Comparison of clinical utilities of the platelet count and platelet-lymphocyte ratio for predicting survival in patients with cervical cancer : a single institutional study and literature review

Objective: To compare the clinical utilities of the platelet count and plateletlymphocyte ratio (PLR) for predicting survival in patients with cervical cancer. Results: Multivariate analyses demonstrated that thrombocytosis and elevated PLR were found to be independent prognostic factors for progression-free survival (PFS, P = 0.0077, P = 0.044) and overall survival (OS, P = 0.025, P = 0.019) in separate Multivariate analyses. In the ROC analysis, the platelet count showed a significantly greater area under the ROC curve (AUC) value than that of PLR for predicting patient recurrence (0.5941 versus 0.5331, p = 0.018) and survival (0.6139 versus 0.5468, p = 0.029). In patients without thrombocytosis, elevated PLR correlated with shorter survival (PFS, P = 0.041; OS, P = 0.017). In contrast, PLR in patients with thrombocytosis did not provide prognostic information. We divided patients into 3 prognostic groups using platelet counts and PLR: high-risk (thrombocytosis with any PLR); intermediate-risk (elevated PLR without thrombocytosis); low-risk (none of the above), which allowed for individualized and accurate survival estimates. Materials and Methods: The baseline characteristics and clinical outcomes of cervical cancer patients were identified. Patients were grouped according to their pretreatment platelet counts or PLR, and clinicopathological characteristics and patient survival were then compared between these groups. The clinical utilities of the platelet count and PLR were compared using a time-dependent receiver operating characteristic (ROC) analysis. Conclusions: Pretreatment thrombocytosis and elevated PLR were identified as independent predictors in cervical cancer patients. Platelet counts were superior to PLR for predicting the prognosis of uterine cervical cancer patients. Our prognostic model consisting of platelet counts and PLR offers individualized survival estimates.


INTRODUCTION
Cervical cancer is the second most common type of cancer affecting women worldwide and has an annual incidence of 530,000 new cases. Although current standard treatments for invasive cervical cancer are potentially curative, a significant number of patients develop recurrence and die of disease progression, with approximately 250,000 deaths being reported globally each year [1].
The identification of new prognostic factors for cervical cancer will improve our understanding of cervical cancer biology, contribute to the stratification of patients into risk groups, and identify those at a high risk of recurrence after the standard initial treatment.

Research Paper
Oncotarget 55395 www.impactjournals.com/oncotarget Since then, an increasing number of studies have reported thrombocytosis in patients with cancer from various origins, and demonstrated that it is associated with poor patient prognosis [20][21][22][23]. To the best of our knowledge, 13 studies have investigated the prognostic implications of thrombocytosis in cervical cancer patients, with about half suggesting that thrombocytosis is an independent prognostic factor in cervical patients (Table 1). A recent study on ovarian cancer indicated that paraneoplastic thrombocytosis is due to the enhancements induced in hepatic thrombopoietin synthesis by tumor-derived IL-6. Moreover, the inhibition of thrombopoietin and IL-6 expression abrogated thrombocytosis in tumor-bearing mice and significantly enhanced the therapeutic efficacy of paclitaxel in mouse models of epithelial ovarian cancer [24]. Thus, thrombocytosis is now regarded not only as a prognostic indicator, but also as a potential therapeutic target in human cancers.
As shown in Table 1, 4 studies have investigated the prognostic implications of increased PLR in cervical cancer patients, with 3 studies suggesting that increased PLR is an independent predictor of survival. However, since most of the studies described above only included surgically-treated early-stage cervical cancer patients, the prognostic significance of PLR in cervical cancer remains unclear. Moreover, the clinical utilities of platelet counts and PLR have not yet been compared, and there is currently no information on how physicians may distinguish between platelet counts and PLR in the management of cervical cancer.
In the present study, we first investigated the prognostic significance of elevated platelet counts and PLR in patients with FIGO stage IA-IVA cervical cancer. Then, we compared the clinical utilities of platelet counts and PLR for predicting the survival of patients with cervical cancer. Finally, we established a prognostic model using platelet counts and PLR to predict patient survival.

