2020 Volume 2 Issue 3 Pages 75-83
BACKGROUND
Unlike the recommendations made in many other countries, Japanese guidelines equally recommend radical hysterectomy or concurrent chemoradiotherapy for treatment of stage IIB cervical carcinoma. The main study objective was to compare the overall mortality of hysterectomy versus concurrent chemoradiotherapy as primary treatment in patients with localized or regionally extended cervical cancer.
METHODS
Using Diagnosis Procedure Combination database combined with population-based cancer registry data in Osaka Prefecture, Japan, we conducted a retrospective cohort study. All adult patients who had been diagnosed with cervical cancer, registered in the population-based cancer registry from January 1, 2010 to December 31, 2015 were included. To compare overall mortality between patients who received radical hysterectomy and concurrent chemoradiotherapy as primary treatment, we performed a Cox regression analysis of the original cohort, and Kaplan-Meier analysis with stabilized inverse probability of treatment weights using propensity score.
RESULTS
Among 740 eligible patients, 564 patients were included in the hysterectomy group and 176 patients were included in the concurrent chemoradiotherapy group. Primary hysterectomy was not independently associated with overall mortality (adjusted HR 0.70, 95% CI 0.46–1.07) by the Cox regression analysis. The Kaplan-Meier analysis with stabilized inverse probability of treatment weights did not show a significant difference in overall mortality between the two groups (P = 0.096).
CONCLUSIONS
This study indicates that primary treatment type (hysterectomy versus concurrent chemoradiotherapy) was not statistically associated with overall mortality among patients diagnosed with localized or regionally extended cervical cancer.
Three types of optimal treatment are broadly recommended for Stage IB to IIA cervical cancer: radical hysterectomy, external beam radiation therapy, or concurrent chemoradiotherapy (CCRT). National Comprehensive Cancer Network (NCCN) guidelines published in 2019 shows radical hysterectomy and radiotherapy as equal options for the optimal treatment of Stage IB1 and IIA1, and radical hysterectomy and CCRT for the optimal treatment of Stage IB2 and IIA2 [1]. The Clinical Practice Guideline of the American Society of Clinical Oncology recommends the following: (i) CCRT is standard treatment for women with stage IB to IVA cervical carcinoma, and (ii) a hysterectomy is recommended for patients with stage IA2, IB1, or IIA1 disease [2].
Previous studies have suggested that primary surgery among women with advanced cervical cancer could not be recommended because surgical cure rate is generally low, and most of patients with advanced cancer will require adjuvant treatment associated with a high incidence of several complications [3, 4]. In contrast, the 2017 Japan Society of Gynecologic Oncology guideline emphasizes surgery as optimal treatment for locally invasive cervical carcinoma [5]. The guideline recommends both radical hysterectomy or CCRT for the patients with stage IIB squamous cell cervical carcinoma equally, and a recent report noted that, in Japan, surgical treatment including radical hysterectomy has been performed more frequently as the primary treatment [5]. Even in stage IIB lesions of cervical cancer, surgery as primary treatment is the main approach in 30–50% of patients [6, 7]. Therefore, there may be worse surgical outcomes in Japanese clinical settings with preference for surgery. However, there have been no large or well-designed studies to analyze actual clinical outcome among patients with large-sized local invasive lesions, including stage IIB who received surgical treatment in a real-world setting in Japan.
The aim of the current multi-center study was to compare overall mortality between primary hysterectomy and CCRT in patients diagnosed with localized or regionally extended cervical cancer in Japan.
In the current study, we combined two data sources to create a large consolidated database with long-term mortality and detailed clinical information: the Osaka Cancer Registry (OCR) and the Diagnosis Procedure Combination (DPC) database.
OCR is a population-based cancer registry and collects information on diagnoses of cancer and mortality in residents of Osaka Prefecture, Japan. Data on patient demographics from the OCR include sex, age at cancer diagnosis, vital status (alive or dead), and date of death or the last follow-up for vital status. Specific information on cancer include cancer site; Surveillance, Epidemiology, and End Results (SEER) summary stage at diagnosis [8]; and date of cancer diagnosis. Cancer stage was categorized as follows: localized, regional to lymph nodes, regional by direct extension, distant, or unknown. Also, patients diagnosed with cancer between 2010 and 2015 were followed up in December 2018 using official resident registries to verify vital status.
