Pretreatment Thrombocytosis as a Prognostic Factor in Women with Gynecologic Malignancies : a Meta-analysis

The clinical observation of thrombocytosis (defined as a platelet counts >400×109/L) and malignancies was described over a century ago (Trousseau et al., 1867; Riess et al., 1872). According to the experimental evidence, it suggested that platelets actively promote cancer cell dissemination by protection of circulating cancer cells from immune surveillance, negotiation of cancer-cell arrest in the microvasculature, and stimulation of angiogenesis (Nieswandt et al., 1999; Borsig et al., 2008). Recent studies have addressed the prevalence and prognostic impact of thrombocytosis in various malignancies, including stomach, lung, kidney, uterus, and gynecologic malignancies (Zeimet et al., 1993; Menczer et al., 1998; Tomita et al., 2008; Heras et al., 2010; Cho et al., 2011). However, up to now, there was no metaanalysis comprehensively analyzing the prognostic value of thrombocytosis in cancer patients. The aim of this study was to comprehensively and quantitatively summarize the evidence for the use of pretreatment or preoperative thrombocytosis to evaluate its prognostic value for women with gynecologic malignancies.


Introduction
The clinical observation of thrombocytosis (defined as a platelet counts >400×10 9 /L) and malignancies was described over a century ago (Trousseau et al., 1867;Riess et al., 1872).According to the experimental evidence, it suggested that platelets actively promote cancer cell dissemination by protection of circulating cancer cells from immune surveillance, negotiation of cancer-cell arrest in the microvasculature, and stimulation of angiogenesis (Nieswandt et al., 1999;Borsig et al., 2008).Recent studies have addressed the prevalence and prognostic impact of thrombocytosis in various malignancies, including stomach, lung, kidney, uterus, and gynecologic malignancies (Zeimet et al., 1993;Menczer et al., 1998;Tomita et al., 2008;Heras et al., 2010;Cho et al., 2011).However, up to now, there was no metaanalysis comprehensively analyzing the prognostic value of thrombocytosis in cancer patients.The aim of this study was to comprehensively and quantitatively summarize the evidence for the use of pretreatment or preoperative thrombocytosis to evaluate its prognostic value for women with gynecologic malignancies.

Identification and Eligibility of Relevant Studies
We identified all studies, published or not, respectively targeting all thrombocytosis in patients with gynecologic malignancies, by an electronic search using online PubMed (MEDLINE) and EMBASE, with the search strategies based on combinations of "ovarian carcinoma", "endometrial carcinoma", "cervical carcinoma", "vulvar carcinoma", "platelet count", "thrombocytosis" and "prognosis".Last query was updated on April 6, 2012.References of retrieved articles were also screened to find out any studies missed by the search strategies.
After reading by two independent reviewers (Yu M. and Ma XL.), the candidate articles were identified for the meta-analysis studies based on title and abstract.Fulltext review (Liang CS.) was retrieved when it cannot be categorized and abstract review was restricted to English, and the full-texts not in English were excluded.Reported data required for meta-analysis were then identified.Prespecified quality-related inclusion or exclusion criteria were not used and each study had not been weight by a quality score because no such score had received general agreement for meta-analyses of observational studies (Altman et al., 2001).We made an effort to contact investigators by e-mail to get unpublished data regarding thrombocytosis, gynecologic malignancies and survival.

Definitions and Standardizations
We defined thrombocytosis as platelet counts >400×10 9 /L in patients with gynecologic malignancies.For statistical analysis, the value of 400×10 9 was used as a cut-off to separate the patients with high versus low platelet count, while the others did not mention.
The main reported data required for our meta-analysis was 5-year overall survival (OS).All of these patients were followed up until death or for at least 60 months.We also listed age, tumor location, number of patients got thrombocytosis, clinical stage and p values.

Data Extraction
Two reviewers (Yu M. and Liu L.) independently extracted data from all primary studies.In particular, we recorded each report using a standardized data collection form, with the following items: year of publication, the first author, country of origin, tumor location, number of patients analyzed, number of patients got thrombocytosis, mean or median age, clinical stage, p value, number of events in, OS or cumulative survival and survival curves.Disagreements were also resolved by consensus between the two reviewers, and studies were all retrospective.

