- Email: [email protected]
Facility volume and postoperative outcomes for malignant pleural mesothelioma: A National Cancer Data Base analysis
Lung Cancer 120 (2018) 7–13
Contents lists available at ScienceDirect
Lung Cancer journal homepage: www.elsevier.com/locate/lungcan
Facility volume and postoperative outcomes for malignant pleural mesothelioma: A National Cancer Data Base analysis
T
Vivek Vermaa, Christopher A. Ahernb, Christopher G. Berlindb, William D. Lindsayb, ⁎ Surbhi Groverc, Melissa J. Culligand, Joseph S. Friedbergd, Charles B. Simone IIe, a
Department of Radiation Oncology, Allegheny General Hospital, Pittsburgh, PA, USA Oncora Medical, Philadelphia, PA, USA c Department of Radiation Oncology, University of Pennsylvania, Philadelphia, PA, USA d Department of Surgery, Division of Thoracic Surgery, University of Maryland Medical Center, Baltimore, MD, USA e Department of Radiation Oncology, University of Maryland Medical Center, Baltimore, MD, USA b
A R T I C LE I N FO
A B S T R A C T
Keywords: Mesothelioma Extrapleural pneumonectomy Pleurectomy Surgery Facility volume
Purpose: This study of a large, contemporary national database evaluated postoperative outcomes and overall survival (OS) for malignant pleural mesothelioma (MPM) by facility volume. Methods: The National Cancer Database was queried for newly-diagnosed non-metastatic MPM undergoing definitive surgery (extrapleural pneumonectomy (EPP) or pleurectomy/decortication (P/D)). Patients were dichotomized into those receiving therapy at a high-volume facility (HVF), defined a priori at the 90th percentile of case volume, with all others categorized as lower-volume facilities (LVFs). Statistics included multivariable logistic regression, Kaplan-Meier analysis, propensity-matching, and multivariable Cox proportional hazards modeling. Sensitivity analysis varied the dichotomized HVF-LVF cutoff and evaluated effects on postoperative outcomes and OS. Results: Of 1307 patients, 621 (48%) were treated at LVFs and 686 (52%) at HVFs. HVFs were more often in the Middle/South Atlantic regions, and less likely in New England, South, and Midwest. Notably, 75% of procedures at HVFs were P/Ds, versus 84% at LVFs (p < 0.001). Patients treated at HVFs experienced shorter length of postoperative hospitalization (p = 0.035), lower 30-day readmission rates (4.6% vs. 6.1%, p = 0.021), and lower 90-day mortality rates (10.0% vs. 14.6%, p = 0.029). Median OS for respective groups were 18 versus 15 months (p = 0.010), which were not significant following propensity-matching (p = 0.540). On multivariable analysis, facility volume did not independently predict for OS. Sensitivity analyses confirmed the postoperative outcomes and OS findings. Conclusions: This is the largest investigation to date assessing facility volume and outcomes following surgery for MPM. Although no independent effects on OS were observed, postoperative outcomes were more favorable at HVFs. These findings have implications for postoperative management, patient counseling, referring providers, and cost-effectiveness.
1. Introduction Although malignant pleural mesothelioma (MPM) is relatively uncommon, it is a highly aggressive neoplasm associated with a very poor prognosis. Gross macroscopic resection is perhaps the most important aspect of management along with chemotherapy, provided technical and medical candidacy for surgery [1]. Definitive surgery is most commonly performed using two approaches: extrapleural pneumonectomy (EPP) or extended pleurectomy/decortication (P/D). Both of these procedures are technically challenging and can cause
serious postoperative complications, irreversible morbidities, and mortality. Postoperative complications occur in 13–38% of patients, with postoperative 30-day mortality occurring in 3–8% [2–5]; owing to publication bias, these rates could very well underestimate the true incidences. In the randomized MARS trial, postoperative complications in the surgery (EPP) arm occurred in 69% of patients, with a mortality rates of 13% (intention to treat) to 16% (any patient in whom surgery was attempted) [6]. Owing to these and other reasons, it may be hypothesized that receiving surgery at a high-volume facility (HVF) may be advantageous
⁎ Corresponding author at: University of Maryland School of Medicine, Department of Radiation Oncology, Maryland Proton Treatment Center, 850 W. Baltimore St., Baltimore, MD 21201, USA. E-mail address: [email protected] (C.B. Simone).
https://doi.org/10.1016/j.lungcan.2018.03.019 Received 2 February 2018; Received in revised form 2 March 2018; Accepted 20 March 2018 0169-5002/ © 2018 Elsevier B.V. All rights reserved.
Lung Cancer 120 (2018) 7–13
V. Verma et al.
over lower-volume facilities (LVFs) in terms of postoperative complications and possibly even survival. Improvements in postoperative outcomes and/or survival at HVFs have been shown for numerous surgical procedures, including sarcoma excision [7], pancreaticoduodenectomy [8], breast reconstruction [9], low anterior/abdominoperineal resection [10], lobectomy/pneumonectomy [11], esophagectomy [12], and colectomy [13]. To date, such an analysis has not been performed for MPM. Such a study could have important consequences considering the technical expertise required for a major surgical procedure and the particularly high perioperative mortality rate associated with MPM surgery. Investigating the large, contemporary National Cancer Data Base (NCDB), this study sought to address the influence of facility volume on differences in the four postoperative outcomes given by the NCDB (length of hospital stay, 30-day readmission, 30-day mortality, and 90day mortality), and secondarily, overall survival (OS).
