Prognostic value of alkaline phosphatase in hormone-sensitive prostate cancer: a systematic review and meta-analysis

Purpose To assess the prognostic value of alkaline phosphatase in patients with hormone-sensitive prostate cancer. Methods A systematic review and meta-analysis was performed using the PUBMED, Web of Science, Cochrane Library, and Scopus in April 2019 according to the Preferred Reporting Items for Systematic Review and Meta-analysis statement. Studies were deemed eligible if they compared hormone-sensitive prostate cancer patients with high vs. low alkaline phosphatase to determine its predictive value for overall survival, cancer-specific survival, and progression-free survival. We performed a formal meta-analysis of these outcomes. Results 42 articles with 7938 patients were included in the systematic review and 28 studies with 5849 patients for the qualitative assessment. High alkaline phosphatase was associated with worse overall survival (pooled HR 1.72; 95% CI 1.37−2.14) and progression-free survival (pooled HR 1.30; 95% CI 1.10−1.54). In subgroup analyses of patients with “high-volume” and “low-volume”, alkaline phosphatase was associated with the overall survival (pooled HR 1.41; 95% CI 1.21−1.64 and pooled HR 1.64; 95% CI, 1.06−2.52, respectively). Conclusions In this meta-analysis, elevated serum levels of alkaline phosphatase were associated with an increased risk of overall mortality and disease progression in patients with hormone-sensitive prostate cancer. In contrast, those were not associated with an increased risk of cancer-specific mortality. Alkaline phosphatase was independently associated with overall survival in both patients with “high-volume” and “low-volume” hormone-sensitive prostate cancer. Alkaline phosphatase may be useful for being integrated into prognostic tools that help guide treatment strategy, thereby facilitating the shared decision making process.


Introduction
Prostate cancer (PC) is not only the most common solid cancer, but also the second most common cause of cancer-related death in men [1]. Following the results of the CHAARTED trial and the LATITUDE trial, the treatment of patients with metastatic hormone-sensitive prostate cancer (HSPC) has changed substantially in the recent years [2,3]. However, systemic therapy based on androgen deprivation remains the standard primary treatment strategy in patients with metastatic HSPC. Despite adequate therapy, the disease eventually progresses to a castration-resistant prostate cancer (CRPC) [4]. To improve PC outcomes, prognostic tools have been developed to help in the daily clinical decision making and patient counselling [5][6][7][8]. These tools include standard clinical features and biomarkers [9], such as alkaline phosphatase (ALP) in patients with CRPC, but not yet in patients with HSPC.
ALP is a glycoprotein derived from bones, liver, kidney, or placenta that has been shown to be elevated and of prognostic value for various malignancies [10][11][12][13]. In PC, ALP has been shown to be of prognostic value in CRPCreflecting disease outcome, independent of therapy [14]. In patients with CRPC, high-baseline ALP levels have been shown to be associated with worse outcomes, including skeletal complications and decreased survival [15][16][17]. Moreover, elevated ALP was also been shown to be correlated with the extent of metastatic bone disease [17,18]. Serum ALP is deemed a simple and inexpensive test that could serve as an objective prognostic parameter that helps improve daily oncologic clinical practice, plan follow-up, and counsel regarding outcomes, thus facilitating the shared decision making process with the patient. Unfortunately, to date, the prognostic value of ALP in HSPC remains insufficiently investigated.
The aim of the current study was to summarize the available data to test the hypothesis that ALP has a strong prognostic value for oncologic outcomes in HSPC patients. To this end, we performed a systematic review and a meta-analysis.

Search strategy
This systematic review and meta-analysis was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement [19]. We searched the electronic databases PUBMED, Web of Science, Cochrane Library and Scopus on April 2019, investigating the prognostic value of ALP in HSPC.
After the first screening based on study title and abstract, all papers were assessed based on full text and excluded with reasons when inappropriate; a further check of the appropriateness of the papers based on full text revision which was performed after the data extraction. Two investigators carried out this process independently. Disagreements were resolved by a consensus meeting with a third investigator. The following keywords were used in our search strategy: (prostate cancer OR prostate carcinoma OR prostate tumor OR prostatic carcinoma OR prostatic cancer OR prostatic tumor NOT resistant) AND (Alkaline Phosphatase OR ALP) AND (survival OR outcome OR prognostic OR mortality OR progression OR recurrence OR OS OR CSS OR PFS OR RFS OR MFS). The primary outcome of interest was overall survival (OS) and secondary outcomes were cancer-specific survival (CSS) and progression-free survival (PFS).

