- Email: [email protected]
Risk assessment model for overall survival in patients with locally advanced cervical cancer treated with definitive concurrent chemoradiotherapy
Gynecologic Oncology 128 (2013) 54–59
Contents lists available at SciVerse ScienceDirect
Gynecologic Oncology journal homepage: www.elsevier.com/locate/ygyno
Risk assessment model for overall survival in patients with locally advanced cervical cancer treated with definitive concurrent chemoradiotherapy Seung-Hyuk Shim a, Shin-Wha Lee a, Jeong-Yeol Park a, Young Seok Kim b, Dae-Yeon Kim a, Jong-Hyeok Kim a, Yong-Man Kim a, Young-Tak Kim a, Joo-Hyun Nam a,⁎ a b
Department of Obstetrics and Gynecology, University of Ulsan College of Medicine, Asan Medical Center, Seoul, South Korea Department of Radiation Oncology, University of Ulsan College of Medicine, Asan Medical Center, Seoul, South Korea
H I G H L I G H T S ► Five year survival prediction model was derived from 209 patients with locally advanced cervical cancer treated with definitive concurrent chemoradiotherapy. ► Histology, tumor size, and paraaortic lymph node metastasis were incorporated in the model. ► The predictive ability for survival of the model proved to be superior to that of the FIGO staging system.
a r t i c l e
i n f o
Article history: Received 7 August 2012 Accepted 30 September 2012 Available online 9 October 2012 Keywords: Locally advanced cervical cancer Nomogram Chemoradiation Survival
a b s t r a c t Objective. To develop a nomogram for predicting the probability of 5 year survival after definitive concurrent chemoradiotherapy (CCRT) in locally advanced cervical cancer (LACC). Methods. Between 1998 and 2008, 209 patients with LACC were treated with definitive CCRT. Multivariate analysis using Cox proportional hazards regression model was performed. A nomogram based on this Cox model was developed and internally validated by bootstrapping. Its performance was assessed by using the concordance index and a calibration curve. Results. The median age was 55 years (range, 26–78). Of the patients, 9, 16, 129, 3, 42 and 10 had FIGO stage IB2, IIA, IIB, IIIA, IIIB, and IVA disease, respectively. Histology revealed that 190, 13, 4, and 2 patients had squamous, adenocarcinoma, adenosquamous, and small cell cervical cancer, respectively. In 91 patients, PET/CT was performed before CCRT. The median follow-up period was 51 months (range, 6–151) and there were 50 (23.9%) disease-related deaths. Multivariate regression analysis revealed that histology, tumor size, and paraaortic lymph node metastasis (defined by MRI), but not PET/CT before CCRT, were independent predictors of overall survival. A nomogram for predicting the 5 year survival incorporating these three significant variables was constructed. The concordance index was 0.69. The predictive ability of the nomogram proved to be superior to that of the FIGO staging system (pb 0.05). Conclusions. The nomogram was a better predictive model for overall survival than the FIGO staging system. If externally validated, it could be used to counsel patients with LACC who must choose additional treatment modalities after definitive CCRT. © 2012 Elsevier Inc. All rights reserved.
Introduction Cervical cancer is the third most commonly diagnosed cancer and the fourth leading cause of cancer death in females worldwide. In 2008, it accounted for 9% (529,800) of all new cancer cases and 8% (275,100) of all cancer deaths among females [1]. In Korea, cervical cancer is the fifth most common female malignancy, with over 3,000
⁎ Corresponding author at: Department of Obstetrics and Gynecology, University of Ulsan College of Medicine, Asan Medical Center, 388-1, Pungnap-dong, Songpa-gu, Seoul 138-736, South Korea. Fax: +82 2 476 7331. E-mail address: [email protected] (J.-H. Nam). 0090-8258/$ – see front matter © 2012 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ygyno.2012.09.033
women being diagnosed annually; it accounts for 9.8% of new female cancer cases [2]. The treatment of cervical cancer mainly depends on the International Federation of Gynecology and Obstetrics (FIGO) stage. Five phase III randomized trials of concurrent chemoradiotherapy (CCRT) have demonstrated that CCRT improves the disease-free and overall survival of patients with locally advanced cervical cancer (LACC) [3–7]. Based on these results, the US National Cancer Institute announced that CCRT should be the standard treatment for LACC [8]. However, despite the significant improvement in the disease-free and overall survival of patients with LACC due to CCRT, the survival of this group remains unsatisfactory [9,10]. To improve this, adjunctive treatment options can be considered for patients with poor prognostic factors. Traditionally, the
S.-H. Shim et al. / Gynecologic Oncology 128 (2013) 54–59
prognosis for LACC has been based on the FIGO stage alone. However, it has been shown that the survival of individual patients with the same stage of disease varies [11,12]. Furthermore, other prognostic factors that are not included in FIGO staging system have been identified [13–17]. Therefore, to predict individual prognosis more accurately, it may be necessary to consider prognostic factors other than FIGO stage alone. Nomograms involve the integration of multiple independent prognostic factors that each carry their own relative risk. Nomograms can quantify overall risk and create a simple graphical representation by combining weighted prognostic factors instead of treating each factor as a discrete entity [18]. The aim of the present study was to develop a risk assessment model, or nomogram, that could predict overall survival in patients with LACC who were treated with definitive CCRT. Methods Patients The patients in this retrospective study were identified from a database of patients who were diagnosed with cervical cancer in Asan Medical Center between 1998 and 2008. The study was approved by our institutional review board (S2012-1239-0002). The eligibility criteria were as follows: patients with pathologically confirmed cervical cancer who were >18 years old and b80 years old; patients who were clinically diagnosed with FIGO stage IB2-IVA cervical cancer; patients who underwent definitive platinum-based CCRT and did not receive consolidation chemotherapy after CCRT; in the case of patients who underwent fluorine-18 fluorodeoxyglucose (18F-FDG) positron emission tomography/computed tomography (PET/CT) before CCRT, patients without evidence of distant metastasis on PET/CT; patients with Eastern Cooperative Oncology Group performance status 0 to 2; patients without a history of prior chemotherapy or radiotherapy; and patients without underlying disease that would affect survival. From a database of 4221 patients with cervical cancer, 209 patients satisfied the eligibility criteria. All the patients underwent exact staging, including physical examination, complete blood count, liver function test, serum squamous cell carcinoma antigen (SCC Ag), urinalysis, chest X-ray, intravenous pyelography, cystoscopy, sigmoidoscopy, and pelvic magnetic resonance imaging (MRI). In 91 patients, PET/CT was performed in the 4 weeks prior to CCRT.
55
Radiotherapy The uterine cervical cancer treatment protocol at Asan Medical Center has been described previously [19]. Briefly, an external beam was delivered with a 15 MV X-ray from a linear accelerator (Clinac 1800, 2100C/D, 21EX, Varian Medical System, Palo Alto, California, USA) by using a four-field box technique (AP-PA and two lateral fields). Patients with tumors that were stage IIB or less received 23 fractions of 1.8 Gy, yielding a total dose of 41.4 Gy. Patients with stage IIIA or higher tumors received 28 fractions of 1.8 Gy, yielding a total dose of 50.4 Gy. Prophylactic irradiation of the paraaortic lymph node area was not performed. However, in 29 patients with suspected paraaortic lymph node metastasis, a total 59.4 Gy irradiation was delivered to the paraaortic area from L1 to L5 by using a three-dimensional conformal boost. After external irradiation of the whole pelvis, each patient underwent high-dose intracavity brachytherapy three times a week using an Ir-192 brachytherapy unit (microSelectron®, Nucletron, Veenendaal, The Netherlands). Patients with stage ≥IIIA tumors received 30 Gy in six sessions to point A, whereas patients with stage IB2, IIA, or IIB tumors received 35 Gy over seven sessions. During the intracavity brachytherapy period, a parametrial boost by external irradiation with conventional fractionation was increased to 65 Gy on the thickened side and to 60 Gy on the unthickened parametrial region twice a week. Chemotherapy All patients underwent platinum-based chemotherapy that was determined by the physician; this treatment was concurrent with extended-field radiotherapy. Thus, of the women, 73 received monthly 5-fluorouracil (1,000 mg/m 2/d) plus cisplatin (20 mg/m 2/d), 114 received weekly cisplatin (30 mg/m 2), 19 received paclitaxel (135 mg/m2/d) plus cisplatin (75 mg/m2/d) at 3 week intervals, and three received paclitaxel (135 mg/m2/d) plus carboplatin (area under the curve, 5) at 3 week intervals. During radiotherapy, or before the initiation of chemotherapy, weekly physical examinations, complete blood counts, and liver and renal function tests were performed. If the absolute neutrophil count was b1,000/mm3 or the platelet count was b100,000/mm3, CCRT was delayed or interrupted until the patient recovered. Treatment-related toxicities were categorized according to the National Cancer Institute Common Terminology Criteria for Adverse Events version 3.0. Follow-up evaluation
Variables The end point was overall survival, which was defined as time to the occurrence of disease-specific death. To analyze the prognostic variables for overall survival, the following factors were included because previous studies have shown they may have some prognostic value [12–17]: age, FIGO stage, histological type, tumor size, parametrial involvement, pelvic lymph node metastasis, paraaortic lymph node metastasis, pretreatment serum SCC Ag, pretreatment hemoglobin, hydronephrosis, and bladder or rectal invasion. The tumor size, the parametrial involvement, and the pelvic/paraaortic lymph node metastases were determined by MRI. The MRI criteria for presumed lymph node metastasis included lymph nodes with a short-axis diameter of ≥1 cm, lobulated or speculated margins, and the presence of necrotic portions, heterogeneous enhancement, or loss of fatty hilum. Bladder/rectal invasion were confirmed by cystoscopic biopsy or sigmoidoscopic biopsy. It was speculated that distant metastasis was less likely to be missed in the patients who underwent pretreatment PET/CT than in the patients without pretreatment PET/CT. To test this notion, the effect of pre-CCRT PET/CT evaluation on overall survival was also included as a factor.