Prognostic significance of platelet counts
The clinicopathological characteristics of patients according to platelet counts are shown in Table 2. Among 684 patients, 87 (12.7%) had platelet counts equal to or greater than 350,000/ml (the thrombocytosis group) at the time of the initial diagnosis. Patients with thrombocytosis were significantly younger (P < 0.001) and presented with a more advanced clinical stage (P = 0.036) than those without thrombocytosis. Thrombocytosis correlated with significantly shorter PFS (P < 0.001) and OS (P < 0.001) in the univariate analysis and Kaplan-Meier analysis (Tables 3, 4, Figure 1A). In the multivariate analysis (Tables 3, 4), in addition to an advanced clinical stage, non-SCC histology, pelvic node metastasis, and larger tumor size, an elevated platelet count (> 350,000/μl) was found to be an independent prognostic factor of PFS (HR, 1.63; 95% CI, 1.14-2.28; P = 0.0077) and OS (HR, 1.56; 95% CI, 1.06-2.24; P = 0.025).

Prognostic significance of PLR
ROC curves were described to select the optimal cut-off value for PLR (Supplementary Figure 1). The cut-off values of PLR for PFS and OS were 125.23 and 131.44, respectively. The clinicopathological characteristics of patients according to PLR are shown in Table 2 and Supplementary Table 1. Among 684 patients, 300 (43.9%) and 263 (38.5%) displayed PLR equal to or greater than 125.23 and 131.44, respectively. Patients with elevated PLR were significantly younger (P = 0.0059, P = 0.0013) and presented with a more advanced clinical stage (P = 0.014, P = 0.035) than those with normal PLR. As shown in Figure 1B, elevated PLR correlated with significantly shorter PFS and OS (PFS: P = 0.0045, OS: P = 0.0022). In the multivariate analysis, in addition to an advanced clinical stage, non-SCC histology, pelvic node metastasis, and larger tumor size, elevated PLR remained an independent prognostic factor of PFS (

Comparison of utilities of platelet counts versus PLR
In order to compare the clinical utilities of platelet counts and PLR for predicting patient prognoses, ROC curves for platelet counts and PLR were generated and compared ( Figure 2A). The area under the ROC curve (AUC) for predicting recurrence using platelet counts and PLR were 0.5941 (95% CI, 0.5448-0.6415) and 0.5331 (95% CI, 0.4833-0.5822), respectively. The AUC for predicting survival using platelet counts and PLR were 0.6139 (95% CI, 0.5552-0.6695) and 0.5468 (95% CI, 0.4889-0.6034), respectively. Platelet counts showed significantly greater AUC values than PLR for predicting recurrence (p = 0.018) and survival (p = 0.029).

Prognostic models using platelet counts and PLR
In order to establish a model for the prediction of life expectancy, PFS and OS were first assessed according to platelet counts and PLR. As shown in Figure 2B, in patients without thrombocytosis, elevated PLR correlated with shorter PFS (P = 0.041) and OS (P = 0.017). In contrast, in patients with thrombocytosis, survival was not influenced by PLR, indicating that it does not provide any prognostic information in this patient population ( Figure 2C).
Based on these results, we finally established a prognostic model in which patients were divided into 3 prognostic groups ( Figure 3A(i), 3B(i)): high-risk (patients Oncotarget 55396 www.impactjournals.com/oncotarget with thrombocytosis regardless of PLR); intermediaterisk (patients with elevated PLR without thrombocytosis); low-risk (none of the above). As shown in Supplementary Table 2, differential treatment outcomes were observed in association with the risk classifications. When PFS and OS were compared between the groups, patients in the high-risk group showed significantly lower PFS and OS rates than those in the intermediate-risk group (Figure 3A