Inclusion criteria and exclusion criteria
Studies were included if they investigated whether patients with high ALP treated for HSPC (patients) who had received systemic therapy (intervention) as compared to those who had low ALP (comparison) to assess the independent predictive value of ALP on OS, CSS, and PFS (outcome) utilizing multivariate Cox regression analysis (study design) in nonrandomized observational, or randomize or cohort studies. We excluded reviews, letters, editorials, meeting abstracts, replies from author, case reports, and articles not published in English. In case of duplicate publications, either the higher quality or the most recent publication was selected. References of included manuscripts were scanned for additional studies of interest.

Data extraction
Two investigators independently extracted the information from the included articles. The information contained the following characteristics: first author's name, publication year, recruitment country, period of patient recruitment, number of patients, age, study design, disease stage, therapy type, oncological outcome, follow-up duration, conclusion, and ALP cut-off. Subsequently, the hazard ratios (HR) and 95% confidence intervals (CI) of ALP associated with each of the outcomes were retrieved. The HRs were extracted from the multivariate analyses. All discrepancies regarding data extraction were resolved by consensus with a third investigator.

Quality assessment
The Newcastle-Ottawa Scale (NOS) was used to assess the quality of the included studies according to the Cochrane Handbook for systematic reviews of interventions for included non-randomized studies [20,21]. The scale focuses on the three factors: Selection (1−4), Comparability (1−2) and Exposure (1−3). The total score ranges from 0 (lowest) to 9 (highest). The main confounders were identified as the important prognostic factors of OS, CSS, and PFS. The presence of confounders was determined by consensus and review of the literature. We identified as "high-quality" choices those with scores more than 6.

Statistical analyses
We performed a forest plot to assess the HRs from the multivariate analyses of individual studies and obtained a summary HR of the value of ALP on OS, CSS, and PFS. Disease progression includes symptomatic or radiographic or biochemical progression in this analysis. Studies with Kaplan-Meier log-rank, univariate Cox proportional hazard regression, or general logistic regression analyses were not considered for the meta-analysis. In case there were only HR and P value, we calculated 95% CI [22,23]. We also performed subgroup analyses in HSPC patients with "high-volume" and "low-volume" disease. We classified as low-volume (lesions < 4 sites and within pelvis-vertebral column) or high-volume disease (lesions ≥ 4 sites and at least one lesion beyond the pelvis-vertebral column) according to the CHAARTED classification [2]. Again, of all the HSPC patients from the studies providing information on EOD scores or Soloway scores, those with EOD scores 2 or higher or those with Soloway scores 2 or higher were defined as high-volume disease [24]. With high-volume disease thus defined, all studies in which those with high-volume disease accounted for 60% or more or less than 60% of all patients were included for the current analysis as "high-volume disease" and "low-volume disease" studies, respectively.
Heterogeneity among the outcomes of the included studies in this meta-analysis was evaluated using Cochrane Q test and I 2 statistic. Significant heterogeneity was indicated by a P < 0.05 in Cochrane Q tests and a ratio > 50% in I 2 statistics, which led to the use of random-effect models. We used fixed effect models for calculation of pooled HRs for nonheterogeneous results [25][26][27]. Publication bias was assessed by funnel plots. Statistical analyses were performed using Stata/MP 14.2 (Stata Corp., College Station, TX); statistical significance level was set at P < 0.05.

Study selection and characteristics
Our initial search identified 2245 records. After removal of duplicates, 2016 remained (Fig. 1). After screening of the titles and abstracts, 1816 articles were excluded. Then we assessed 200 full texts for further selection. After selection, 42 articles with 7938 patients were included in the systematic review and 28 studies with 5849 patients for qualitative meta-analysis . The baseline characteristics of the 42 studies are outlined in Table 1. All included studies were published between 1995 and 2019 with 15 being from Europe, and 27 from Asia. Median age ranged from 63 to 77 years, 10 studies included non-metastatic HSPC. Studies were heterogeneous regarding cut-off value for ALP ranging from 67 to 620 for OS, from 115 to 683.4 for CSS, and from 114.56 to 400 for PFS; follow-up ranged from 14.4 to 156 months.

Association of ALP with OS in HSPC
Sixteen studies including 3747 patients provided data on the association of ALP with OS in HSPC. The forest plot (Fig. 2a) showed that ALP was significantly associated with OS in HSPC (pooled HR 1.72; 95% CI 1.37 − 2.14; z = 4.76). The Cochrane Q test (χ 2 = 85.73; P = 0.000) and I 2 test (I 2 = 81.3%) showed significant heterogeneity. The funnel plot identified nine studies over the pseudo 95% CI (Fig. 2a).

Association of ALP with CSS in HSPC
Ten studies including 2225 patients provided data on the association of ALP with CSS in HSPC. The forest plot (Fig. 2b) showed that ALP was not significantly associated with CSS in HSPC (pooled HR 1.00; 95% CI 1.00-1.01; z = 1.55). The Cochrane Q test (χ 2 = 80.97; P = 0.000) and I 2 test (I 2 = 88.9%) showed significant heterogeneity. The funnel plot identified four studies over the pseudo 95% CI (Fig. 2b).