After CCRT completion, all patients were evaluated by a gynecologic oncologist and radiation oncologist at 1 month, then at 3 month intervals for 2 years, and every 6 months thereafter. A complete response was defined as the disappearance of gross tumor clinically or radiologically (as determined by follow-up pelvic MRI at 3 months), while a partial response was defined as a >50% reduction in the initial tumor volume. Local recurrence was defined as persistent disease or any recurrence within the radiation field, and distant metastasis was defined as recurrence outside the field. Statistical methods Overall survival was calculated in months from the initial date of CCRT to death, or the date of the last follow-up for surviving patients. Patients were censored in the overall survival analysis if they were alive at last contact or died without disease. Overall survival was estimated by using the Kaplan–Meier method. Of the variables that were tested, age, pretreatment hemoglobin, and pretreatment SCC Ag were modeled as continuous variables for statistical analysis. Regarding the tumor size, the patients were categorized into three groups: ≤ 4 cm, >4 and ≤ 5 cm, and > 5 cm. Parametrial involvement, pelvic lymph
56
S.-H. Shim et al. / Gynecologic Oncology 128 (2013) 54–59
node metastasis, paraaortic lymph node metastasis, hydronephrosis, and bladder or rectal invasion were considered as dichotomous variables. Similarly, histological type was stratified as squamous cell type or non-squamous cell type. Multivariate analyses were performed by using Cox proportional hazards regression. The proportional hazards assumption was verified by testing correlations with time and by examining residual plots. This Cox model formed the basis of the survival-predictive model. All decisions with respect to the grouping of categorical variables were made before modeling. The nomogram was constructed as described previously [18]. The validation of the nomogram that was constructed included three steps. First, discrimination was quantified with the concordance index, which means that for two randomly selected patients, the patient who dies first has a higher probability of death in the nomogram. This was calculated by bootstrapping 1000 samples from the original 209 patients to provide an unbiased and robust model. Second, the calibration was performed. The patients were grouped into quartiles based on nomogram-predicted probabilities and compared with observed Kaplan–Meier probabilities of overall survival. Finally, the concordance index of the nomogram was compared to the concordance index of the FIGO staging system. All analyses were performed by using SPSS version 13.0 (SPSS, Chicago, IL) or R version 2.14.0 (http://cran.r-project.org/mirrors.html). Results The characteristics of the 209 enrolled patients are summarized in Table 1. The median follow-up period was 50.7 months (range, 6–151 months). During follow-up, 50 patients died of disease, and the 3 year and 5 year overall survival rates were 83% and 75%, respectively (Fig. 1). Table 1 Characteristics of the model derivation cohort. Characteristics Age, years
FIGO stage, n (%)
Histology, n (%)
Tumor size by MRI, cm
Pretreatment hemoglobin, g/dL Pretreatment SCC Ag, ng/mL PM involvement by MRI, n (%)
PLN metastasis by MRI, n (%) PALN metastasis by MRI, n (%) Hydronephrosis, n (%) Bladder/rectal invasion, n (%) PET/CT evaluation before CCRT, n (%)
Patients (n = 209) Median (range) b55 years, n (%) ≥55 years, n (%) IB2 IIA IIB IIIA IIIB IVA Squamous cell Adenocarcinoma Adenosquamous Small cell Median (range) ≤4 cm, n (%) >4 cm and ≤5 cm, n (%) >5 cm, n (%) Median (range) Median (range) No Yes, unilateral Yes, bilateral No Yes No Yes No Yes No Yes No Yes
55 (26–78) 94 (45) 115 (55) 9 (4.3) 16 (7.7) 129 (61.7) 3 (1.4) 42 (20.1) 10 (4.8) 190 (90.8) 13 (6.2) 4 (1.9) 2 (1.0) 4.6 (1–12) 75 (35.9) 62 (29.7) 72 (34.4) 12 (4.7–15.1) 6.4 (0–319) 28 (13.4) 115 (55.0) 66 (31.6) 111 (53.1) 98 (46.9) 180 (86.1) 29 (13.9) 174 (83.3) 35 (16.7) 199 (95.2) 10 (4.8) 118 (56.5) 91 (43.5)
FIGO, International Federation of Gynecology and Obstetrics; MRI, magnetic resonance imaging; SCC Ag, squamous cell carcinoma antigen; PM, parametrial; PLN, pelvic lymph node; PALN, paraaortic lymph node; PET/CT, positron emission tomography/computed tomography CCRT, concurrent chemoradiotherapy.
Fig. 1. Overall survival rate after definitive concurrent chemoradiotherapy in the study cohort with locally advanced cervical cancer. During follow-up, 50 patients died of disease, and the 3 year and 5 year overall survival rates were 83% and 75%, respectively. CCRT = concurrent chemoradiotherapy.
Table 2 shows the results of the Cox proportional hazards regression model for predicting overall survival in LACC. Univariate analysis revealed that histology (p b 0.001), tumor size measured by MRI (p = 0.022), pelvic lymph node metastasis defined by MRI (p = 0.003), and paraaortic lymph node metastasis defined by MRI (p = 0.001) are significantly associated with overall survival. After multivariate analysis, histology (p = 0.001), tumor size (p = 0.004), and paraaortic lymph node metastasis (p = 0.008) remained as significant prognostic factors. A nomogram was constructed on the basis of this Cox proportional hazards model. It was composed of the three independent factors described above (Fig. 2). The point value assigned to each factor was proportionate to the hazard ratio that was derived from its own beta coefficients by regression analysis. Internal validation was performed by using bootstrapping resampling. The bootstrap-corrected concordance index for the model was 0.69. Fig. 3 shows the calibration of the constructed nomogram. The dashed line represents the performance of an ideal nomogram, in which predicted 100% survival matches the actuarial survival. The solid line represents the performance of the actual nomogram. The filled dots were obtained from a subcohort of the present database, and the check marks indicate bootstrap-corrected predictions of the nomogram's performance with 1000 resamples. When plotting the survival probabilities predicted by the current nomogram against the actuarial survival, the calibration curve lay close to the dashed line. The predictive ability of the nomogram was compared to that of the FIGO staging system. The concordance index for the FIGO staging system was 0.59, which means that the predictive ability of the nomogram was superior to that of the FIGO staging system (pb 0.05). Fig. 4 shows the heterogeneous nature of disease stratified by FIGO stage. Patients with the same stage disease had a broad spectrum of survival duration. The different stages also overlapped in terms of survival distribution, especially IIA, IIB, and IIIA. Discussion For both the patient with carcinoma and her (or his) physician, the question “will I die from cancer?” is a major concern. A prediction tool that shows who is at high risk for cancer-specific death will allow the physician to select ideal targets for more aggressive therapy. In recent years, such statistical prediction models have been developed for various types of cancers [20–22]. One such risk prediction model is the nomogram, which creates a simple graphical representation of a statistical predictive tool that generates a numerical probability of a clinical event.