Table 1 Demographic characteristics of the overall cohort and factors associated with receiving treatment at a high volume facility in the final multivariable logistic regression model. Parameter
2. Materials and methods The NCDB is a joint project of the Commission on Cancer (CoC) of the American College of Surgeons and the American Cancer Society that consists of information regarding tumor characteristics, patient demographics, and patient survival for approximately 70% of the United States population [14–19]. The NCDB contains information not included in the Surveillance, Epidemiology, and End Results (SEER) database, including details pertaining to systemic therapy usage. The data used in this study were derived from a de-identified NCDB file. The American College of Surgeons and the CoC have not verified and are neither responsible for the analytic or statistical methodology employed nor the conclusions of the authors drawn from these data. As all patient information in the NCDB database is de-identified, this study was exempt from institutional review board evaluation. The NCDB Participant User File corresponding to mesothelioma (2004–2012) was utilized for this study. Inclusion criteria for this investigation were patients with newly-diagnosed MPM who received definitive surgery (EPP or P/D). Definitive surgery was defined by a board-certified thoracic surgeon with notable mesothelioma surgical experience (J.S.F.) as surgery of the primary intrathoracic site with codes 20–23, 30, 33, 40, 45–48, and 50 for P/D and codes 55–56, 60, 66, and 70 for EPP [20]. Patients with all other types of surgery (including those with ambiguous surgical codes/labels) were eliminated. Patients without proper TNM staging were also removed, as were patients designated in the NCDB as metastatic (stage M1) or receiving palliative care. In accordance with the variables in NCDB files, information collected on each patient broadly included demographic, clinical, and treatment data. The definition of a HVF was similar to other established work [21,22]. Briefly, facility volume was dichotomized into HVF or LVF based on a threshold corresponding to the 90th percentile of patient numbers treated per facility over the time period. Similar to previously published work, this cutoff was utilized in order to evaluate a roughly 1:1 ratio of patients [21,22]. However, in order to evaluate whether this a priori definition was significant at other cutoffs, sensitivity analysis was performed post hoc to evaluate whether altering the HVF definition affected the association with postoperative outcomes and OS. This was performed by repeating the multivariable analyses for each given threshold of HVF definition. Statistical analysis was performed with R [23]. Tests were twosided, with a threshold of p < 0.05 for statistical significance. First, clinical characteristics between HVF and LVF groups were tabulated. Multivariable logistic regression analysis was performed to ascertain factors independently associated with treatment at a HVF. The chisquared or Fisher’s exact tests evaluated differences in 30-day readmission, 30-day mortality, and 90-day mortality. The Mann-Whitney U test assessed differences between groups in length of postoperative hospitalization. Kaplan-Meier curves were calculated to evaluate OS,
LVF (N = 621)
HVF (N = 686)
Age (years) Median (IQR)
69 (61–75)
67 (61–74)
Gender Male Female
489 (79%) 132 (21%)
541 (79%) 145 (21%)
Race White Black Other Unknown
585 (94%) 24 (4%) 3 (0%) 9 (1%)
644 (94%) 19 (3%) 16 (2%) 7 (1%)
Charlson Deyo score 0 1
431 (69%) 148 (24%)
534 (78%) 129 (19%)
42 (7%)
23 (3%)
Insurance type Private Medicare Medicaid Other government Uninsured Unknown
225 (36%) 355 (57%) 14 (2%) 6 (1%) 11 (2%) 10 (2%)
272 (40%) 380 (55%) 10 (1%) 8 (1%) 4 (1%) 12 (1%)
Income (US dollars/ year) < $30,000 $30,000–$34,999 $35,000–$45,999 ≥$46,000 Unknown
87 (14%) 134 (22%) 164 (26%) 219 (35%) 17 (3%)
65 (9%) 105 (15%) 151 (22%) 343 (50%) 22 (3%)
≥2
Percentage of adults in zip code without high school diploma ≥21% 83 (13%) 54 (8%) 13–20.9% 142 (23%) 143 (21%) 7–12.9%
227 (37%)
227 (33%)
< 7%
152 (24%)
240 (35%)
Unknown
17 (3%)
22 (3%)
Patient residence Urban Metro Rural Unknown
94 (15%) 490 (79%) 11 (2%) 26 (4%)
62 (9%) 584 (85%) 8 (1%) 32 (5%)
Facility location East North Central East South Central
117 (19%) 40 (6%)
128 (19%) 11 (2%)
Middle Atlantic
84 (14%)
232 (34%)
Mountain
24 (4%)
17 (2%)
New England
62 (10%)
10 (1%)
Pacific
78 (13%)
86 (13%)
South Atlantic
118 (19%)
162 (24%)
West North Central
56 (9%)
16 (2%)
West South Central
35 (6%)
15 (2%)
Unknown
7 (1%)
9 (1%)
166 (27%) 447 (72%)
579 (84%) 98 (14%)
Facility type Academic Community
Final multivariable model OR (95% CI)
p-value
REF 0.777 (0.568–1.061) 0.487 (0.260–0.896)
REF 0.112
REF 1.745 (1.083–2.825) 1.496 (0.947–2.374) 2.708 (1.685–4.376) –
REF 0.023
REF 0.361 (0.162–0.757) 2.536 (1.729–3.737) 0.510 (0.239–1.068) 0.131 (0.059–0.266) 1.040 (0.672–1.612) 1.543 (1.055–2.263) 0.245 (0.125–0.460) 0.579 (0.278–1.168) –
REF 0.009
0.022
0.085 < 0.001 –
< 0.001 0.077 < 0.001 0.860 0.026 < 0.001 0.134 –
(continued on next page)
8
Lung Cancer 120 (2018) 7–13
V. Verma et al.
(range, 0–162) versus 6 days (range, 0–61)), it was statistically lower for the former (p = 0.035). Thirty-day readmission rates were 4.6% in the HVF cohort, as compared to 6.1% for LVFs (p = 0.021). Although there were no differences in 30-day mortality (4.7% vs. 4.6%, p = 0.656), a lower 90-day mortality rate was found in patients treated at HVFs (10.0% vs. 14.6%, p = 0.029). Because propensity matching was specifically designed for the secondary OS endpoint, it was not statistically indicated to compare postoperative outcomes between the matched populations, similar to other work [20]. Nevertheless, Supplementary Table 1 displays results of sensitivity analysis when varying the threshold to define a HVF. Findings related to 30-day readmission and 90-day mortality persisted with all tested cutoffs, whereas length of hospital stay was statistically significant above and including a base threshold of 87th percentile of case volume. The median follow-up was 15 months (range, 0–125 months). Kaplan-Meier estimates comparing OS between HVFs and LVFs are illustrated in Fig. 2. In the overall population, patients treated at HVFs experienced higher median OS (18 months (95% CI 17–19 months) versus 15 months (95% CI 13–17 months), p = 0.010). However, after propensity matching, there were no statistical OS differences between groups (17 months versus 15 months, p = 0.540). When adjusting for potential confounders, there were several predictors of OS on multivariable analysis in both the matched and unmatched populations (Table 3). Factors independently predictive of poorer OS were advanced age, male gender, urban residence, presence of pleural effusion, increasing T classification, and sarcomatoid histology (p < 0.05 for all). Improved OS was associated with receipt of chemotherapy and increasing radiation dose (p < 0.05 for both). Facility volume as a dichotomized variable was eliminated in stepwise selection (owing to lack of statistical significance) for the Cox multivariable model, which was true for several different cutoffs tested on sensitivity analysis in both the overall and matched populations (Supplementary Table 2).
Table 1 (continued) Parameter
LVF (N = 621)
HVF (N = 686)
Final multivariable model OR (95% CI)
Unknown
8 (1%)
9 (1%)
Distance to treating facility (mi) Median (IQR)
10 (4–21)
26 (10–90)
Year of diagnosis 2004–2008 2009–2012
276 (44%) 345 (56%)
289 (42%) 397 (58%)
p-value
Statistically significant p-values are in bold. Only statistically significant parameters are shown on multivariable analysis. OR, odds ratio; CI, confidence interval; IQR, interquartile range.
defined as the interval between the date of diagnosis and the date of death or censored at last contact. The univariate association of each covariate with OS was assessed using a Cox proportional hazards model. Covariates that were significantly associated with OS (p < 0.05) were included in a multivariable model, and backwards stepwise selection was performed with α = 0.20. Lastly, in an attempt to minimize selection and indication biases, patients in both cohorts underwent propensity matching. To estimate the propensity score for each patient, the univariate association of each covariate with treatment type was assessed using a simple logistic regression model. Covariates that were significantly associated with treatment type (p < 0.05) were included in a multivariable logistic regression model, and backwards stepwise selection was performed with α = 0.20. One-to-one matching was performed using a “greedy” nearest neighbor algorithm with a caliper 0.2 times the standard deviation of the logit propensity score [24,25]. The standardized difference between groups after matching was less than 0.1, indicating sufficient balancing [26]. Propensity matching took into account all variables (listed in Table 1) except those not meeting the α = 0.20 threshold on backwards stepwise selection. Kaplan-Meier curves were calculated on the propensity matched patients, and a multivariable Cox model was fit on the propensity matched dataset by the same method described above.