The DPC database is a national database for acute-care inpatients in Japan [9, 10]. Academic hospitals are obliged to participate in the database, while community hospitals are voluntary on participation. We collected DPC data from 36 hospitals in the Osaka cancer medical cooperation council which are designated as cancer care hospitals by the national or prefectural government. The hospitals treated approximately 50% of all cancer patients within the prefecture from 2010 through 2015. The database contains discharge abstracts and administrative reimbursement data for inpatient episodes, together with the following: unique identifiers of hospitals; dates of admission and discharge; type of admission (planned or emergency); basic patient characteristics (age, sex, body height and weight); smoking status (nonsmoker or current/past smoker); diagnoses (primary and secondary); pre-existing comorbidities on admission and complications during hospitalization; medical procedures performed; dates and types of surgery (Japanese original codes); medications and devices used; and in-hospital mortality. Diagnoses, comorbidities, and complications are registered using International Classification of Diseases Tenth Revision (ICD-10) codes and text data in Japanese. Laboratory data and results on obstetric and gynecologic examination (including findings of ultrasound echogram or magnetic resonance imaging and vaginal bacteriological culture). To link the diagnoses accurately with reimbursement for health care costs, the attending physicians are obliged to register the diagnoses with reference to medical records. Several studies have been conducted using the DPC database in the field of obstetrics and gynecology [11, 12]. The validity of the diagnostic records in the DPC database was reported to be generally high; the sensitivity and specificity of the primary diagnoses were 50–80% and 96%, respectively [13]. The sensitivity and specificity of procedures performed during hospitalization were found to exceed 90% [13].
The two databases were linked at the patient level using the patient identification number of each hospital as the linkage key. Approximately 98% of the patients in the OCR database at the 36 hospitals were linked with their corresponding inpatient information in the DPC database.
PATIENTSThis retrospective cohort study included all adult patients who had been diagnosed with cervical cancer according to the topographical codes of the International Classification of Diseases for Oncology, Third Edition (C53.x) at the 31 subject hospitals and had been included in the OCR between January 1, 2010 and December 31, 2015. From this pool, we excluded patients with the following: missing vital status, age at diagnosis, or cancer stage; a diagnosis of early stage (carcinoma in situ), metastasis or unspecified stage; and primary treatment by palliative surgery. The STROBE guidelines were followed to outline the current retrospective cohort study. The current study was approved by the Institutional Review Board of Osaka International Cancer Institute. The Institutional Review Board of Osaka International Cancer Institute waived the need for informed consent given the anonymous nature of the current study.
VARIABLESAs baseline patient characteristics, we extracted information on age, body mass index, smoking status, calendar year of cancer diagnosis, clinical cancer stage based on SEER summary staging, multiple cancer, consciousness level, activities of daily living, comorbidities at diagnosis, and overall survival.
We categorized years of diagnosis as 2010–2011, 2012–2013, and 2014–2015. Because the database did not contain detailed information on FIGO stage, cancer stage was categorized into “localized” or “regional extension” (including regional lymph nodes or adjacent organs) in the current study. Therefore, we could not analyze only for patients with stage IIB. The DPC database also contains Japan Coma Scale (JCS) scores, which correlates well with the Glasgow Coma Scale [14]. Consciousness scored at 100 points on the JCS is almost equivalent to a score of 6–9 on the Glasgow Coma Scale. We identified patients with disturbance of consciousness at first admission to hospital (JCS = 1–300). Activities of daily living at hospital admission of each patient was included as variable (dependent or independent).
Comorbidities at first admission after cancer diagnosis were identified. The comorbidities included hypertension (ICD-10 code, I10), type 2 diabetes mellitus (E11, E110-119, E14), chronic kidney disease (N189), old cerebral infarction (I693), and composite cardiac complication (chronic heart failure; I500, I501, I509, I110, angina pectoris; I200, I201, I208, I209, I238, atrial fibrillation; I48, I480–482, I489, aneurysm; I712).