S.-H. Shim et al. / Gynecologic Oncology 128 (2013) 54–59
57
Table 2 Univariate and multivariate Cox proportional hazards model for predicting overall survival in locally advanced cervical cancer. Variables
n
Univariate analysis HR (95% CI)
Age (as continuous variable) Age FIGO stage
Histology Tumor size by MRI (as continuous variable) Tumor size by MRI
Pretreatment hemoglobin (as continuous variable) Pretreatment hemoglobin
Pretreatment SCC Ag (as continuous variable) PM involvement by MRI
PLN metastasis by MRI PALN metastasis by MRI Hydronephrosis Bladder/rectal invasion PET/CT evaluation before CCRT
b55 years ≥55 years IB, IIA IIB, IIIA IIIB, IVA Squamous cell Non-squamous
94 115 25 132 52 190 13
≤4 cm >4 cm and ≤5 cm >5 cm b8 g/dL ≥8 g/dL and b12 g/dL ≥12 g/dL
75 62 72 209 13 91 105
No Yes, unilateral Yes, bilateral No Yes No Yes No Yes No Yes No Yes
28 115 66 111 98 180 29 174 35 199 10 118 91
0.981 (0.954–1.008) 1.000 0.784 (0.450–1.366) 1 1.235 (0.432–3.530) 2.701 (0.918–7.947) 1 3.894 (1.814–8.358) 1.193 (1.026–1.388) 1 1.880 (0.834–4.235) 3.212 (1.548–6.664) 0.896 (0.784–1.025) 2.150 (0.817–5.658) 1.218 (0.678–2.186) 1 0.988 (0.968–1.008) 1 0.895 (0.364–2.198) 1.699 (0.685–4.216) 1 2.386 (1.338–4.253) 1 3.127 (1.632–5.993) 1 1.448 (0.723–2.898) 1 2.126 (0.764–5.916) 1 0.920 (0.509–1.663)
Multivariate analysis p-Value
HR (95% CI)
p-Value
1.000 3.605 (1.674–7.764)
0.001
1 1.973 (0.873–4.460) 2.902 (1.394–6.044)
0.102 0.004
1 2.458 (1.260–4.798)
0.008
0.161 0.391 0.694 0.071 b0.001 0.022 0.128 0.002 0.110 0.121 0.509 0.238 0.808 0.253 0.003 0.001 0.296 0.149 0.782
FIGO, International Federation of Gynecology and Obstetrics; MRI, magnetic resonance imaging; SCC Ag, squamous cell carcinoma antigen; PM, parametrial; PLN, pelvic lymph node; PALN, paraaortic lymph node; PET/CT, positron emission tomography/computed tomography CCRT, concurrent chemoradiotherapy.; HR, hazard risk; CI, confidence interval.
Traditionally, the prognoses of patients with LACC who are treated with definitive CCRT have been estimated by using the FIGO staging system. However, previous studies have shown that each FIGO stage is associated with a broad spectrum of survival probabilities [11,12]. The present study confirmed the heterogeneity of overall survival for each stage and the overlapping of survival distributions between different stages. These findings suggest that FIGO stage alone is not entirely satisfactory with regard to providing an accurate prediction of individual outcome. Moreover, many prognostic factors other than FIGO stage have been found to affect survival [13–17]. In the present study, a three-variable risk assessment model that predicted the 5 year survival probability of patients with LACC who were treated with definitive CCRT was developed and internally validated. The selected variables for the nomogram were histology, tumor size, and paraaortic lymph node metastasis. Age, parametrial involvement, pelvic lymph node metastasis, pretreatment SCC Ag, pretreatment hemoglobin, hydronephrosis, and bladder or rectal invasion, which have been reported as prognostic factors in other studies, were not shown to be of significant prognostic value in the present study. In terms of predicting overall survival, the discriminating power of the model constructed in the present study was superior to that of the FIGO staging system alone. Tseng et al. analyzed the outcome of 251 patients with LACC who were treated with CCRT in Taiwan between 1999 and 2006, and developed a seven-variable nomogram for survival [11]. The variables that were incorporated were age, SCC Ag, tumor size, parametrial invasion, hydronephrosis, bladder/rectal invasion, and lymph node metastasis. As the authors mentioned, the study was limited by the fact that consolidation chemotherapy was performed in 23% of patients after CCRT, and about 70% of the study subjects were staged by CT. Notably, although the nomogram was based on seven variables, the concordance index was 0.69, which is not different from the concordance of the nomogram developed in the present study.
Individualized prediction based on the nomogram in the present study could help the physician to select adjunctive treatment after CCRT and to counsel the patient. Possible adjunctive treatments may be consolidation chemotherapy or hysterectomy. Two meta-analyses showed that consolidation chemotherapy in LACC after CCRT was effective and well tolerated [9,23]. Recently, Duenas-Gonzalez et al. reported the results of a phase III trial that showed that gemcitabine plus cisplatin chemoradiotherapy followed by adjuvant gemcitabine/cisplatin chemotherapy improved survival outcomes [24]. Hysterectomy may also be considered as an adjunctive treatment option for high-risk patients after CCRT, although this possibility is only supported by limited evidence [25–27]. PET/CT is known to be superior in detecting distant metastasis compared to conventional CT or MRI [28]. It was speculated in the present study that distant metastasis was more likely to be missed in patients without pretreatment PET/CT than in patients who did receive PET/CT. This would mean that the latter patients would have a better prognosis than patients who did not undergo pretreatment PET/CT. However, pretreatment PET/CT did not associate significantly with overall survival in the present study. Nevertheless, PET/CT seems to be better at identifying extrapelvic disease and nodal metastasis than conventional CT or MRI [29]. A recent study by Kang et al. introduced a risk assessment model that predicted distant recurrence in LACC [30]. The authors selected four variables, including histology, tumor size, and pelvic/paraaortic lymph node metastasis on PET/CT. As described in Methods section, we defined pelvic and paraaortic lymph node metastases using MRI as pretreatment PET/CT was performed in only 91 cases whereas MRI was performed in all enrolled patients. The concordance index of the study by Kang et al. was 0.72, which is better than the concordance index of the nomogram constructed in the present study. It remains possible that the more accurate prediction of the model of Kang et al. may reflect the more accurate evaluation of nodal status by PET/CT.
58
S.-H. Shim et al. / Gynecologic Oncology 128 (2013) 54–59
Fig. 2. Nomogram for predicting 3 year and 5 year survival probability in patients with locally advanced cervical cancer who were treated with definitive concurrent chemoradiotherapy. The nomogram was constructed by incorporating three variables. For each level of the prognostic variables, points were allocated according to the scale shown here. Total score was determined by adding up the individual parameter points, and these were then used to calculate the 3 year and 5 year survival probability. SCC = squamous cell carcinoma, MRI = magnetic resonance imaging, PALN = paraaortic lymph node.
The present study possesses all the limitations that are inherent in retrospective chart reviews. While all patients were given platinum-based chemotherapy, the chemotherapy regimens were not uniform. While this may cause a bias, the chemotherapy regimens did not differ significantly in terms of survival. In addition, some known prognostic factors such as human papilloma virus genotyping were not included in present study because of a respectable amount of missing value and over-fitness problem. Another known prognostic factor, namely tumor volume regression rate after CCRT, could also be a nomogram variable [31]. If this
Fig. 3. Calibration curve for the 5 year survival probability-predictive nomogram. Dashed line, the ideal reference value where predicted probabilities would match the observed 5 year survival; solid line, performance of the current nomogram; filled dots, calculated from a subcohort of the present database; check marks, bootstrap-corrected predictions of the performance of the nomogram; vertical bar, 95% confidence interval.
time-dependent variable can be incorporated into the present nomogram by using a time-dependent Cox regression model, it could increase the performance index of the nomogram. Furthermore, the nodal status was assessed by MRI, not surgical staging, which means that the real nodal metastasis status was not known. This is important because nodal metastasis in women with LACC is one of the strongest prognostic factors for survival [32]. While a tool that is more accurate than MRI (such as PET/CT) could be used to evaluate nodal metastasis, false negative results have been recorded in 12–22% patients who were evaluated by PET/CT [29]. To eliminate this bias, it is essential to determine the pathology-proven nodal status. However, the advantages of staging surgery must always be balanced against its potential morbidity, although the complication rate seems to be low when laparoscopy is performed by trained teams [29]. Finally, the present model was derived from
Fig. 4. The boxplot reveals the distribution of the observed overall survival for each FIGO stage. Note the heterogeneous nature of the overall survival associated with each stage, and how the different stages overlap in terms of survival distribution. • indicates a specific value above the upper fence, which is 1.5 times the interquartile range (above the 75th percentile). ★ indicates a extreme value above the upper fence, which is 3.0 times the interquartile range (above the 75th percentile).
S.-H. Shim et al. / Gynecologic Oncology 128 (2013) 54–59
patients attending a single institution. To determine the generalizability of the model, it must be validated externally. In conclusion, a risk assessment model for predicting the probability of 5 year survival after definitive CCRT in LACC was developed and internally validated. The derived model seems to predict the outcome more accurately than the FIGO staging system. Although this nomogram must be externally validated before it can be applied, it may be valuable in terms of choosing adjunctive treatment, counseling patients, and planning clinical trials. Conflict of interest statement The authors declare that there is no conflict of interest.