4. Discussion This novel investigation of a large, contemporary national database regarding the influence of facility volume on postoperative outcomes and survival for MPM most notably illustrates that patients treated at HVFs may experience more favorable postoperative outcomes than those treated at LVFs. Although true for length of postoperative hospitalization, this is most notably true for 30-day readmission and 90day mortality at a wide variety of HVF definition cutoffs. These findings have implications for postoperative management, patient counseling, referring providers, and cost-effectiveness. It is very important to initially note that, similar to other work [22], although the vast majority of HVFs were academic institutions, the two terms are not synonymous. This is also in concordance with a prior publication demonstrating independently higher OS at academic centers [20], in contrast to lack thereof with HVFs in this study. Although this could imply that outcomes at high-volume community centers could have “diluted” the findings between that study and this one, the low sample sizes of such cases prevents formal exploratory analysis. Next, it is also important to consider that although P/D techniques were more often utilized at LVFs, this does not imply that P/D itself was associated with worse postoperative outcomes. In fact, previous analysis showed that although EPPs were more commonly performed at academic centers, there were no appreciable differences in postoperative outcomes or OS following EPP versus P/D [20]. Similar conclusions have also been reached elsewhere [4,5,27]. There are several reasons to explain the lack of OS differences in this study, in addition to the differentiation between HVFs and academic centers detailed above. Overall, MPM is associated with very poor prognosis, and delineating subtle differences in OS are thus quite difficult. Second, sensitivity analysis showed that defining HVFs are the 77th or 82nd percentiles would have resulted in inclusion into
3. Results A patient selection diagram is provided in Fig. 1. In total, 1307 patients treated at 476 different facilities met study criteria. Of those, 686 (52%) patients received treatment at HVFs, and 621 (48%) at LVFs. Notably, 84% of procedures at LVFs were P/Ds, versus 75% at HVFs (p < 0.001). Tables 1 and 2 display notable clinical characteristics of each cohort. Most patients were male, Caucasian, and lived in a metro area. HVFs were more likely located in the Middle and South Atlantic areas (NJ, NY, PA, DC, DE, FL, GA, MD, NC, SC, VA, WV), but less so in New England, parts of the South (AL, KY, MS, TN), and parts of the Midwest (IA, KS, MN, MO, ND, NE, SD) (p < 0.05 for all). Differences between groups were also observed for educational status of residence and comorbidity index; HVFs were less likely to operate on sarcomatoid patients and more likely to deliver chemotherapy (p < 0.05 for all). There were no independent correlations between facility type and socioeconomic status or staging. Although the vast majority of HVFs were academic (84%) and LVFs were community facilities (72%), these two variables did not independently correlate following multivariable adjustment. Likewise, despite patients traveling over twice the distances for therapy at HVFs (26 vs. 10 miles), this was also not independently associated. Postoperative outcomes were first analyzed between groups. Although the overall median length of postoperative hospitalization was numerically similar between the HVF and LVF groups (6 days 9
Lung Cancer 120 (2018) 7–13
V. Verma et al.
Fig. 1. Patient selection diagram. *TNM, tumor-node-metastasis.
herein. Additionally, HVFs were less likely to treat patients with higher comorbidities and sarcomatoid histology. This could represent more careful patient selection at these centers. Despite the NCDB not uniformly coding histology, HVFs (most of which were academic centers) may more closely follow national guidelines [1]. Additionally, although these institutions could better recognize that sarcomatoid patients have a more limited prognosis and thus may not appreciably benefit from definitive surgery, it is also possible that referring providers recognize this element as well and hence less commonly refer patients to academic centers. Our novel findings regarding postoperative outcomes are similar to existing literature. The 30-day mortality of 5% herein is comparable to the findings in multiple studies [2–5], although definitions of “postoperative” versus 30-day mortality are not uniform from publication to publication. Although the high (10–15%) observed 90-day mortality is higher than the few studies that have reported this parameter [4], the median 6 days of postoperative hospitalization in this investigation is notably shorter than other reports [2,5]. However, caveats to any of
multivariable stepwise selection, but this was not true for 90th percentile or higher. It is hence possible that the highest volume centers (e.g. p = 0.458 for the 99th percentile threshold in Supplementary Table 1) may have experienced unexpectedly lower OS, possibly relating to resecting more aggressive disease and/or losing out-of-town patients to suboptimal follow-up (although there is a comparison in Table 1, this does not represent the highest-volume centers). Third, several postulations of OS benefits at HVFs in analogous studies are purely speculative (e.g. streamlined diagnostic processes, coordination of multimodality care, aggressive clinical monitoring, and/or willingness to deliver salvage therapy), and there is limited evidence that those measures improve OS of a near-universally-fatal neoplasm such as MPM. Instead of speculation, it is helpful to cite evidence that (despite known caveats) well-selected patients treated by experienced clinicians can indeed experience prolonged OS [28]. In addition to geographic differences between patients treated at HVFs versus LVFs, it is important that the former were more likely to deliver chemotherapy, which independently correlated with higher OS 10
Lung Cancer 120 (2018) 7–13
V. Verma et al.
Table 2 Clinical characteristics of the overall cohort and factors associated with receiving treatment at a high volume facility in the final multivariable logistic regression model. Parameter
LVF (N = 621)
HVF (N = 686)
Tumor size, greatest dimension (mm) Median (IQR)
50 (30–76)
44 (19–78)
Presence of pleural effusion Yes No Unknown
508 (82%) 90 (14%) 23 (4%)
555 (81%) 101 (15%) 30 (4%)
Clinical T classification T0 T1 T2 T3 T4
2 (0%) 186 (30%) 199 (32%) 110 (18%) 124 (20%)
3 (0%) 182 (27%) 230 (34%) 183 (27%) 88 (13%)
Clinical N classification N0 N1 N2 N3
506 (81%) 38 (6%) 66 (11%) 11 (2%)
523 (76%) 45 (7%) 112 (16%) 6 (1%)
Clinical group stage I II III IV Unknown
167 (27%) 156 (25%) 159 (26%) 134 (22%) 5 (1%)
163 (24%) 190 (28%) 228 (33%) 96 (14%) 9 (1%)
Histology Epithelioid Biphasic
272 (44%) 67 (11%)
414 (60%) 105 (15%)
Sarcomatoid
82 (13%)
52 (8%)
Not otherwise specified
200 (32%)
115 (17%)
Type of surgery Extrapleural pneumonectomy Pleurectomy/ decortication
99 (16%)
172 (25%)
522 (84%)
514 (75%)
Surgical margins Negative Positive Unknown
157 (25%) 186 (30%) 278 (45%)
289 (42%) 165 (24%) 232 (34%)
Receipt of radiation therapy Yes No Unknown
85 (14%) 509 (82%) 27 (4%)
160 (23%) 498 (73%) 28 (4%)
Receipt of chemotherapy Yes
350 (56%)
488 (71%)
232 (37%) 39 (6%)
162 (24%) 36 (5%)
No Unknown
Final multivariable model OR (95% CI)
p-value
REF 1.197 (0.815–1.769) 0.592 (0.370–0.944) 0.371 (0.267–0.511)
REF 0.363
1.800 (1.354–2.399) REF –
Fig. 2. (A) Overall survival based on facility volume for the entire cohort. (B) Overall survival based on facility volume in the propensity matched population. Shaded areas represent 95% confidence intervals.
0.028 < 0.001
receipt of a particular treatment, technique, or paradigm. First, the parity between high-volume and low-volume centers was immense, and more than analogous studies in the literature of other neoplasms. Each center treated an average of 2.7 patients (range, 1.0–38.3 patients) over the entire 2004–2012 time period, raising the question of whether competency for a major surgical procedure could be maintained with so few cases. In fact, prior to the application of inclusion/exclusion criteria in this investigation (starting with 23,414 patients treated at 1248 different facilities), the top 3 (0.2%) facilities treated 861 (3.7%) patients and the bottom 48 (3.8%) facilities each treated a single patient (0.2%). Second, facility volume is considerably related to the overall population in the geographic area (“encatchment radius”) and not necessarily an accurate surrogate for surgeon experience or technical expertise. Third, although the NCDB includes data for 70% of the United States population, only CoC-accredited facilities contribute data. As such, these findings may not necessarily be representative of the entire United States population. For instance, pertaining to this study, it is difficult to rationalize that there were statistically fewer HVFs in the New England region. Fourth, the NCDB does not list specific postoperative complications, which is essential to making conclusions related to facility volume. Next, missing values in the NCDB for histology and surgical margins make for difficult interpretation, although negative surgical margins were interestingly numerically higher in the HVF group. Lastly, the NCDB does not keep track of several noteworthy variables, such as pulmonary function tests, salvage and intraoperative therapies [29], as well as enrollment into clinical trials, all which could affect OS and confound conclusions of the current study. It also does not record other endpoints, such as causes of postoperative mortality, tolerance of therapy, cancer-specific survival, and local/regional control.
