Analysis of complications and revisions after spine surgery in 270 multiple myeloma patients with spinal involvement

Patients with multiple myeloma (MM) are at increased risk of infections and suffer from poor bone quality due to their disseminated malignant bone disease. Therefore, postoperative complications may occur following surgical treatment of MM lesions. In this study, we aimed to determine the incidence of postoperative complications and retreatments after spinal surgery in MM patients. Additionally, we sought to identify risk factors associated with complications and retreatments. Retrospective cohort study. In total, 270 patients with MM who received surgical treatment for spinal involvement between 2008 and 2021 were included. The incidence of perioperative complications within 6 weeks and reoperations within 2.5 years and individual odds ratios for factors associated with these complications and reoperations. Data were collected through manual chart review. Hosmer and Lemeshow’s purposeful regression method was used to identify risk factors for complications and reoperations. The median age of our cohort was 65 years (SD = 10.8), and 58% were male (n = 57). Intraoperative complications were present in 24 patients (8.9%). The overall 6-week complication rate after surgery was 35% (n = 95). The following variables were independently associated with 6-week complications: higher Genant grading of a present vertebral fracture (OR 1.41; 95% CI 1.04–1.95; p = .031), receiving intramuscular or intravenous steroids within a week prior to surgery (OR 3.97; 95% CI 1.79–9.06; p = .001), decompression surgery without fusion (OR 6.53; 95% CI 1.30–36.86; p = .026), higher creatinine levels (OR 2.18; 95% CI 1.19–5.60; p = .014), and lower calcium levels (OR 0.58; 95% CI 0.37–0.88; p = .013). A secondary surgery was indicated for 53 patients (20%), of which 13 (4.8%) took place within two weeks after the initial surgery. We additionally discovered factors associated with retreatments, which are elucidated within the manuscript. The goal of surgical treatment for MM bone disease is to enhance patient quality of life and reduce symptom burden. However, postoperative complication rates remain relatively high after spine surgery in patients with MM, likely attributable to both inherent characteristics of the disease and patient comorbidities. The risk for complications and secondary surgeries should be explored and a multidisciplinary approach is crucial.


Introduction
Multiple myeloma (MM) is the second most common hematological malignancy and arises from the antibodyproducing plasma cells inside the bone marrow [1].MM can occur in any part of the skeleton but typically affects the spine, ribs, skull, and pelvis.Pathological vertebral compression fractures (VCFs) are common in MM [2].Approximately 30-60% of patients present with a VCF at the time of diagnosis, and over 60% of patients develop new VCFs later on [3].VCFs can cause intractable back pain, reduced mobility, and spinal cord compression, potentially leading to rapid neurologic decline, thereby severely increasing the burden of disease [4].
MM patients are primarily treated with chemotherapy, stem cell transplantation, and immunomodulatory drugs under the supervision of a (hemato)oncologist.For some patients however, surgical intervention may be warranted [5].Indications for surgery include gross spinal instability, compression of neural elements, and mechanical or neuropathic pain caused by an (impending) fracture or epidural tumor extension [6,7].Previous studies have shown that surgical intervention can improve patients' neurologic function and quality of life [7,8].However, MM patients are at an increased risk of (postoperative) infections due to a multifactorial immunodeficiency caused by the disease itself and due to the treatment regimens [9].Furthermore, they suffer from overall poor bone quality due to disseminated malignant bone disease.Specifically, the osteoclast differentiation promoted by monoclonal immunoglobulins can lead to weakening of bones and consequently a higher rate of pseudoarthrosis and hardware failure [10].Therefore, it is critical to ensure that the benefits of surgical management outweigh the associated risks.
To date, little is known about the incidence of and predisposing factors for postoperative complications and retreatments after spinal surgery in MM.Therefore, in this study, we aimed to determine the incidence of postoperative complications within 6 weeks and the incidence of retreatments within 2.5 years after spinal surgery in MM patients.Additionally, we sought to identify factors associated with such complications and retreatments.

Study design
This retrospective cohort study was approved by our institutional review board (registration number 2018P000688).The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines were used for adequate reporting [11].