Disclosures Drs. Seung-Hyuk Shim, Shin-Wha Lee, Jeong-Yeol Park, Young Seok Kim, Dae-Yeon Kim, Jong-Hyeok Kim, Yong-Man Kim, Young-Tak Kim, and Joo-Hyun Nam have no conflicts of interest or financial ties to disclose. Acknowledgment We thank Hwa Jung Kim, M.D. and Sun Ok Kim, Department of Clinical Epidemiology and Biostatistics, University of Ulsan College of Medicine, Asan Medical Center for statistical consultation. References [1] Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D. Global cancer statistics. CA Cancer J Clin 2011;61:69-90. [2] Kim YT. Current status of cervical cancer and HPV infection in Korea. J Gynecol Oncol 2009;20:1-7. [3] Keys HM, Bundy BN, Stehman FB, Muderspach LI, Chafe WE, Suggs 3rd CL, et al. Cisplatin, radiation, and adjuvant hysterectomy compared with radiation and adjuvant hysterectomy for bulky stage IB cervical carcinoma. N Engl J Med 1999;340:1154-61. [4] Morris M, Eifel PJ, Lu J, Grigsby PW, Levenback C, Stevens RE, et al. Pelvic radiation with concurrent chemotherapy compared with pelvic and para-aortic radiation for high-risk cervical cancer. N Engl J Med 1999;340:1137-43. [5] Rose PG, Bundy BN, Watkins EB, Thigpen JT, Deppe G, Maiman MA, et al. Concurrent cisplatin-based radiotherapy and chemotherapy for locally advanced cervical cancer. N Engl J Med 1999;340:1144-53. [6] Whitney CW, Sause W, Bundy BN, Malfetano JH, Hannigan EV, Fowler Jr WC, et al. Randomized comparison of fluorouracil plus cisplatin versus hydroxyurea as an adjunct to radiation therapy in stage IIB-IVA carcinoma of the cervix with negative para-aortic lymph nodes: a Gynecologic Oncology Group and Southwest Oncology Group study. J Clin Oncol 1999;17:1339-48. [7] Peters 3rd WA, Liu PY, Barrett 2nd RJ, Stock RJ, Monk BJ, Berek JS, et al. Concurrent chemotherapy and pelvic radiation therapy compared with pelvic radiation therapy alone as adjuvant therapy after radical surgery in high-risk early-stage cancer of the cervix. J Clin Oncol 2000;18:1606-13. [8] McNeil C. New standard of care for cervical cancer sets stage for next questions. J Natl Cancer Inst 1999;91:500-1. [9] Chemoradiotherapy for Cervical Cancer Meta-Analysis C. Reducing uncertainties about the effects of chemoradiotherapy for cervical cancer: a systematic review and meta-analysis of individual patient data from 18 randomized trials. J Clin Oncol 2008;26:5802-12. [10] Pearcey R, Miao Q, Kong W, Zhang-Salomons J, Mackillop WJ. Impact of adoption of chemoradiotherapy on the outcome of cervical cancer in Ontario: results of a population-based cohort study. J Clin Oncol 2007;25:2383-8. [11] Tseng JY, Yen MS, Twu NF, Lai CR, Horng HC, Tseng CC, et al. Prognostic nomogram for overall survival in stage IIB-IVA cervical cancer patients treated with concurrent chemoradiotherapy. Am J Obstet Gynecol 2010;202:174e1-7. [12] Seo Y, Yoo SY, Kim MS, Yang KM, Yoo HJ, Kim JH, et al. Nomogram prediction of overall survival after curative irradiation for uterine cervical cancer. Int J Radiat Oncol Biol Phys 2011;79:782-7.
59
[13] Yalman D, Aras AB, Ozkok S, Duransoy A, Celik OK, Ozsaran Z, et al. Prognostic factors in definitive radiotherapy of uterine cervical cancer. Eur J Gynaecol Oncol 2003;24:309-14. [14] Hong JH, Chen MS, Lin FJ, Tang SG. Prognostic assessment of tumor regression after external irradiation for cervical cancer. Int J Radiat Oncol Biol Phys 1992;22:913-7. [15] Parker K, Gallop-Evans E, Hanna L, Adams M. Five years' experience treating locally advanced cervical cancer with concurrent chemoradiotherapy and high-dose-rate brachytherapy: results from a single institution. Int J Radiat Oncol Biol Phys 2009;74: 140-6. [16] Lai CH, Chang CJ, Huang HJ, Hsueh S, Chao A, Yang JE, et al. Role of human papillomavirus genotype in prognosis of early-stage cervical cancer undergoing primary surgery. J Clin Oncol 2007;25:3628-34. [17] Kidd EA, Siegel BA, Dehdashti F, Grigsby PW. The standardized uptake value for F-18 fluorodeoxyglucose is a sensitive predictive biomarker for cervical cancer treatment response and survival. Cancer 2007;110:1738-44. [18] Iasonos A, Schrag D, Raj GV, Panageas KS. How to build and interpret a nomogram for cancer prognosis. J Clin Oncol 2008;26:1364-70. [19] Kim YS, Shin SS, Nam JH, Kim YT, Kim YM, Kim JH, et al. Prospective randomized comparison of monthly fluorouracil and cisplatin versus weekly cisplatin concurrent with pelvic radiotherapy and high-dose rate brachytherapy for locally advanced cervical cancer. Gynecol Oncol 2008;108:195-200. [20] International Bladder Cancer Nomogram C, Bochner BH, Kattan MW, Vora KC. Postoperative nomogram predicting risk of recurrence after radical cystectomy for bladder cancer. J Clin Oncol 2006;24:3967-72. [21] Karakiewicz PI, Briganti A, Chun FK, Trinh QD, Perrotte P, Ficarra V, et al. Multi-institutional validation of a new renal cancer-specific survival nomogram. J Clin Oncol 2007;25:1316-22. [22] Wierda WG, O'Brien S, Wang X, Faderl S, Ferrajoli A, Do KA, et al. Prognostic nomogram and index for overall survival in previously untreated patients with chronic lymphocytic leukemia. Blood 2007;109:4679-85. [23] Green JA, Kirwan JM, Tierney JF, Symonds P, Fresco L, Collingwood M, et al. Survival and recurrence after concomitant chemotherapy and radiotherapy for cancer of the uterine cervix: a systematic review and meta-analysis. Lancet 2001;358: 781-6. [24] Duenas-Gonzalez A, Zarba JJ, Patel F, Alcedo JC, Beslija S, Casanova L, et al. Phase III, open-label, randomized study comparing concurrent gemcitabine plus cisplatin and radiation followed by adjuvant gemcitabine and cisplatin versus concurrent cisplatin and radiation in patients with stage IIB to IVA carcinoma of the cervix. J Clin Oncol 2011;29:1678-85. [25] Walji N, Chue AL, Yap C, Rogers LJ, El-Modir A, Chan KK, et al. Is there a role for adjuvant hysterectomy after suboptimal concurrent chemoradiation in cervical carcinoma? Clin Oncol (R Coll Radiol) 2010;22:140-6. [26] Colombo PE, Bertrand MM, Gutowski M, Mourregot A, Fabbro M, Saint-Aubert B, et al. Total laparoscopic radical hysterectomy for locally advanced cervical carcinoma (stages IIB, IIA and bulky stages IB) after concurrent chemoradiation therapy: surgical morbidity and oncological results. Gynecol Oncol 2009;114: 404-9. [27] Morice P, Rouanet P, Rey A, Romestaing P, Houvenaeghel G, Boulanger JC, et al. Results of the GYNECO 02 study, an FNCLCC phase III trial comparing hysterectomy with no hysterectomy in patients with a (clinical and radiological) complete response after chemoradiation therapy for stage IB2 or II cervical cancer. Oncologist 2012;17:64-71. [28] Liu FY, Yen TC, Chen MY, Lai CH, Chang TC, Chou HH, et al. Detection of hematogenous bone metastasis in cervical cancer: 18F-fluorodeoxyglucose-positron emission tomography versus computed tomography and magnetic resonance imaging. Cancer 2009;115:5470-80. [29] Gouy S, Morice P, Narducci F, Uzan C, Gilmore J, Kolesnikov-Gauthier H, et al. Nodal-staging surgery for locally advanced cervical cancer in the era of PET. Lancet Oncol 2012;13:e212-20. [30] Kang S, Nam BH, Park JY, Seo SS, Ryu SY, Kim JW, et al. Risk assessment tool for distant recurrence after platinum-based concurrent chemoradiation in patients with locally advanced cervical cancer: a Korean gynecologic oncology group study. J Clin Oncol 2012;30:2369-74. [31] Nam H, Park W, Huh SJ, Bae DS, Kim BG, Lee JH, et al. The prognostic significance of tumor volume regression during radiotherapy and concurrent chemoradiotherapy for cervical cancer using MRI. Gynecol Oncol 2007;107: 320-5. [32] Duenas-Gonzalez A, Orlando M, Zhou Y, Quinlivan M, Barraclough H. Efficacy in high burden locally advanced cervical cancer with concurrent gemcitabine and cisplatin chemoradiotherapy plus adjuvant gemcitabine and cisplatin: prognostic and predictive factors and the impact of disease stage on outcomes from a prospective randomized phase III trial. Gynecol Oncol 2012;126:334-40.