STATISTICAL ANALYSISThe primary outcome was overall mortality. Duration of survival was defined as the time interval between the date of cancer diagnosis and the date of death or the censor date of the last follow-up for vital status. The primary objective of the current analysis was to examine the association of primary treatment (hysterectomy versus CCRT) and overall mortality among patients diagnosed with cervical cancer (local or organ-invasive lesion). We used the Kaplan-Meier method in order to construct survival curves, and assessed differences by the log-rank test. Cox proportional hazard regression models was used to assess the association of primary treatment and survival on multivariable analysis using all the variables. The magnitude of statistical significance was expressed with hazard ratio (HR) and 95% confidence intervals (CIs). We assessed the proportional hazard assumption by the Schoenfeld Residuals Test and there was no interaction with time (data not shown).
We conducted a stabilized inverse probability of treatment weight (IPTW) analysis using propensity scores to compare the outcomes between patients who received hysterectomy and those who received CCRT as primary treatment to account for differences in patient background characteristics. Stabilized IPTW employs propensity scores and adjusts for measured potential confounding while preserving sample size [15]. We estimated propensity scores with a logistic regression model for the likelihood of undergoing primary hysterectomy as a function of all variables of patient characteristics including age, body mass index, smoking status, calendar year of cancer diagnosis, cancer stage, multiple cancers, consciousness level, activities of daily living, and comorbidities at diagnosis [16]. After weighting, we assessed the balance of baseline variables in the both groups with the standardized difference with a threshold of 0.2 to denote significant between-group differences [17]. Mortality risk over time was assessed with weighted Kaplan-Meier cumulative incidence plots. We compared overall mortality between the hysterectomy and CCRT groups using the log-rank test.
Categorical variables were described using numbers and percentages and were compared using Fisher’s exact test. Continuous variables were described using mean and standard deviation or median and interquartile range and were compared using Student’s t test or the Mann-Whitney U test. Risk differences and 95% CIs for the primary outcome were also reported.
All statistical analyses were performed using Stata software version 16.0 (StataCorp LP, College Station, TX, USA). All tests were two-sided, and p values less than 0.05 was considered statistically significant for all tests.
During the study period, we identified 1,552 registered patients with a diagnosis of cervical cancer in 31 hospitals and selected 740 of these patients as eligible. Of them, 564 (76.2%) had undergone primary radical hysterectomy (surgery group) and 176 (23.8%) had received primary CCRT (CCRT group). The stabilized IPTW analysis created a surgery group of 567 patients and a CCRT group of 169 patients (Fig. 1).
Table 1 shows the baseline characteristics of all patients (N = 740) and patients after stabilization by IPTW estimation (N = 736). Among all patients, the surgery group was significantly more likely to have younger patients, smokers, and patients with cancer at the local lesion stage whereas the CCRT group had a larger proportion of patients with diabetes mellitus. Among the surgery group, 19.9% received only radiotherapy and 47.2% of patients received only chemotherapy. Combined radiotherapy and the chemotherapy was given to 28.6% of patients who received primary surgery. The distributions of the variables in the patients after stabilized IPTW were well balanced.
| Characteristics | Eligible patients before weighting | Eligible patients after weighting | ||||
|---|---|---|---|---|---|---|
| Surgery group (n = 564) | CCRT group (n = 176) | ASD, % | Surgery group (n = 567) | CCRT group (n = 169) | ASD, % | |
| Age, yearsa | 48.7 (13.4) | 56.9 (13.6) | 61.0 | 50.6 (13.9) | 52.4 (13.2) | 13.0 |
