calculating the AUC value of 0.849, in order to improve the clinical reference value and more intuitively reflect the continuous variables in the nomogram model to predict the postoperative adverse risk of ARCR, the AUC value and cutoff value corresponding to the two variables of preoperative UCLA score and the number of rivets used during the operation were calculated (see Table 6). The cutoff value of the number of rivets was calculated to be 3.5, and the cutoff value of preoperative UCLA score was 12.5. This shows that when the preoperative UCLA score is less than 12.5, the risk of postoperative adverse ARCR will increase. When the number of rivets used during the operation exceeds 3.5, the risk of postoperative adverse ARCR will increase. Finally, the C-index of the predicted nomogram is calculated to be 0.900 (95% CI: 0.845 ∼ 0.955), and the corrected C-index was 0.836 after internal validation by self-random sampling validation, which indicated that the nomogram model had good discriminative ability and showed good predictive ability in predicting the risk of poor postoperative outcome in patients with rotator cuff injury treated with ARCR.

Fig. 2

Poor postoperative outcomes in patients with rotator cuff injuries treated with ARCR in cohort Risk Calibration curve of nomogram

Note: The dashed line represents the ideal prediction model. The solid line represents the actual performance of the nomogram. When the solid line is closer to the dashed line, the prediction is indicated the higher the accuracy performance

Fig. 3

ROC curve of nomogram of risk of poor postoperative outcome in patients with rotator cuff injury treated with ARCR

Note: AUC is shaded area

Table 6 AUC values and cut-off values of continuous variables in the Prediction Model

Decision curve analysis

The nomogram model was made clinically useful by constructing a decision curve (DCA) for the nomogram of the risk of poor outcome after arthroscopic rotator cuff repair, which showed that, if the threshold probabilities for the patient and clinician were > 2% and < 91%, the nomogram predicted that the risk of rotator cuff injury patients treated with ARCR increased more than the risk of intervention with all or no intervention. Within this range, the net benefits were comparable according to the nomogram model of the risk of poor outcome after rotator cuff injury patients treated with ARCR, but there were several overlaps (Fig. 4).

Fig. 4

Decision curve analysis of nomogram of risk of poor postoperative outcome in patients with rotator cuff injuries treated with ARCR

Discussion

Arthroscopic techniques have become the gold standard for minimally invasive surgery in patients with rotator cuff tears who fail to respond to conservative treatment [15]. Compared with younger patients, rotator cuff tears in older patients are usually larger and accompanied by more tendon retraction. Older age and comorbidities are also thought to reduce the healing ability of the tendon-bone interface. It is reported that patients over 65 years of age the recurrence rate is 57% [16]. Therefore, this article specifically reviewed the postoperative outcomes of the elderly population ≥ 65 years old. We finally found that ARCR is satisfactory for patients with rotator cuff injury, whether it is postoperative pain or postoperative shoulder function, including only for the 65-year-old elderly population. The clinical results are equally impressive, and we found through regression analysis that the age of the patient over 65 years old is not a factor affecting the poor postoperative efficacy of ARCR. Similar to us, Razmjou H et al. [17] was also found that increasing age was not a negative predictor of patient satisfaction after rotator cuff repair. Our study also found that frozen shoulder, huge rotator cuff tear, number of rivets, and preoperative UCLA score were all influencing factors for poor shoulder function after ARCR in patients with rotator cuff injury. Rotator cuff injury with frozen shoulder, rotator cuff injury with huge rotator cuff injury, and rotator cuff injury with a large number of rivets required for surgical repair of rotator cuff significantly increased the risk of poor postoperative outcome after ARCR, while preoperative UCLA score of shoulder joint significantly reduced the risk of poor postoperative outcome in patients with rotator cuff injury treated arthroscopically. Dai Fei et al. [18] and Haviv B et al. [19] the study also found that larger rotator cuff tear size was a predictor of poor shoulder function after arcr. Malavolta et al. [20] and Jenssen et al. [21] studies have also shown that a higher preoperative shoulder function score is a better predictor of shoulder function improvement after ARCR, which is similar to our findings. Patients with massive rotator cuff tears generally have significant preoperative pain and poor function because of massive tear area or cumulative 2 or more tendons. For patients with massive rotator cuff tears, ARCR takes longer and more difficult than general tears, and sometimes the rotator cuff cannot completely cover the footprint area. The more rivets used in the operation may also mean that patients have larger and more complex rotator cuff injuries. Therefore, even experienced orthopedic surgeons are difficult to grasp the postoperative effect in patients with massive rotator cuff injuries. At the same time, Lee, Y. S et al. [22] was also found initial rotator cuff tear size to be an independent risk factor for rotator cuff retear in a cohort study. Therefore, clinicians should be vigilant for large-scale or multi-tendon rotator cuff injuries found in preoperative MRI, do adequate preoperative planning, and develop personalized rehabilitation programs for patients. Frozen shoulder is a musculoskeletal disease characterized by pain and limited mobility of the glenohumeral joint [23]. It usually occurs in middle-aged women. It is a sterile inflammation of the soft tissue around the shoulder joint. It has a certain self-limiting nature. The etiology is not clear. Clinically, some patients with rotator cuff injury have severe shoulder function damage, so they are finally diagnosed. For rotator cuff injury, it is often combined with frozen shoulder. Studies have shown that preoperative shoulder stiffness is a risk factor for postoperative stiffness [24]. Jeong JY et al. [25] a retrospective cohort study also found that rotator cuff tears combined with frozen shoulder had a negative impact on most functional outcomes (including ROM) at 6 months and 1 year after ARCR, so we should pay enough attention to patients with rotator cuff injuries combined with frozen shoulder before surgery. For example, patients with rotator cuff injuries combined with frozen shoulder should be more careful in preoperative diagnosis to avoid missed diagnosis, and patients should be more active in rehabilitation training after surgery.

