Abstract
Objective:
To determine whether process and outcome measures varied for patients with early-stage cervical cancer based on hospital surgical volume.
Methods:
Using the National Cancer Database, we identified women with stages IA2 – IB1 cervical cancer (2011–2013). Annual hospital volume was calculated using number of hysterectomies performed in the prior year and grouped into patient level-quartiles. Centers in the highest quartile of volume were defined as HVCs; those in the lowest quartile, as LVCs. Demographics, type/mode of hysterectomy, lymph node assessment, NCCN-compliant surgery (radical hysterectomy (RH) with LND), and survival outcomes were compared across quartiles of hospital volume. Cox Proportional Hazards model was performed to determine impact of volume on mortality.
Results:
We identified 3,469 women treated at 598 different hospitals. RH was more likely at HVCs versus LVCs (68.9% vs. 59.6%, p<0.001). LND was more likely at HVCs versus LVCs (96.1% vs 87.3%, p< 0.001). Patients treated at HVCs were 11.4% more likely to receive guideline-compliant surgery compared to LVCs (67.8% vs. 56.4%, p<0.001). There was no difference in 5-year survival, 90-day survival, all-cause mortality across volume quartiles. Thirty-day mortality was significantly lower at HVCs (0 deaths in 880 patients) versus LVCs (1 in 1,058 (0.1%; p=0.02). Age >=80, Medicaid and Medicare insurance, Hispanic race, and poorly differentiated histology were independent predictors of mortality. Hospital volume was not found to be an independent predictor of mortality (p=0.95).
Conclusions:
HVCs demonstrated higher rates of NCCN-recommended surgery for early-stage cervical cancer. There was no association between hospital volume and survival.
INTRODUCTION
There will be an estimated 13,170 new cases of cervical cancer diagnosed in 2019 [1]. For women with International Federation of Gynecology and Obstetrics (FIGO) 2009 Stage IA2 and 1B1 disease, which includes tumors that are less than 4 cm and confined to the cervix, the 5-year survival rate is 92% [1]. The recommended surgical treatment for Stage IA2 and IB1 disease includes radical hysterectomy (or radical trachelectomy in cases where fertility preservation is desired) with pelvic lymph node sampling [2]. An acceptable alternative to surgery is pelvic external beam radiotherapy with brachytherapy [2].
Radical hysterectomy is a complex surgical procedure that involves removal of the uterus, cervix, upper part of the vagina, and parametrial tissue. In a recent multicenter retrospective study, Matsui et al., reported that patients with Stage IB1 - IIA cervical cancer who underwent radical hysterectomy and received treatment at higher volume centers had decreased rates of local recurrence and all-cause mortality [3]. Additional studies involving women with locally advanced cervical cancer have demonstrated improvements in guideline compliance and survival when care is delivered at high volume centers (HVCs) [4,5,6]. Robin et al. reported improved guideline compliance at HVCs and a corresponding survival benefit when guidelines were followed [4]. Wright et al. evaluated patients who were treated with radiation therapy and found that rates of guideline adherence were higher at HVCs, but that hospital volume did not affect survival [5]. Lastly, Richard et al. studied patients with stage IIB – IIIB cervical cancer who received radiation therapy and found that treatment at HVCs was independently associated with better adherence to guidelines and improved survival [6].
For women with early-stage cervical cancer, it remains unclear how often guideline recommended care is delivered and whether hospital volume affects guideline compliance and survival outcomes. In this analysis, we explore surgical treatment patterns in women diagnosed with early-stage cervical cancer, and whether hospital volume affects surgical guideline compliance and survival outcomes.
MATERIALS AND METHODS
Data source and patient selection
The National Cancer Data Base (NCDB) is a nationwide hospital registry of cancer patients sponsored by the American Cancer Society and the American College of Surgeons [7]. The NCDB gathers information on approximately 70% of all new invasive cancer diagnoses from more than 1,500 Commission on Cancer-affiliated hospitals within the United States. Data captured include hospital factors, patient demographics, disease stage, first course of treatment, and overall survival (OS).
This study was deemed to be non-human subjects research and approved by the Columbia University Institutional Review Board. We queried the NCDB for all patients diagnosed with cervical cancer from 2010 to 2013 (n=38,545). We included all women with Stage IA2 and IB1 cervical cancer, with squamous cell, adenocarcinoma, or adenosquamous histologies, who underwent radical or simple hysterectomy for their initial treatment (n=5,263). We excluded all patients with an unknown surgical modality (n=453), patients for whom this was not a first or only cancer diagnosis (n=209), those without a microscopic confirmation of cancer (n=6), and those who received chemotherapy or radiotherapy prior to surgery (n=48). The original cohort included 4,547 patients from 661 hospitals.
Defining hospital volume
Hospital volume was computed using the number of simple and radical hysterectomies (Codes: 50, 51, 52, 53, 54, 60, 61, 62) performed at a given hospital in the previous year on patients with suspected, confirmed, or occult Stage IA2 or IB1 cervical cancer (C53.0 – C53.9).. Hospital volume was analyzed as a categorical variable and divided into patient-level volume quartiles. Based on patient-level quartiles from procedures performed in the previous year, low volume was defined as an annual volume of ≤1, medium-low as a volume of 2, medium-high as a volume of 3 – 5, and high as a volume of 6 – 20 cases in the prior year (Table 1). As seen in Table 1, the volume of patients in each quartile are not evenly distributed as they are based on volume from the previous year. Sensitivity analyses were performed using volume as a continuous variable. The final analytical cohort excluded patients from 2010 due to the absence of surgical modality, leaving a total of 3,469 patients for the analysis.
Table 1:
Patient-level quartiles of prior year volume
| Volume Category | Number of patients (N=3469) | Annual Cervical Cancer Hysterectomy Volume |
|---|---|---|
| Low | 1,058 | 0–1 |
| Medium-Low | 519 | 2 |
| Medium-High | 1,012 | 3–5 |
| High | 880 | 6–20 |
Data collection
For the 3,469 patients included in the analysis, demographic characteristics including age at diagnosis (<40, 40–49, 50–59, 60–69, 70–79, ≥80 years), race (white, black, Hispanic, other, unknown), year of diagnosis (2011, 2012, 2013), insurance status (uninsured, private, Medicaid, Medicare, other/unknown), and income (<$38,000, $38,000-$47,999, $48,000-$62,999, ≥$63,000, unknown) were evaluated.
Location was classified as metropolitan, urban, or rural, determined by matching zip codes to rural-urban continuum codes from the United States Department of Agriculture Economic Research Service. Comorbidity status was described using the Deyo classification of the Charlson Comorbitidy score (0, 1, ≥2) [8]. Facility type was based on classification by the American College of Surgeons Commission on Cancer Accreditation program as a community cancer program, comprehensive community cancer program, academic/research program, or integrated network cancer center [5]. The region of the treating facility was classified as Northeast, Midwest, South, or West. Tumor characteristics included stage based on American Joint Committee on Cancer (AJCC) clinical stage, FIGO stage, or AJCC pathologic stage (IA2 or IB1), histology (squamous, adenocarcinoma, or adenosquamous), and grade (well, moderate, poorly differentiated, or unknown). Surgical modality was classified as intent-to-treat and reported as minimally invasive surgery (MIS) or open.
Defining outcomes and process measures
The primary process measure evaluated was National Comprehensive Cancer Network (NCCN) guideline-compliant surgical management. NCCN guideline compliance was defined as a radical hysterectomy with lymph node assessment. The secondary outcome of interest was 30-day mortality, 90-day mortality, and OS, reported as all-cause mortality, which includes death from cancer and other causes.
Statistical analysis
We sought to examine the effect of hospital volume on NCCN guideline-compliant surgical management and OS. An unadjusted analysis was performed to compare demographics and outcomes of interest across volume quartiles using X2 test. The results of these tests were described in the text to highlight a comparison of differences between centers in the highest quartile of volume (HVCs) and centers in the lowest quartile of volume (LVCs). However, all statistical tests were conducted across volume quartiles.
OS was compared across volume quartiles using the Kaplan-Meir method and compared using the Log-rank test. Thirty-day and 90-day mortality were compared across volume quartiles using X2 test. A marginal multivariable Cox Proportional Hazards model was used to determine factors associated with all-cause mortality, accounting for hospital-level clustering. Results from the Cox Proportional Hazards models are reported as hazards ratios (HRs) with 95% confidence intervals (CIs). A sensitivity analysis comparing the effect of volume as a continuous variable was performed using a similar model. All hypothesis tests were two-sided. A P-value of <0.05 was considered statistically significant. Analyses were conducted using SAS version 9.4 (SAS institute Inc., Cary, NC).
RESULTS
We identified 3,469 women treated at 598 different hospitals, including 1,058 (30.5%) at LVCs, 519 (15.0%) at medium-low volume centers, 1,012 (29.2%) at medium-high volume centers, and 880 (25.4%) at HVCs (Table 1). Most patients had stage 1B1 disease (86.3%), underwent radical hysterectomy (64.7%), and received MIS (57.6%). Additionally, most patients were white (66.8%), had private insurance (57.3%), and lived in metropolitan areas (82.5%).
Hospital volume was found to be associated with race, year of diagnosis, insurance type, median household income, patient’s residence area, facility type, facility location, stage, grade, hysterectomy type, and lymphadenectomy (Table 2). Uninsured patients comprised a larger proportion of patients at HVCs (9.8% vs 5.6%) and Medicaid patients comprised a larger proportion of patients at LVCs (21.3% vs. 17.6%) (P=0.03). Patients residing in zip codes with a median household income <$38,000 were more likely to receive care at HVCs (20.1% vs. 16.6%) (P=0.02). Patients living in metropolitan areas were more likely to receive care at LVCs (85.0% vs 78.9%) and those living in urban areas were more likely to receive care at HVCs (16.0% vs 11.4%) (P<0.001). HVCs were more frequently described as academic/research facilities (71.1%) whereas LVCs were more frequently described as comprehensive community cancer centers (46.1%) (P<0.001).
Table 2.
Characteristics stratified by hospital volume quartiles
| Characteristics | Low | Medium Low | Medium High | High | P-value | ||||
|---|---|---|---|---|---|---|---|---|---|
| N | % | N | % | N | % | N | % | ||
| Number of patients | 1,058 | (30.5) | 519 | (15.0) | 1,012 | (29.2) | 880 | (25.4) | |
| Number of hospitals | |||||||||
| Age of diagnosis | 0.47 | ||||||||
| <40 | 323 | (30.5) | 174 | (33.5) | 342 | (33.8) | 303 | (34.4) | |
| 40–49 | 349 | (33.0) | 153 | (29.5) | 311 | (30.7) | 296 | (33.6) | |
| 50–59 | 214 | (20.2) | 106 | (20.4) | 209 | (20.7) | 164 | (18.6) | |
| 60–69 | 199 | (11.2) | 66 | (12.7) | 111 | (11.0) | 76 | (8.6) | |
| 70–79 | 41 | (3.9) | 15 | (2.9) | 29 | (2.9) | 33 | (3.8) | |
| >/=80 | 12 | (1.1) | * | * | 10 | (1.0) | * | * | |
| Race | 0.01 | ||||||||
| White | 723 | (68.3) | 351 | (67.6) | 642 | (63.4) | 602 | (68.4) | |
| Black | 112 | (10.6) | 59 | (11.4) | 113 | (11.2) | 100 | (11.4) | |
| Hispanic | 149 | (14.1) | 70 | (13.5) | 176 | (17.4) | 123 | (14.0) | |
| Other | 54 | (5.1) | 36 | (6.9) | 75 | (7.4) | 50 | (5.7) | |
| Unknown | 20 | (1.9) | * | * | * | * | * | * | |
| Year of diagnosis | 0.001 | ||||||||
| 2011 | 353 | (33.4) | 182 | (35.1) | 332 | (32.8) | 293 | (33.3) | |
| 2012 | 366 | (34.6) | 186 | (35.8) | 295 | (29.2) | 320 | (36.4) | |
| 2013 | 339 | (32.0) | 151 | (29.1) | 385 | (38.0) | 267 | (30.3) | |
| Insurance | 0.03 | ||||||||
| Uninsured | 59 | (5.6) | 36 | (6.9) | 86 | (8.5) | 86 | (9.8) | |
| Private | 611 | (57.8) | 303 | (58.4) | 566 | (55.9) | 509 | (57.8) | |
| Medicaid | 225 | (21.3) | 111 | (21.4) | 199 | (19.7) | 155 | (17.6) | |
| Medicare | 124 | (11.7) | 55 | (10.6) | 126 | (12.5) | 90 | (10.2) | |
| Other government/unknown | 39 | (3.7) | 14 | (2.7) | 35 | (3.5) | 40 | (4.5) | |
| Median household income | 0.02 | ||||||||
| <$38,000 | 176 | (16.6) | 88 | (17.0) | 220 | (21.7) | 177 | (20.1) | |
| $38,000-$47,999 | 258 | (24.4) | 114 | (22.0) | 268 | (26.5) | 213 | (24.2) | |
| $48,000-$62,999 | 308 | (29.1) | 150 | (28.9) | 245 | (24.2) | 243 | (27.6) | |
| $63,000+ | 315 | (29.8) | 164 | (31.6) | 278 | (27.5) | 244 | (27.7) | |
| Unknown | * | * | * | * | * | * | * | * | |
| Patient’s Residence Area | <0.001 | ||||||||
| Metropolitan | 899 | (85.0) | 432 | (83.2) | 836 | (82.6) | 694 | (78.9) | |
| Urban | 121 | (11.4) | 61 | (11.8) | 135 | (13.3) | 141 | (16.0) | |
| Rural | 17 | (1.6) | 16 | (3.1) | 14 | (1.4) | * | * | |
| Unknown | 21 | (2.0) | 10 | (1.9) | 27 | (2.7) | 37 | (4.2) | |
| Comorbidity score | 0.65 | ||||||||
| 0 | 910 | (86.0) | 445 | (85.7) | 891 | (88.0) | 766 | (87.0) | |
| 1 | 120 | (11.3) | 62 | (11.9) | 105 | (10.4) | 93 | (10.6) | |
| >/= 2 | 28 | (2.6) | 12 | (2.3) | 16 | (1.6) | 21 | (2.4) | |
| Facility type | <0.001 | ||||||||
| Community cancer | 78 | (7.4) | 17 | (3.3) | 14 | (1.4) | * | * | |
| Comprehensive community Ca | 488 | (46.1) | 192 | (37.0) | 341 | (33.7) | 193 | (21.9) | |
| Academic/Research | 356 | (33.6) | 211 | (40.7) | 513 | (50.7) | 626 | (71.1) | |
| Integrated network cancer | 136 | (12.9) | 99 | (19.1) | 144 | (14.2) | 61 | (6.9) | |
| Facility location | <0.001 | ||||||||
| Northeast | 200 | (18.9) | 114 | (22.0) | 161 | (15.9) | 108 | (12.3) | |
| Midwest | 289 | (27.3) | 102 | (19.7) | 196 | (19.4) | 221 | (25.1) | |
| South | 356 | (33.6) | 183 | (35.3) | 440 | (43.5) | 396 | (45.0) | |
| West | 213 | (20.1) | 120 | (23.1) | 215 | (21.2) | 155 | (17.6) | |
| Stage | <0.001 | ||||||||
| 1A2 | 190 | (18.0) | 78 | (15.0) | 102 | (10.1) | 106 | (12.0) | |
| 1B1 | 868 | (82.0) | 441 | (85.0) | 910 | (89.9) | 774 | (88.0) | |
| Histology | 0.66 | ||||||||
| Squamous cell | 634 | (59.9) | 312 | (60.1) | 595 | (58.8) | 502 | (57.0) | |
| Adenocarcinoma | 376 | (35.5) | 176 | (33.9) | 360 | (35.6) | 325 | (36.9) | |
| Adenosquamous | 48 | (4.5) | 31 | (6.0) | 57 | (5.6) | 53 | (6.0) | |
| Grade | 0.002 | ||||||||
| Well | 203 | (19.2) | 103 | (19.8) | 153 | (15.1) | 129 | (14.7) | |
| Moderate | 482 | (45.6) | 228 | (43.9) | 437 | (43.2) | 436 | (49.5) | |
| Poorly | 279 | (26.4) | 142 | (27.4) | 324 | (32.0) | 221 | (25.1) | |
| Unknown | 94 | (8.9) | 46 | (8.9) | 98 | (9.7) | 94 | (10.7) | |
| Hysterectomy type | <0.001 | ||||||||
| Simple | 427 | (40.4) | 178 | (34.3) | 347 | (34.3) | 274 | (31.1) | |
| Radical | 631 | (59.6) | 341 | (65.7) | 665 | (65.7) | 606 | (68.9) | |
| Route of surgery | 0.35 | ||||||||
| Open | 429 | (40.5) | 217 | (41.8) | 432 | (42.7) | 392 | (44.5) | |
| Minimally invasive | 629 | (59.5) | 302 | (58.2) | 580 | (57.3) | 488 | (55.5) | |
| Lymphadenectomy | <0.001 | ||||||||
| No | 134 | (12.7) | 39 | (7.5) | 62 | (6.1) | 34 | (3.9) | |
| Yes | 924 | (87.3) | 480 | (92.5) | 950 | (93.9) | 846 | (96.1) | |
Fewer than 10 patients
Women in the Northeast were more likely to receive care at low- or medium-low volume centers (53.9%) and those in the South were more likely to receive care at high- or medium-high volume centers (60.8%) (P<0.001). Patients with Stage IA2 disease were more likely to receive care at LVCs (39.9%) while patients with Stage IB1 disease were more likely to receive care at medium-high and HVCs (30.4%) (P<0.001).
The majority of patients across quartiles of hospital volume underwent radical hysterectomy; performance of radical hysterectomy was 59.6% at LVCs-, 65.7% at medium-low, 65.7% at medium-high and 68.9% at HVCs (P<0.001). Volume was not found to be associated with whether the procedure was done via an open or minimally invasive approach, with 59.5% of women at LVCs undergoing MIS compared to 55.5% at HVCs (P=0.35). Lymph node assessment was more likely to occur at HVCs compared to LVCs (96.1% vs 87.3%, p< 0.001).
We then examined NCCN guideline compliance defined as both radical hysterectomy and lymph node assessment across volume quartiles. In this analysis, patients at HVCs were significantly more likely to receive NCCN guideline-compliant surgical management compared to patients seen at LVCs (Table 3). Patients treated at HVCs were 11.4% more likely to receive NCCN-compliant surgical management compared to patients treated at LVCs (67.8% vs 56.4%, p<0.001).
Table 3.
Comparing NCCN guideline compliance across quartiles of volume
| Characteristics | Low | Medium Low | Medium High | High | P-value | ||||
|---|---|---|---|---|---|---|---|---|---|
| N | % | N | % | N | % | N | % | ||
| NCCN Guideline Compliance | <0.001 | ||||||||
| No | 461 | (43.6) | 191 | (36.8) | 363 | (35.9) | 283 | (32.2) | |
| Yes | 597 | (56.4) | 328 | (63.2) | 649 | (64.1) | 597 | (67.8) | |
NCCN compliance was defined as having radical hysterectomy and lymph node dissection
In the initial unadjusted survival analysis, 5-year survival, 30-day mortality, and 90-day mortality were compared across volume quartiles (Table 4). There was no difference in 5-year survival across volume quartiles (Figure 1). Thirty-day mortality was significantly lower at HVCs, with 0 deaths in 880 patients at HVCs compared to 1 death in 1,058 patients (0.1%) at LVCs (p=0.02). There was no difference in 90-day mortality across volume quartiles (p=0.09).
Table 4.
Comparing survival outcomes across quartiles of volume
| Survival Characteristics | Low | Medium Low | Medium High | High | P-value | ||||
|---|---|---|---|---|---|---|---|---|---|
| N | % | N | % | N | % | N | % | ||
| All | 1,058 | 519 | 1,012 | 880 | |||||
| 5-year survival | 91.8 | 92.2 | 89.5 | 90.2 | 0.97 | ||||
| 95% CI | (89.4–93.8) | (89.1–94.4) | (86.3–91.9) | (86.9–92.8) | |||||
| 30-day mortality | 1 | 0.1 | 0 | 0.0 | 2 | 0.2 | 0 | 0.0 | 0.02* |
| 90-day mortality | 2 | 0.2 | 0 | 0.0 | 3 | 0.3 | 1 | 0.1 | 0.09* |
Chi-Square test performed in lieu of Fisher’s Exact test
Figure 1.
Overall survival expressed as months from diagnosis grouped by hospital volume quartiles. P=0.97 by log rank test (n=3,469)
In the adjusted survival model, hospital volume as a categorical variable was not an independent predictor of mortality (Table 5). A sensitivity analysis was performed using volume as a continuous variable, and results were unchanged (Supplementary Table S1). In the Cox Proportional Hazards model, age >=80, Medicaid and Medicare insurance, and having a poorly differentiated cervical cancer were independent predictors of increased mortality, and Hispanic race was an independent predictor of decreased mortality.
Table 5:
Multivariable Cox Proportional Hazards model of mortality controlling for quartiles of volume
| Characteristics | Adjusted Hazards Ratio |
|---|---|
| Number of patients | 3,469 |
| Number of hospitals | 598 |
| Age of diagnosis | |
| <40 | Referent |
| 40–49 | 0.83 (0.55–1.24) |
| 50–59 | 1.31 (0.88–1.95) |
| 60–69 | 1.20 (0.75–1.92) |
| 70–79 | 1.45 (0.69–3.06) |
| >/=80 | 5.90 (2.61–13.33)* |
| Race | |
| White | Referent |
| Black | 1.02 (0.67–1.56) |
| Hispanic | 0.38 (0.23–0.64)* |
| Other | 0.88 (0.46–1.71) |
| Unknown | 0.62 (0.08–4.51) |
| Year of diagnosis | |
| 2011 | Referent |
| 2012 | 1.01 (0.74–1.38) |
| 2013 | 0.91 (0.63–1.31) |
| Insurance | |
| Private | Referent |
| Medicaid | 1.47 (1.02–2.10)* |
| Medicare | 1.85 (1.21–2.81)* |
| Other government/unknown | 1.34 (0.66–2.73) |
| Uninsured | 1.07 (0.57–2.02) |
| Median household income | |
| <$38,000 | Referent |
| $38,000-$47,999 | 0.92 (0.60–1.40) |
| $48,000-$62,999 | 0.93 (0.63–1.38) |
| $63,000+ | 0.70 (0.44–1.10) |
| Unknown | 1.83 (0.19–17.12) |
| Comorbidity score | |
| 0 | Referent |
| 1 | 1.06 (0.72–1.57) |
| >/= 2 | 0.46 (0.14–1.52) |
| Facility type | |
| Academic/Research | Referent |
| Community cancer | 1.83 (0.84–3.95) |
| Comprehensive community ca | 1.36 (0.99–1.86) |
| Integrated network cancer | 1.10 (0.67–1.82) |
| Stage | |
| 1A2 | 0.54 (0.28–1.03) |
| 1B1 | Referent |
| Histology | |
| Squamous cell | Referent |
| Adenocarcinoma | 0.71 (0.49–1.03) |
| Adenosquamous | 1.39 (0.81–2.40) |
| Grade | |
| Well | Referent |
| Moderate | 1.07 (0.63–1.83) |
| Poorly | 2.36 (1.34–4.16)* |
| Unknown | 0.80 (0.38–1.70) |
| Hysterectomy type | |
| Simple | Referent |
| Radical | 0.97 (0.72–1.30) |
| Route of surgery | |
| Open | Referent |
| Minimally invasive | 1.27 (0.97–1.67) |
| Lymphadenectomy | |
| No | Referent |
| Yes | 1.18 (0.65–2.16) |
| Prior year volume, quartiles | |
| Low | Referent |
| High | 1.07 (0.70–1.64) |
| Medium High | 0.99 (0.70–1.40) |
| Medium Low | 1.03 (0.67–1.60) |
p-value <0.05
DISCUSSION
These findings suggest that patients with early-stage cervical cancer are more likely to receive guideline-recommended surgery and less likely to die within 30 days of surgery when treated at HVCs. However, treatment at HVCs does not appear to influence 90-day or all-cause mortality.
This is one of the first analyses to specifically investigate the relationship between volume and outcomes in patients surgically treated for Stage IA2 and IB1 cervical cancer. Several previously published reports have examined the association between volume and outcomes in different populations of patients with cervical cancer. An early perspective database study of 1,500 women who underwent radical hysterectomy for cervical cancer, surgeon volume was found to influence rates of postoperative medical complications, length of stay, and transfusion requirements. While surgeon volume had a strong impact on outcomes, hospital volume did not appear to have an independent effect on outcomes [9]. In a recent study of 5,964Japanese women with FIGO stage IB1 - IIB cervical cancer treated with radical hysterectomy, the authors demonstrated lower rates of local recurrence and lower rates of all-cause mortality when patients were treated at HVCs [3]. In our cohort of patients with IA2 and IB1 cervical cancer, we did not find any differences in all-cause mortality across hospital volume settings. However, there are notable differences between our study and the Japanese study which might explain this. First, in our analysis and the perspective database study, the highest-volume hospital cohort included far lower-volume hospitals than those in the highest-volume cohort in the Japanese study [3, 9]. For example, the perspective database study defined high volume as hospitals with an average annual volume of >7, and we defined high volume as hospitals with a prior year volume of between 6 and 20; whereas the Japanese study defined high volume as hospitals with an average annual volume of ≥21. This may indicate that, unlike the Japanese cohort, very few centers in the U.S. are truly high-volume enough to demonstrate mortality differences. Second, we included all patients with IA2 and IB1 cervical cancer who underwent hysterectomy, regardless of the type of hysterectomy performed. Extrafascial (simple) hysterectomy is technically less complex and carries lower risk of surgical complications; therefore, in our cohort volume would be less likely to affect mortality. Additionally, we did not include patients with FIGO Stage IB2 – IIB disease who might require an even more technically complex surgery for complete tumor removal and are often treated with radiation in lieu of surgery. By only evaluating radical hysterectomy across volume settings and by including higher-risk patients, differences in immediate surgical outcomes and longer-term oncologic outcomes may be more profound and therefore show a mortality difference.