Patient selection
We identified 1,850 potentially eligible patients through the ICD-9 and ICD-10 codes for 'Multiple Myeloma' (203.0 and C90), and 'Secondary malignant neoplasm of bone and bone marrow' (198.5 and C79.5) or 'Fracture of vertebrae' (733.13, 805.0, 805.1, 805.2, M48.5, M84.5, S12.0, S22.0, and S32.0).We manually screened medical records on surgical treatment.The following inclusion criteria were applied: (1) adult patients with MM or plasmacytoma, (2) received spinal surgery for spinal MM-lesions between January 2010 and June 2021 in one of two tertiary academic medical centers that operate under a single umbrella institution.Patients were excluded if (1) no information was available, (2) if they had a history of another (major) malignant or neurologic disease for which they were treated, or (3) if they had prior surgery in the same area.Two-hundred-seventy patients met these criteria.A flowchart of selection can be found in the Appendix, Fig. 1.

Variables
Our outcome measures were (1) postoperative complication within 6 weeks after surgery, (2) retreatments within 2.5 years, and (3) risk factors associated with complications and retreatments expressed in individual odds ratios and subdistribution hazard ratios, respectively.The following baseline characteristics from patients' medical records were extracted: age, gender, body mass index (BMI, in kg/m 2 ), International Staging System (ISS) disease stage [12], ageadjusted Charlson comorbidity index (ACCI) [13], Eastern Cooperative Oncology Group (ECOG) performance status [14], back pain (yes/no), chemotherapy regimen, administration of perioperative steroids (minimum 6-10mg every 6-8 h intramuscular or intravenous within a week from surgery), treated area (cervical, thoracic, lumbar, sacral, or combined), and discharge disposition.Radiological variables were spread of myeloma disease (local: 1 affected vertebrae, multiple: 2-5 affected vertebrae, disseminated: widespread disease), Epidural Spinal Cord Compression (ESCC) scale [15], number of vertebral levels affected by the ESCC (≤ 2 or ≥ 3), highest Spinal Instability Neoplastic Score (SINS) score of treated area [16], number of vertebral fractures in the treated area, highest Genant fracture grading [17].Surgical variables were type of surgery, estimated blood loss during surgery, duration of surgery, and intraoperative complications.Date of death was extracted from the charts (linked to the National Death Index), and for non-deceased patients, the last follow-up date was set on the date of their last hospital visit.

Surgical complications and retreatments
Postoperative complications were graded according to the Clavien-Dindo classification, which is based on the therapy used to treat or correct a specific complication [18].Grade I complications were defined as any deviation from the normal postoperative course without the need for pharmacological treatment (with the exception of analgesics, antipyretics, diuretics, antiemetics, and electrolytes).Grade II complications were defined as postoperative deviations requiring any pharmacological treatment other than those allowed for grade I complications.Blood transfusions are also included in this grade.Grade III complications were defined as postoperative deviations requiring either surgical, endoscopic or radiological intervention.Grade IV complications were defined as life-threatening postoperative deviations warranting admission to the Intensive Care Unit.Grade V complications were defined as any postoperative deviations that led to the death of a patient [18].We recorded all postoperative complications separately and additionally dichotomized complications into minor complications (Grade I or II) and major complications (Grade III, IV, or V).Intraoperative complications were also recorded but were not graded according to the Clavien-Dindo classification.Retreatments were defined as unplanned surgical interventions directly related to the initial surgery or unplanned radiation treatments to the operated area.Patients who died within 6 weeks from a cause that was unknown or presumably unrelated to the initial surgery or complications were excluded (n = 2).

Statistical analysis
Categorical variables were presented as count (%) and continuous data as median with interquartile ranges (IQR) as they were non-parametric after inspecting histograms.Baseline differences were assessed using Chi-square tests or Fisher's exact tests for categorical variables and Mann-Whitney U tests for continuous variables.Independent risk factors associated with postoperative complications were assessed using multivariable regression, and the Fine-Gray competing risk regression method was applied for retreatment analyses to address the situation where there is a competing risk of death within 2.5 years [19].The purposeful selection approach described by Hosmer and Lemeshow was used to select variables of interest, where covariates are removed from the model if they are non-significant and not a confounder [20].Factors with a p-value below 0.25 in bivariate analysis are entered in a multivariable regression model.Next, variables with a p-value below 0.10 in the multivariable model are entered into a new model together with variables that are considered clinically important.Then, other covariates are added back to the model one by one, to identify potential confounders that, by themselves, are not significantly related to the outcome but make an important contribution in the presence of other variables.A change in a parameter estimate above 20% indicates that the excluded variable was important in the sense of providing a needed adjustment for one or more of the variables remaining in the model [20].The steps of this iterative process of deleting, refitting, and verifying are outlined in the Appendix, Tables 8, 9, and 10.Patients with missing values were excluded from the analysis; however, laboratory variables with more than 30% missing data were not included in the multivariable analyses due to the substantial loss of patients when retaining these factors (Alkaline Phosphatase