In order to better help clinicians predict the clinical outcome of patients with rotator cuff injury who need ARCR treatment, we developed and validated a nomogram prediction model for the risk of poor postoperative outcome in patients with rotator cuff injury treated with ARCR. This is also the first study to apply nomograms to predict the prognosis of patients with rotator cuff injury after arthroscopic treatment. The nomogram has been used as a prognostic tool for tumor prognosis. It not only has a very friendly digital interface, but also has higher accuracy and clearer prognosis judgment, which can help patients make better clinical decisions [26]. Through multivariate regression analysis, we finally combined frozen shoulder, huge rotator cuff tear, number of rivets, preoperative UCLA score 4 factors identified as our nomogram model composition, these factors are easy to obtain, clinicians only need to calculate the total score by simply adding the variable scores of each factor, and then find the corresponding risk value to predict the risk of poor ARCR results in patients with rotator cuff injury, which means that the use of nomogram becomes easier to operate, At the same time, it is beneficial for clinicians to use this model to predict the postoperative adverse risk of ARCR patients. In addition, this study provides a relatively accurate prediction tool for ARCR patients with rotator cuff injury. Internal validation in the cohort shows good discrimination and calibration. In particular, the high adjusted C index in our own random sample validation indicates that the nomogram can be widely and accurately applied to other samples, and predicting the risk of adverse ARCR in patients with rotator cuff injury in advance can better help clinicians to take interventions and allow patients to actively adjust their lifestyle, which is a benefit to both patients and doctors.

Limitations of this study: First, this study is a single-center study, and the selected samples may not be representative enough. Second, because this study is a retrospective study, some influencing factors cannot be obtained, so the analysis of risk factors affecting postoperative adverse arthroscopic rotator cuff repair does not include all potential influencing factors. In addition, the nomogram model constructed in this study is only internally validated, and more external validation support is still needed.

Disclosure statement

The authors report no conflicts of interest.

Conclusion

Arthroscopic single-row rivet rotator cuff repair resulted in significant improvement in shoulder pain and function (including in the elderly population ≥ 65 years of age). Coexisting frozen shoulders, massive rotator cuff tears, and increased number of rivets during surgery were all factors associated with poor outcome after arthroscopic rotator cuff repair, while higher preoperative UCLA scores were factors associated with good outcome after arthroscopic rotator cuff repair. This study provides clinicians with a new, relatively accurate nomogram model to accurately assess the risk of poor outcome in patients with rotator cuff injuries requiring arthroscopic rotator cuff repair at the beginning of treatment. By estimating individual risk, clinicians can proactively take more favorable interventions before, during, and after arthroscopic rotator cuff repair.