DISCUSSION
In the present study, we showed that an elevated platelet count (≥ 350,000/μl) was an independent predictor of shorter PFS and OS in cervical cancer patients. These results are consistent with previous findings (Table 1).
Among our study population, 12.7% of patients displayed elevated platelet counts (≥ 350,000/μl), which correlated with a younger age and advanced clinical stage. We also observed elevated PLR (≥ 125.23 for PFS and ≥ 131.44 for OS) in approximately 40% of patients, and identified it as an independent predictor of shorter PFS and OS in cervical cancer patients. This result is also consistent with previous findings (Table 1). However, since most of the studies that previously investigated the significance of increased PLR only included surgically-treated early-stage cervical cancer patients (Table 1), the present study provides a novel insight into PLR in cervical cancer treatment: elevated PLR at the initial diagnosis is prognostically important regardless of the clinical stage and treatment modality. Moreover, our ROC analysis demonstrated, for the first time, that platelet counts are significantly superior to PLR for the prediction of patient prognoses. We consider this result to be clinically important because it suggests that platelet counts need to be preferentially examined in patients with cervical cancer. There are currently no universally accepted risk classifications that may be applied to all cervical cancer patients: i.e. patients treated with surgery, definitive radiotherapy, and chemotherapy. Thus, the results of the present study may have valuable clinical implications. Since the present study includes stage IA-IVA cervical cancer patients treated with surgery, definitive radiotherapy, or chemotherapy, the prognostic model proposed herein may be applied to all cervical cancer patients. Moreover, our prognostic model requires only low-cost peripheral blood examinations to identify a group of patients at high risk of recurrence. As shown in Figure 3, we demonstrated that it was possible to divide patients into 3 prognostic groups using platelet counts and PLR: high-risk (patients with thrombocytosis regardless of PLR); intermediate-risk (patients with elevated PLR without thrombocytosis); low-risk (none of the above). This prognostic model may have advantages that are relevant to clinical practices: this simple model offers individualized survival estimates ( Figure 3A (ii), 3B (ii)).
In addition, this model may enable physicians to offer closer follow-ups for patients in the intermediate-and high-risk groups. The results shown in Figures 2-3 also provide important information on the clinical applications of platelet counts and PLR: we recommend that platelet counts be initially examined for survival estimations in cervical cancer patients. PLR may then be evaluated in patients without thrombocytosis only because it did not provide prognostic information on patients with thrombocytosis.
Based on the poor prognosis of cervical cancer patients who display elevated platelet counts or PLR, novel treatment strategies need to be developed. The mechanisms responsible for increased platelet production in cervical cancer and subsequent increases in the aggressiveness of the disease remain poorly understood. However, theoretically, treatments targeting thrombopoiesis-stimulating cytokines or growth factors, their receptors, or their downstream effectors may exhibit therapeutic efficacy in cervical cancer patients displaying pretreatment thrombocytosis. In a previous study, the inhibition of thrombopoietin or IL-6 prevented the development of thrombocytosis in mice and significantly enhanced the therapeutic efficacy of paclitaxel in mouse models of epithelial ovarian cancer [24]. Thus, in order to obtain a clearer understanding of platelet count alterations and advance the development of novel treatments, further mechanistic investigations on cervical cancer are warranted.
The limitations of our study need to be addressed. The first limitation is that the present study was conducted at a single institution. We intend to verify our clinical findings in collaborative multi-institutional studies in the future. Another limitation is the retrospective nature of the Oncotarget 55398 www.impactjournals.com/oncotarget present study. The significance of elevated platelet counts, PLR, and our prognostic model consisting of platelet counts and PLR need to be prospectively evaluated in future studies. The second limitation is the cut-off values used for thrombocytosis. In the present study, we defined elevated platelet counts as greater than or equal to 350,000/μl. The cut-off values for thrombocytosis in previous studies that investigate the significance of thrombocytosis in cervical cancer ranged between 200,000/μl and 400,000/μl, with 400,000/μl being the most popular cut-off value (Table 1). However, most of the studies listed in Table 1 were from countries other than Japan. In studies on various malignant tumors from Japanese institutions, the cutoff values for thrombocytosis were lower: most studies employed a cut-off value between 22,000/μl and 370,000/μl to define thrombocytosis [23,[25][26][27][28][29]. The reason why a lower cut-off value was employed in studies from Japanese institutions currently remains unknown; however, the   baseline platelet count in cancer patients may differ due to ethnicity. We also showed that patients with thrombocytosis were significantly younger (P < 0.001) than those without thrombocytosis. This result is consistent with a recent finding on cervical cancer [17]. However, the reason for this phenomenon remains unknown. Thus, the optimal platelet threshold for diagnosing thrombocytosis and the underlying mechanisms responsible for the development of thrombocytosis need to be investigated in future studies.
In conclusion, thrombocytosis and elevated PLR at the time of the initial diagnosis were identified as independent predictors of PFS and OS in FIGO stage IA-IVA cervical cancer patients. Platelet counts were significantly superior to PLR for predicting patient prognoses. Our proposed prognostic model consisting of platelet counts and PLR offers individualized and accurate survival estimates.

