1.1 Background

Metastatic bone disease is the third most common site of metastasis in cancer patients.¹ Especially in patients with lung, breast, or prostate cancer skeletal metastases frequently occur.^1,2 Skeletal metastases limit mobility, cause pain, lead to pathologic fractures, and decrease quality of life significantly.³ Decision making on the appropriateness of surgical management in this population is partly dependent on the preoperative estimations of expected survival.¹ Several prognostic factors have previously been identified and multiple prognostic models have been created to improve decision making.^4–6 In a recently developed machine learning model for the same purpose, alkaline phosphatase was selected by random forest algorithms as a predictor.⁷ The aim of the current study was to shed more light on its utility as a prognostic marker in patients with extremity metastatic bone disease.

1.2 Rationale

Serum alkaline phosphatase is released by both liver and bone. Elevated serum alkaline phosphatase can be the result of increased osteoblastic activity in bone disease and/or increased synthesis by hepatocytes in liver disease.^8,9 These mechanisms suggest that serum alkaline phosphatase may be a useful marker for metastatic burden and a prognostic marker for survival in patients with extremity metastatic bone disease.

1.3 Purpose of the study

The purpose of this retrospective cohort study was to investigate the value of serum alkaline phosphatase in the prognostication of patients with metastatic disease of the extremities. Specifically, we examined the association of this marker with three-month and one-year mortality in this population. Additionally, we sought to determine the relationship between serum alkaline phosphatase and baseline metastatic tumor burden.

2.1 Study design and setting

The current study was carried out according to the STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) guidelines.¹⁰ Our institutional review board approved the study. Retrospective review at two tertiary care centers was used to identify patients with who underwent surgical management for extremity metastatic bone disease.

2.2 Participants/study subjects

The following inclusion criteria were used to identify eligible patients: (1) 18 years or older, (2) surgical management for metastatic long bone disease, (3) date of surgery between January 1999 and January 2017. Surgical management included the following operations: endoprosthetic reconstruction, plate-screw fixation, intramedullary nailing and dynamic hip screw. Only the first surgery per patient was included, if patients underwent more than one operation on different occasions to prevent the violation of the statistical assumption of independence. Patients that did not receive their original treatment at one of the two centers were excluded. Overall, 1090 patients were identified with three-month and one-year mortality rates were of 29% (n = 305) and 62% (n = 639), respectively. The majority of patients was female (56% versus 44%) and most were under the age of 65 (55%).

2.3 Outcome

The primary outcome was overall survival after surgery and the date of death was obtained from clinical documentation in electronic medical records and/or the Social Security Death Index.¹¹

2.4 Covariates

The following demographic variables were extracted: sex, age, body mass index (kg/m²), and any other Charlson comorbidity in addition to metastatic cancer. Disease factors that were collected were: primary tumor histology in the groupings proposed by Katagiri et al.⁴ [slow growth, moderate growth, rapid growth]. Additional factors were anatomic location [lower extremity, upper extremity], presence of multiple bone metastases (defined as metastatic bone foci at other sites than the primary operative site), spine metastases, visceral metastases (defined as lung and or liver metastases), brain metastases, previous systemic therapy, and prior local radiation to the affected long bone. Preoperative laboratory values that were collected were hemoglobin (g/dL), platelet count (x10³/μL), white blood cell count (x10³/μL), creatinine (mg/dL), calcium (mg/dL), platelet-to-lymphocyte ratio, neutrophil-to-lymphocyte ratio, preoperative albumin (g/dL), and alkaline phosphatase (IU/L).

2.5 Statistical analysis

For baseline characteristics, categories are displayed as frequencies and percentages. The optimal cutoff value of serum alkaline phosphatase was determined using López-Ratón et al.‘s optimal cutpoint method with the area under the receiver operating curve metric.¹²

Bivariate analyses for serum alkaline phosphatase with baseline characteristics and outcomes were performed with univariable logistic regression. Boxplots were created to visualize the association between elevated serum alkaline phosphatase and metastatic burden and Kaplan-Meier curves were generated to visualize the association between serum alkaline phosphatase and survival up to one year. Multiple imputation was used for variables with less than 30% missing data.¹³ Multivariable logistic regression with bootstrapped (100 replications with replacement) backward stepwise elimination was used to determine whether serum alkaline phosphatase was associated with three-month and one-year mortality.

The confidence intervals given are 95% confidence intervals and p-values <0.05 were deemed significant.

3.1 Alkaline phosphatase and baseline characteristics

Baseline characteristics are displayed in Table 1. Fig. 1 shows the boxplots of alkaline phosphatase with metastatic burden. Serum alkaline phosphatase was associated with [A] other bone metastasis (p < 0.001), [B] spinal metastasis (p < 0.001), and [C] visceral metastasis (p < 0.001) but not [D] brain metastasis (p = 0.2). The optimal cutoff for alkaline phosphatase was determined to be 100 IU/L using the area under the receiver operating curve metric. Table 2 shows the association of serum alkaline phosphatase with baseline characteristics after being dichotomized at this threshold. Factors associated with this marker included older age (p = 0.04), local radiation (0.001), hemoglobin (p = 0.04), creatinine (p = 0.002), platelet-to-lymphocyte ratio (p = 0.04), and albumin (p < 0.001).