Contents lists available at SciVerse ScienceDirect
Gynecologic Oncology journal homepage: www.elsevier.com/locate/ygyno
Risk assessment model for overall survival in patients with locally advanced cervical cancer treated with definitive concurrent chemoradiotherapy Seung-Hyuk Shim a, Shin-Wha Lee a, Jeong-Yeol Park a, Young Seok Kim b, Dae-Yeon Kim a, Jong-Hyeok Kim a, Yong-Man Kim a, Young-Tak Kim a, Joo-Hyun Nam a,⁎ a b
Department of Obstetrics and Gynecology, University of Ulsan College of Medicine, Asan Medical Center, Seoul, South Korea Department of Radiation Oncology, University of Ulsan College of Medicine, Asan Medical Center, Seoul, South Korea
H I G H L I G H T S ► Five year survival prediction model was derived from 209 patients with locally advanced cervical cancer treated with definitive concurrent chemoradiotherapy. ► Histology, tumor size, and paraaortic lymph node metastasis were incorporated in the model. ► The predictive ability for survival of the model proved to be superior to that of the FIGO staging system.
a r t i c l e
i n f o
Article history: Received 7 August 2012 Accepted 30 September 2012 Available online 9 October 2012 Keywords: Locally advanced cervical cancer Nomogram Chemoradiation Survival
a b s t r a c t Objective. To develop a nomogram for predicting the probability of 5 year survival after definitive concurrent chemoradiotherapy (CCRT) in locally advanced cervical cancer (LACC). Methods. Between 1998 and 2008, 209 patients with LACC were treated with definitive CCRT. Multivariate analysis using Cox proportional hazards regression model was performed. A nomogram based on this Cox model was developed and internally validated by bootstrapping. Its performance was assessed by using the concordance index and a calibration curve. Results. The median age was 55 years (range, 26–78). Of the patients, 9, 16, 129, 3, 42 and 10 had FIGO stage IB2, IIA, IIB, IIIA, IIIB, and IVA disease, respectively. Histology revealed that 190, 13, 4, and 2 patients had squamous, adenocarcinoma, adenosquamous, and small cell cervical cancer, respectively. In 91 patients, PET/CT was performed before CCRT. The median follow-up period was 51 months (range, 6–151) and there were 50 (23.9%) disease-related deaths. Multivariate regression analysis revealed that histology, tumor size, and paraaortic lymph node metastasis (defined by MRI), but not PET/CT before CCRT, were independent predictors of overall survival. A nomogram for predicting the 5 year survival incorporating these three significant variables was constructed. The concordance index was 0.69. The predictive ability of the nomogram proved to be superior to that of the FIGO staging system (pb 0.05). Conclusions. The nomogram was a better predictive model for overall survival than the FIGO staging system. If externally validated, it could be used to counsel patients with LACC who must choose additional treatment modalities after definitive CCRT. © 2012 Elsevier Inc. All rights reserved.
Introduction Cervical cancer is the third most commonly diagnosed cancer and the fourth leading cause of cancer death in females worldwide. In 2008, it accounted for 9% (529,800) of all new cancer cases and 8% (275,100) of all cancer deaths among females [1]. In Korea, cervical cancer is the fifth most common female malignancy, with over 3,000
⁎ Corresponding author at: Department of Obstetrics and Gynecology, University of Ulsan College of Medicine, Asan Medical Center, 388-1, Pungnap-dong, Songpa-gu, Seoul 138-736, South Korea. Fax: +82 2 476 7331. E-mail address: [email protected] (J.-H. Nam). 0090-8258/$ – see front matter © 2012 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ygyno.2012.09.033
women being diagnosed annually; it accounts for 9.8% of new female cancer cases [2]. The treatment of cervical cancer mainly depends on the International Federation of Gynecology and Obstetrics (FIGO) stage. Five phase III randomized trials of concurrent chemoradiotherapy (CCRT) have demonstrated that CCRT improves the disease-free and overall survival of patients with locally advanced cervical cancer (LACC) [3–7]. Based on these results, the US National Cancer Institute announced that CCRT should be the standard treatment for LACC [8]. However, despite the significant improvement in the disease-free and overall survival of patients with LACC due to CCRT, the survival of this group remains unsatisfactory [9,10]. To improve this, adjunctive treatment options can be considered for patients with poor prognostic factors. Traditionally, the
S.-H. Shim et al. / Gynecologic Oncology 128 (2013) 54–59
prognosis for LACC has been based on the FIGO stage alone. However, it has been shown that the survival of individual patients with the same stage of disease varies [11,12]. Furthermore, other prognostic factors that are not included in FIGO staging system have been identified [13–17]. Therefore, to predict individual prognosis more accurately, it may be necessary to consider prognostic factors other than FIGO stage alone. Nomograms involve the integration of multiple independent prognostic factors that each carry their own relative risk. Nomograms can quantify overall risk and create a simple graphical representation by combining weighted prognostic factors instead of treating each factor as a discrete entity [18]. The aim of the present study was to develop a risk assessment model, or nomogram, that could predict overall survival in patients with LACC who were treated with definitive CCRT. Methods Patients The patients in this retrospective study were identified from a database of patients who were diagnosed with cervical cancer in Asan Medical Center between 1998 and 2008. The study was approved by our institutional review board (S2012-1239-0002). The eligibility criteria were as follows: patients with pathologically confirmed cervical cancer who were >18 years old and b80 years old; patients who were clinically diagnosed with FIGO stage IB2-IVA cervical cancer; patients who underwent definitive platinum-based CCRT and did not receive consolidation chemotherapy after CCRT; in the case of patients who underwent fluorine-18 fluorodeoxyglucose (18F-FDG) positron emission tomography/computed tomography (PET/CT) before CCRT, patients without evidence of distant metastasis on PET/CT; patients with Eastern Cooperative Oncology Group performance status 0 to 2; patients without a history of prior chemotherapy or radiotherapy; and patients without underlying disease that would affect survival. From a database of 4221 patients with cervical cancer, 209 patients satisfied the eligibility criteria. All the patients underwent exact staging, including physical examination, complete blood count, liver function test, serum squamous cell carcinoma antigen (SCC Ag), urinalysis, chest X-ray, intravenous pyelography, cystoscopy, sigmoidoscopy, and pelvic magnetic resonance imaging (MRI). In 91 patients, PET/CT was performed in the 4 weeks prior to CCRT.
55
Radiotherapy The uterine cervical cancer treatment protocol at Asan Medical Center has been described previously [19]. Briefly, an external beam was delivered with a 15 MV X-ray from a linear accelerator (Clinac 1800, 2100C/D, 21EX, Varian Medical System, Palo Alto, California, USA) by using a four-field box technique (AP-PA and two lateral fields). Patients with tumors that were stage IIB or less received 23 fractions of 1.8 Gy, yielding a total dose of 41.4 Gy. Patients with stage IIIA or higher tumors received 28 fractions of 1.8 Gy, yielding a total dose of 50.4 Gy. Prophylactic irradiation of the paraaortic lymph node area was not performed. However, in 29 patients with suspected paraaortic lymph node metastasis, a total 59.4 Gy irradiation was delivered to the paraaortic area from L1 to L5 by using a three-dimensional conformal boost. After external irradiation of the whole pelvis, each patient underwent high-dose intracavity brachytherapy three times a week using an Ir-192 brachytherapy unit (microSelectron®, Nucletron, Veenendaal, The Netherlands). Patients with stage ≥IIIA tumors received 30 Gy in six sessions to point A, whereas patients with stage IB2, IIA, or IIB tumors received 35 Gy over seven sessions. During the intracavity brachytherapy period, a parametrial boost by external irradiation with conventional fractionation was increased to 65 Gy on the thickened side and to 60 Gy on the unthickened parametrial region twice a week. Chemotherapy All patients underwent platinum-based chemotherapy that was determined by the physician; this treatment was concurrent with extended-field radiotherapy. Thus, of the women, 73 received monthly 5-fluorouracil (1,000 mg/m 2/d) plus cisplatin (20 mg/m 2/d), 114 received weekly cisplatin (30 mg/m 2), 19 received paclitaxel (135 mg/m2/d) plus cisplatin (75 mg/m2/d) at 3 week intervals, and three received paclitaxel (135 mg/m2/d) plus carboplatin (area under the curve, 5) at 3 week intervals. During radiotherapy, or before the initiation of chemotherapy, weekly physical examinations, complete blood counts, and liver and renal function tests were performed. If the absolute neutrophil count was b1,000/mm3 or the platelet count was b100,000/mm3, CCRT was delayed or interrupted until the patient recovered. Treatment-related toxicities were categorized according to the National Cancer Institute Common Terminology Criteria for Adverse Events version 3.0. Follow-up evaluation
Variables The end point was overall survival, which was defined as time to the occurrence of disease-specific death. To analyze the prognostic variables for overall survival, the following factors were included because previous studies have shown they may have some prognostic value [12–17]: age, FIGO stage, histological type, tumor size, parametrial involvement, pelvic lymph node metastasis, paraaortic lymph node metastasis, pretreatment serum SCC Ag, pretreatment hemoglobin, hydronephrosis, and bladder or rectal invasion. The tumor size, the parametrial involvement, and the pelvic/paraaortic lymph node metastases were determined by MRI. The MRI criteria for presumed lymph node metastasis included lymph nodes with a short-axis diameter of ≥1 cm, lobulated or speculated margins, and the presence of necrotic portions, heterogeneous enhancement, or loss of fatty hilum. Bladder/rectal invasion were confirmed by cystoscopic biopsy or sigmoidoscopic biopsy. It was speculated that distant metastasis was less likely to be missed in the patients who underwent pretreatment PET/CT than in the patients without pretreatment PET/CT. To test this notion, the effect of pre-CCRT PET/CT evaluation on overall survival was also included as a factor.