Patients
Permission to proceed with data acquisition and analyses was obtained from the Institutional Review Board of Osaka University Hospital. A list of patients diagnosed with FIGO stage IA-IVA cervical cancer and treated at Osaka University Hospital between November 1993 and December 2011 was generated from our institutional tumor registry, and their clinical data were retrospectively analyzed. Patients who had been diagnosed with other types of cancers within the past 5 years, had a history of splenectomy, myeloproliferative disorders, or acute inflammatory disease were excluded. Of the 684 patients included in the present study, 286 had been examined in a previous clinical study [13]. www.impactjournals.com/oncotarget

Treatment and post-treatment follow-up
Patients were treated in accordance with institutional treatment guidelines. Briefly, patients with FIGO stage IA2-IIB cervical cancer and younger than 70 years were treated with radical hysterectomy plus pelvic lymphadenectomy with or without adjuvant radiotherapy as described previously [30]. Adjuvant radiotherapy with or without platinum-based concurrent chemotherapy, was indicated when a patient's pathological report displayed any one of the following 'high-risk' prognostic factors: parametrial invasion, pelvic lymph node metastasis, or a positive surgical margin, or one of the following 'intermediate-risk' prognostic factors: deep stromal invasion, lymphovascular space invasion, or a large tumor (more than 4 cm in diameter), as reported previously [30]. Patients with FIGO stage III-IV disease, patients with FIGO stage I-II disease and older than 70 years, or patients with FIGO stage IA2-IIB disease and younger than 70 years who desired definitive radiotherapy rather than surgery were treated with definitive radiotherapy consisting of external beam radiation therapy followed by high-doserate intracavitary brachytherapy with or without platinum-based concurrent chemotherapy as described previously [31]. Patients with systemic disease were primarily treated with platinum-based chemotherapy as described previously [32,33]. Follow-up examinations performed after the initial treatment were conducted by gynecological oncologists or/and radiation oncologists at regular intervals in an outpatient clinic, as reported previously [31,34].

Definition of elevated platelet counts and PLR
During the period between the first presentation and the start day of the initial treatment, all patients underwent at least 2 blood tests including complete blood counts. Thrombocytosis was defined as platelet counts equal to or greater than 350,000/μl on at least 2 separate occasions, as described previously [13]. Elevated PLR for predicting progression-free survival (PFS) or overall survival (OS) were defined as PLR equal to or greater than 125.23 or 131.44, respectively (Supplementary Figure 1). The cutoff values for PLR were defined based on the maximum Youden index (i.e. sensitivity+specificity-1) in the timedependent receiver operating characteristic (ROC) curve for PFS and OS, as reported previously [35,36].

Statistical analysis
PFS was defined as the time from the date of therapy to the date of the first physical or radiographical evidence of disease progression. OS was defined as the time from the date of therapy to the date of death.
Time-dependent ROC curves were generated to evaluate the diagnostic performance of platelet counts and PLR for predicting recurrence or death at 3 years after the treatment. Differences in AUCs were analyzed according to the methods described in a previous study [37].
Continuous data were compared between the groups using the Student's t-test or Log-rank test, where appropriate. Frequency counts and proportions were compared between the groups using the chi-squared test or a two-tailed Fisher's exact test, where appropriate. The survival analysis was based on the Kaplan-Meier method and was compared by the Wilcoxon test. Cox's proportional hazards regression analysis was performed to identify significant independent prognostic factors for survival. P-values of < 0.05 were considered to be significant. All analyses were performed using the