Close

Fig. 1 Serum alkaline phosphatase was associated with preoperative presence of A) other bone metastasis (p < 0.001), B) spine metastasis (p < 0.001), and C) visceral metastasis (p < 0.001). It was not associated with D) brain metastasis (p = 0.2).

Export to PPT

3.2 Alkaline phosphatase and survival

For patients with serum alkaline phosphatase below the threshold of 100 IU/L, three-month and one-year mortality rates were 23% (n = 84) and 55% (n = 196), respectively. In comparison, for patients with elevated serum alkaline phosphatase, three-month and one-year mortality rates were 39% (n = 148) and 74% (n = 281) respectively. Fig. 2 shows the Kaplan-Meier predicted survival curves by strata of serum alkaline phosphatase; elevated serum alkaline phosphatase was associated with lower survival at all time points up to one year. In bivariate analysis, elevated alkaline phosphatase was associated with three-month (Odds ratio [OR] = 1.6, 95% confidence interval [CI] = 1.2–2.1) and one-year (OR = 1.6, CI = 1.3–2.1) mortality. On multivariate analysis, serum alkaline phosphatase remained associated with three-month (OR = 1.5, CI = 1.1–2.1), and one-year (OR = 1.5, CI = 1.1–2.0) mortality (Tables 3 and 4).

Close

Fig. 2 This shows the Kaplan-Meier curve by strata of serum alkaline phosphatase over a one year period.

Export to PPT

4.1 Summary

Serum alkaline phosphatase is associated with baseline metastatic tumor burden (visceral metastasis, multiple bone metastases, and spine metastasis) in cancer patients undergoing surgical management for extremity metastatic bone disease. Elevated preoperative serum alkaline phosphatase is an independent prognostic marker for ninety-day and one-year survival in this population.

4.2 Alkaline phosphatase and metastatic disease

Alkaline phosphatase has previously been identified as a marker for the presence of liver and bone metastases in different types of cancer, including breast cancer,¹⁴ lung cancer,¹⁵ and bladder cancer.¹⁶ Chen et al. studied 2133 patients with breast cancer of which 327 (15%) had bone metastases. They found high alkaline phosphatase to be an independent risk factor for the development of a bone metastasis.¹⁴ Zhou et al. studied a total of 2021 patients with lung cancer of whom 483 (24%) had bone metastasis. Elevated alkaline phosphatase was found to be independently associated with bone metastasis in this patient group.¹⁵

4.3 Alkaline phosphatase and survival in metastatic cancer

Previous studies have found elevated alkaline phosphatase to be a prognostic marker for poorer survival in metastatic cancer.^17–20 A study consisting of 1790 patients with spinal metastatic disease from the National Surgical Quality Improvement Program (NSQIP) found an elevated alkaline phosphatase as a marker for thirty-day mortality on multivariable analysis.¹⁹ Another study consisting of 732 patients with spinal metastatic disease confirmed those findings for ninety –day and one-year survival.¹⁸

4.4 Prognostic algorithms for bone metastatic disease

Several prognostic factors have been identified in extremity metastatic bone disease and multiple prognostic models have been developed in order to aid survival estimation and surgical decision making.^4–7 In the prognostic models that currently exist for extremity metastasis, hemoglobin is the most frequently used laboratory factor.^5–7 Forsberg et al. also studied absolute lymphocyte count⁶ and Katagiri et al. studied clusters of laboratory factors which grouped as “abnormal” (CRP ≥ 0.4 mg/dL, LDH ≥ 250 IU/L, or serum albumin <3.7 g/dL) or “critical” (platelets <100,000/L, serum calcium ≥10.3 mg/dL, or total bilirubin ≥1.4 mg/dL).⁴ We previously developed machine learning algorithms in which multiple laboratory factors were included, among which alkaline phosphatase.⁷ The purpose of that study to explore the use of machine learning for the development of prognostic algorithms, and the variables were selected by random forest algorithms. In the current study we more thoroughly explored the use of alkaline phosphatase as a prognostic marker in extremity metastatic bone disease and our results suggest that the addition of serum alkaline phosphatase may improve prognostication in this population.

4.5 Limitations

There are several limitations to this study. First, this was a retrospective study and future prospective studies are needed to confirm the findings presented here. Additionally, only patients that were surgically managed were included and the prognostic value of serum alkaline phosphatase in patients managed non-operatively remains to be determined. Furthermore, the patient cohort from this study was drawn from two affiliated institutions from a single healthcare system. Future studies must determine whether the same results apply to a more heterogeneous population.