Risk of Cancer Following Lumbar Fusion Surgery With Recombinant Human Bone Morphogenic Protein-2 (rhBMP-2): An Analysis Using a Commercially Insured Patient Population

INTRODUCTION

Bone morphogenic proteins (BMPs) are growth factors that are known, among other properties, to induce bone formation and thus have been evaluated as an alternative to iliac crest bone grafting at the time of fusion of the lumbar spine.¹ Recombinant human BMP-2 (rhBMP-2) is indicated for anterior lumbar fusion and is administered via an absorbable collagen sponge carrier known as the Infuse Bone Graft (Medtronic Inc, Memphis, Tennessee). In addition, rhBMP-7 is available as a mixture with bovine collagen and after reconstitution with saline is administered as a paste. BMPs are thought to play a role in apoptosis as well as cell growth and differentiation, and receptors for BMP are found on multiple cell types, including cancer cells.² A review of the preclinical literature concluded that whereas BMP-2 likely does not cause de novo cancers, it may have potential to enhance tumor function, and thus more definitive research is needed.³

Although randomized clinical trial data did not suggest any association of rhBMP with development of cancers,^4,5 additional analyses of trial data found a greater frequency of malignancy in patients who received rhBMP compared with those who received bone grafts,^6–8 with two analyses achieving statistical significance.^7,8 In addition, observational studies using both Medicare^9–11 and commercial insurance claims^12,13 data did not show an increased cancer risk, but they were limited by relatively short duration of follow-up after surgery and/or questions of generalizability to younger patients. A recently published study that used a linked tumor-Medicare database found no risk of second primary cancers or cancer recurrence,¹⁴ and a single-center study of over 500 patients also did not show an increased cancer incidence.¹⁵ Finally, a review¹⁶ of the clinical data found there was no conclusive evidence that rhBMP resulted in a higher risk of subsequent cancer but that the potential risk should be considered for each patient. However, because published studies typically had follow-up of 3 to 4 years and as little as under 2 years,¹³ delayed carcinogenic effects may not have been apparent.

Given the conflicting data about cancer risk, we performed a retrospective cohort study in a commercially insured population of patients less than 65 years of age, which would complement previous studies on the Medicare population^9–11 and evaluate a population that was at lower baseline cancer risk. In addition, we restricted our analysis to patients with at least 3 years of follow-up. We hypothesized that the incidence of cancer in follow-up after surgery would be similar in the rhBMP-treated and untreated patients.

METHODS

RESULTS

From the MarketScan database, we identified 356 306 patients who underwent lumbar spinal fusion. We then excluded 112 164 patients with surgery before October 2003 or after December 2011; 124 154 patients without continuous enrollment for at least 2 years prior to surgery; 47 450 with less than 3 years of follow-up; 12 890 with a history of cancer; and 20 200 who were under 18 or over age 64. The remaining 39 448 patients were the subject of this analysis.

The characteristics of 39 448 patients are shown in Table 1. The mean age was 51.7 ± 7.8 years, 53.6% were women, 46.4% were men, and most patients had low comorbidity scores. There was evidence of rhBMP administration in 2345 patients (5.9%). Compared with others, those who received rhBMP tended to be younger (51.2 ± 8.1 years versus 51.7 ± 7.7 years, P < .0001), to be women, to have higher comorbidity indices, and were also more likely to undergo anterior procedures as well as multiple level or redo procedures. The use of rhBMP increased over the duration of the study. The mean and median length of follow-up were 4.90 and 4.87 years, respectively, in the rhBMP-treated patients (total of 11 246 person years) and 5.00 and 5.04 years, respectively, in others (total of 187 033 person years). All persons in both groups had a minimum of 3 years follow-up and a maximum of 9.2 years follow-up.

A cancer diagnosis in follow-up was observed in 1121 patients, corresponding to an incidence of 0.57/100 person years. A total of 49 cancers were observed in the BMP-treated patients (incidence, 0.43/100 person years) and 1072 in the nontreated patients (0.58/100 person years). This corresponded to an incidence rate ratio of 0.75 (95% CI, 0.56–0.99), which indicates a slightly lower risk of cancer in the rhBMP-treated patients. Of note, in the age- and gender-matched general population, the incidence rate is 0.45/100 person years. The incidence of individual cancers by site is shown in Table 2. The most frequently observed sites were breast, non-Hodgkins lymphoma, melanoma, lung, prostate, myeloma, renal, and colorectal cancers. Although the incidence of lung cancer was somewhat higher in the rhBMP-treated patients (0.04/100 versus 0.02/100 person years), the difference was not statistically significant (P = .13), and the incidence of other cancer types in rhBMP-treated patients never exceeded that of others. In a Kaplan-Meier analysis (Figure), patients who received rhBMP were at a risk for cancer development that was similar to those who did not (P = .2687 by log-rank test).

• Download figure
• Open in new tab
• Download powerpoint

Figure

Kaplan-Meier plot of malignant neoplasia risk in recombinant human bone morphogenic protein (rhBMP)-treated and untreated patients. Through a follow-up period of up to 8 years, patients receiving rhBMP were at similar risk to receive a cancer diagnosis as untreated patients (P = .2687 by log-rank test).

We also compared the observed with expected incidence of malignant neoplasms in an age- and gender-matched population according to SEER data. The standardized incidence ratio for the entire sample was 0.80 (95% CI, 0.74–0.87). For patients treated with rhBMP, the standardized incidence ratio was 0.42 (95% CI, 0.30–0.58), compared with a standardized incidence ratio of 0.83 (95% CI, 0.77–0.90) in others. These findings indicate a somewhat lower cancer risk in both groups compared with the general population, but especially in the BMP treated patients.