Association of ALP with PFS in HSPC
Seven studies including 1547 patients provided data on the association of ALP with PFS in HSPC. The forest plot (Fig. 2c) showed that ALP was significantly associated with PFS in HSPC (pooled HR 1.30; 95% CI 1.10−1.54; z = 3.04). The Cochrane Q test (χ 2 = 40.49; P = 0.000) and I 2 test (I 2 = 85.2%) showed significant heterogeneity. The funnel plot identified four studies over the pseudo 95% CI (Fig. 2c).

Association of ALP with OS in HSPC with "high volume"
Five studies including 1509 patients provided data on the association of ALP with OS in HSPC with "high-volume" disease. The forest plot (Fig. 3a) showed that ALP was significantly associated with OS in HSPC with "highvolume" disease (pooled HR 1.41; 95% CI 1.21−1.64; z = 4.47). The Cochrane Q test (χ 2 = 7.25; P = 0.123) and I 2 test (I 2 = 44.8%) showed no significant heterogeneity. The funnel plot identified no studies over the pseudo 95% CI (Fig. 3a).

Association of ALP with OS in HSPC with "low volume"
Six studies including 1039 patients provided data on the association of ALP with OS in HSPC with "low-volume" disease. The forest plot (Fig. 3b) showed that ALP was significantly associated with OS in HSPC with "low-volume"

Discussion
In this systematic review and meta-analysis, we investigated the prognostic value of ALP in HSPC by assessing its impact on PFS, CSS, and OS. We found that the HSPC patients with elevated ALP have significantly worse OS and PFS compared to their counterparts with normal ALP levels. In other words, pre-treatment ALP values may be a useful biomarker in the choice of treatment, even in early metastatic PC. The prognostic value of ALP has been shown in various solid malignancies with bone metastasis [11][12][13]. However, while there is a biological rationale underlying this association, the exact mechanism remains unclear. A potential explanation is that when cancer starts to metastasize, ALP reflects bone turnover, osteoblast activity, and osteoid formation in the adjacent bone tissues [70]. Thus, ALP may be an indicator of bone metastatic tumor load. Accordingly, ALP has been shown to be elevated in cancer patients with bone metastasis, as the current literature shows, ALP is already among the biomarkers included in the tools used for prognosticating outcomes in CRPC patients [5][6][7][8].
Interestingly, ALP was significantly associated with worse OS in metastatic HSPC patients not only with "high-volume" disease, but also with "low-volume" disease, suggesting that ALP is an indirect sensitive measure of metastatic tumor burden which could not be captured by conventional imaging. It is likely that the elevated ALP reflects micro metastases despite negative findings on conventional imaging. Although few studies have assessed this patient subgroup, ALP could be used to select patients who may benefit more from intensive therapy such as upfront docetaxel or abiraterone in addition to standard androgen deprivation therapy. Moreover, ALP could also be used as a response/monitoring marker for these therapies as well as bone-targeting therapies such as bisphosphonate.
Despite showing a strong association of ALP with mortality and progression in HSPC patients, this systematic review and meta-analysis has some limitations. There is a reporting bias, as some studies with negative results may not have been published. Further, many included studies were retrospective, leading to a patient selection bias. Second, unknown pretreatment conditions (i.e., physical conditions, comorbidities, obstructive jaundice, bone disease, hyperthyroidism and hepatitis, medication, and lifestyle habits) could have altered ALP values leading to a systematic bias. Third, heterogeneity was detected for OS, CSS, and PFS analyses limiting the value of these results.
Although the random effect model takes into account the heterogeneity among studies, the conclusions should be interpreted with caution. Fourth, there is no established cut-off value for ALP among the included studies; most investigators chose the cut-off based on the statistical methods assessing for the highest sensitivity and specificity, using the upper limit of normal, or using literature predefined ALP cut-offs. Only three studies investigated ALP as a continuous variable. Regardless of these limitations, ALP is a fast and readily available biomarker. Well-designed prospective studies with longer follow-up are needed to validate the prognostic value of ALP and its potential value in risk stratification of patients with HSPC using clinical decision-analytical tools.

Conclusions
In this meta-analysis, high serum ALP was associated with an increased risk of overall mortality and disease progression in patients with HSPC. In contrast, high serum ALP was not associated with an increased risk of cancer-specific mortality. Furthermore, ALP was an independent risk factor for OS in HSPC patients with both "high-volume" and "low-volume" metastatic disease. ALP may be useful for clinical decision making regarding treatment selection, as well as for patient counselling. However, considering the limitations including heterogeneity, the conclusions should be interpreted with caution.