Baseline characteristics
Among 270 included patients, the mean age was 65 years (SD 10.8), and 157 (58%) were male (Table 1).Most patients had an ECOG score of 2 or lower (n = 235, 91%), and an ASIA impairment score E (n = 176, 66%).The median time from diagnosis to surgery was 57 months (IQR 0-438).Prior to surgery, 259 (96%) patients experienced back pain and 106 (41%) had neurologic symptoms.One hundred thirty-nine (57%) patients were started on chemotherapy prior to surgery, and 76 (29%) received perioperative steroids.The thoracic spine was the most commonly treated region (n = 133, 49%) and 88 patients had an ESCC grade of 3 (37%).One hundred seventythree patients (71%) had a potentially unstable vertebra, and 59 had an unstable vertebra according to the SINS scoring system, warranting surgical consultation [16].There were 196 fractures in our cohort, of which 133 (52%) were classified as severe.One hundred and thirty patients were treated with a vertebral augmentation (48%), 64 had a decompression with fusion (24%), 55 patients had a corpectomy with fusion (20%), and 21 had a posterior decompression without fusion (8%).Other surgical variables and laboratory data are displayed in the Appendix, Tables 6 and 7, respectively.Patients receiving less invasive surgeries (augmentations and decompression surgery without fusion) were, compared with patients receiving more invasive procedures, generally older at the time of surgery (67 vs 64 years, p = 0.061), in a higher ISS disease stage (30% ISS 3 vs 11% ISS 3, p < 0.001),