< 0.001 REF –
Statistically significant p-values are in bold. Only statistically significant parameters are shown on multivariable analysis. OR, odds ratio; CI, confidence interval; IQR, interquartile range.
these comparisons include the lack of standardization in surgical technique, time period/era of treatment, and available supportive therapies (both inpatient and outpatient). Although the NCDB provides a unique means to study these novel issues, this investigation is not without weaknesses, in addition to the fact that no retrospective study can ever completely control for biases in 11
Lung Cancer 120 (2018) 7–13
V. Verma et al.
Table 3 Multivariable Cox proportional hazards model for overall survival. Parameter
Overall Population
Matched Population
HR
95% CI
p-value
HR
95% CI
Age (continuous)
1.017
Gender (male vs. female)
1.324
Patient residence (reference: metro) Urban Rural
p-value
1.009–1.024
< 0.001
1.020
1.010–1.030
< 0.001
1.103–1.589
0.003
1.410
1.143–1.740
0.001
1.317 1.482
1.069–1.622 0.889–2.445
0.010 0.123
1.279 1.288
0.957–1.701 0.728–2.278
0.096 0.384
Pleural effusion (yes vs. no)
1.199
1.000–1.437
0.050
1.305
1.029–1.656
0.028
Clinical T classification (reference: T0) T1 T2 T3 T4
1.845 1.762 1.940 2.696
0.532–6.391 0.509–6.104 0.556–6.767 0.772–9.412
0.334 0.371 0.299 0.120
1.318 1.342 1.340 1.483
1.243–1.414 1.266–1.440 1.258–1.448 1.360–1.646
< 0.001 < 0.001 < 0.001 < 0.001
Histology (reference: biphasic) Epithelioid Sarcomatoid
0.618 1.548
0.508–0.752 1.126–2.128
< 0.001 0.007
0.617 1.707
0.466–0.818 1.152–2.526
< 0.001 0.008
Total radiation dose (continuous)
1.000
1.000–1.000
< 0.001
1.000
1.000–1.000
0.018
Receipt of chemotherapy (yes vs. no)
0.801
0.685–0.936
0.005
0.811
0.664–0.989
0.039
Statistically significant p values are in bold. HR, hazard ratio; CI, confidence interval.
Further investigation to corroborate the conclusions of these hypothesis-generating data is necessary.
consecutive patients, J. Thorac. Oncol. 9 (2014) 390–396. [5] B.M. Burt, R.B. Cameron, N.M. Mollberg, et al., Malignant pleural mesothelioma and the Society of Thoracic Surgeons Database: an analysis of surgical morbidity and mortality, J. Thorac. Cardiovasc. Surg. 148 (2014) 30–35. [6] T. Treasure, L. Lang-Lazdunski, D. Waller, et al., Extra-pleural pneumonectomy versus no extra-pleural pneumonectomy for patients with malignant pleural mesothelioma: clinical outcomes of the Mesothelioma and Radical Surgery (MARS) randomised feasibility study, Lancet Oncol. 12 (2011) 763–772. [7] J.C. Gutierrez, E.A. Perez, F.L. Moffat, et al., Should soft tissue sarcomas be treated at high-volume centers? An analysis of 4205 patients, Ann. Surg. 245 (2007) 952–958. [8] V. Ho, M.J. Heslin, Effect of hospital volume and experience on in-hospital mortality for pancreaticoduodenectomy, Ann. Surg. 237 (2003). [9] C.T. Tuggle, A. Patel, N. Broer, et al., Increased hospital volume is associated with improved outcomes following abdominal-based breast reconstruction, J. Plast. Surg. Hand Surg. 48 (2014) 382–388. [10] J.A. Meyerhardt, J.E. Tepper, D. Niedzwiecki, et al., Impact of hospital procedure volume on surgical operation and long-term outcomes in high-risk curatively resected rectal cancer: findings from the Intergroup 0114 study, J. Clin. Oncol. 22 (2004) 166–174. [11] J.E. Rosen, J.G. Hancock, A.W. Kim, F.C. Detterbeck, D.J. Boffa, Predictors of mortality after surgical management of lung cancer in the National Cancer Database, Ann. Thorac. Surg. 98 (2014) 1953–1960. [12] P.J. Speicher, B.R. Englum, A.M. Ganapathi, et al., Traveling to a high-volume center is associated with improved survival for patients with esophageal cancer, Ann. Surg. 265 (2017). [13] C.J. Balentine, A.D. Naik, C.N. Robinson, et al., Association of high-volume hospitals with greater likelihood of discharge to home following colorectal surgery, JAMA Surg. 149 (2014) 244–251. [14] K. Bilimoria, A. Stewart, D. Winchester, C. Ko, The National Cancer Data Base: a powerful initiative to improve cancer care in the United States, Ann. Surg. Oncol. 15 (2008) 683–690. [15] J.M. Stahl, C.D. Corso, V. Verma, et al., Trends in stereotactic body radiation therapy for stage I small cell lung cancer, Lung Cancer 103 (2017) 11–16. [16] V. Verma, M.T. McMillan, S. Grover, C.B. Simone 2nd, Stereotactic body radiation therapy and the influence of chemotherapy on overall survival for large (≥5 centimeter) non-small cell lung cancer, Int. J. Radiat. Oncol. Biol. Phys. 97 (2017) 146–154. [17] M.T. McMillan, E. Ojerholm, V. Verma, et al., Radiation treatment time and overall survival in locally advanced non-small cell lung cancer, Int. J. Radiat. Oncol. Biol. Phys. 98 (2017) 1142–1152. [18] V. Verma, C.B. Simone 2nd, C. Lin, Human papillomavirus and nasopharyngeal cancer, Head Neck 40 (4 (April)) (2018) 696–706, http://dx.doi.org/10.1002/hed. 24978. [19] V. Verma, J.M. Ryckman, C.B. Simone 2nd, C. Lin, Patterns of care and outcomes with the addition of chemotherapy to radiation therapy for stage I nasopharyngeal cancer, Acta Oncol. 57 (2 (Feb.)) (2018) 257–261, http://dx.doi.org/10.1080/ 0284186X.2017.1351039. [20] V. Verma, C.A. Ahern, C.G. Berlind, et al., National cancer database report on pneumonectomy versus lung-sparing surgery for malignant pleural mesothelioma, J. Thorac. Oncol. 12 (2017) 1704–1714. [21] W. Haque, V. Verma, E.B. Butler, B.S. Teh, Definitive chemoradiation at high volume facilities is associated with improved survival in glioblastoma, J. Neurooncol.