After CCRT completion, all patients were evaluated by a gynecologic oncologist and radiation oncologist at 1 month, then at 3 month intervals for 2 years, and every 6 months thereafter. A complete response was defined as the disappearance of gross tumor clinically or radiologically (as determined by follow-up pelvic MRI at 3 months), while a partial response was defined as a >50% reduction in the initial tumor volume. Local recurrence was defined as persistent disease or any recurrence within the radiation field, and distant metastasis was defined as recurrence outside the field. Statistical methods Overall survival was calculated in months from the initial date of CCRT to death, or the date of the last follow-up for surviving patients. Patients were censored in the overall survival analysis if they were alive at last contact or died without disease. Overall survival was estimated by using the Kaplan–Meier method. Of the variables that were tested, age, pretreatment hemoglobin, and pretreatment SCC Ag were modeled as continuous variables for statistical analysis. Regarding the tumor size, the patients were categorized into three groups: ≤ 4 cm, >4 and ≤ 5 cm, and > 5 cm. Parametrial involvement, pelvic lymph
56
S.-H. Shim et al. / Gynecologic Oncology 128 (2013) 54–59
node metastasis, paraaortic lymph node metastasis, hydronephrosis, and bladder or rectal invasion were considered as dichotomous variables. Similarly, histological type was stratified as squamous cell type or non-squamous cell type. Multivariate analyses were performed by using Cox proportional hazards regression. The proportional hazards assumption was verified by testing correlations with time and by examining residual plots. This Cox model formed the basis of the survival-predictive model. All decisions with respect to the grouping of categorical variables were made before modeling. The nomogram was constructed as described previously [18]. The validation of the nomogram that was constructed included three steps. First, discrimination was quantified with the concordance index, which means that for two randomly selected patients, the patient who dies first has a higher probability of death in the nomogram. This was calculated by bootstrapping 1000 samples from the original 209 patients to provide an unbiased and robust model. Second, the calibration was performed. The patients were grouped into quartiles based on nomogram-predicted probabilities and compared with observed Kaplan–Meier probabilities of overall survival. Finally, the concordance index of the nomogram was compared to the concordance index of the FIGO staging system. All analyses were performed by using SPSS version 13.0 (SPSS, Chicago, IL) or R version 2.14.0 (http://cran.r-project.org/mirrors.html). Results The characteristics of the 209 enrolled patients are summarized in Table 1. The median follow-up period was 50.7 months (range, 6–151 months). During follow-up, 50 patients died of disease, and the 3 year and 5 year overall survival rates were 83% and 75%, respectively (Fig. 1). Table 1 Characteristics of the model derivation cohort. Characteristics Age, years
FIGO stage, n (%)
Histology, n (%)
Tumor size by MRI, cm
Pretreatment hemoglobin, g/dL Pretreatment SCC Ag, ng/mL PM involvement by MRI, n (%)
PLN metastasis by MRI, n (%) PALN metastasis by MRI, n (%) Hydronephrosis, n (%) Bladder/rectal invasion, n (%) PET/CT evaluation before CCRT, n (%)
Patients (n = 209) Median (range) b55 years, n (%) ≥55 years, n (%) IB2 IIA IIB IIIA IIIB IVA Squamous cell Adenocarcinoma Adenosquamous Small cell Median (range) ≤4 cm, n (%) >4 cm and ≤5 cm, n (%) >5 cm, n (%) Median (range) Median (range) No Yes, unilateral Yes, bilateral No Yes No Yes No Yes No Yes No Yes
55 (26–78) 94 (45) 115 (55) 9 (4.3) 16 (7.7) 129 (61.7) 3 (1.4) 42 (20.1) 10 (4.8) 190 (90.8) 13 (6.2) 4 (1.9) 2 (1.0) 4.6 (1–12) 75 (35.9) 62 (29.7) 72 (34.4) 12 (4.7–15.1) 6.4 (0–319) 28 (13.4) 115 (55.0) 66 (31.6) 111 (53.1) 98 (46.9) 180 (86.1) 29 (13.9) 174 (83.3) 35 (16.7) 199 (95.2) 10 (4.8) 118 (56.5) 91 (43.5)
FIGO, International Federation of Gynecology and Obstetrics; MRI, magnetic resonance imaging; SCC Ag, squamous cell carcinoma antigen; PM, parametrial; PLN, pelvic lymph node; PALN, paraaortic lymph node; PET/CT, positron emission tomography/computed tomography CCRT, concurrent chemoradiotherapy.
Fig. 1. Overall survival rate after definitive concurrent chemoradiotherapy in the study cohort with locally advanced cervical cancer. During follow-up, 50 patients died of disease, and the 3 year and 5 year overall survival rates were 83% and 75%, respectively. CCRT = concurrent chemoradiotherapy.
Table 2 shows the results of the Cox proportional hazards regression model for predicting overall survival in LACC. Univariate analysis revealed that histology (p b 0.001), tumor size measured by MRI (p = 0.022), pelvic lymph node metastasis defined by MRI (p = 0.003), and paraaortic lymph node metastasis defined by MRI (p = 0.001) are significantly associated with overall survival. After multivariate analysis, histology (p = 0.001), tumor size (p = 0.004), and paraaortic lymph node metastasis (p = 0.008) remained as significant prognostic factors. A nomogram was constructed on the basis of this Cox proportional hazards model. It was composed of the three independent factors described above (Fig. 2). The point value assigned to each factor was proportionate to the hazard ratio that was derived from its own beta coefficients by regression analysis. Internal validation was performed by using bootstrapping resampling. The bootstrap-corrected concordance index for the model was 0.69. Fig. 3 shows the calibration of the constructed nomogram. The dashed line represents the performance of an ideal nomogram, in which predicted 100% survival matches the actuarial survival. The solid line represents the performance of the actual nomogram. The filled dots were obtained from a subcohort of the present database, and the check marks indicate bootstrap-corrected predictions of the nomogram's performance with 1000 resamples. When plotting the survival probabilities predicted by the current nomogram against the actuarial survival, the calibration curve lay close to the dashed line. The predictive ability of the nomogram was compared to that of the FIGO staging system. The concordance index for the FIGO staging system was 0.59, which means that the predictive ability of the nomogram was superior to that of the FIGO staging system (pb 0.05). Fig. 4 shows the heterogeneous nature of disease stratified by FIGO stage. Patients with the same stage disease had a broad spectrum of survival duration. The different stages also overlapped in terms of survival distribution, especially IIA, IIB, and IIIA. Discussion For both the patient with carcinoma and her (or his) physician, the question “will I die from cancer?” is a major concern. A prediction tool that shows who is at high risk for cancer-specific death will allow the physician to select ideal targets for more aggressive therapy. In recent years, such statistical prediction models have been developed for various types of cancers [20–22]. One such risk prediction model is the nomogram, which creates a simple graphical representation of a statistical predictive tool that generates a numerical probability of a clinical event.