Discussion
Clinical management of MM patients with spinal involvement remains challenging due to a lack of clear treatment guidelines.Patients with MM are more susceptible to postoperative complications as compared to other spinal metastases, due to their immunodeficient state, disseminated bone disease, and diminished bone healing [9,10].This study describes a series of spinal procedures in 270 patients with MM and shows that 35% of patients suffered from postoperative complications within 6 weeks after surgery and 20% of patients received unplanned additional surgical/radiotherapy treatment.As expected, due to the inherent characteristics of the disease, this is slightly higher than complication rates found in patients undergoing surgery for spinal metastases originating from solid tumors, where reported rates vary between 17 and 32% [21][22][23].Furthermore, risk factors were identified for both postoperative complications and retreatments that can assist in improving patient-centered treatment strategies in a multidisciplinary approach.MM causes hypogammaglobulinemia, T-cell dysfunction, and granulocytopenia, which results in a higher risk of infectious complications [9], [24].This reflects in our cohort as the most common complication was related to infection (either systemic or wound infection).Additionally, patients with MM are specifically more at risk for thromboembolic complications compared to the general population since MM  results in a hypercoagulable state [24].Our results show that 11% of patients with a complication had a thromboembolic complication.Furthermore, it has been well established that preoperative high-dose steroid therapy is associated with an increased risk of postoperative complications.Patients receiving steroids perioperatively are especially vulnerable to infectious complications such as surgical site infections, wound dehiscence, and pneumonia [25,26].It has been shown that patients receiving steroids were more than twice as likely to experience pulmonary embolism [27], and have an almost fourfold higher mortality risk [28].Our results underscore these findings, as the administration of steroids was strongly associated with postoperative complications.Since steroids play an integral role in the treatment of MM, it is important that patients are counseled preoperatively about the higher likelihood of complications.Consideration of perioperative infection mitigation strategies and/or antithrombotic prophylaxis is essential.Postoperative monitoring of patients should be done with extra vigilance.
In contrast to common expectations, our data showed that more patients treated with vertebral augmentation reported severe pain within the first six weeks postoperatively when compared to those who underwent invasive surgery.One might anticipate that the more invasive procedures would result in more pain.However, this was not reflected in our findings.The explanation for this counterintuitive result may lie in the nature of the procedures themselves.While cement augmentation will often address the fracture pain itself, residual discomfort could persist due to unresolved mechanical stresses and global spinal imbalances [29].Our findings suggest that the etiology of pain in these patients is complex and likely multifactorial, extending beyond the fracture itself.Therefore, to effectively manage pain, some patients may require a more comprehensive approach that includes spinal instrumentation.Vertebral augmentation plays a vital role in managing MM patients with spinal involvement, but our results highlight the importance of a tailored treatment approach that considers each patient's individual needs and circumstances.
Several laboratory values seemed to play an important role in both complication risk analysis and retreatment risk analysis.Elevated serum creatinine levels before surgery have been associated with increased complication rates and are often attributed to muscle wasting or poor nutritional status [30].Elevated serum creatinine levels are also indicative of active myeloma disease and renal failure, which is a common complication in MM and has a strong impact on both morbidity and mortality [31,32].Also in this study, elevated creatinine levels are strongly associated with postoperative complications.Therefore, it is important to be alert to the increased complication risk when a patient with elevated creatinine levels or advanced disease stage is indicated for surgery.
Adjacent-level fractures were the most common reason for revision surgery.These fractures occurring after augmentation procedures or long spinal segment instrumentation have been discussed in the literature for patients with MM, osteoporosis, or after scoliosis surgery [33][34][35][36].The procedures fuse the target segments and thereby cause increased stress on the adjacent vertebrae.Specifically, patients with decreased bone density (i.e., patients with MM with widespread bone destruction), have an increased risk of adjacent-level fractures after surgery [36].Caution is warranted when patients with widespread MM are indicated for surgical intervention.Measures to prevent these fractures, like augmentation of the first mobile vertebra or establishing less rigid constructs can help to prevent these in multilevel fusion surgeries [37], [38], and could be considered in myeloma surgery as well.
Decompression surgery without fusion was found to be a risk factor for postoperative complications.This is in contrast to previous literature, showing that the incidence of complications is lower after decompression surgery when compared to spinal fusion surgery [39,40].The use of spinal instrumentation requires more extensive tissue disruption and often results in prolonged surgical time and increased blood loss [26].Many studies have proven that both surgical time and blood loss are correlated with a higher risk of perioperative complications [41,42].We assume that our findings are a result of a selection bias where surgeons deliberately chose not to perform invasive spinal fusions on sicker patients, as they were already at a higher risk of developing complications.It is worth noting that the different types of surgical intervention, be it vertebral augmentation or open surgery, may have different complication profiles.We aimed to account for this in our regression analysis by including the type of surgery as a covariate.While it is plausible that separate analyses for each surgical type might provide more specific insights, such an approach would substantially reduce our sample size for each group, limiting our ability to analyze all variables of interest comprehensively.We aimed to offer a clearer understanding of the unique risks associated with each procedure by separating them in the provided table.The same goes for the seemingly protective effect of doing a corpectomy for retreatment risk.When a patient was in a good condition and therefore selected for a more invasive surgical approach, they may have been less likely to experience problems postoperatively because of their general fitness level before surgery.From the results, we can see that patients who were in a worse disease stage and had their MM diagnosis for a longer period of time were generally treated with less invasive surgeries like augmentations and decompression surgery without fusion.This is also reflected by the effect of elevated calcium levels.Similar to elevated creatinine levels, high calcium levels are an indicator of further disease progression in MM.In later stages of the disease, induction of osteoclasts causes bone destruction which in turn leads to hypercalcemia.Hypercalcemia is a sign of further disease progression and is associated with renal failure, anemia, and poor survival [43].It is present in 30% of MM patients at the time of diagnosis [44].It is expected that patients with elevated calcium levels would experience more complications.However in our study, the opposite was true.We interpret this as a reflection of the selection bias for less invasive treatments rather than a reflection of the true effect of elevated calcium levels.

Limitations
Our study had several limitations.Firstly, due to the retrospective nature of the study, complications were assessed by screening discharge summaries and postoperative visit notes.Minor complications (Clavien-Dindo grade 1 and 2) might not have always been reported in such notes.It is also possible that some complications and reoperations were treated in other facilities and did not end up in our institution's medical records.This might have led to an underestimation of true complication and retreatment rates.Furthermore, because of the retrospective design, when a revision surgery was done 2 years from the initial surgery, it was difficult to distinguish whether this was done because of (instrument) failure or disease progression if there was no evident myelomatous disease in that vertebra initially.Second, the retrospective design of our study might have introduced some selection bias.The choice to operate on a patient and the subsequent type of surgery depended on the surgical indication and on the surgeon and we did not document the indications that led to these decisions.As mentioned before, surgeons may have chosen to only perform highly invasive surgery such as corpectomy on the healthiest patients and refrained from implanting hardware in the sickest patients.Third, certain variables in our analysis were subjective in nature (such as ECOG performance status) and are subject to interobserver variability.Fourth, we decided to exclude patients with missing values from analysis, which might also have introduced selection bias.Additionally, to assess individually associated factors with retreatments, the Fine-Gray model was used.We chose the Fine-Gray model since it refers to the pattern of actually performed retreatments and not to the retreatment risk that could be expected for a patient living long enough.It does not censor deaths but treats death as a competing risk [45].However, caution should be taken with the interpretation of the documented SHR.The SHR can be interpreted in a similar manner to an HR from a Cox regression, but an SHR accounted for the risk of death.This means that an SHR is often smaller than a Cox HR, since part of the cohort already died before they could have a retreatment [45].To overcome this, we also provided the Cox Regression HRs for easier interpretation and to allow comparison of results across different studies.Despite the abovementioned limitations, considering the fact that we analyzed one of the biggest surgical MM cohorts, we believe that these results provide valuable knowledge for postoperative risk assessment in patients with MM facing a spinal intervention.Furthermore, we believe that this work will raise surgeons' awareness of the relatively high risk of postoperative complications and retreatments of MM patients.This awareness can help them better plan patient care and co-manage patients with internal medicine specialists for better outcomes.