5. Conclusions This contemporary NCDB dataset, the largest of its kind to date, illustrates the influence of facility volume on postoperative outcomes and survival for MPM. Patients treated at HVFs may experience more favorable postoperative outcomes than those treated at LVFs, particularly for length of postoperative hospitalizations, 30-day readmission rates, and 90-day mortality rates. Although these data must be corroborated by other investigations, they have implications for postoperative management, patient counseling, referring providers, and cost-effectiveness. Declaration There are no acknowledgements. There was no funding for this study. This study has not been presented or published in part or full form elsewhere. CAA, CGB, and WDL report personal fees from Oncora Medical during the conduct of the study and stock ownership in Oncora Medical. All other authors declare that conflicts of interest do not exist. All other authors declare no conflicts of interest. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.lungcan.2018.03.019. References [1] National Comprehensive Cancer Network, Malignant Pleural Mesothelioma. Version 2, (2017) https://www.nccn.org/professionals/physician_gls/pdf/mpm.pdf . (Accessed 30 October 2017). [2] D.J. Sugarbaker, R.M. Flores, M.T. Jaklitsch, et al., Resection margins, extrapleural nodal status, and cell type determine postoperative long-term survival in trimodality therapy of malignant pleural mesothelioma: results in 183 patients, J. Thorac. Cardiovasc. Surg. 117 (1999) 54–65. [3] V.W. Rusch, K. Rosenzweig, E. Venkatraman, et al., A phase II trial of surgical resection and adjuvant high-dose hemithoracic radiation for malignant pleural mesothelioma, J. Thorac. Cardiovasc. Surg. 122 (2001) 788–795. [4] P. Bovolato, C. Casadio, A. Bille, et al., Does surgery improve survival of patients with malignant pleural mesothelioma? A multicenter retrospective analysis of 1365
12
Lung Cancer 120 (2018) 7–13
V. Verma et al.
Stat. 10 (2011) 150–161. [26] S.L.T. Normand, M.B. Landrum, E. Guadagnoli, et al., Validating recommendations for coronary angiography following an acute myocardial infarction in the elderly: a matched analysis using propensity scores, J. Clin. Epidemiol. 54 (2001) 387–398. [27] H.F. Batirel, M. Metintas, H.B. Caglar, et al., Adoption of pleurectomy and decortication for malignant mesothelioma leads to similar survival as extrapleural pneumonectomy, J. Thorac. Cardiovasc. Surg. 151 (2016) 478–484. [28] J.S. Friedberg, C.B. Simone 2nd, M.J. Culligan, et al., Extended pleurectomy-decortication-based treatment for advanced stage epithelial mesothelioma yielding a median survival of nearly three years, Ann. Thorac. Surg. 103 (2017) 912–919. [29] C.B. Simone 2nd, K.A. Cengel, Photodynamic therapy for lung cancer and malignant pleural mesothelioma, Semin. Oncol. 41 (2014) 820–830.
135 (1 (Oct.)) (2017) 173–181, http://dx.doi.org/10.1007/s11060-017-2563-0. [22] V. Verma, P.K. Allen, C.B. Simone 2nd, H.A. Gay, S.H. Lin, Impact of treatment at high-volume facilities on survival in patients receiving chemoradiotherapy for nasopharyngeal cancer, JAMA Otolaryngol. Head Neck Surg. (Nov) (2017), http://dx. doi.org/10.1001/jamaoto.2017.1874 [Epub ahead of print] No abstract available. PMID: 29098291. [23] R Core Team, R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria, 2017 URL https://www.Rproject.org/. [24] D.E. Ho, I. Kosuke, G. King, E.A. Stuart, MatchIt: nonparametric preprocessing for parametric causal inference, J. Stat. Softw. 42 (2011) 1–28. [25] P.C. Austin, Optimal caliper widths for propensity-score matching when estimating differences in means and differences in proportions in observational studies, Pharm.
13
Contents lists available at ScienceDirect
Lung Cancer journal homepage: www.elsevier.com/locate/lungcan
Facility volume and postoperative outcomes for malignant pleural mesothelioma: A National Cancer Data Base analysis
T
Vivek Vermaa, Christopher A. Ahernb, Christopher G. Berlindb, William D. Lindsayb, ⁎ Surbhi Groverc, Melissa J. Culligand, Joseph S. Friedbergd, Charles B. Simone IIe, a
Department of Radiation Oncology, Allegheny General Hospital, Pittsburgh, PA, USA Oncora Medical, Philadelphia, PA, USA c Department of Radiation Oncology, University of Pennsylvania, Philadelphia, PA, USA d Department of Surgery, Division of Thoracic Surgery, University of Maryland Medical Center, Baltimore, MD, USA e Department of Radiation Oncology, University of Maryland Medical Center, Baltimore, MD, USA b
A R T I C LE I N FO
A B S T R A C T
Keywords: Mesothelioma Extrapleural pneumonectomy Pleurectomy Surgery Facility volume
Purpose: This study of a large, contemporary national database evaluated postoperative outcomes and overall survival (OS) for malignant pleural mesothelioma (MPM) by facility volume. Methods: The National Cancer Database was queried for newly-diagnosed non-metastatic MPM undergoing definitive surgery (extrapleural pneumonectomy (EPP) or pleurectomy/decortication (P/D)). Patients were dichotomized into those receiving therapy at a high-volume facility (HVF), defined a priori at the 90th percentile of case volume, with all others categorized as lower-volume facilities (LVFs). Statistics included multivariable logistic regression, Kaplan-Meier analysis, propensity-matching, and multivariable Cox proportional hazards modeling. Sensitivity analysis varied the dichotomized HVF-LVF cutoff and evaluated effects on postoperative outcomes and OS. Results: Of 1307 patients, 621 (48%) were treated at LVFs and 686 (52%) at HVFs. HVFs were more often in the Middle/South Atlantic regions, and less likely in New England, South, and Midwest. Notably, 75% of procedures at HVFs were P/Ds, versus 84% at LVFs (p < 0.001). Patients treated at HVFs experienced shorter length of postoperative hospitalization (p = 0.035), lower 30-day readmission rates (4.6% vs. 6.1%, p = 0.021), and lower 90-day mortality rates (10.0% vs. 14.6%, p = 0.029). Median OS for respective groups were 18 versus 15 months (p = 0.010), which were not significant following propensity-matching (p = 0.540). On multivariable analysis, facility volume did not independently predict for OS. Sensitivity analyses confirmed the postoperative outcomes and OS findings. Conclusions: This is the largest investigation to date assessing facility volume and outcomes following surgery for MPM. Although no independent effects on OS were observed, postoperative outcomes were more favorable at HVFs. These findings have implications for postoperative management, patient counseling, referring providers, and cost-effectiveness.
1. Introduction Although malignant pleural mesothelioma (MPM) is relatively uncommon, it is a highly aggressive neoplasm associated with a very poor prognosis. Gross macroscopic resection is perhaps the most important aspect of management along with chemotherapy, provided technical and medical candidacy for surgery [1]. Definitive surgery is most commonly performed using two approaches: extrapleural pneumonectomy (EPP) or extended pleurectomy/decortication (P/D). Both of these procedures are technically challenging and can cause
serious postoperative complications, irreversible morbidities, and mortality. Postoperative complications occur in 13–38% of patients, with postoperative 30-day mortality occurring in 3–8% [2–5]; owing to publication bias, these rates could very well underestimate the true incidences. In the randomized MARS trial, postoperative complications in the surgery (EPP) arm occurred in 69% of patients, with a mortality rates of 13% (intention to treat) to 16% (any patient in whom surgery was attempted) [6]. Owing to these and other reasons, it may be hypothesized that receiving surgery at a high-volume facility (HVF) may be advantageous
⁎ Corresponding author at: University of Maryland School of Medicine, Department of Radiation Oncology, Maryland Proton Treatment Center, 850 W. Baltimore St., Baltimore, MD 21201, USA. E-mail address: [email protected] (C.B. Simone).
https://doi.org/10.1016/j.lungcan.2018.03.019 Received 2 February 2018; Received in revised form 2 March 2018; Accepted 20 March 2018 0169-5002/ © 2018 Elsevier B.V. All rights reserved.