We then used proportional hazards models to examine the association of rhBMP with cancer risk. For all cancers combined, use of rhBMP was associated with a similar risk of cancer in both univariate (hazard ratio [HR], 0.80; 95% CI, 0.54–1.19, P = .276) and mulitivariate analyses (HR, 0.81; 95% CI, 0.54–1.20, P = .283). Other factors associated with cancer risk in multivariate analysis included older age (ages 40–49 years: HR, 1.50; 95% CI, 1.01–2.23, P = .05; ages 50–59 years: HR, 2.78; 95% CI, 1.91–4.06, P < .0001; ages 60–64 years: HR, 12.02; 95% CI, 7.53–19.20; P < .0001 compared with ages 18–39 years) and increased comorbidity score (1 comorbidity HR, 1.46; 95% CI, 1.23–1.73, P < .0001, ≥ 2 comorbidities HR, 1.12; 95% CI, 0.60–2.09, P = .730). Data for individual sites are shown in Table 3. In both univariate and multivariate analyses, there was no association of rhBMP use with risk of any of the malignant neoplasms, though for many sites, the infrequent number of events precluded the calculation of reliable estimates.

In order to adjust for potential selection bias in rhBMP treatment allocation, we developed a propensity score to predict the probability of receipt of rhBMP. The propensity score had good discrimination in predicting the likelihood of receiving rhBMP, as evidenced by a receiver operating characteristic curve area of 0.810, which indicates a much better likelihood of prediction than by chance alone (0.50). After adjustment for the propensity score, the use of rhBMP was associated with a similar risk of malignant neoplasia as that of untreated patients (HR, 0.76; 95% CI, 0.51–1.14). Finally, we performed a secondary proportional hazards model that was limited to patients who underwent multiple level or redo procedures, which are often associated with higher rhBMP doses. In this cohort, the multivariate HR for rhBMP exposure was 0.71 (95% CI, 0.46–1.11; P = .134) and thus not associated with cancer risk.

DISCUSSION

The effect of rhBMP on cancer risk is controversial. Although a meta-analysis failed to demonstrate increased risk of malignant tumors with rhBMP,²¹ a tumor-promotion effect of rhBMP cannot be excluded on a molecular level.³ In the current study, which included a relatively long duration of follow-up, we did not demonstrate any cancer risk associated with rhBMP administration. The findings were robust when we attempted to adjust for confounding factors, including the likelihood of receiving rhBMP, as well as when restricted to indications that may be associated with higher rhBMP doses. The observed cancer incidence (0.43 case/100 person years versus 0.58 case/100 person years in non-BMP patients) was lower than in studies that included a Medicare-aged population (1.7–2.1 cases/100 person years)^11,14 and in the same range as in non-BMP-treated patients from clinical trial data (0.50 case/100 person years).⁸ The latter study⁸ also found a much higher incidence in BMP-treated patients (3.37 cases/100 person years).

Receptors for BMP are found on a variety of cancer cells,² and thus there is potential concern for BMP in the promotion of tumor growth both locally and at metastatic sites. Although at the cellular level, BMP has been shown to promote angiogenesis,^22,23 cell growth,²⁴ bone metastases,²⁵ and malignant cell motility and invasiveness,²⁶ BMP is also capable of inhibiting proliferation and growth and thus could have potential antineoplastic effects.^25,26 However, given at least the potential concern for BMP promoting progression, rhBMP is not indicated in the vicinity of a resected or extant cancer or in those receiving treatment for malignancy.²⁷

The methodology and data sources that we used have several strengths and limitations. Our study had a very large sample and consisted of a wide range of practices and captured multiple cancer diagnoses. The data were limited by the absence of clinical detail, including factors such as smoking, alcohol use, obesity, family history of cancer, and differences in intraoperative technique that may have been associated with rhBMP use and/or cancer risk. However, consistent results were observed in an analysis that included a propensity score for likelihood of rhBMP administration. Also, in a previous study⁹ of pancreatic carcinoma after rhBMP exposure, medical record review found no association of rhBMP with other cancer risk factors such as obesity and smoking. We also could not rule out differences in treatment allocation, although patients at increased cancer risk at baseline were preferentially not given rhBMP. In addition, although the patients receiving rhBMP were somewhat younger and therefore at lower baseline cancer risk, the differences were maintained after adjustment for age as well as gender and comorbidity. We also could not measure the actual rhBMP dose, but in analyses limited to procedures typically associated with higher doses, there was no association with malignancy. Although the follow-up was at least 3 years after surgery, with some patients followed as long as 8 years, if the potential risk of rhBMP is mutagenesis rather than tumor promotion, an even longer follow-up period may be required to definitively exclude its malignant potential. We also ascertained previous and subsequent cancers through the use of ICD-9-CM codes, which were developed for reimbursement and not for research. However, the algorithm that was used included fairly stringent criteria to define the presence of malignant tumors. Our study was limited to lumbar fusion procedures in adult patients. We recognize that the product is not uncommonly used off label and therefore included posterior or transverse procedures, neither of which were approved indications. However, we did exclude patients with contraindications to BMP including age <18 years or previous cancer diagnoses, and no women in the sample were known to be pregnant. Despite the large sample, we did not have sufficient power to detect any differences in the incidence of rare tumors and did not capture neoplasms not contained in SEER such as nonmelanoma skin cancers. Finally, because we used a procedure code as a measure of rhBMP administration, there is the potential for misclassification. However, in previous work,⁹ the specificity of the procedure code for receipt of rhBMP-2 (compared with rhBMP-7) was 95% and the positive predictive value of the code was 100%.

CONCLUSIONS

In this large sample of commercially insured patients, we found that treatment with rhBMP during fusion of the lumbar spine did not increase the subsequent risk of cancer. Although some previous studies did show an increased cancer risk, the findings of this and other database studies should provide reassurance to both patients and providers.