S.-H. Shim et al. / Gynecologic Oncology 128 (2013) 54–59
57
Table 2 Univariate and multivariate Cox proportional hazards model for predicting overall survival in locally advanced cervical cancer. Variables
n
Univariate analysis HR (95% CI)
Age (as continuous variable) Age FIGO stage
Histology Tumor size by MRI (as continuous variable) Tumor size by MRI
Pretreatment hemoglobin (as continuous variable) Pretreatment hemoglobin
Pretreatment SCC Ag (as continuous variable) PM involvement by MRI
PLN metastasis by MRI PALN metastasis by MRI Hydronephrosis Bladder/rectal invasion PET/CT evaluation before CCRT
b55 years ≥55 years IB, IIA IIB, IIIA IIIB, IVA Squamous cell Non-squamous
94 115 25 132 52 190 13
≤4 cm >4 cm and ≤5 cm >5 cm b8 g/dL ≥8 g/dL and b12 g/dL ≥12 g/dL
75 62 72 209 13 91 105
No Yes, unilateral Yes, bilateral No Yes No Yes No Yes No Yes No Yes
28 115 66 111 98 180 29 174 35 199 10 118 91
0.981 (0.954–1.008) 1.000 0.784 (0.450–1.366) 1 1.235 (0.432–3.530) 2.701 (0.918–7.947) 1 3.894 (1.814–8.358) 1.193 (1.026–1.388) 1 1.880 (0.834–4.235) 3.212 (1.548–6.664) 0.896 (0.784–1.025) 2.150 (0.817–5.658) 1.218 (0.678–2.186) 1 0.988 (0.968–1.008) 1 0.895 (0.364–2.198) 1.699 (0.685–4.216) 1 2.386 (1.338–4.253) 1 3.127 (1.632–5.993) 1 1.448 (0.723–2.898) 1 2.126 (0.764–5.916) 1 0.920 (0.509–1.663)
Multivariate analysis p-Value
HR (95% CI)
p-Value
1.000 3.605 (1.674–7.764)
0.001
1 1.973 (0.873–4.460) 2.902 (1.394–6.044)
0.102 0.004
1 2.458 (1.260–4.798)
0.008
0.161 0.391 0.694 0.071 b0.001 0.022 0.128 0.002 0.110 0.121 0.509 0.238 0.808 0.253 0.003 0.001 0.296 0.149 0.782
FIGO, International Federation of Gynecology and Obstetrics; MRI, magnetic resonance imaging; SCC Ag, squamous cell carcinoma antigen; PM, parametrial; PLN, pelvic lymph node; PALN, paraaortic lymph node; PET/CT, positron emission tomography/computed tomography CCRT, concurrent chemoradiotherapy.; HR, hazard risk; CI, confidence interval.
Traditionally, the prognoses of patients with LACC who are treated with definitive CCRT have been estimated by using the FIGO staging system. However, previous studies have shown that each FIGO stage is associated with a broad spectrum of survival probabilities [11,12]. The present study confirmed the heterogeneity of overall survival for each stage and the overlapping of survival distributions between different stages. These findings suggest that FIGO stage alone is not entirely satisfactory with regard to providing an accurate prediction of individual outcome. Moreover, many prognostic factors other than FIGO stage have been found to affect survival [13–17]. In the present study, a three-variable risk assessment model that predicted the 5 year survival probability of patients with LACC who were treated with definitive CCRT was developed and internally validated. The selected variables for the nomogram were histology, tumor size, and paraaortic lymph node metastasis. Age, parametrial involvement, pelvic lymph node metastasis, pretreatment SCC Ag, pretreatment hemoglobin, hydronephrosis, and bladder or rectal invasion, which have been reported as prognostic factors in other studies, were not shown to be of significant prognostic value in the present study. In terms of predicting overall survival, the discriminating power of the model constructed in the present study was superior to that of the FIGO staging system alone. Tseng et al. analyzed the outcome of 251 patients with LACC who were treated with CCRT in Taiwan between 1999 and 2006, and developed a seven-variable nomogram for survival [11]. The variables that were incorporated were age, SCC Ag, tumor size, parametrial invasion, hydronephrosis, bladder/rectal invasion, and lymph node metastasis. As the authors mentioned, the study was limited by the fact that consolidation chemotherapy was performed in 23% of patients after CCRT, and about 70% of the study subjects were staged by CT. Notably, although the nomogram was based on seven variables, the concordance index was 0.69, which is not different from the concordance of the nomogram developed in the present study.
Individualized prediction based on the nomogram in the present study could help the physician to select adjunctive treatment after CCRT and to counsel the patient. Possible adjunctive treatments may be consolidation chemotherapy or hysterectomy. Two meta-analyses showed that consolidation chemotherapy in LACC after CCRT was effective and well tolerated [9,23]. Recently, Duenas-Gonzalez et al. reported the results of a phase III trial that showed that gemcitabine plus cisplatin chemoradiotherapy followed by adjuvant gemcitabine/cisplatin chemotherapy improved survival outcomes [24]. Hysterectomy may also be considered as an adjunctive treatment option for high-risk patients after CCRT, although this possibility is only supported by limited evidence [25–27]. PET/CT is known to be superior in detecting distant metastasis compared to conventional CT or MRI [28]. It was speculated in the present study that distant metastasis was more likely to be missed in patients without pretreatment PET/CT than in patients who did receive PET/CT. This would mean that the latter patients would have a better prognosis than patients who did not undergo pretreatment PET/CT. However, pretreatment PET/CT did not associate significantly with overall survival in the present study. Nevertheless, PET/CT seems to be better at identifying extrapelvic disease and nodal metastasis than conventional CT or MRI [29]. A recent study by Kang et al. introduced a risk assessment model that predicted distant recurrence in LACC [30]. The authors selected four variables, including histology, tumor size, and pelvic/paraaortic lymph node metastasis on PET/CT. As described in Methods section, we defined pelvic and paraaortic lymph node metastases using MRI as pretreatment PET/CT was performed in only 91 cases whereas MRI was performed in all enrolled patients. The concordance index of the study by Kang et al. was 0.72, which is better than the concordance index of the nomogram constructed in the present study. It remains possible that the more accurate prediction of the model of Kang et al. may reflect the more accurate evaluation of nodal status by PET/CT.
58
S.-H. Shim et al. / Gynecologic Oncology 128 (2013) 54–59
Fig. 2. Nomogram for predicting 3 year and 5 year survival probability in patients with locally advanced cervical cancer who were treated with definitive concurrent chemoradiotherapy. The nomogram was constructed by incorporating three variables. For each level of the prognostic variables, points were allocated according to the scale shown here. Total score was determined by adding up the individual parameter points, and these were then used to calculate the 3 year and 5 year survival probability. SCC = squamous cell carcinoma, MRI = magnetic resonance imaging, PALN = paraaortic lymph node.
The present study possesses all the limitations that are inherent in retrospective chart reviews. While all patients were given platinum-based chemotherapy, the chemotherapy regimens were not uniform. While this may cause a bias, the chemotherapy regimens did not differ significantly in terms of survival. In addition, some known prognostic factors such as human papilloma virus genotyping were not included in present study because of a respectable amount of missing value and over-fitness problem. Another known prognostic factor, namely tumor volume regression rate after CCRT, could also be a nomogram variable [31]. If this
Fig. 3. Calibration curve for the 5 year survival probability-predictive nomogram. Dashed line, the ideal reference value where predicted probabilities would match the observed 5 year survival; solid line, performance of the current nomogram; filled dots, calculated from a subcohort of the present database; check marks, bootstrap-corrected predictions of the performance of the nomogram; vertical bar, 95% confidence interval.
time-dependent variable can be incorporated into the present nomogram by using a time-dependent Cox regression model, it could increase the performance index of the nomogram. Furthermore, the nodal status was assessed by MRI, not surgical staging, which means that the real nodal metastasis status was not known. This is important because nodal metastasis in women with LACC is one of the strongest prognostic factors for survival [32]. While a tool that is more accurate than MRI (such as PET/CT) could be used to evaluate nodal metastasis, false negative results have been recorded in 12–22% patients who were evaluated by PET/CT [29]. To eliminate this bias, it is essential to determine the pathology-proven nodal status. However, the advantages of staging surgery must always be balanced against its potential morbidity, although the complication rate seems to be low when laparoscopy is performed by trained teams [29]. Finally, the present model was derived from
Fig. 4. The boxplot reveals the distribution of the observed overall survival for each FIGO stage. Note the heterogeneous nature of the overall survival associated with each stage, and how the different stages overlap in terms of survival distribution. • indicates a specific value above the upper fence, which is 1.5 times the interquartile range (above the 75th percentile). ★ indicates a extreme value above the upper fence, which is 3.0 times the interquartile range (above the 75th percentile).
S.-H. Shim et al. / Gynecologic Oncology 128 (2013) 54–59
patients attending a single institution. To determine the generalizability of the model, it must be validated externally. In conclusion, a risk assessment model for predicting the probability of 5 year survival after definitive CCRT in LACC was developed and internally validated. The derived model seems to predict the outcome more accurately than the FIGO staging system. Although this nomogram must be externally validated before it can be applied, it may be valuable in terms of choosing adjunctive treatment, counseling patients, and planning clinical trials. Conflict of interest statement The authors declare that there is no conflict of interest.
Disclosures Drs. Seung-Hyuk Shim, Shin-Wha Lee, Jeong-Yeol Park, Young Seok Kim, Dae-Yeon Kim, Jong-Hyeok Kim, Yong-Man Kim, Young-Tak Kim, and Joo-Hyun Nam have no conflicts of interest or financial ties to disclose. Acknowledgment We thank Hwa Jung Kim, M.D. and Sun Ok Kim, Department of Clinical Epidemiology and Biostatistics, University of Ulsan College of Medicine, Asan Medical Center for statistical consultation. References [1] Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D. Global cancer statistics. CA Cancer J Clin 2011;61:69-90. [2] Kim YT. Current status of cervical cancer and HPV infection in Korea. J Gynecol Oncol 2009;20:1-7. [3] Keys HM, Bundy BN, Stehman FB, Muderspach LI, Chafe WE, Suggs 3rd CL, et al. Cisplatin, radiation, and adjuvant hysterectomy compared with radiation and adjuvant hysterectomy for bulky stage IB cervical carcinoma. N Engl J Med 1999;340:1154-61. [4] Morris M, Eifel PJ, Lu J, Grigsby PW, Levenback C, Stevens RE, et al. Pelvic radiation with concurrent chemotherapy compared with pelvic and para-aortic radiation for high-risk cervical cancer. N Engl J Med 1999;340:1137-43. [5] Rose PG, Bundy BN, Watkins EB, Thigpen JT, Deppe G, Maiman MA, et al. Concurrent cisplatin-based radiotherapy and chemotherapy for locally advanced cervical cancer. N Engl J Med 1999;340:1144-53. [6] Whitney CW, Sause W, Bundy BN, Malfetano JH, Hannigan EV, Fowler Jr WC, et al. Randomized comparison of fluorouracil plus cisplatin versus hydroxyurea as an adjunct to radiation therapy in stage IIB-IVA carcinoma of the cervix with negative para-aortic lymph nodes: a Gynecologic Oncology Group and Southwest Oncology Group study. J Clin Oncol 1999;17:1339-48. [7] Peters 3rd WA, Liu PY, Barrett 2nd RJ, Stock RJ, Monk BJ, Berek JS, et al. Concurrent chemotherapy and pelvic radiation therapy compared with pelvic radiation therapy alone as adjuvant therapy after radical surgery in high-risk early-stage cancer of the cervix. J Clin Oncol 2000;18:1606-13. [8] McNeil C. New standard of care for cervical cancer sets stage for next questions. J Natl Cancer Inst 1999;91:500-1. [9] Chemoradiotherapy for Cervical Cancer Meta-Analysis C. Reducing uncertainties about the effects of chemoradiotherapy for cervical cancer: a systematic review and meta-analysis of individual patient data from 18 randomized trials. J Clin Oncol 2008;26:5802-12. [10] Pearcey R, Miao Q, Kong W, Zhang-Salomons J, Mackillop WJ. Impact of adoption of chemoradiotherapy on the outcome of cervical cancer in Ontario: results of a population-based cohort study. J Clin Oncol 2007;25:2383-8. [11] Tseng JY, Yen MS, Twu NF, Lai CR, Horng HC, Tseng CC, et al. Prognostic nomogram for overall survival in stage IIB-IVA cervical cancer patients treated with concurrent chemoradiotherapy. Am J Obstet Gynecol 2010;202:174e1-7. [12] Seo Y, Yoo SY, Kim MS, Yang KM, Yoo HJ, Kim JH, et al. Nomogram prediction of overall survival after curative irradiation for uterine cervical cancer. Int J Radiat Oncol Biol Phys 2011;79:782-7.
59
[13] Yalman D, Aras AB, Ozkok S, Duransoy A, Celik OK, Ozsaran Z, et al. Prognostic factors in definitive radiotherapy of uterine cervical cancer. Eur J Gynaecol Oncol 2003;24:309-14. [14] Hong JH, Chen MS, Lin FJ, Tang SG. Prognostic assessment of tumor regression after external irradiation for cervical cancer. Int J Radiat Oncol Biol Phys 1992;22:913-7. [15] Parker K, Gallop-Evans E, Hanna L, Adams M. Five years' experience treating locally advanced cervical cancer with concurrent chemoradiotherapy and high-dose-rate brachytherapy: results from a single institution. Int J Radiat Oncol Biol Phys 2009;74: 140-6. [16] Lai CH, Chang CJ, Huang HJ, Hsueh S, Chao A, Yang JE, et al. Role of human papillomavirus genotype in prognosis of early-stage cervical cancer undergoing primary surgery. J Clin Oncol 2007;25:3628-34. [17] Kidd EA, Siegel BA, Dehdashti F, Grigsby PW. The standardized uptake value for F-18 fluorodeoxyglucose is a sensitive predictive biomarker for cervical cancer treatment response and survival. Cancer 2007;110:1738-44. [18] Iasonos A, Schrag D, Raj GV, Panageas KS. How to build and interpret a nomogram for cancer prognosis. J Clin Oncol 2008;26:1364-70. [19] Kim YS, Shin SS, Nam JH, Kim YT, Kim YM, Kim JH, et al. Prospective randomized comparison of monthly fluorouracil and cisplatin versus weekly cisplatin concurrent with pelvic radiotherapy and high-dose rate brachytherapy for locally advanced cervical cancer. Gynecol Oncol 2008;108:195-200. [20] International Bladder Cancer Nomogram C, Bochner BH, Kattan MW, Vora KC. Postoperative nomogram predicting risk of recurrence after radical cystectomy for bladder cancer. J Clin Oncol 2006;24:3967-72. [21] Karakiewicz PI, Briganti A, Chun FK, Trinh QD, Perrotte P, Ficarra V, et al. Multi-institutional validation of a new renal cancer-specific survival nomogram. J Clin Oncol 2007;25:1316-22. [22] Wierda WG, O'Brien S, Wang X, Faderl S, Ferrajoli A, Do KA, et al. Prognostic nomogram and index for overall survival in previously untreated patients with chronic lymphocytic leukemia. Blood 2007;109:4679-85. [23] Green JA, Kirwan JM, Tierney JF, Symonds P, Fresco L, Collingwood M, et al. Survival and recurrence after concomitant chemotherapy and radiotherapy for cancer of the uterine cervix: a systematic review and meta-analysis. Lancet 2001;358: 781-6. [24] Duenas-Gonzalez A, Zarba JJ, Patel F, Alcedo JC, Beslija S, Casanova L, et al. Phase III, open-label, randomized study comparing concurrent gemcitabine plus cisplatin and radiation followed by adjuvant gemcitabine and cisplatin versus concurrent cisplatin and radiation in patients with stage IIB to IVA carcinoma of the cervix. J Clin Oncol 2011;29:1678-85. [25] Walji N, Chue AL, Yap C, Rogers LJ, El-Modir A, Chan KK, et al. Is there a role for adjuvant hysterectomy after suboptimal concurrent chemoradiation in cervical carcinoma? Clin Oncol (R Coll Radiol) 2010;22:140-6. [26] Colombo PE, Bertrand MM, Gutowski M, Mourregot A, Fabbro M, Saint-Aubert B, et al. Total laparoscopic radical hysterectomy for locally advanced cervical carcinoma (stages IIB, IIA and bulky stages IB) after concurrent chemoradiation therapy: surgical morbidity and oncological results. Gynecol Oncol 2009;114: 404-9. [27] Morice P, Rouanet P, Rey A, Romestaing P, Houvenaeghel G, Boulanger JC, et al. Results of the GYNECO 02 study, an FNCLCC phase III trial comparing hysterectomy with no hysterectomy in patients with a (clinical and radiological) complete response after chemoradiation therapy for stage IB2 or II cervical cancer. Oncologist 2012;17:64-71. [28] Liu FY, Yen TC, Chen MY, Lai CH, Chang TC, Chou HH, et al. Detection of hematogenous bone metastasis in cervical cancer: 18F-fluorodeoxyglucose-positron emission tomography versus computed tomography and magnetic resonance imaging. Cancer 2009;115:5470-80. [29] Gouy S, Morice P, Narducci F, Uzan C, Gilmore J, Kolesnikov-Gauthier H, et al. Nodal-staging surgery for locally advanced cervical cancer in the era of PET. Lancet Oncol 2012;13:e212-20. [30] Kang S, Nam BH, Park JY, Seo SS, Ryu SY, Kim JW, et al. Risk assessment tool for distant recurrence after platinum-based concurrent chemoradiation in patients with locally advanced cervical cancer: a Korean gynecologic oncology group study. J Clin Oncol 2012;30:2369-74. [31] Nam H, Park W, Huh SJ, Bae DS, Kim BG, Lee JH, et al. The prognostic significance of tumor volume regression during radiotherapy and concurrent chemoradiotherapy for cervical cancer using MRI. Gynecol Oncol 2007;107: 320-5. [32] Duenas-Gonzalez A, Orlando M, Zhou Y, Quinlivan M, Barraclough H. Efficacy in high burden locally advanced cervical cancer with concurrent gemcitabine and cisplatin chemoradiotherapy plus adjuvant gemcitabine and cisplatin: prognostic and predictive factors and the impact of disease stage on outcomes from a prospective randomized phase III trial. Gynecol Oncol 2012;126:334-40.