Abstract
Aim: The current study aimed to determine the prognostic significance of thrombocytosis in patients with colorectal cancer (CRC) by a meta-analysis of the literature. Patients and Methods: The meta-analysis followed the 2009 guidelines of Preferred Reporting Items for Systematic Reviews and Meta-Analyses. A systematic literature review was conducted from PubMed and Web of Science for articles published up to May 15, 2015. Sixteen studies with a total of 5,619 patients met the inclusion criteria. Hazard ratios and 95% confidence intervals were retrieved from the original articles, calculated from the published Kaplan–Meier survival curves, or the corresponding authors were contacted for additional information. Heterogeneity was assessed using the I2 statistic and Chi-square tests. Publication bias was assessed by Begg's funnel plot, Egger's linear regression test and trim-and-fill method. Sensitivity analysis was performed to validate the reliability. Results: Thrombocytosis is associated with shorter overall, disease-free and cancer-specific survival. Overall survival is reduced in patients with thrombocytosis regardless of their clinical tumor stage, and ethnicity. Shortened disease-free survival is associated with elevated platelet count in the non-specific stage (I-IV), localized tumor (stage I-III), and in the Asian patient population. Thrombocytosis is further associated with reduced cancer-specific survival in the non-specific stage and in Asian patients. Finally, thrombocytosis is significantly related to female patients, colon tumor location, T3-4 stage, lymph node positivity, metastasis, undifferentiated histology and lymphatic involvement. Conclusion: Thrombocytosis portends adverse prognosis in CRC, and may serve as a clinically useful marker to facilitate risk stratification and guide postoperative management.
Colorectal cancer (CRC) is one of the most common malignancies worldwide. In the United States, CRC is the third most commonly diagnosed cancer and the third leading cause of cancer-related death, accounting for approximately 140,000 new cases and 50,000 deaths in 2014 (1). In recent years, CRC has also become the third most common malignancy in China, with steadily increasing prevalence partially owing to the lack of public awareness and health authority support for early screening (2). In the past decade, there has been continued progress in the early detection and surgical management of CRC, resulting in a gradual decline in cancer mortality. However, the 5-year recurrence of CRC is persistently high, ranging between 10-40% (1). Patients with relapse have dismal outcomes despite chemoradiotherapy. Therefore, some experts advocate neoadjuvant chemotherapy or immediate postoperative chemotherapy for early-stage disease (stage I and II) (3). This may in theory reduce recurrence; however, it is associated with significant systemic toxicity. It is, thus, critical to establish reliable prognostic factors for risk stratifying the patients where the benefits of chemotherapy outweigh the possible risks.
Paraneoplastic thrombocytosis is defined as an abnormally elevated platelet count (usually >400×103/μl) in the blood circulation of patients with cancer (4). This phenomenon has been known for over 100 years. In fact, according to an early study, approximately one-third of patients with cancer may have thrombocytosis at the time of diagnosis, and 40% of patients with an elevated platelet count may also have cancer (5). Thrombocytosis has been studied in a variety of solid tumor types, including gynecological, lung, renal, gastric and breast carcinomas (6-11). Most studies in solid tumors have suggested that thrombocytosis is associated with reduced patient survival. Given that platelet measurement is standardized and readily available in every clinical laboratory, it is an attractive and practical surrogate marker for predicting disease course and for guiding clinical therapy.
Thrombocytosis in gastrointestinal cancer, including CRC, has also been widely reported (9, 12-14). Epidemiological studies investigating the association between thrombocytosis and CRC, however, failed to reach a consensus. For example, a Hungarian study concluded that patients with CRC with thrombocytosis and hepatic metastatic disease had worse overall (OS) and disease-free (DFS) survival (15). By contrast, a British study did not find any correlation between thrombocytosis and survival in a cohort of 630 patients (16).
Against this background, we performed a meta-analysis of all relevant published studies on this subject. Meta-analysis is a statistical approach of combining results from different studies into a common metric, and exploring the connection between study characteristics and findings. With this systematic approach, we wanted to better define the prognostic significance of thrombocytosis in the survival of patients with CRC.
