Edited by: Imtiaz Ahmad Siddiqui, University of Wisconsin-Madison, United States
Reviewed by: Ashish Solanki, Medical University of South Carolina, United States; Hedwin Kitdorlang Dkhar, University of Florida, United States; Ying Chao Zhao, Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, China; Takeshi Yuasa, Japanese Foundation for Cancer Research, Japan
This article was submitted to Cancer Epidemiology and Prevention, a section of the journal Frontiers in Oncology
†These authors have contributed equally to this work
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Cancer is the second major cause of morbidity and mortality worldwide behind cardiovascular diseases, with an estimated over 8.7 million deaths in 2015 (
Bisphosphonates are commonly-prescribed medications widely prescribed for the prevention and treatment of osteoporosis. They are generally well-tolerated with the most common side effects being renal toxicity, gastrointestinal reactions, osteonecrosis of the jaw, atypical fractures, atrial fibrillation, musculoskeletal complaints, and esophageal cancer (
Several observational studies have found potential associations between bisphosphonate exposure and cancer risk/survival including gastrointestinal cancer, colorectal cancer, endometrial and ovarian cancer, and these studies reported inconsistent findings (
This meta-analysis was performed in accordance with the PRISMA guidelines (
Two authors (DW and ZX) reviewed all titles or abstracts of citations identified by the database search to retrieve potentially relevant articles. The full-text for each potentially relevant article was then reviewed by each of the two authors. The articles were cross checked by two independent authors for possible inclusion and disagreements were resolved through discussion or by consensus with a senior author (XW or JX).
Observational studies were satisfied the inclusion criteria and thus included for analysis if they (i) included patients clinically or pathologically diagnosed with solid cancer, (ii) evaluated and clearly defined exposure to any bisphosphonates, (iii) reported survival outcomes in patients with solid cancer, and (iv) reported hazard ratios (HRs) or relative risks (RRs) and their corresponding 95% CIs or provided relevant data for their calculation. If several publications shared the same cohort, we used the article with the most informative one.
Two authors (DW and ZX) independently conducted data abstraction using a standardized form. The following data were summarized from each study: first author, publication year, tumor type, study design, study country, sex, sample size, follow-up duration, HR and 95% confidence intervals (CIs) with fully adjustment for confounding variables. Conflicts were resolved by discussion or by a senior author. One author (XW) assessed the quality of each study by Ottawa–Newcastle scale (NOS). After the resolution of discrepancies, a final NOS score for observational studies was obtained based on selection of involved population, comparability of study groups, and adequacy of outcome assessment (
The primary outcome measure was OS which defined as the time from the enrollment to death from any cause. The secondary outcome measures included CSS, defined as the time from the enrollment to death from a specific cancer type, and RFS, defined as the time from the enrollment to the first recorded disease recurrence or death from any cause, which would occurred first.
Stata® version 12.0 (Stata Corp LP, College Station, Texas, USA) was applied to perform all statistical analyses. We used the DerSimonian and Laird random-effects model to calculate summarized HR and 95% CI (
Sensitivity analyses were performed to further investigate the sources of inter-study heterogeneity in terms of some major study characteristics for OS and CSS subset, which was also used to examine the robustness of the findings. The influence of individual study on the pooled estimates was conducted by omitting one study at a time and recalculating the others. Furthermore, subgroup differences were conducted using meta-regression analysis when necessary. A
The database search yielded a total of 9,742 unique citations published through March, 2018. After title or abstract review, 43 records were potentially appropriate, which were further reviewed through full-text reading, and an additional 35 studies were excluded due to various reasons. Finally, eight studies with a total of 241,634 individuals met the inclusion criteria (Figure
Flow chart for the process of identifying studies included in and excluded from the systematic review.
The characteristics of the included studies are shown in Table
Major baseline features of the included studies.
Korde et al. ( |
Population-based cohort study | Early stage invasive breast cancer | Multiple center | USA | Female | Medical record, interview, and cancer registry data | 1,813 | Median 11.8 years | 8 |
Kwan et al. ( |
Hospital-based cohort study | Breast cancer | Multiple center | USA | Female | Electronic health records and cancer registries | 16,781 | Median 6.4 years | 8 |
Hicks et al. ( |
Prospective population-based cohort study | Colorectal cancer | Multiple center | UK | Male/female | The UK Clinical Practice Research Datalink (CPRD) | 4,791 | Mean 3.3 years | 9 |
Kremer et al. ( |
Community-based cohort study | Breast cancer | Multiple center | Canada | Female | Prescription recorded database | 21,664 | Median 5 years | 9 |
Rennert et al. ( |
Population-based nested case-control study | Breast cancer | Multiple center | Israel | Female | Computerized prescription records | 3,731 | Mean 5.83 years | 8 |
Rennert et al. ( |
Population-based nested case-control study | Breast cancer | Multiple center | Israel | Female | Pharmacy records | 1,706 | Mean 5.8 years | 8 |
Pazianas et al. ( |
Population-based national register-based cohort study, | Colon cancer | Multiple center | Denmark | Male/female | Medical record | 38,118 | Mean 4.9 years | 8 |
Abrahamsen et al. ( |
Population-based national register-based cohort study, | Esophageal and Gastric Cancer | Multiple center | USA, Danmark | Male/female | Record of all hospitalizations and outpatient appointments | 153,030 | Mean 3.5 years | 9 |
Five studies were involved in the analysis of bisphosphonate use and OS. Figure
Summary estimates and 95% CIs for overall survival, cancer-specific survival and recurrence-free survival for associations between bisphosphonate use and survival of patients with solid tumors. Weights are from random effects analysis. CI, confidence interval; HR, hazard ratio; W (random), Weights (random effects model).