Lung Cancer 120 (2018) 7–13
V. Verma et al.
over lower-volume facilities (LVFs) in terms of postoperative complications and possibly even survival. Improvements in postoperative outcomes and/or survival at HVFs have been shown for numerous surgical procedures, including sarcoma excision [7], pancreaticoduodenectomy [8], breast reconstruction [9], low anterior/abdominoperineal resection [10], lobectomy/pneumonectomy [11], esophagectomy [12], and colectomy [13]. To date, such an analysis has not been performed for MPM. Such a study could have important consequences considering the technical expertise required for a major surgical procedure and the particularly high perioperative mortality rate associated with MPM surgery. Investigating the large, contemporary National Cancer Data Base (NCDB), this study sought to address the influence of facility volume on differences in the four postoperative outcomes given by the NCDB (length of hospital stay, 30-day readmission, 30-day mortality, and 90day mortality), and secondarily, overall survival (OS).
Table 1 Demographic characteristics of the overall cohort and factors associated with receiving treatment at a high volume facility in the final multivariable logistic regression model. Parameter
2. Materials and methods The NCDB is a joint project of the Commission on Cancer (CoC) of the American College of Surgeons and the American Cancer Society that consists of information regarding tumor characteristics, patient demographics, and patient survival for approximately 70% of the United States population [14–19]. The NCDB contains information not included in the Surveillance, Epidemiology, and End Results (SEER) database, including details pertaining to systemic therapy usage. The data used in this study were derived from a de-identified NCDB file. The American College of Surgeons and the CoC have not verified and are neither responsible for the analytic or statistical methodology employed nor the conclusions of the authors drawn from these data. As all patient information in the NCDB database is de-identified, this study was exempt from institutional review board evaluation. The NCDB Participant User File corresponding to mesothelioma (2004–2012) was utilized for this study. Inclusion criteria for this investigation were patients with newly-diagnosed MPM who received definitive surgery (EPP or P/D). Definitive surgery was defined by a board-certified thoracic surgeon with notable mesothelioma surgical experience (J.S.F.) as surgery of the primary intrathoracic site with codes 20–23, 30, 33, 40, 45–48, and 50 for P/D and codes 55–56, 60, 66, and 70 for EPP [20]. Patients with all other types of surgery (including those with ambiguous surgical codes/labels) were eliminated. Patients without proper TNM staging were also removed, as were patients designated in the NCDB as metastatic (stage M1) or receiving palliative care. In accordance with the variables in NCDB files, information collected on each patient broadly included demographic, clinical, and treatment data. The definition of a HVF was similar to other established work [21,22]. Briefly, facility volume was dichotomized into HVF or LVF based on a threshold corresponding to the 90th percentile of patient numbers treated per facility over the time period. Similar to previously published work, this cutoff was utilized in order to evaluate a roughly 1:1 ratio of patients [21,22]. However, in order to evaluate whether this a priori definition was significant at other cutoffs, sensitivity analysis was performed post hoc to evaluate whether altering the HVF definition affected the association with postoperative outcomes and OS. This was performed by repeating the multivariable analyses for each given threshold of HVF definition. Statistical analysis was performed with R [23]. Tests were twosided, with a threshold of p < 0.05 for statistical significance. First, clinical characteristics between HVF and LVF groups were tabulated. Multivariable logistic regression analysis was performed to ascertain factors independently associated with treatment at a HVF. The chisquared or Fisher’s exact tests evaluated differences in 30-day readmission, 30-day mortality, and 90-day mortality. The Mann-Whitney U test assessed differences between groups in length of postoperative hospitalization. Kaplan-Meier curves were calculated to evaluate OS,
LVF (N = 621)
HVF (N = 686)
Age (years) Median (IQR)
69 (61–75)
67 (61–74)
Gender Male Female
489 (79%) 132 (21%)
541 (79%) 145 (21%)
Race White Black Other Unknown
585 (94%) 24 (4%) 3 (0%) 9 (1%)
644 (94%) 19 (3%) 16 (2%) 7 (1%)
Charlson Deyo score 0 1
431 (69%) 148 (24%)
534 (78%) 129 (19%)
42 (7%)
23 (3%)
Insurance type Private Medicare Medicaid Other government Uninsured Unknown
225 (36%) 355 (57%) 14 (2%) 6 (1%) 11 (2%) 10 (2%)
272 (40%) 380 (55%) 10 (1%) 8 (1%) 4 (1%) 12 (1%)
Income (US dollars/ year) < $30,000 $30,000–$34,999 $35,000–$45,999 ≥$46,000 Unknown
87 (14%) 134 (22%) 164 (26%) 219 (35%) 17 (3%)
65 (9%) 105 (15%) 151 (22%) 343 (50%) 22 (3%)
≥2
Percentage of adults in zip code without high school diploma ≥21% 83 (13%) 54 (8%) 13–20.9% 142 (23%) 143 (21%) 7–12.9%
227 (37%)
227 (33%)
< 7%
152 (24%)
240 (35%)
Unknown
17 (3%)
22 (3%)
Patient residence Urban Metro Rural Unknown
94 (15%) 490 (79%) 11 (2%) 26 (4%)
62 (9%) 584 (85%) 8 (1%) 32 (5%)
Facility location East North Central East South Central
117 (19%) 40 (6%)
128 (19%) 11 (2%)
Middle Atlantic
84 (14%)
232 (34%)
Mountain
24 (4%)
17 (2%)
New England
62 (10%)
10 (1%)
Pacific
78 (13%)
86 (13%)
South Atlantic
118 (19%)
162 (24%)
West North Central
56 (9%)
16 (2%)
West South Central
35 (6%)
15 (2%)
Unknown
7 (1%)
9 (1%)
166 (27%) 447 (72%)
579 (84%) 98 (14%)
Facility type Academic Community
Final multivariable model OR (95% CI)
p-value
REF 0.777 (0.568–1.061) 0.487 (0.260–0.896)
REF 0.112
REF 1.745 (1.083–2.825) 1.496 (0.947–2.374) 2.708 (1.685–4.376) –
REF 0.023
REF 0.361 (0.162–0.757) 2.536 (1.729–3.737) 0.510 (0.239–1.068) 0.131 (0.059–0.266) 1.040 (0.672–1.612) 1.543 (1.055–2.263) 0.245 (0.125–0.460) 0.579 (0.278–1.168) –
REF 0.009
0.022
0.085 < 0.001 –
< 0.001 0.077 < 0.001 0.860 0.026 < 0.001 0.134 –
(continued on next page)
8
Lung Cancer 120 (2018) 7–13
V. Verma et al.
(range, 0–162) versus 6 days (range, 0–61)), it was statistically lower for the former (p = 0.035). Thirty-day readmission rates were 4.6% in the HVF cohort, as compared to 6.1% for LVFs (p = 0.021). Although there were no differences in 30-day mortality (4.7% vs. 4.6%, p = 0.656), a lower 90-day mortality rate was found in patients treated at HVFs (10.0% vs. 14.6%, p = 0.029). Because propensity matching was specifically designed for the secondary OS endpoint, it was not statistically indicated to compare postoperative outcomes between the matched populations, similar to other work [20]. Nevertheless, Supplementary Table 1 displays results of sensitivity analysis when varying the threshold to define a HVF. Findings related to 30-day readmission and 90-day mortality persisted with all tested cutoffs, whereas length of hospital stay was statistically significant above and including a base threshold of 87th percentile of case volume. The median follow-up was 15 months (range, 0–125 months). Kaplan-Meier estimates comparing OS between HVFs and LVFs are illustrated in Fig. 2. In the overall population, patients treated at HVFs experienced higher median OS (18 months (95% CI 17–19 months) versus 15 months (95% CI 13–17 months), p = 0.010). However, after propensity matching, there were no statistical OS differences between groups (17 months versus 15 months, p = 0.540). When adjusting for potential confounders, there were several predictors of OS on multivariable analysis in both the matched and unmatched populations (Table 3). Factors independently predictive of poorer OS were advanced age, male gender, urban residence, presence of pleural effusion, increasing T classification, and sarcomatoid histology (p < 0.05 for all). Improved OS was associated with receipt of chemotherapy and increasing radiation dose (p < 0.05 for both). Facility volume as a dichotomized variable was eliminated in stepwise selection (owing to lack of statistical significance) for the Cox multivariable model, which was true for several different cutoffs tested on sensitivity analysis in both the overall and matched populations (Supplementary Table 2).