Patients and Methods
Search strategy. A systematic literature review was conducted to assess the relationship of thrombocytosis and CRC. The search was performed in the PubMed (National Institutes of Health) and Web of Science (Thomson Reuters) databases for articles published up to May 15, 2015. The search terms included the following: “colon” or “colonic” or “rectum” or “rectal” or “colorectum” or “colorectal” or “cecum” or “cecal” or “appendix” or “appendiceal”, and “thrombocytosis” or “thrombocythemia” or “platelet” or “thrombocyte”, and “cancer” or “carcinoma”. The search was conducted in free text in PubMed and in “title” and “topic” in the Web of Science. Furthermore, our search was broadened by the manual screening of the references of the identified publications. All scientific articles were independently reviewed by the Authors via the titles and abstracts. For articles for which inclusion or exclusion could not be determined based on the titles and abstracts, the full-text files were retrieved and analyzed. The disagreements were resolved through discussion. The meta-analysis followed the 2009 guidelines of Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (17).
Study selection. Studies were considered eligible when meeting the following criteria: (i) prospective or retrospective epidemiological study design; (ii) examined the association of pre-treatment thrombocytosis with CRC. The exclusion criteria were as follows: (i) study population fewer than 50 patients; (ii) not published in English; (iii) letters, case reports, meeting abstracts, posters, editorial or review articles; (iv) median follow-up time less than 12 months; (v) cut-off value for defining thrombocytosis was vague; (vi) lack of data critical for evaluating the hazard ratio (HR) and its 95% confidence interval (CI). For duplicated publications from the same research populations, the most comprehensive articles were included in our meta-analysis.
Data extraction. Data were independently extracted from candidate studies by the Authors. In particular, a pre-established form was used to standardize the information recorded, including the name of the first author, year of publication, geographic location, the size of the study population, clinical stage, follow-up period, cut-off value and statistical method of the survival analysis. HR of the survival analyses and 95% CI were retrieved from the original articles as the endpoints of the meta-analysis. For articles lacking HR or 95% CI, the values were calculated using the Kaplan–Meier survival curves from the original articles, or the corresponding authors were contacted for additional information. For articles containing both univariate and multivariate analyses, the latter was chosen owing to combinatorial effects. Within the eligible studies, five articles investigated the relationship between thrombocytosis and pathological features of CRC (14, 18-21). The following features were analyzed by our study to determine their individual effect on survival: gender, tumor location, histology, TNM classification, lymphatic and venous involvement.
Statistical analysis. Heterogeneity of HR was assessed using the I2 statistic and Chi-square tests. When I2 was great than 50%, heterogeneity was considered present and the random-effects model was used; otherwise the fixed-effects model was used. A pooled odds ratio (OR) greater than 1 indicated that a significant association existed between thrombocytosis and pathological characteristics of CRC. Publication bias was accessed by Begg's funnel plot, Egger's linear regression test and trim-and-fill method. sensitivity analysis was performed to validate the reliability. Two-sided p-values of less than 0.05 were considered statistically significant. In our study, heterogeneity analysis, publication bias, sensitivity analysis and pooled ORs were conducted or analyzed in STATA 12.0 software (STATCorp, College Station, TX, USA).
Results
Eligible studies. We searched the PubMed and Web of Science databases with the noted key words and identified 918 candidate articles. Of these, we excluded 149 duplicates, 13 case reports, 13 reviews, eight meeting abstracts, eight letters, one poster and one editorial opinion. We then screened the titles and abstracts of the remaining 725 articles. Based on the inclusion criteria as outlined in the Materials and Methods, we selected 35 publications to further determine their eligibility (13-16, 18-45). We included 30 studies for the meta-analysis given that they fulfilled the exclusion criteria: full-text article with a study population great than 50; explicit cut-off value for defining thrombocytosis; and more than 12 months follow-up (13-15, 18-22, 25-28, 30-35, 37-45). Of these 30 studies, however, we had to eliminate 14 due to their lack of survival analysis or because data were insufficient for calculating HRs. For example, we excluded the 2015 study of Jósa et al. (33) because we realized that Baranyai et al. (15) studied the same population and included the same results. We illustrate the search process in Figure 1. Finally, we chose 16 studies with a total of 5,619 patients for our meta-analysis of thrombocytosis in patients with CRC (14, 15, 18-21, 37-45).