Conclusion
The goal of surgical treatment for MM bone disease is to enhance patient quality of life, spinal stability, and relieve symptoms.However, postoperative complication rates remain high after spine surgery in patients with MM.We found an overall complication rate of 35% and a retreatment rate of 20% which might still be an underestimation.These relatively high rates are likely attributable to the inherent characteristics of the disease.The risk factors identified that are associated with complication and retreatment can assist future efforts for improving patient-centered treatment strategies in a multidisciplinary approach.Patient risk for complications and subsequent surgery should be explored, and multidisciplinary therapy including other medical disciplines is crucial to mitigate these risks.10 Analysis of confounding.The third step in the purposeful regression approach is an assessment whether variables removed in steps #1 and #2 are potential confounders.Variables are added back to the model if they are.The recommendation is that if the changes in beta coefficients (log odds) exceed 20%, then the excluded variable could be a confounder and should be included in the reduced model due to its effect on the other estimates [14] ** Indicates a potential confounding effect (change in log odds > 25%)

Table 1
Values are presented as number (%) and mean[standard deviation] for normally distributed variables or median [interquartile range] for non-normally distributed variables MM, Multiple Myeloma; BMI, Body Mass Index; ISS, International Staging System; ECOG, Eastern Cooperative Oncology Group; CCI, Charlson Comorbidity score; ASIA, American Spinal Cord Injury Association; RVD, Revlimid-Velcade-Dexamethasone; ESCC, Epidural Spinal Cord Compression; SINS, Spinal Instability Neoplastic Score; VCF, vertebral compression fracture ⋄ Minimum 6-10mg every 6-8 h intramuscular or intravenous within a week from surgery P-values are shown for differences between groups using chi-squared tests for categorical variables and Kruskal-Wallis tests for continuous variables *Indicates statistical significance ^Indicates the group with the highest proportion in the highest category (only for the statistically significant differences) ˅Indicates the group with the highest proportion in the lowest category (only for the statistically significant differences in categorical variables)

Table 5
Results from Cox and fine and gray regression analysis for factors associated with revision surgery or unplanned radiation HR, Hazard Ratio; SHR, Subdistribution Hazard Ratio; CI, Confidence Interval; ECOG, Eastern Cooperative Oncology Group; ESCC, Epidural Spinal Cord Compression; mg, milligram; dL, deciliter; L, liter * Indicates statistical significance ^ Re-entered and included because of confounding effect ° Retained because of clinical relevance

Table 6
Surgical data RT, Radiotherapy; mL, milliliter; CSF, cerebrospinal fluid P-values are shown for differences between groups using chi-squared tests for categorical variables and Kruskal-Wallis tests for continuous variables *Indicates statistical significance ^Indicates the group with the highest proportion in the highest category (only for the statistically significant differences) v Indicates the group with the highest proportion in the lowest category (only for the statistically significant differences in categorical variables)

Table 8
[14]ariable analysis.A simple logistic regression is fir for each variable.Significance is judged at a higher level for inclusion (suggested 0.25).Then, factors with a p-value below .25 in bivariate analysis are entered in a multivariable regression model (step 2)[14]

Table 9
Step 2 of purposeful selection approach (Complications analysis): Multivariable analysis

Table 9
Multivariable analysis.A multivariable model was fitted and the importance of each covariate was assessed with the purpose to get to a smaller, reduced model.Starting with all covariates identified for inclusion in Step #1 in the full model, where the reduced model contains only covariates that are below a 0.05 (*) cutoff for significance or that have strong clinical reasons to stay in the model.