Statistical Analyses
REVMAN, version 5.1 was used for our analysis and publication bias was evaluated using the Begg's funnel plot and tested by STATA 11.0 (STATA Corporation, College Station, TX) (Yu M., Niu ZM. and Wu YK.).A study was considered significant when the p value was less than 0.05 in univariate analysis.For the quantitative aggregation of survival results, we measured the impact of thrombocytosis on survival by estimating the risk ratio (RR) between the thrombocytosis and normal (defined as a platelet count <400×10 9 /L) groups.An observed RR>1 indicated worse outcome and would be considered statistically significant if the 95%CI did not overlap 1, with p<0.05.The study was termed "positive" when pretreatment or preoperative thrombocytosis predicted poorer survival while "negative" did not.The simplest method to get RR and their 95% confidence interval (CI) was to find in the original article, while none of our articles consist.Thus, the numbers of events in groups with and without thrombocytosis and the numbers of patients in each group were used to determine the RR estimate.When the data of the total numbers of events was not available directly in the article but as graphical survival plots, and the number of steps on the curves equaled the number of events given in the publication, the calculations were done, presuming that the rate of censored patients was constant during the study follow-up (Parmar et al., 1998), and we read Kaplan-Meier curves (Yu M. and Qin X.) by Engauge Digitizer version 4.1 (free software down-loaded from http://sourceforge.net).Considering the many sources of heterogeneity between studies and consequently between their individual RR estimates, a random-effect model (Der Simonian and Laird's method) was used to calculate the overall RR.

Eligible Studies
Our electronic search algorithm retrieved a total of 315 references for pretreatment or preoperative thrombocytosis and gynecologic malignancies, and 22 full texts were evaluated.20 reports were identified with thrombocytosis in patients with gynecologic malignancies while one of excluded was an editorial for another article included (Rodriguez et al., 1994;Rodriguez et al., 1994) and the other was in German (Zeimet et al., 1993).Of all the remaining 20 reports, 6 were excluded: one used 300×10 9 to define the thrombocytosis (Lukas et al., 2000), and five lacked informative clinical data (Zeimet et al., 1994;Menczer et al., 1998;Hernandez et al., 2000;Tamussino et al., 2001;Kuyumcuoglu et al., 2010).
Characteristics of the 14 eligible studies are listed in Table 1.6 reports originated from U.S., 2 from England, 2 from Turkey, 1 from Australia, 1 from China, 1 from Norway and 1 from Korea.All of the eligible studies were observational retrospective studies.According to the information, the median or mean age of the enrolled patients ranged from 45.5 to 68.1 years.In the 14 studies included for meta-analysis, only 1 study was about vulvar cancer, 3 were about endometrial cancer, 5 were cervical cancer and 5 were ovarian cancer.Meanwhile, the tumor was located in the cervix in 1691 patients (48.5%), endometrium in 223 patients (6.4%), ovary in 1375 patients (39.4%), and vulva in 201 patients (5.8%).Of the 3490 women with gynecologic malignancies, 709 (20.3%) had thrombocytosis at primary diagnosis, indicating platelet counts >400×10 9 /L.The incidence of thrombocytosis was 14.9% (30/201) for vulva cancer, 17.5% (39/223) for endometrial cancer, 12.6% (213/1691) for cervical cancer and 31.1% (427/1375) for ovarian cancer, respectively.