Six studies were involved in the analysis of bisphosphonate use and CSS. The result showed that bisphosphonate use was associated with prolonged CSS (HR 0.73, 95% CI 0.58–0.90). We noted that there was evidence of heterogeneity (I2 = 80.9%).
Three studies were involved in the analysis of bisphosphonate use and RFS. The result showed that bisphosphonate use was associated with prolonged RFS (HR 0.72, 95% CI 0.53–0.96). We noted that there was evidence of heterogeneity (I2 = 75.1%; Table
Meta-analysis of associations between bisphosphonate use and survival of patients with solid tumors.
Overall survival | 5 | 0.84 (0.76–0.93) | < 0.001 | 81.5, < 0.001 |
Cancer-specific survival | 6 | 0.73 (0.58–0.90) | < 0.001 | 80.9, < 0.001 |
Recurrence-free survival | 3 | 0.72 (0.53–0.96) | < 0.001 | 75.1, < 0.001 |
Colorectal cancer | 2 | 0.74 (0.43–1.27) | 0.02 | 81.4, < 0.001 |
Breast cancer | 2 | 0.81 (0.63–1.04) | < 0.001 | 90.7, < 0.001 |
Gastroesophageal cancer | 1 | 0.62 (0.40–0.98) | < 0.001 | 81.5, < 0.001 |
Colorectal cancer | 3 | 0.71 (0.44–1.15) | 0.005 | 80.9, < 0.001 |
Breast cancer | 3 | 0.73 (0.55–0.95) | 0.002 | 79.4, < 0.001 |
The results of subgroup analyses for the associations between bisphosphonate use and OS or CSS were presented in Tables
Subgroup analyzes for the associations between bisphosphonate use and overall survival of patients with solid tumors.
Geographical location | < 0.001 | ||||
Europe | 3 | 0.79 (0.57–1.12) | 48.1 | 0.146 | |
America | 2 | 0.72 (0.63–0.81) | 0 | 0.553 | |
Asia | 2 | 0.75 (0.42–1.35) | 84 | 0.012 | |
Study quality | < 0.001 | ||||
Low risk | 3 | 0.94 (0.88–1.01) | 76.3 | 0.015 | |
High risk | 4 | 0.71 (0.63–0.80) | 0 | 0.674 | |
Number of cases | |||||
< 10,000 | 2 | 0.75 (0.42–1.35) | 84 | 0.012 | 0.006 |
≥10,000 | 5 | 0.77 (0.63–0.94) | 78.4 | 0.001 | |
Type of cohort | 0.011 | ||||
Retrospective | 4 | 0.71 (0.48–1.05) | 72.1 | 0.013 | |
Prospective | 3 | 0.80 (0.65–0.99) | 86.7 | 0.001 | |
Gender | 0.41 | ||||
Female | 6 | 0.74 (0.59–0.92) | 84.2 | < 0.001 | |
Both gender | 1 | 0.93 (0.89–0.97) | N/A | N/A | |
Tumor stage | < 0.001 | ||||
I–IV | 5 | 0.92 (0.85–1.00) | 68.6 | 0.013 | |
I–III | 2 | 0.72 (0.63–0.81) | 0 | 0.553 | |
Source of exposure | N/A | ||||
Prescription database | 7 | 0.84 (0.76–0.93) | 81.4 | < 0.001 | |
Medical records | 0 | N/A | N/A | N/A | |
Statistical analysis | 0.07 | ||||
Time-dependent | 5 | 0.85 (0.77–0.94) | 85.8 | < 0.001 | |
Non time-dependent | 2 | 0.62 (0.40–0.98) | 0 | 0.361 | |
Adjustment for major confounders | 0.303 | ||||
Yes (Age, stage and grade) | 4 | 0.76 (0.60–0.97) | 89 | < 0.001 | |
No (Age, stage and grade) | 3 | 0.79 (0.57–1.12) | 48.1 | 0.146 |
Subgroup analyses for the associations between bisphosphonate use and cancer-specific survival of patients with solid tumors.