Flowchart of article selection.
As shown in Table I, all studies were published within the past 5 years. The study cohort included one study each from China and Mexico, two studies each from Korea, Hungary, United Kingdom and the United States of America, five studies from Japan, and one joint study from Japan and Korea. The number of patients per study varied, with a median of 238 patients and range of 1,463. It is noteworthy that the study by Wan et al. included 1,513 patients with a median follow-up of 46.7 months (40). The cut-off value for defining thrombocytosis also varied, with 60% of the studies (9/15) applying a value of 400×103/μl.
Meta-analysis
Heterogeneity. We evaluated the prognostic significance of thrombocytosis by comparing the survival of patients with thrombocytosis to that of patients with a normal platelet count. Of the selected studies, 13 examined OS, eight examined DFS and four examined cancer-specific survival (CSS) (Figure 2). With regard to OS, defined as the period from cancer diagnosis or treatment to death, our results indicate that thrombocytosis is associated with shorter OS (HR=1.61, 95% CI=1.19-2.04). of note, we used a random-effects model in the presence of heterogeneity (I2=66.5%, p<0.001). The studies proposing a strong poor prognostic role of thrombocytosis in CRC included those of Cravioto-Villanueva et al. (43), Kaneko et al. (20), Jósa et al. (45), and Toiyama et al. (21). For DFS, a greater HR was observed in our study cohort (HR=1.96, 95% CI=1.23-2.69) with a random-effects model (I2=69.3%, p=0.002). For CSS, we applied a fixed-effects model in the absence of heterogeneity (I2=20.0%, p=0.290), finding a HR of 1.82 and 95% CI=1.04-2.59.
Next, we stratified the analysis according to the clinical tumor stage: non-specified (all stages), localized tumor (stage I-III) and metastatic disease (stage IV) to identify the source of heterogeneity. For the non-specified stage (Figure 3A), seven articles studied OS, three studied DFS and three studied CSS, reporting poorer OS (HR=1.61, 95% CI=1.36-1.87), DFS (HR=2.33, 95% CI=1.65-3.00) and CSS (HR=1.69, 95% CI=1.04-2.34). We found no significant heterogeneity in the non-specified stage. For the localized tumor (stage I-III), three articles each reported OS and DFS. Elevated thrombocyte count was associated with shorter OS (HR=1.92, 95% CI=0.64-4.47) and DFS (HR=1.42, 95% CI=0.61-2.24). We found no heterogeneity when analyzing the OS for patients with localized tumor (I2=25.9%, p=0.259). On the contrary, we identified heterogeneity in the DFS of those with localized tumor (I2=56.4%, p=0.101; Figure 3B). Furthermore, with regard to the metastatic stage (stage IV), all three articles reported OS and showed an elevated platelet count portended worse OS (HR=3.60. 95% CI=1.62-5.58) with no heterogeneity present (I2=0.0%, p=0.762; Figure 3C).
Lastly, we investigated the impact of ethnicity on patient survival. We divided the study cohort into Asian, British and American, and other (Mexico and Hungary). In the Asian group, there were five OS, four DFS and three CSS studies, respectively. Meta-analysis of the Asian population showed that an elevated thrombocyte count was associated with poorer OS (HR=1.52, 95% CI=1.06-1.97), DFS (HR=2.29, 95% CI=1.65-2.93) and CSS (HR=2.22, 95% CI=1.19-3.25) with no significant heterogeneity revealed (Figure 4A). The British and American groups included four OS, two DFS and one CSS study, and only the OS was sufficient for analysis (Figure 4B). Here we observed worse OS (HR=1.42, 95% CI=0.87-1.97, I2=82.6%, p=0.001) with thrombocytosis. The last category contained four OS and two DFS (not analyzed due to data insufficiency) studies conducted in Mexico and Hungary. Our analysis showed an elevated platelet count was associated with poor OS (HR=2.07, 95% CI=1.22-2.92, I2=2.7%, p=0.379; Figure 4C).
Thrombocytosis and clinicopathological features of CRC. Among the included 16 studies, five investigated pathological features such as tumor location, histology, TNM classification, and lymphatic and venous involvement (4, 18-21). We performed a meta-analysis of these variables. As shown in Table II, thrombocytosis had a significant association with female gender, tumor location in the colon, higher pathological T-stage, pathological positive N-stage, pathological positive M-stage, undifferentiated histology, and lymphatic involvement. Meanwhile, we noted no heterogeneity except for lymphatic and venous involvement (Table II).
Major characteristics of the eligible studies.
Meta-analysis of the association between thrombocytosis and clinicopathological features of colorectal cancer.
Publication bias and sensitivity analysis. Following the criteria by the Cochrane Handbook for Systematic Reviews of Interventions (46), we analyzed the publication bias in the OS group with more than 10 studies. Our Begg's test result revealed an asymmetric funnel plot. Egger's regression test showed that the bias coefficient was −0.005 (95% CI=−0.010 to −0.001, t-value=−2.63 and p=0.024). To this end, we performed a trim-and-fill analysis to further assess the publication bias. The test showed a symmetrical funnel plot: this implies a “well-behaved” data set in which publication bias is unlikely (Figure 5).
Lastly, we removed each study stepwise to perform the sensitivity analysis. Our results showed no significant change, indicating the reliability of our meta-analysis.
Discussion
Platelets play a crucial role in a variety of biological processes, ranging from hemostasis and vascular integrity through immunity to inflammation (47). An elevated platelet count, termed thrombocytosis, has been found in various pathological conditions. Platelets may rise in myeloproliferative disorders (primary thrombocytosis), or become elevated secondary to events such as infection, inflammation, trauma, surgery, and iron deficiency (secondary thrombocytosis).
Forest plot of patient survival. Meta-analysis indicates thrombocytosis is associated with shorter overall (OS), disease-free (DFS) and cancer-specific (CSS) survival in patients with colorectal carcinoma. HR: Hazard ratio; CI: confidence interval.
A special form of secondary thrombocytosis, termed paraneoplastic thrombocytosis, is associated with cancer. First reported in 1964 (5), paraneoplastic thrombocytosis phenomenon is now recognized as a poor prognostic indicator in many types of cancer including lung, gastric, breast, renal and gynecological cancer (6-11). It is only beginning to be understood how thrombocytosis is generated by cancer and how this may impair patient survival. Neoplasms are known to synthesize cytokines including interleukin (IL6). IL6 can enhance thrombopoietin production in the liver, which, in turn, may stimulate megakaryocytes in the marrow to produce platelets in excess (48). Circulating platelets may promote tumor growth via secretion of angiogenic or tumor growth factors, such as platelet-derived epidermal growth factor, thrombospondin and vascular endothelial growth factor (48). In addition, platelets may protect circulating tumor cells from attacks by the host immune system, thereby facilitating cancer metastasis (49).
There is increasing evidence that patients with CRC with thrombocytosis have worse OS due to a combination of higher recurrence and metastatic disease. In node-positive disease, the importance of thrombocytosis was reported to be equivalent to lymph node positivity (26). In node-negative patients, thrombocytosis was independently associated with DFS and OS in multivariate analysis, together with tumor depth, grade and lymphatic invasion (13). Patients with CRC with preoperative thrombocytosis prior to chemoradiotherapy had lower rates of radiographic and pathological response (44).
Forest plot analysis of patient survival according to clinical tumor stage of colorectal cancer. Thrombocytosis is associated with poorer overall (OS), disease-free (DFS) and cancer-specific (CSS) survival in patients with tumor of non-specified stage (all stages) (A), with shorter OS and DFS in patients with localized tumor (stage I-III) (B), and with worse OS in the metastatic group (stage IV). HR: Hazard ratio; CI: confidence interval.
Forest plot analysis of patient survival according to ethnicity. Thrombocytosis is associated with shorter overall (OS), disease-free (DFS) and cancer-specific (CSS) survival in the Asian group (A), with worse OS in the British and American group (B), and poorer OS in the group of other ethnicity (Mexico and Hungary) (C). HR: Hazard ratio; CI: confidence interval.
Hazard ratio funnel plots with pseudo 95% confidence interval of studies investigating the association between thrombocytosis and overall survival. Thirteen datasets (circles) represent the studies included in this analysis. Four datasets (circles with outer squares) represent the virtual studies with which the analysis shows a symmetrical funnel plot.
Although most studies implicated thrombocytosis as a negative prognostic factor in CRC, other studies were unable to replicate this observation. For example, Nyasavajjala et al. analyzed 630 consecutive patients with CRC from 2004 to 2007 and concluded that preoperative thrombocytosis was not a prognostic indicator of survival in patients with CRC regardless of pathological stage (16). This inconsistency in the literature prompted us to conduct this meta-analysis. Meta-analysis with systematic review of the literature allows the conclusive synthesis of existing scientific evidence. To many, it represents the top of the hierarchy of research methods.
Our meta-analysis of 16 independent studies with 5,619 patients unequivocally supports an adverse prognostic significance for thrombocytosis. The included studies contained multiple pathological and demographic features that allowed us to analyze the impact of thrombocytosis on different subgroups. Our study suggests that thrombocytosis is reliable predictor of unfavorable prognosis (OS and DFS) regardless of tumor clinical stage and ethnicity. The five studies in Table II allowed us to further individually analyze the clinicopathological features of CRC. We found CRC with thrombocytosis was remarkably associated with pathological T3, T4, N and M stages, undifferentiated histology, lymphatic involvement, colon tumor location (versus rectum) and the female population. Venous involvement, on the contrary, was not associated with paraneoplastic thrombocytosis.
Clearly, meta-analysis has its own limitations (50). For instance, publication bias could compromise the quality of the meta-analysis. We detected publication bias in our study when investigating the association of thrombocytosis and OS. For example, the important negative study by Nyasavajjala et al. was excluded from our analysis due to the lack of a survival curve and HRs (16). publication bias can also be introduced by excluding studies with smaller study populations, or articles published in languages other than English. Studies with negative results may never have been published. The limited number of studies precluded us from conducting an adequate assessment of publication bias and sensitivity analysis.
Heterogeneity is another common concern in meta-analysis (51). After detecting heterogeneity in the analysis, we conducted subgroup studies to validate the results and hoped to pinpoint the source of heterogeneity. The subgroup analysis revealed that DFS for patients with localized disease, and OS and DFS in the British and American populations could be a source of heterogeneity. In addition, American (and British) patient populations can be racially mixed. We also speculated two other potential factors that might have contributed to the heterogeneity. Firstly, there was great difference in the follow-up time, ranging from 17 to 65 months. secondly, there was no agreed definition of thrombocytosis. The cut-off values for defining thrombocytosis were likely determined by the hospital committees and local populations. This appears to the same issue that has been faced by the meta-analyses of thrombocytosis in other cancer types (52). Lastly, our meta-analysis is an observational study of retrospective studies. It is more likely to be affected by inherent recall and selection biases than prospective randomized controlled studies.
In summary, our meta-analysis suggests that thrombocytosis heralds a worse prognosis in patients with CRC. Thus, thrombocytosis may serve as a clinically useful marker for facilitating risk stratification and guiding postoperative management. To our knowledge, the present study represents the first systematic review and meta-analysis of thrombocytosis in CRC.
Footnotes
This article is freely accessible online.
Conflicts of Interest
None.
Funding
The Authors received no funding for this work.
- Received June 23, 2017.
- Revision received July 17, 2017.
- Accepted July 20, 2017.
- Copyright© 2017, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.