Discussion
There has been great interest in identifying prognostic markers for patients with gynecologic as these markers can help guide clinical decision-making regarding therapy and outcomes.In this paper, we examined the correlation of thrombocytosis (platelet counts >400×10 9 /L) with OS before any curative treatment or surgery of gynecologic malignancies.Our meta-analysis for the 14 primary studies from the published literature according to the survival data in patients with gynecologic malignancies, showed that the thrombocytosis (platelet counts >400×10 9 /L) at primary diagnosis is associated with worse survival (RR=1.62)and proved to be a significant but weaker prognostic factor for patients in stage Ⅲ to Ⅳ in comparison with the entire group (RR=1.29),although it had evidence that thrombocytosis is more frequent in advanced disease in patients with ovarian cancer, endometrial cancer and cervical cancer (Gücer et al., 1998;Gücer et al., 2004;Gorelick et al., 2009).For the cervical cancer patients in stage ⅠB, especially, it showed a significant effect on the survival too (RR= 1.73).When adjusted for different gynecologic malignancies, all three of the meta-analyses for endometrial cancer (RR= 2.09), cervical cancer (RR = 1.86) and ovarian cancer (RR = 1.52) gave statistically significant results, favoring a link between thrombocytosis and poor survival, while the vulvar cancer did not (RR = 0.43).
findings were consistent with the outcome on the link between thrombocytosis and survival in patients with another common cancer for women, breast cancer, which showed an incidence of 3.7% (161/4300) and a strong relation (n=4300,RR= 2.04,], p<0.0001) (Taucher et al., 2003).We finally got the only study about breast cancer and it got the maximum sample size (n=4300).To avoid bias, four malignancies were ultimately included in our study, as endometrial, cervical, ovarian and vulvar cancer.However, only one study had evaluated the correlation between thrombocytosis and survival for patients with vulva cancer, and reported that thrombocytosis had no effect on 5-year survival rate in 201 patients with primarily surgical treatment (Lavie et al., 1999).Compared to the others, the studies about prognosis of patients with ovarian cancer lagged behind (2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012), and the earliest studies were all about cervical cancer (1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999).The overall statistical link we got between thrombocytosis and survival was rather weak, with a global RR of 1.62.While as a rule of the thumb, a prognostic factor with a RR <2 is of limited practical use (Hayes et al., 2001), only the breast and endometrial cancer subgroup got a RR>2.
There were no studies which had analyzed the prognostic value of thrombocytosis in tumors yet, and thrombocytosis has repeatedly been identified as a marker of gynecologic malignancies with independent prognostic value.Our study aimed to summarize the evidence for the use of pretreatment or preoperative thrombocytosis to evaluate its prognostic value for women with gynecologic malignancies.According to the International Federation of Gynecologists and Obstetricians (FIGO) guidelines, all patients had received surgery, on the basis of primary surgical staging with the intent of optimal tumor cytoreduction, radiotherapy or chemotherapy.
The precise mechanisms underlying the cancerassociated thrombocytosis are not clearly understood.Although the thrombocytosis should not be thought of as an isolated factor involved in the cancer growth and metastatic process, it might represents an important therapeutic prospect.For example, heparin given as thromboembolic prophylaxis in the perioperative period may incidentally act to reduce the assistance of platelets to tumor spread (Nash et al., 2002).However, it could also results in a lowering of the platelet count by administration of cytotoxic drugs in chemotherapy and is currently unclear whether decreasing platelet counts has a place in the treatment of malignant disease.
Though we attempted to minimize publication bias by searching completely, it is unavoidable that some data was missing for various reasons such as publishing language only in German (Zeimet et al., 1993), failing to get the survival data in the full-text or the authors (Zeimet et al., 1994;Menczer et al., 1998;Hernandez et al., 2000;Tamussino et al., 2001;Kuyumcuoglu et al., 2010), or unpublished or ignored studies.All of our studies are retrospective observational studies, which are more prone to many biases than prospective randomized controlled studies (Grimes et al., 2002).Between-study heterogeneity is significant in our study, but we tried to reduce the variability by screening the literature using the same standard, such as the same experimental design and definition of thrombocytosis in prognostic meta-analysis.Furthermore, results should be interpreted with caution

Figure 1 .
Figure 1.Meta-analysis of the Association Between Thrombocytosis and the Risk of Death at 5 Years.Each study is shown by the name of the lead author and the RR with 95% CIs.The summary RR and 95% CIs (according to random effect calculations) are also shown (ALL)

Figure 4 .
Figure 4. A. Meta-analysis of the association between thrombocytosis and the risk of death at 5 years for patients with cervical cancer.B. Meta-analysis of the association between thrombocytosis and the risk of death at 5 years for patients with ovarian cancer.C. Meta-analysis of the association between thrombocytosis and the risk of death at 5 years for patients with endometrial cancer.D. Meta-analysis of the association between thrombocytosis and the risk of death at 5 years for patients with cervical cancer in stage IB.Each study is shown by the name of the lead author and the RR with 95% CIs.The summary RR and 95% CIs (according to random effect calculations) are also shown