Geographical location | < 0.001 | ||||
Europe | 2 | 0.82 (0.46–1.45) | 88.4 | 0.003 | |
America | 3 | 0.65 (0.52–0.82) | 25.3 | 0.262 | |
Asia | 2 | 0.70 (0.34–1.43) | 76.0 | 0.041 | |
Study quality | 0.255 | ||||
Low risk | 4 | 0.64 (0.45–0.92) | 87.5 | < 0.001 | |
High risk | 3 | 0.82 (0.60–1.11) | 66.5 | 0.050 | |
Number of cases | < 0.001 | ||||
< 10,000 | 4 | 0.76 (0.54–1.07) | 77.3 | 0.004 | |
≥10,000 | 3 | 0.66 (0.57–0.76) | 0 | 0.558 | |
Type of cohort | < 0.001 | ||||
Retrospective | 2 | 0.70 (0.34–1.43) | 76.0 | 0.041 | |
Prospective | 5 | 0.71 (0.56–0.89) | 66.7 | 0.017 | |
Gender | 0.037 | ||||
Female | 5 | 0.69 (0.52–0.91) | 78.2 | 0.001 | |
Both gender | 2 | 0.82 (0.46–1.45) | 88.4 | 0.003 | |
Tumor stage | 0.002 | ||||
I–IV | 4 | 0.79 (0.59–1.06) | 83.9 | < 0.001 | |
I–III | 3 | 0.65 (0.52–0.82) | 25.3 | 0.262 | |
Source of exposure | 0.009 | ||||
Prescription database | 6 | 0.77 (0.62–0.95) | 79.7 | < 0.001 | |
Medical records | 1 | 0.48 (0.30–0.76) | N/A | N/A | |
Statistical analysis | N/A | ||||
Time-dependent | 7 | 0.73 (0.58–0.90) | 80.9 | < 0.001 | |
Non time-dependent | 0 | N/A | N/A | N/A | |
Adjustment for major confounders | 0.001 | ||||
Yes (Age, stage and grade) | 6 | 0.75 (0.60–0.95) | 75.0 | 0.001 | |
No (Age, stage and grade) | 1 | 0.62 (0.50–0.76) | N/A | N/A |
For individual cancer type, we found that bisphosphonate use was significantly associated with longer OS for patients with gastroesophageal cancer (HR 0.62, 95% CI 0.40–0.98), but not for patients with breast (HR 0.74, 95% CI 0.43–1.27) or colorectal cancer (HR 0.81, 95% CI 0.63–1.04; Figure
Summary estimates and 95% CIs for overall survival associations between bisphosphonate use and survival of patients with solid tumors according to tumor location. Weights are from random effects analysis. CI, confidence interval; HR, hazard ratio; W (random), Weights (random effects model).
Summary estimates and 95% CIs for cancer-specific survival of associations between bisphosphonate use and survival of patients with solid tumors according to tumor location. Weights are from random effects analysis. CI, confidence interval; HR, hazard ratio; W (random), Weights (random effects model).
As there were limited number of studies (< 10) included in each study outcome subset, we did not assess publication bias. Sensitivity analyses by omitting one study at a time and recalculating the summary HRs for the remaining studies yielded consistent results for each outcome of OS, CSS and RFS subset (Figures
In this meta-analysis of eight studies investigating the prognostic effect of bisphosphonate use on the survival of patients with solid cancer, we found that bisphosphonate use was associated with improved survival for patients with solid cancer in terms of OS, CSS and RFS. However, we did not find improved OS or CSS in all type of solid cancers. We noted considerable inter-study heterogeneity that could be explained by study design, tumor stage or sample size.
This meta-analysis is the first one to provide evidence for substantial prognostic role of bisphosphonate use in patients with solid cancer. Although several previous studies showed no prognostic association of bisphosphonate use in solid cancers (
Despite the above limitations, the present study also has several strengths. Firstly, we developed a set of comprehensive database search strategy for search without language limits, which could reduce the risk of missing studies to a great deal. Secondly, this is the largest meta-analysis with more than 153,000 participants involved which provided the most powerful evidence regarding this subject ever to date. Thirdly, at least two authors independently did the literature search, study selection and study quality assessment and the results were cross-checked by a senior author, which guaranteed the objective assessment of the systematic review.
Studies have indicated that several major side effects or adverse events are closely related to bisphosphonate use. It seems that people with bisphosphonate use have a higher risk of developing acute-phase reactions (fever, myalgias, lymphopenia, etc.), hypocalcemia, iritis, renal toxicity and osteonecrosis of the jaw (
X-YW and JX study concept and design. D-TW, ZX, M-LX, G-RL, W-LZ, and X-FL acquisition of data. D-TW, X-YW, and JX analysis and interpretation of data. D-TW and X-YW drafting of the manuscript. All authors critical revision of the manuscript for important intellectual content. X-YW and JX study supervision.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The Supplementary Material for this article can be found online at: