Abstract
Background
Prior studies assessing colorectal cancer survival have reported better outcomes when operations are performed at high-volume centers. These studies have largely been cross-sectional, making it difficult to interpret their estimates. We aimed to assess the effect of facility volume on survival following proctectomy for rectal cancer.
Methods
Using data from the National Cancer Database, we included all patients with complete baseline information who underwent proctectomy for non-metastatic rectal cancer between 2004 and 2016. Facility volume was defined as the number of rectal cancer cases managed at the treating center in the calendar year prior to the patient’s surgery. Overall survival estimates were obtained for facility volumes ranging from 10 to 100 cases/year. Follow-up began on the day of surgery and continued until loss to follow-up or death.
Results
A total of 52,822 patients were eligible. Patients operated on at hospitals with volumes of 10, 30, and 50 cases/year had similar distributions of grade, clinical stage, and neoadjuvant therapies. 1-, 3-, and 5-year survival all improved with increasing facility volume. One-year survival was 94.0% (95% CI: 93.7, 94.3) for hospitals that performed 10 cases/year, 94.5% (95% CI: 94.2, 94.7) for 30 cases/year, and 94.8% (95% CI: 94.5, 95.0) for 50 cases/year. Five-year survival was 68.9% (95% CI: 68.0, 69.7) for hospitals that performed 10 cases/year, 70.8% (95% CI: 70.1, 71.5) for 30 cases/year, and 72.0% (95% CI: 71.2, 72.8) for 50 cases/year.
Conclusions
Treatment at a higher volume facility results in improved survival following proctectomy for rectal cancer, though the small benefits are less profound than previously reported.
Keywords: Rectal, Cancer, Proctectomy, Volume, Survival
Introduction
The management of rectal cancer has become progressively more complex, requiring a multidisciplinary approach. According to the National Comprehensive Cancer Network (NCCN) guidelines, management of stage I–III rectal cancer typically involves proctectomy with either abdominoperineal resection or low anterior resection in order to achieve total mesorectal excision.1 The treatment course may additionally include neoadjuvant chemoradiation and adjuvant chemotherapy, depending on tumor depth and lymph node involvement at the time of diagnostic workup.1
Colorectal surgery is technically complex. When it comes to rectal cancer surgery, refined surgical technique is required to consistently achieve a complete total mesorectal excision with negative margins and to appropriately meet the standard of care. In fact, a large Cochrane review looking at the impact of surgeon specialty training and individual surgeon volume on colorectal surgery outcomes found better overall 5-year survival at high-volume hospitals and among high-volume surgeons and colorectal specialists.2
There has been a growing body of literature to support centralization of care, which is the concept of funneling care toward experienced facilities and providers in order to improve quality of care, utilization of resources, and cost-effectiveness.3 Centralization of surgical services may improve outcomes for high-risk cancer surgery, including colorectal surgery.4 Some investigators have found no association between centralization and improved outcomes following proctectomy.5 Conversely, other studies assessing survival of patients with colorectal cancer have reported better outcomes with lower complication and mortality rates when operations are performed at high-volume centers.2, 4, 6–8 The majority of these studies have been cross-sectional in design, which requires strong assumptions (e.g., that facility volume is constant over the entire study period) in order to make the inferences necessary for decision-making.
Given the increased discussion surrounding centralization of surgical services and its possible benefit for rectal cancer surgery outcomes, there is a need for non-cross-sectional evidence. Using data from a national cancer database and a definition of hospital volume that can, at least in principle, be replicated in a trial setting, we aimed to estimate the effect of hospital volume on survival following proctectomy for rectal cancer. We hypothesize that higher hospital volume will result in improved survival.
Materials and Methods
Data Source and Patient Selection
The National Cancer Database (NCDB) is a large database containing demographic and treatment information for a wide range of malignancies collected from accredited cancer centers across the USA.9 We queried the NCDB participant user files for the years 2004–2016 to identify patients who had undergone proctectomy for rectal cancer (Fig. 1). We included all patients with complete baseline disease, treatment, and facility volume information who underwent proctectomy for non-metastatic rectal cancer. The Brigham and Women’s Hospital institutional review board deemed the study exempt.
Fig. 1.
Flowchart of patient selection from the National Cancer Database (NCDB) and volume analysis parameters
Facility Volume
For each individual undergoing proctectomy, facility volume was defined as the number of cases performed by that individual’s treating facility in the calendar year prior to their surgery. We define facility volume in this way in order to avoid making the assumption that volume remains constant over time, and to avoid the possibility of “reverse causation” (i.e., that prior outcomes may impact future facility volumes). For analysis purposes, overall survival estimates were obtained for facility volumes in 10 cases/year increments, from 10 cases/year up to 100 cases/year. In order to compare distinct overall survival point estimates between different volume levels (i.e., low, intermediate, and high), we chose distinct facility volumes that corresponded to approximately the 25th, 50th, and 75th percentiles of facility volume (Fig. 1). These volumes were 10 cases/year, 30 cases/year, and 50 cases/year, respectively. These volume levels corresponded closely with cutoffs used in prior studies assessing rectal cancer volume-outcome relationships. Specifically, a prior study utilized an annual volume standard of 15 cases for proctectomy, previously established by the Leapfrog Group.5, 10, 11 Another study divided volume into quintiles of annual surgical volume which for rectal cancer resulted in a high volume threshold of > 34 cases and a low volume threshold of < 10 cases.12
Covariates
Since facility volume was not randomly assigned in our study, we adjusted for several confounders. These included age, sex, race, Charlson Comorbidity Index, insurance status, income, education, facility type, tumor grade, clinical stage, receipt of neoadjuvant chemotherapy, and receipt of neoadjuvant radiation. For facility type, the academic group included patients with a treatment facility that was categorized as academic/research. The community group included patients with a treatment facility that was categorized as one of the following: community cancer program, comprehensive community cancer program, or integrated network cancer program.
Follow-up and Outcomes
Follow-up began on the day of surgery and continued until loss to follow-up defined as last reported contact by the NCDB hospital or death. We calculate mean number of days between diagnosis and definitive surgery to compare between the 3 distinct facility volume groups (10 cases/year, 30 cases/year, and 50 cases/year). We estimated overall survival curves as well as 1-year, 3-year, and 5-year probabilities of survival.
Statistical Analysis
Demographic, disease, and treatment variables are summarized using descriptive statistics for the overall population studied and by the chosen distinct volume levels (10 cases/year, 30 cases/year, and 50 cases/year). To estimate the (discrete-time) hazard of death, we fit pooled (over time) logistic regression models which flexibly modelled time and hospital volume using restricted cubic splines and also adjusted for all the aforementioned confounding variables. The model was used to compute the conditional probability of survival for each individual under each level of hospital volume at every month following surgery, up to 5 years. Overall survival curves were estimated by taking the cumulative probability of survival up to each month of follow-up and averaging over all individuals; thus, standardizing to the distribution of baseline covariates of the sample population. For the previously chosen volume levels (10 cases/year, 30 cases/year, and 50 cases/year), point estimates are provided with 95% confidence intervals (95% CI) obtained using a nonparametric bootstrap. Using these estimates, we also plot overall survival over time as well as overall survival by volume at 1, 3, and 5 years, comparing the 3 distinct volume levels.
As described previously, we defined facility volume in a way that avoids assuming a constant volume over time and avoids the possibility of reverse causation. In order to examine how relevant these issues might have been if had we used the typical definition of facility volume (i.e. averaged over the entire study period), we plotted the average facility volume by year for all hospitals. We also plotted facility volume by year for 10 randomly selected facilities. These plots provide some evidence that an assumption of constant volume over time is not well-supported. We also plot change in facility volume as a function of 30-day mortality in the previous calendar year to demonstrate how facility volume may respond to prior outcomes (introducing the possibility of reverse causation).
Statistical analyses were performed using STATA statistical software, edition 16.1 (StataCorp, College Station, Texas) and R version 3.6.3 (The R Foundation for Statistical Computing).
Results
A total of 52,822 patients were eligible, with 1,176 in the 10 cases/year group, 811 in the 30 cases/year group, and 441 in the 50 cases/year group (Table 1). The mean number of days between diagnosis and definitive surgery was similar between the 3 volume levels: 123.9 days (std. dev 53.6 days) for the 10 cases/year group, 135.5 days (std. dev. 54.2 days) for the 30 cases/year group, and 128.3 days (std. dev 47.9 days) for the 50 cases/year group.
Table 1.
Demographic characteristics for all patients and by specific facility volumes (10, 30, 50 cases/year)
| Variable | Facility volume (cases/year) | All N (%) (N total = 52,822) |
||
|---|---|---|---|---|
| 10 N (%) (N total = 1,176) |
30 N (%) (N total = 811) |
50 N (%) (N total = 441) |
||
| Age (years) | ||||
| ≤ 50 | 241 (20.5) | 184 (22.7) | 120 (27.2) | 12,306 (22.8) |
| 51–60 | 349 (29.7) | 259 (31.9) | 110 (24.9) | 15,440 (29.2) |
| 61–70 | 316 (26.9) | 218 (26.9) | 122 (27.7) | 14,586 (27.6) |
| 71–80 | 216 (18.4) | 121 (14.9) | 76 (17.2) | 8,599 (16.3) |
| 81–90 | 54 (4.6) | 29 (3.6) | 13 (3.0) | 2,161 (4.1) |
| Sex | ||||
| Male | 721 (61.3) | 510 (62.9) | 265 (60.1) | 32,713 (61.9) |
| Female | 455 (38.7) | 301 (37.1) | 176 (39.9) | 20,109 (38.1) |
| Race | ||||
| White Non-Hispanic | 943 (80.2) | 646 (79.7) | 329 (74.6) | 43,131 (81.7) |
| Black Non-Hispanic | 99 (8.4) | 54 (6.7) | 34 (7.7) | 4,117 (7.8) |
| Hispanic | 74 (6.3) | 60 (7.4) | 33 (7.5) | 3,007 (5.7) |
| Asian | 43 (3.7) | 26 (3.2) | 38 (8.6) | 1,761 (3.3) |
| Other/unknown | 17 (1.5) | 25 (3.1) | 7 (1.6) | 806 (1.5) |
| Charlson Comorbidity Index | ||||
| 0 | 908 (77.2) | 648 (79.9) | 368 (83.5) | 41,533 (78.6) |
| 1 | 213 (18.1) | 127 (15.7) | 56 (12.7) | 8,823 (16.7) |
| ≥ 2 | 55 (4.7) | 36 (4.4) | 17 (3.9) | 2,466 (4.7) |
| Insurance | ||||
| Private | 580 (49.3) | 403 (49.7) | 238 (54.0) | 27,099 (51.3) |
| Medicaid | 90 (7.7) | 65 (8.0) | 36 (8.2) | 4,474 (8.5) |
| Medicare | 443 (37.7) | 274 (33.8) | 151 (34.2) | 18,207 (34.5) |
| Uninsured | 55 (4.7) | 51 (6.3) | 14 (3.2) | 2,357 (4.5) |
| Unknown | 8 (0.7) | 18 (2.2) | 2 (0.5) | 685 (1.3) |
| Income | ||||
| < $38,000 | 224 (19.1) | 135 (16.7) | 91 (20.6) | 9,130 (17.3) |
| $38,000–47,999 | 306 (26.0) | 197 (24.3) | 87 (19.7) | 13,188 (25.0) |
| $48,000–62,999 | 303 (25.8) | 243 (30.0) | 119 (27.0) | 14,397 (27.3) |
| > $63,000 | 335 (28.5) | 232 (28.6) | 143 (32.4) | 15,906 (30.1) |
| Unknown | 8 (0.68) | 4 (0.5) | 1 (0.2) | 201 (0.4) |
| Education (% no high school degree) | ||||
| < 7% | 239 (20.3) | 187 (23.1) | 81 (18.4) | 12,228 (23.2) |
| 7–12.9% | 394 (33.5) | 257 (31.7) | 129 (29.3) | 17,500 (33.1) |
| 13–20.9% | 315 (26.8) | 212 (26.1) | 124 (28.1) | 14,049 (26.6) |
| > 21% | 222 (18.9) | 152 (18.7) | 106 (24.0) | 8,868 (16.8) |
| Unknown | 6 (0.5) | 3 (0.4) | 1 (0.2) | 177 (0.3) |
| Facility type | ||||
| Academic | 102 (8.7) | 226 (27.9) | 214 (48.5) | 17,307 (32.8) |
| Non-academic | 1,043 (88.7) | 552 (68.1) | 198 (44.9) | 33,006 (62.5) |
| Unknown | 31 (2.6) | 33 (4.1) | 29 (6.6) | 2,509 (4.8) |
Patient Characteristics
Patient demographic information for the overall study population and for the distinct facility volume levels is listed in Table 1. When comparing frequencies between facility volume groups, age was similarly distributed though patients in the 10 cases/year volume group tended to be slightly older compared to the other groups, while those in the 50 cases/year group tended to be slightly younger. Distribution of sex was similar in the 3 groups. Race was also similarly distributed between groups, though the 10 cases/year group demonstrated a slightly higher proportion of Non-Hispanic Black patients, while the 50 cases/year group demonstrated a higher proportion of Asian patients. Patients in the 50 cases/year group had Charlson Comorbidity scores of 0 more frequently than patients in the other two groups, while patients in the 10 cases/year group had scores of 1 more frequently than the other two groups. Patients in the 50 cases/year volume were more frequently privately insured and tended to be distributed in the higher income categories compared to the other groups. Conversely, patients in the 50 cases/year volume group demonstrated slightly lower levels of education compared to the other two groups. Finally, patients in the 50 cases/year group were more frequently treated at academic facilities, compared to those in the 10 and 30 cases/year groups. That is, for the 10 cases/year group, only 8.7% were treated at academic facilities and 88.7% were treated at non-academic facilities; in the 30 cases/year group, 27.9% were treated at academic facilities and 68.1% at non-academic facilities; and for the 50 cases/year group, 48.5% were treated at academic facilities and 44.9% at non-academic facilities.
Disease and Treatment Characteristics
Disease and treatment characteristics for the entire study population and by volume groups are listed in Table 2. Tumor grade distribution was similar between the 3 volume groups. Clinical stage was also similar between volume groups with the majority of patients either clinical stage II or III, though patients in the 30 cases/year group had the highest proportion of patients with clinical stage III. In terms of neoadjuvant therapy, though the distribution was similar between groups, the 30 cases/year group demonstrated the highest proportion of both neoadjuvant chemotherapy and radiation compared to the other groups. A total of 89.9% of patients in the 30 cases/year group received neoadjuvant chemotherapy, compared to 86.1% and 88.2% of those in the 10 cases/year and 50 cases/year groups, respectively. A total of 90.8% of patients in the 30 cases/year group received neoadjuvant radiation, compared to 87.4% and 89.6% of patients in the 10 cases/year and 50 cases/year groups, respectively.
Table 2.
Disease and treatment characteristics for all patients and by specific facility volumes (10, 30, 50 cases/year)
| Variable | Facility volume (cases/year) | All N (%) (N total = 52,822) |
||
|---|---|---|---|---|
| 10 N (%) (N total = 1,176) |
30 N (%) (N total = 811) |
50 N (%) (N total = 441) |
||
| Grade | ||||
| Grade 1 | 111 (9.4) | 66 (8.1) | 26 (5.9) | 4,550 (8.6) |
| Grade 2 | 893 (75.9) | 628 (77.4) | 351 (79.6) | 41,014 (77.7) |
| Grade 3 | 162 (13.8) | 107 (13.2) | 58 (13.2) | 6,580 (12.5) |
| Grade 4 | 10 (0.9) | 10 (1.2) | 6 (1.4) | 678 (1.3) |
| Stage | ||||
| Stage 0 | 3 (0.3) | 4 (0.5) | 1 (0.2) | 201 (0.4) |
| Stage I | 136 (11.6) | 75 (9.3) | 50 (11.3) | 5,512 (10.4) |
| Stage II | 509 (43.3) | 318 (39.2) | 180 (40.8) | 20,983 (39.7) |
| Stage III | 528 (44.9) | 414 (51.1) | 210 (47.6) | 26,126 (49.5) |
| Neoadjuvant chemotherapy | ||||
| No | 164 (14.0) | 82 (10.1) | 52 (11.8) | 5,675 (10.7) |
| Yes | 1,012 (86.1) | 729 (89.9) | 389 (88.2) | 47,147 (89.3) |
| Neoadjuvant radiation | ||||
| No | 148 (12.6) | 75 (9.3) | 46 (10.4) | 4,875 (9.2) |
| Yes | 1,028 (87.4) | 736 (90.8) | 395 (89.6) | 47,947 (90.8) |
Survival Outcomes
Adjusted 1-, 3-, and 5-year survival estimates for the three distinct facility volume groups are summarized in Table 3. There was an increase in 1-, 3-, and 5-survival for treatment at hospitals with increasing levels of volume from 10 cases/year to 30 cases/year to 50 cases/year. Survival at all time points was highest for treatment at hospitals that performed 50 cases/year, followed by those that performed 30 cases/year, and then those that performed 10 cases/year. One-year survival was 94.0% (95% CI: 93.7, 94.3) for treatment at hospitals that performed 10 cases/year, 94.5% (95% CI: 94.2, 94.7) for 30 cases/year, and 94.8% (95% CI: 94.5, 95.0) for 50 cases/year. Five-year survival was 68.9% (95% CI: 68.0, 69.7) for treatment at hospitals that performed 10 cases/year, 70.8% (95% CI: 70.1, 71.5) for 30 cases/year, and 72.0% (95% CI: 71.2, 72.8) for 50 cases/year. The confidence intervals for all survival point estimates were narrow. The relationship of survival over time comparing the 3 distinct volume groups is represented in Fig. 2. As follow-up time increases, the difference in survival between the 3 groups increases, with the 50 cases/year group consistently demonstrating the highest estimated survival and the 10 cases/year group demonstrating the lowest estimated survival among the 3 groups, all the way through the 5-year follow-up time point.
Table 3.
Survival outcomes for non-metastatic rectal cancer patients following proctectomy, by specific facility volumes (10, 30, 50 cases/year)
| Facility volume (cases/year) | 1-year survival % (95% CI) | 3-year survival % (95% CI) | 5-year survival % (95% CI) |
|---|---|---|---|
| 10 | 94.0 (93.7–94.3) | 80.7 (80.0–81.2) | 68.9 (68.0–69.7) |
| 30 | 94.5 (94.2–94.7) | 81.9 (81.4–82.4) | 70.8 (70.1–71.5) |
| 50 | 94.8 (94.5–95.0) | 82.8 (82.2–83.3) | 72.0 (71.2–72.8) |
Fig. 2.
Survival over time following proctectomy for rectal cancer, for specific facility volumes (10, 30, 50 cases/year)
When looking at volume from 10 cases/year up to 100 cases/year, in increments of 10 cases/year, 1-, 3-, and 5-year survival all increased progressively with increasing facility volume. The adjusted survival-volume relationships for 1-, 3-, and 5-years are presented in Fig. 3a–c. For each follow-up time demonstrated, there appears to a positive curvilinear relationship between increasing facility volume and improved survival estimates. The volume-survival relationship demonstrates significant improvements initially comparing 20 cases/year to 40 cases/year. From 40 cases/year, the survival improvement then becomes a more gradual increase with increasing volume.
Fig. 3.
a–c Survival by volume following proctectomy for rectal cancer. Panel a: 1-year survival vs. volume. Panel b: 3-year survival vs. volume. Panel c: 5-year survival vs. volume
Facility Volume Trends
Figure 4 demonstrates average facility volume by year for all hospitals, over the entire study period. On average, facility volume is not constant over time, and overall increases over the study period. Facility volume by year over the study period for 10 randomly selected facilities is demonstrated in Fig. 5. As illustrated in the figure, volume is not static over time, and is rather highly variable within and between facilities. Figure 6 demonstrates change in facility volume as a function of past mortality, based on 30-day mortality of the previous calendar year. When mortality increases, facility volume decreases slightly, suggesting that facility volume responds to outcomes.
Fig. 4.
Average facility volume (cases/year) by year for all hospitals over study period
Fig. 5.
Facility volume (cases/year) by year for 10 randomly selected and distinct facilities over study period
Fig. 6.
Change in facility volume as a function of 30-day mortality of previous calendar year
Discussion
We utilized a large national database to analyze the effect of facility volume on survival following proctectomy for rectal cancer. In our study, we estimated that treatment at a higher volume facility would result in improved survival for individuals undergoing proctectomy for non-metastatic rectal cancer. Differences in survival were minor and less profound than previously reported in cross-sectional studies. Adjusted survival at 1, 3, and 5 year progressively increased when proceeding from facilities performing 10 cases/year up to 100 cases/year. The relationship between facility volume and survival appears to be a positive curvilinear relationship, with an initial sharp improvement in survival and a subsequent gradual increase with higher facility volume. All confidence intervals were narrow, suggesting that within the NCDB population, in the absence of significant structural biases, the data is compatible with the conclusion that surgical treatment of rectal cancer at higher volume hospitals results in slight improvement in survival.
For volumes of 10 cases/year, 30 cases/year, and 50 cases/year, we found 5-year survival rates of 68.9%, 70.8%, and 72.0%, respectively. A prior NCDB study from 2008 looking at survival outcomes following surgery for various cancers by comparing quintiles of hospital volume found that patients who underwent surgery for non-metastatic rectal cancer demonstrated 5-year unadjusted overall survival rates of 53.4%, 56.5%, and 60.7% for low, moderate, and high volume hospitals, respectively.12 A study looking at the centralization outcomes for various complex cancer surgeries, with centralization defined as “the proportion of operations performed by the highest-volume center within each system in a given year,” calculated a mean degree of centralization of only 33.4% for proctectomy, compared to the highest mean degree of centralization of 71.2% for pancreatectomy.5 This study found that centralization was not associated with better outcomes for proctectomy, with no significant difference in post-operative 30-day mortality, complications, or readmissions.5 This study looked at 30-day mortality without assessing longterm survival outcomes, and perhaps the low degree of centralization may have impacted the association identified. Interestingly, 52.9% of systems included in that study had no hospitals that met the volume threshold of 15 for proctectomy.5
There are likely many factors influencing the effect seen between higher volume and improved survival following proctectomy that we identified in our study. To start, the management of rectal cancer has evolved significantly over the period of our study, with evidence supporting preoperative chemoradiation therapy, total neoadjuvant therapy, and more recently a watch-and-wait nonoperative approaches.13–15 Thus, it is possible that high-volume centers adopted evidence-based practices sooner than low-volume centers as a result of greater familiarity and expertise. Further, higher volume hospitals may have more experienced established multidisciplinary care teams, pathways, and decision-making. It is possible, though, that improved dissemination and adherence to national guidelines are contributing to the decreased impact of hospital volumeon outcomes that we observed.
Next, though studies have reported a hospital volume-outcome relationship following colorectal surgery, they have also investigated other factors that may be driving or influencing these outcomes, including individual surgeon volume and specialty training. A meta-analysis of the association between hospital or surgeon volume on colorectal cancer surgery outcomes found that for rectal cancer surgery, both hospital and surgeon volume were associated with reduced rates of in-hospital mortality, local recurrence, and anastomotic leak; hospital volume was associated with reduced 30-day in-hospital mortality, while surgeon volume was associated with improved 5-year survival.4 A Cochrane review found that for rectal cancer surgeries, high volume hospitals were associated with improved 5-year survival, while surgeon volume was not.2 Though not specific to rectal cancer, the study also demonstrated that surgeon specialization was associated with improved 5-year overall and disease-specific survival for colorectal cancer, as well as lower odds of inpatient and 30-day mortality.2 Another study looking at laparoscopic rectal surgery outcomes found that both high-volume laparoscopic rectal surgeons and high-volume open rectal surgeons reduced all-cause morbidity.16 Overall, the existing evidence suggests that hospital volume, surgeon volume, and surgeon specialization can all positively contribute to improved outcomes for rectal cancer surgery, though it is unclear which component may be most important.3 Therefore, when considering centralization for rectal cancer, it is important to think about ways in which to optimize these three components: hospital volume, surgeon volume, and surgeon specialization. It is also important to concomitantly optimize multidisciplinary care and compliance with treatment guidelines. For instance, there are National Cancer Institute (NCI) Designated Cancer Centers across the USA that meet rigorous research standards and deliver up-to-date and novel oncologic treatments.17 These NCI-designated cancer centers could set the facility standard when thinking about or formally implementing centralization for rectal cancer surgical care. When considering how to appropriately implement centralization, it is crucial that policy makers are cautious that this does not worsen disparities in access to care.
Ideally, decisions about centralization of rectal cancer surgery would be informed by evidence from randomized controlled trials which assign patients to treatment arms defined by different facility volumes. In the absence of such trials, observational studies designed to mimic randomized trials as closely as possible are the next best source of evidence.18–20 In the current study, we used observational data from the NCDB to mimic a hypothetical trial comparing outcomes following proctectomy for rectal cancer between facilities of different volumes, with time zero defined as the day of surgery, and facility volume defined in the year prior to patient assignment to a particular facility.18, 21 Unfortunately, most available studies assessing volume-outcome relationships have not used a definition of facility volume that would be possible to replicate in a prospective trial setting. For example, for a particular individual in a study, the volume of their treating facility has typically been defined as that facility’s total or yearly average volume over the entire study period, including the time after which the individual has undergone surgery. Such a definition could not correspond to any treatment arm in a randomized controlled trial. The consequence is an inability to rule out “reverse causation” that survival outcomes at a given hospital dictate its future volume levels. In general, profound effects have been reported for the association between volume and patient outcomes following various surgical procedure types.22, 23 However, reverse causation has likely contributed to the pronounced associations reported between hospital volume and patient survival. In fact, we demonstrated that volume over time for all facilities as well as for randomly selected individual facilities is not static, varying, sometimes considerably, from 1 year to the next. Importantly, we demonstrate that facility volume decreased in response to increased mortality in the previous calendar year, providing evidence that reverse causation may be an issue for studies analyzing volume-outcome relationships with volume defined as the average facility volume over the entire study period.
In contrast, in the current study’s analysis, for every individual, we defined facility volume as the number of cases performed by that individual’s treating facility in the calendar year prior to their surgery. Our definition directly corresponds to a trial in which patients with rectal cancer are randomly assigned to treatment at a facility with a particular volume and such facilities are identified by their volume in the prior calendar year.18, 21 This study design approach is unique and unlike existing studies on volume-outcome relationships for rectal cancer which, due to their cross-sectional design, cannot estimate the effect of hospital volume on survival without strong assumptions (e.g., that hospital volume is constant over the entire study period). It is not surprising that, in our study, the estimated benefits of treatment at higher volume hospitals are less striking than those previously reported.
This study has a few important limitations. It is an observational study and, as with all observational studies, there may be bias due to unmeasured or residual confounding. For example, while we adjusted for receipt of neoadjuvant therapy, the specific type and duration of therapy administered was not known, which may result in residual confounding. Further, in our study, we began follow-up on the day of surgery and assumed that there were no substantial differences in the time from diagnosis to surgery for patients treated at facilities with different volumes. This was not an unreasonable assumption given the similar mean number of days from diagnosis to definitive surgery between the 3 distinct volume levels. Additionally, though we were able to determine the volume of proctectomies performed at each unique facility, we did not consider the interplay between hospital and surgeon volume which adds important nuance to the hospital volume-outcome relationship. Finally, the NCDB does not provide data on cancer-specific survival, so such cause-specific outcomes were not estimated.
Conclusion
Utilizing a national database, the current study identifies improved survival following proctectomy for rectal cancer with increasing facility volume. Though facility volume may be an important driver of the observed effect, there are likely other influencing factors to consider, including individual surgeon volume, surgeon specialty training, and compliance with treatment guidelines. The volume-outcome relationship following proctectomy is supported by our study, but benefits are less profound than previously reported. We demonstrated that facility volume responds to prior outcomes. This challenges prior methodological designs which have treated volume as constant over time. Next steps include optimizing centralization of rectal cancer surgery, possibly with improved referral patterns and distribution of trained specialists.
Footnotes
An earlier version of this work was presented as a display e-poster at the Massachusetts Chapter of the American College of Surgeons (MCACS) 67th Annual Meeting (held virtually December 5, 2020), and as a Virtual Quickshot Presentation at the 16th Annual Academic Surgical Congress (ASC) (held virtually February 2–4, 2021).
Conflicts of Interest The authors declare no competing interests.
References
- 1.NCCN. NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines): Rectal Cancer June 2020.
- 2.Archampong D, Borowski D, Wille-Jørgensen P, Iversen LH. Workload and surgeońs specialty for outcome after colorectal cancer surgery. Cochrane Database of Systematic Reviews. 2012(3). doi: 10.1002/14651858.CD005391.pub3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Vonlanthen R, Lodge P, Barkun JS, Farges O, Rogiers X, Soreide K et al. Toward a Consensus on Centralization in Surgery. Ann Surg. 2018;268(5):712–24. doi: 10.1097/sla.0000000000002965. [DOI] [PubMed] [Google Scholar]
- 4.Huo YR, Phan K, Morris DL, Liauw W. Systematic review and a meta-analysis of hospital and surgeon volume/outcome relationships in colorectal cancer surgery. Journal of Gastrointestinal Oncology. 2017;8(3):534–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Sheetz KH, Dimick JB, Nathan H. Centralization of High-Risk Cancer Surgery Within Existing Hospital Systems. Journal of Clinical Oncology. 2019;37(34):3234–42. doi: 10.1200/jco.18.02035. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Hamidi M, Hanna K, Omesiete P, Cruz A, Ewongwo A, Pandit V et al. Does it matter where you get your surgery for colorectal cancer? International Journal of Colorectal Disease. 2019;34(12): 2121–7. doi: 10.1007/s00384-019-03436-6. [DOI] [PubMed] [Google Scholar]
- 7.Borowski DW, Bradburn DM, Mills SJ, Bharathan B, Wilson RG, Ratcliffe AA et al. Volume-outcome analysis of colorectal cancer-related outcomes. Br J Surg. 2010;97(9):1416–30. doi: 10.1002/bjs.7111. [DOI] [PubMed] [Google Scholar]
- 8.van Gijn W, Gooiker GA, Wouters MW, Post PN, Tollenaar RA, van de Velde CJ. Volume and outcome in colorectal cancer surgery. Eur J Surg Oncol. 2010;36 Suppl 1:S55–63. doi: 10.1016/j.ejso.2010.06.027. [DOI] [PubMed] [Google Scholar]
- 9.ACS. National Cancer Database. American College of Surgeons. https://www.facs.org/quality-programs/cancer/ncdb. Accessed November 17 2020. [Google Scholar]
- 10.Leapfrog Ratings: Complex Adult Surgery. The Leapfrog Group. 2020. https://ratings.leapfroggroup.org/measure/hospital/complex-adult-surgery. Accessed February 21 2021. [Google Scholar]
- 11.Urbach DR. Pledging to Eliminate Low-Volume Surgery. N Engl J Med. 2015;373(15):1388–90. doi: 10.1056/NEJMp1508472. [DOI] [PubMed] [Google Scholar]
- 12.Bilimoria KY, Bentrem DJ, Feinglass JM, Stewart AK, Winchester DP, Talamonti MS et al. Directing Surgical Quality Improvement Initiatives: Comparison of Perioperative Mortality and Long-Term Survival for Cancer Surgery. Journal of Clinical Oncology. 2008;26(28):4626–33. doi: 10.1200/jco.2007.15.6356. [DOI] [PubMed] [Google Scholar]
- 13.Sauer R, Becker H, Hohenberger W, Rödel C, Wittekind C, Fietkau R et al. Preoperative versus Postoperative Chemoradiotherapy for Rectal Cancer. New England Journal of Medicine. 2004;351(17): 1731–40. doi: 10.1056/NEJMoa040694. [DOI] [PubMed] [Google Scholar]
- 14.Habr-Gama A, Gama-Rodrigues J, São Julião GP, Proscurshim I, Sabbagh C, Lynn PB et al. Local recurrence after complete clinical response and watch and wait in rectal cancer after neoadjuvant chemoradiation: impact of salvage therapy on local disease control. Int J Radiat Oncol Biol Phys. 2014;88(4):822–8. doi: 10.1016/j.ijrobp.2013.12.012. [DOI] [PubMed] [Google Scholar]
- 15.Appelt AL, Pløen J, Harling H, Jensen FS, Jensen LH, Jørgensen JC et al. High-dose chemoradiotherapy and watchful waiting for distal rectal cancer: a prospective observational study. Lancet Oncol. 2015;16(8):919–27. doi: 10.1016/s1470-2045(15)00120-5. [DOI] [PubMed] [Google Scholar]
- 16.Lee JK, Doumouras AG, Springer JE, Eskicioglu C, Amin N, Cadeddu M et al. Examining the transferability of colon and rectal operative experience on outcomes following laparoscopic rectal surgery. Surgical Endoscopy. 2020;34(3):1231–6. doi: 10.1007/s00464-019-06885-w. [DOI] [PubMed] [Google Scholar]
- 17.NCI. NCI-Designated Cancer Centers. In: Cancer Research Infrastructure. National Cancer Institute. 2019. https://www.cancer.gov/research/infrastructure/cancer-centers. Accessed January 31 2021. [Google Scholar]
- 18.Hernán MA, Robins JM. Using Big Data to Emulate a Target Trial When a Randomized Trial Is Not Available. Am J Epidemiol. 2016;183(8):758–64. doi: 10.1093/aje/kwv254. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.García-Albéniz X, Hsu J, Hernán MA. The value of explicitly emulating a target trial when using real world evidence: an application to colorectal cancer screening. Eur J Epidemiol. 2017;32(6):495–500. doi: 10.1007/s10654-017-0287-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Lodi S, Phillips A, Lundgren J, Logan R, Sharma S, Cole SR et al. Effect Estimates in Randomized Trials and Observational Studies: Comparing Apples With Apples. Am J Epidemiol. 2019;188(8): 1569–77. doi: 10.1093/aje/kwz100. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Hernán MA, Hernández-Díaz S, Robins JM. A structural approach to selection bias. Epidemiology. 2004;15(5):615–25. doi: 10.1097/01.ede.0000135174.63482.43. [DOI] [PubMed] [Google Scholar]
- 22.Reames BN, Ghaferi AA, Birkmeyer JD, Dimick JB. Hospital volume and operative mortality in the modern era. Ann Surg. 2014;260(2):244–51. doi: 10.1097/sla.0000000000000375. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Birkmeyer JD, Siewers AE, Finlayson EVA, Stukel TA, Lucas FL, Batista I et al. Hospital Volume and Surgical Mortality in the United States. New England Journal of Medicine. 2002;346(15):1128–37. doi: 10.1056/NEJMsa012337. [DOI] [PubMed] [Google Scholar]