By including all surgically treated patients in our analysis, we were able to demonstrate higher rates of guideline-compliant surgery in patients treated at HVCs. In our analysis, 62.6% of patients received guideline-compliant surgery, a figure that increased to 67.8% at centers in the highest quartile of volume. This finding raises the question of whether performing simple hysterectomy in patients with Stage IA2 and IB1 cervical cancer results in higher rates of local recurrence. We were unable to evaluate rates of local recurrence due to limitations of the NCDB dataset. A recent study used the NCDB to evaluate survival differences associated with use of simple compared to radical hysterectomy in patients with IA2 and IB1 (2 cm or less). The authors found no survival difference between simple and radical hysterectomy for patients with Stage IA2 cancers. However, in patients with stage IB1 (2 cm or less) disease, simple hysterectomy was associated with a 55% increased risk of death compared to radical hysterectomy [10].
Though there are few studies evaluating volume-outcomes in patients with early-stage cervical cancer, several studies have evaluated the association between volume and outcomes in patients with locally advanced cervical cancer [4,8,9]. A prior NCDB study of 15,194 women with locally advanced cervical cancer showed improved guideline-compliant treatment using chemotherapy, external beam radiotherapy, and brachytherapy at higher-volume centers [4]. In this patient cohort, 44.3% received guideline-compliant treatment and guideline-compliance was associated with improved survival. However, the impact of volume on survival outcomes was not evaluated [4]. An NCDB study of 20,766 women with stage IIB-IVA cervical cancer examined factors associated with survival and found that while hospital volume did not affect survival, adherence to treatment guidelines, and the specific hospital in which patients received care, impacted survival [5]. In fact, the specific hospital in which patients received care proved to be the strongest predictor of survival in that analysis. Lastly, an NCDB study of 27,660 women with stage IIB-IIIB cervical cancer who received radiation found that treatment at HVCs was independently associated with improved adherence to standard therapy and improved survival [6].
There are some notable limitations associated with our analysis. As is the case with any administrative dataset, we were unable to determine the factors that contributed to patients’ or physicians’ treatment decisions. A limitation specific to our analysis and use of the NCDB is that we were unable to determine whether rates of local recurrence differ across volume settings, which has important implications in this population. While initial treatment information is captured in the NCDB, subsequent treatment data is not chronologically catalogued, rendering analysis of local recurrence not feasible. We evaluated 5-year survival and all-cause mortality and found no difference across volume settings, which may indicate that, if differences in local recurrence are present, they do not appear to affect survival. Additionally, the cancer cases in NCDB include both suspected then confirmed cases and occult malignancies diagnosed after surgery on pathology alone. Amongst the 3,469 patients, 167 patients lacked information regarding clinical or FIGO stage and instead were only pathologically staged raising suspicion that this group may represent occult malignancies not suspected at time of surgery.
There are several additional limitations associated with analyzing hospital volume. First, within each hospital are potentially wide ranges of different provider volumes. To account for this, researchers can perform provider volume and hospital volume analyses in parallel or perform provider-level cluster analysis to mitigate the influence of any dominant provider on overall hospital volume. The NCDB does not provide data at the provider level, limiting our ability to perform these analyses. Second, there is the well-known potential for overestimation of volume-based differences when volume is evaluated as a categorical variable. To account for this, we included a sensitivity analysis for the multivariate model, which included volume as a continuous variable (Supplementary Table S1). In our analysis, volume was not a significant predictor of mortality as either a categorical or continuous variable. Moreover, the fact that hospital volume as a categorical variable was not significant in the adjusted survival analysis provides further reassurance that hospital volume is not a significant predictor of survival in these patients.
Lastly, when considering the impact of high- and low-volume centers on patients with early-stage cervical cancer, fertility-preserving surgery is an important consideration that we did not incorporate into our analysis. There may be volume-dependent differences in the safety and efficacy of radical trachelectomy and other fertility-sparing options that were, therefore, beyond the scope of the current study.
CONCLUSIONS
In summary, these findings provide important direction for future efforts aimed at elevating the quality of care delivered to patients with early-stage cervical cancer. Notably, only 54.6% of patients treated at LVCs in our cohort received guideline-concordant surgery, with an 11.4% increase at HVCs. Though statistically significant, the difference in compliance between low and high volume centers did not lead to a difference in survival based on the high and low volume cutoffs generated by the dataset. Additional studies, focused on determining if there is a high-volume threshold that leads to a survival difference, are warranted. Additionally, public reporting of compliance rates, and focused quality improvement efforts that aim to identify and disseminate processes at high-performing hospitals, are needed to meaningfully improve the disparities in guideline compliance found across hospital volume settings.
Supplementary Material
Supplementary Table S1. Multivariable Cox Proportional-Hazards model of mortality controlling for volume as continuous variable
Highlights:
Higher hospital procedural volume is associated with higher rates of indicated radical hysterectomy
Higher hospital volume is associated with higher rates of indicated lymph node assessment
Hospital volume is not associated with differences in mortality
FUNDING:
This study was funded in part through the NIH/NCI Support Grant P30 CA008748 (Dr. Leitao).
DISCLOSURES:
Dr. Wright reports personal fees from Clovis Oncology, personal fees from Tesaro, grants from Merck, outside the submitted work.
Dr. Leitao is a consultant for Intuitive Surgical, outside the submitted work.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
CONFLICTS OF INTEREST: None declared.
REFERENCES
- [1].Siegel RL, Miller KD, Jemal A. Cancer Statistics, 2019. CA Cancer J Clin 2019;69:7–34. [DOI] [PubMed] [Google Scholar]
- [2].National Comprehensive Cancer Network Clinical Practice Guideline in Oncology. Cervical Cancer. Available at: https://www.nccn.org/professionals/physician_gls/pdf/cervical.pdf. Date accessed: August 14, 2019.
- [3].Matsuo K, Shimada M, Yamaguchi S, et al. Association of Radical Hysterectomy Surgical Volume and Survival for Early-Stage Cervical Cancer. Obstet Gynecol. 2019;133:1086–1098. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Robin TP, Amini A, Schefter TE, et al. , Disparities in standard of care treatment and associated survival decrement in patients with locally advanced cervical cancer. Gynecol Oncol. 2016;143:319–325. [DOI] [PubMed] [Google Scholar]
- [5].Wright JD, Huang Y, Ananth CV. Influence of treatment center and hospital volume on survival for locally advanced cervical cancer. Gynecol Oncol. 2015;139:506–512. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].Lin JF, Berger JL, Krivak TC, et al. , Impact of facility volume on therapy and survival for locally advanced cervical cancer. Gynecol Oncol. 2014;132: 416–422. [DOI] [PubMed] [Google Scholar]
- [7].Bilimoria KY, Stewart AK, Winchester DP, et al. The National Cancer Data Base: a powerful initiative to improve cancer care in the United States. Ann Surg Oncol. 15:683–90, 2008 [DOI] [PMC free article] [PubMed] [Google Scholar]
- [8].Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992;45:613–619. [DOI] [PubMed] [Google Scholar]
- [9].Wright JD, Lewin SN, Deutsch I, et al. The influence of surgical volume on morbidity and mortality of radical hysterectomy for cervical cancer. Am J Obstet Gynecol. 2011;205:225.e1–7. [DOI] [PubMed] [Google Scholar]
- [10].Sia TY, Chen LC, Melamed A, et al. Trends in Use and Effect on Survival of Simple Hysterectomy for Early-Stage Cervical Cancer. Obstet Gynecol. 2019;134:1132–1143. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary Table S1. Multivariable Cox Proportional-Hazards model of mortality controlling for volume as continuous variable