| Body mass index, kg/m2 a | 21.8 (3.3) | 21.8 (4.7) | 0.3 | 21.8 (3.4) | 21.9 (4.7) | 1.2 |
| Disturbance of consciousness on admission (JCS = 1–300) | 2 (0.4) | 2 (1.1) | 3.2 | 4 (0.7) | 0 | 7.0 |
| Dependent in activities of daily living | 513 (91.0) | 150 (85.2) | 17.7 | 508 (89.6) | 156 (92.3) | 6.1 |
| Smoking | 178 (31.6) | 39 (22.2) | 20.5 | 167 (29.5) | 50 (29.6) | 0.3 |
| Cancer stage | ||||||
| Localized | 329 (58.3) | 20 (11.4) | 112.7 | 269 (47.5) | 73 (43.4) | 9.8 |
| Regional extension | 235 (41.7) | 156 (88.6) | 112.7 | 297 (52.5) | 95 (56.2) | 8.7 |
| Multiple cancers | 40 (7.1) | 17 (9.7) | 10.2 | 42 (7.4) | 11 (6.5) | 3.7 |
| Diagnosis year | ||||||
| 2010–2011 | 98 (17.4) | 25 (14.2) | 7.9 | 96 (16.9) | 36 (21.3) | 11.3 |
| 2012–2013 | 171 (30.3) | 63 (35.8) | 10.2 | 171 (30.2) | 37 (21.9) | 17.3 |
| 2014–2015 | 295 (52.3) | 88 (50.0) | 3.8 | 300 (52.9) | 96 (56.8) | 8.0 |
| Comorbidities on cancer diagnosis | ||||||
| Hypertension | 31 (5.5) | 15 (8.5) | 12.1 | 36 (6.3) | 10 (5.9) | 1.5 |
| Diabetes mellitus (type 2) | 14 (2.5) | 13 (7.4) | 23.0 | 19 (3.4) | 5 (3.0) | 1.2 |
| Chronic kidney disease | 1 (0.2) | 0 | 6.0 | 3 (0.5) | 0 | 16.4 |
| Old cerebral infarction | 3 (0.5) | 1 (0.6) | 0.5 | 3 (0.5) | 1 (0.6) | 0.9 |
| Cardiac complication | 6 (1.1) | 3 (1.7) | 5.6 | 7 (1.2) | 2 (1.2) | 1.5 |
Data are shown as n (%)
a Data are shown as mean (standard deviation)
ASD absolute standardized difference, CCRT concurrent chemoradiotherapy, JCS Japan coma scale, IPTW inverse probability of treatment weight
Median follow-up of censored cases was 3.7 years (interquartile range, 2.1–4.4). A total of 122 deaths were noted, 75 in the surgery group and 47 in the CCRT group.
The rate of overall survival at 3.0 years was 88.6% in the surgery group and 72.6% in the CCRT group (difference, −15.9 percentage points; 95% CI, −8.8 to −23.0). At 5.0 years, the difference in overall survival between these groups had increased (82.9% vs 60.9%, difference, −22.0 percentage points; 95% CI, −14.1 to −29.9) (Fig. 2).
Table 2 shows the hazard risks of each variables for mortality among all eligible patients before weighting by the Cox regression analysis. Primary surgery was not an independent prognostic factor for overall mortality (adjusted HR 0.70, 95% CI 0.46–1.07, p = 0.10). In addition, after stabilized IPTW estimation, overall mortality was not significantly different between the surgery group and CCRT group by the Kaplan-Meier method and the log-rank test (p = 0.096) (Fig. 3).
| Variables | Hazard Ratio | 95% Confidence Interval |
|---|---|---|
| Primary treatment | ||
| CCRT | Ref | |
| Surgery | 0.70 | 0.46–1.07 |
| Age, yearsa | 1.00 | 0.99–1.02 |
| Body mass index, kg/m2 a | 0.95 | 0.91–1.00 |
| Disturbance of consciousness on admission (JCS = 1–300) | 9.30 | 1.21–71.4 |
| Activities of daily living | ||
| Dependent | Ref | |
| Independent | 0.78 | 0.47–1.28 |
| Smoking | 1.26 | 0.84–1.89 |
| Cancer stage | ||
| Local | Ref | |
| Organ invasion or lymph node metastasis | 3.47 | 2.15–5.61 |
| Multiple cancer | 2.11 | 1.27–3.53 |
| Diagnosis year | ||
| 2010–2011 | Ref | |
| 2012–2013 | 0.77 | 0.49–1.20 |
| 2014–2015 | 0.55 | 0.33–0.92 |
| Comorbidities on cancer diagnosis | ||
| Hypertension | 1.40 | 0.68–2.88 |
| Diabetes mellitus (type 2) | 1.13 | 0.48–2.69 |
| Chronic kidney disease | 10.0 | 0.53–187.4 |
| Old cerebral infarction | 1.32 | 0.17–10.2 |
| Cardiac complication | 0.98 | 0.13–7.20 |
Variables are categorical
a Variables are numerical
CCRT concurrent chemoradiotherapy, JCS Japan coma scale
In the current study, we combined administrative data with survival status from population-based registry data, and showed that approximately three quarters of patients with local or organ-invasive cervical cancer received hysterectomy as primary treatment in Japan. The Cox regression analysis showed that primary treatment (hysterectomy versus CCRT) was not an independent prognostic factor for overall mortality. Using stabilized IPTW analysis, long-term overall survival was not statistically different between the hysterectomy and CCRT groups.
To our knowledge, this is the largest study on the long-term outcome of localized or regionally extended cervical cancer among patients mainly received hysterectomy as primary treatment. Although the current Japanese guidelines emphasize optimal surgical treatment for locally invasive cervical carcinoma [5], there has been no supporting evidence among Japanese patients. Because our results did not show a statistically worse survival outcome in the hysterectomy group than the CCRT group, for localized or regionally extended cervical cancer they are probably equivalent treatment options in terms of prognosis. However, there may be low statistical power to assess the main results in the current study according to point estimation and 95% confidence intervals of adjusted hazard ratio.
Previous retrospective studies have shown comparable outcomes for International Federation of Gynecology and Obstetrics (FIGO) stage IIB cervical cancer between radical hysterectomy with adjuvant radiotherapy and definitive radiotherapy with or without chemotherapy in Japan [18–20]. Our results were similar, but there was approximately 12% difference in the overall mortality at 5 years between the two groups. Although the baseline characteristics and cancer stage were adjusted by the stabilized IPTW for both groups, possible unmeasured confounders such as socioeconomic status, results of laboratory tests, and size and pathological findings of the tumor might lead to incomplete adjustment in our study. The unmeasured confounders could widen the difference in the overall mortality between the two groups.
In the present study, 76.2% of the eligible patients underwent radical hysterectomy as primary treatment. The result reflected recent Japanese clinical practice, which mainly focuses on surgical treatment, including radical hysterectomy for stage I–II cervical cancer [6, 7]. However, the current study does not have detailed information on FIGO stage. Also, the proportions of adjuvant radiotherapy alone and CCRT in high-risk patients with cervical cancer among the hysterectomy group could not be assessed in our study, although there was a recommendation on adjuvant therapy after primary surgery for patients with high- and intermediate-risk cervical cancer in early stage by NCCN and the Japan Society of Gynecologic Oncology [5, 18].
Compared with the previous studies, the current study had two advantages. First, our study analyzed a larger number of patients from multiple hospitals [19–21]. Second, our study used stabilized IPTW to balance the patient backgrounds between the hysterectomy and CCRT groups in real-world clinical settings. A population-based registry data has an additional strength on high follow-up rate because the vital status of all registered patients is continuously tracked. Vital status is routinely updated using death certificates.
There are several limitations associated with this study. First, although patient background information was obtained from an administrative database, clinical information recorded in the database may have been less well validated than those in clinical databases or medical charts. However, most studies that have examined the association between patient background information including comorbidities and cancer mortality were conducted using administrative data linked with cancer registry data [22]. Second, because the database used in our study did not contain detailed information on FIGO stage, we used categorical cancer stage based on SEER summary staging. Third, we could not determine the physician’s information including surgical skill and experience in each hospital. Radical hysterectomy for cervical cancer is likely to be complicated due to cancer invasion or intrapelvic adhesions. Previous studies have shown that better surgical skill and higher surgical volumes are associated with improved surgical outcomes in gynecologic procedures [23, 24]. Fourth, there was no detailed information on chemotherapy regimen, irradiation dose, surgical approach, pathohistological findings or performance status in the current study. The lack of these information may lead to selection biases. Finally, complications after primary treatment could not be assessed in our study. Some studies have reported that radical hysterectomy itself and adjuvant treatment including chemotherapy and radiotherapy were associated with an increased incidence of serious complications [25–27]. In Japan, a recent paper showed increased risk of grade 3–4 adverse events with postoperative radiotherapy and increased risk of bladder dysfunction with adjuvant chemotherapy among patients with cervical carcinoma of stage IB–IIB [28]. Although the current study did not show a statistically worse survival outcome in the hysterectomy group than the CCRT group, physicians should decide a primary treatment with consideration of complications and quality of life.
In conclusion, the findings of the present retrospective study indicate that primary treatment type (hysterectomy versus CCRT) was not statistically associated with overall mortality among patients diagnosed with localized or regionally extended cervical cancer.
None.
The authors declare that they have no conflict of interests.
This work was supported by a Ministry of Health, Labor and Welfare Sciences Research Grants [H30-Gantaisaku-ippan-009]. These funding organizations had no role in planning research, analysis, interpretation of results, or report writing.