Table 1 (continued) Parameter
LVF (N = 621)
HVF (N = 686)
Final multivariable model OR (95% CI)
Unknown
8 (1%)
9 (1%)
Distance to treating facility (mi) Median (IQR)
10 (4–21)
26 (10–90)
Year of diagnosis 2004–2008 2009–2012
276 (44%) 345 (56%)
289 (42%) 397 (58%)
p-value
Statistically significant p-values are in bold. Only statistically significant parameters are shown on multivariable analysis. OR, odds ratio; CI, confidence interval; IQR, interquartile range.
defined as the interval between the date of diagnosis and the date of death or censored at last contact. The univariate association of each covariate with OS was assessed using a Cox proportional hazards model. Covariates that were significantly associated with OS (p < 0.05) were included in a multivariable model, and backwards stepwise selection was performed with α = 0.20. Lastly, in an attempt to minimize selection and indication biases, patients in both cohorts underwent propensity matching. To estimate the propensity score for each patient, the univariate association of each covariate with treatment type was assessed using a simple logistic regression model. Covariates that were significantly associated with treatment type (p < 0.05) were included in a multivariable logistic regression model, and backwards stepwise selection was performed with α = 0.20. One-to-one matching was performed using a “greedy” nearest neighbor algorithm with a caliper 0.2 times the standard deviation of the logit propensity score [24,25]. The standardized difference between groups after matching was less than 0.1, indicating sufficient balancing [26]. Propensity matching took into account all variables (listed in Table 1) except those not meeting the α = 0.20 threshold on backwards stepwise selection. Kaplan-Meier curves were calculated on the propensity matched patients, and a multivariable Cox model was fit on the propensity matched dataset by the same method described above.
4. Discussion This novel investigation of a large, contemporary national database regarding the influence of facility volume on postoperative outcomes and survival for MPM most notably illustrates that patients treated at HVFs may experience more favorable postoperative outcomes than those treated at LVFs. Although true for length of postoperative hospitalization, this is most notably true for 30-day readmission and 90day mortality at a wide variety of HVF definition cutoffs. These findings have implications for postoperative management, patient counseling, referring providers, and cost-effectiveness. It is very important to initially note that, similar to other work [22], although the vast majority of HVFs were academic institutions, the two terms are not synonymous. This is also in concordance with a prior publication demonstrating independently higher OS at academic centers [20], in contrast to lack thereof with HVFs in this study. Although this could imply that outcomes at high-volume community centers could have “diluted” the findings between that study and this one, the low sample sizes of such cases prevents formal exploratory analysis. Next, it is also important to consider that although P/D techniques were more often utilized at LVFs, this does not imply that P/D itself was associated with worse postoperative outcomes. In fact, previous analysis showed that although EPPs were more commonly performed at academic centers, there were no appreciable differences in postoperative outcomes or OS following EPP versus P/D [20]. Similar conclusions have also been reached elsewhere [4,5,27]. There are several reasons to explain the lack of OS differences in this study, in addition to the differentiation between HVFs and academic centers detailed above. Overall, MPM is associated with very poor prognosis, and delineating subtle differences in OS are thus quite difficult. Second, sensitivity analysis showed that defining HVFs are the 77th or 82nd percentiles would have resulted in inclusion into
3. Results A patient selection diagram is provided in Fig. 1. In total, 1307 patients treated at 476 different facilities met study criteria. Of those, 686 (52%) patients received treatment at HVFs, and 621 (48%) at LVFs. Notably, 84% of procedures at LVFs were P/Ds, versus 75% at HVFs (p < 0.001). Tables 1 and 2 display notable clinical characteristics of each cohort. Most patients were male, Caucasian, and lived in a metro area. HVFs were more likely located in the Middle and South Atlantic areas (NJ, NY, PA, DC, DE, FL, GA, MD, NC, SC, VA, WV), but less so in New England, parts of the South (AL, KY, MS, TN), and parts of the Midwest (IA, KS, MN, MO, ND, NE, SD) (p < 0.05 for all). Differences between groups were also observed for educational status of residence and comorbidity index; HVFs were less likely to operate on sarcomatoid patients and more likely to deliver chemotherapy (p < 0.05 for all). There were no independent correlations between facility type and socioeconomic status or staging. Although the vast majority of HVFs were academic (84%) and LVFs were community facilities (72%), these two variables did not independently correlate following multivariable adjustment. Likewise, despite patients traveling over twice the distances for therapy at HVFs (26 vs. 10 miles), this was also not independently associated. Postoperative outcomes were first analyzed between groups. Although the overall median length of postoperative hospitalization was numerically similar between the HVF and LVF groups (6 days 9
Lung Cancer 120 (2018) 7–13
V. Verma et al.
Fig. 1. Patient selection diagram. *TNM, tumor-node-metastasis.
herein. Additionally, HVFs were less likely to treat patients with higher comorbidities and sarcomatoid histology. This could represent more careful patient selection at these centers. Despite the NCDB not uniformly coding histology, HVFs (most of which were academic centers) may more closely follow national guidelines [1]. Additionally, although these institutions could better recognize that sarcomatoid patients have a more limited prognosis and thus may not appreciably benefit from definitive surgery, it is also possible that referring providers recognize this element as well and hence less commonly refer patients to academic centers. Our novel findings regarding postoperative outcomes are similar to existing literature. The 30-day mortality of 5% herein is comparable to the findings in multiple studies [2–5], although definitions of “postoperative” versus 30-day mortality are not uniform from publication to publication. Although the high (10–15%) observed 90-day mortality is higher than the few studies that have reported this parameter [4], the median 6 days of postoperative hospitalization in this investigation is notably shorter than other reports [2,5]. However, caveats to any of
multivariable stepwise selection, but this was not true for 90th percentile or higher. It is hence possible that the highest volume centers (e.g. p = 0.458 for the 99th percentile threshold in Supplementary Table 1) may have experienced unexpectedly lower OS, possibly relating to resecting more aggressive disease and/or losing out-of-town patients to suboptimal follow-up (although there is a comparison in Table 1, this does not represent the highest-volume centers). Third, several postulations of OS benefits at HVFs in analogous studies are purely speculative (e.g. streamlined diagnostic processes, coordination of multimodality care, aggressive clinical monitoring, and/or willingness to deliver salvage therapy), and there is limited evidence that those measures improve OS of a near-universally-fatal neoplasm such as MPM. Instead of speculation, it is helpful to cite evidence that (despite known caveats) well-selected patients treated by experienced clinicians can indeed experience prolonged OS [28]. In addition to geographic differences between patients treated at HVFs versus LVFs, it is important that the former were more likely to deliver chemotherapy, which independently correlated with higher OS 10
Lung Cancer 120 (2018) 7–13
V. Verma et al.
Table 2 Clinical characteristics of the overall cohort and factors associated with receiving treatment at a high volume facility in the final multivariable logistic regression model. Parameter
LVF (N = 621)
HVF (N = 686)
Tumor size, greatest dimension (mm) Median (IQR)
50 (30–76)
44 (19–78)
Presence of pleural effusion Yes No Unknown
508 (82%) 90 (14%) 23 (4%)
555 (81%) 101 (15%) 30 (4%)
Clinical T classification T0 T1 T2 T3 T4
2 (0%) 186 (30%) 199 (32%) 110 (18%) 124 (20%)
3 (0%) 182 (27%) 230 (34%) 183 (27%) 88 (13%)
Clinical N classification N0 N1 N2 N3
506 (81%) 38 (6%) 66 (11%) 11 (2%)
523 (76%) 45 (7%) 112 (16%) 6 (1%)
Clinical group stage I II III IV Unknown
167 (27%) 156 (25%) 159 (26%) 134 (22%) 5 (1%)
163 (24%) 190 (28%) 228 (33%) 96 (14%) 9 (1%)
Histology Epithelioid Biphasic
272 (44%) 67 (11%)
414 (60%) 105 (15%)
Sarcomatoid
82 (13%)
52 (8%)
Not otherwise specified
200 (32%)
115 (17%)
Type of surgery Extrapleural pneumonectomy Pleurectomy/ decortication
99 (16%)
172 (25%)
522 (84%)
514 (75%)
Surgical margins Negative Positive Unknown
157 (25%) 186 (30%) 278 (45%)
289 (42%) 165 (24%) 232 (34%)
Receipt of radiation therapy Yes No Unknown
85 (14%) 509 (82%) 27 (4%)
160 (23%) 498 (73%) 28 (4%)
Receipt of chemotherapy Yes
350 (56%)
488 (71%)
232 (37%) 39 (6%)
162 (24%) 36 (5%)
No Unknown
Final multivariable model OR (95% CI)
p-value
REF 1.197 (0.815–1.769) 0.592 (0.370–0.944) 0.371 (0.267–0.511)
REF 0.363
1.800 (1.354–2.399) REF –
Fig. 2. (A) Overall survival based on facility volume for the entire cohort. (B) Overall survival based on facility volume in the propensity matched population. Shaded areas represent 95% confidence intervals.
0.028 < 0.001
receipt of a particular treatment, technique, or paradigm. First, the parity between high-volume and low-volume centers was immense, and more than analogous studies in the literature of other neoplasms. Each center treated an average of 2.7 patients (range, 1.0–38.3 patients) over the entire 2004–2012 time period, raising the question of whether competency for a major surgical procedure could be maintained with so few cases. In fact, prior to the application of inclusion/exclusion criteria in this investigation (starting with 23,414 patients treated at 1248 different facilities), the top 3 (0.2%) facilities treated 861 (3.7%) patients and the bottom 48 (3.8%) facilities each treated a single patient (0.2%). Second, facility volume is considerably related to the overall population in the geographic area (“encatchment radius”) and not necessarily an accurate surrogate for surgeon experience or technical expertise. Third, although the NCDB includes data for 70% of the United States population, only CoC-accredited facilities contribute data. As such, these findings may not necessarily be representative of the entire United States population. For instance, pertaining to this study, it is difficult to rationalize that there were statistically fewer HVFs in the New England region. Fourth, the NCDB does not list specific postoperative complications, which is essential to making conclusions related to facility volume. Next, missing values in the NCDB for histology and surgical margins make for difficult interpretation, although negative surgical margins were interestingly numerically higher in the HVF group. Lastly, the NCDB does not keep track of several noteworthy variables, such as pulmonary function tests, salvage and intraoperative therapies [29], as well as enrollment into clinical trials, all which could affect OS and confound conclusions of the current study. It also does not record other endpoints, such as causes of postoperative mortality, tolerance of therapy, cancer-specific survival, and local/regional control.
< 0.001 REF –
Statistically significant p-values are in bold. Only statistically significant parameters are shown on multivariable analysis. OR, odds ratio; CI, confidence interval; IQR, interquartile range.
these comparisons include the lack of standardization in surgical technique, time period/era of treatment, and available supportive therapies (both inpatient and outpatient). Although the NCDB provides a unique means to study these novel issues, this investigation is not without weaknesses, in addition to the fact that no retrospective study can ever completely control for biases in 11
Lung Cancer 120 (2018) 7–13
V. Verma et al.
Table 3 Multivariable Cox proportional hazards model for overall survival. Parameter
Overall Population
Matched Population
HR
95% CI
p-value
HR
95% CI
Age (continuous)
1.017
Gender (male vs. female)
1.324
Patient residence (reference: metro) Urban Rural
p-value
1.009–1.024
< 0.001
1.020
1.010–1.030
< 0.001
1.103–1.589
0.003
1.410
1.143–1.740
0.001
1.317 1.482
1.069–1.622 0.889–2.445
0.010 0.123
1.279 1.288
0.957–1.701 0.728–2.278
0.096 0.384
Pleural effusion (yes vs. no)
1.199
1.000–1.437
0.050
1.305
1.029–1.656
0.028
Clinical T classification (reference: T0) T1 T2 T3 T4
1.845 1.762 1.940 2.696
0.532–6.391 0.509–6.104 0.556–6.767 0.772–9.412
0.334 0.371 0.299 0.120
1.318 1.342 1.340 1.483
1.243–1.414 1.266–1.440 1.258–1.448 1.360–1.646
< 0.001 < 0.001 < 0.001 < 0.001
Histology (reference: biphasic) Epithelioid Sarcomatoid
0.618 1.548
0.508–0.752 1.126–2.128
< 0.001 0.007
0.617 1.707
0.466–0.818 1.152–2.526
< 0.001 0.008
Total radiation dose (continuous)
1.000
1.000–1.000
< 0.001
1.000
1.000–1.000
0.018
Receipt of chemotherapy (yes vs. no)
0.801
0.685–0.936
0.005
0.811
0.664–0.989
0.039
Statistically significant p values are in bold. HR, hazard ratio; CI, confidence interval.
Further investigation to corroborate the conclusions of these hypothesis-generating data is necessary.
consecutive patients, J. Thorac. Oncol. 9 (2014) 390–396. [5] B.M. Burt, R.B. Cameron, N.M. Mollberg, et al., Malignant pleural mesothelioma and the Society of Thoracic Surgeons Database: an analysis of surgical morbidity and mortality, J. Thorac. Cardiovasc. Surg. 148 (2014) 30–35. [6] T. Treasure, L. Lang-Lazdunski, D. Waller, et al., Extra-pleural pneumonectomy versus no extra-pleural pneumonectomy for patients with malignant pleural mesothelioma: clinical outcomes of the Mesothelioma and Radical Surgery (MARS) randomised feasibility study, Lancet Oncol. 12 (2011) 763–772. [7] J.C. Gutierrez, E.A. Perez, F.L. Moffat, et al., Should soft tissue sarcomas be treated at high-volume centers? An analysis of 4205 patients, Ann. Surg. 245 (2007) 952–958. [8] V. Ho, M.J. Heslin, Effect of hospital volume and experience on in-hospital mortality for pancreaticoduodenectomy, Ann. Surg. 237 (2003). [9] C.T. Tuggle, A. Patel, N. Broer, et al., Increased hospital volume is associated with improved outcomes following abdominal-based breast reconstruction, J. Plast. Surg. Hand Surg. 48 (2014) 382–388. [10] J.A. Meyerhardt, J.E. Tepper, D. Niedzwiecki, et al., Impact of hospital procedure volume on surgical operation and long-term outcomes in high-risk curatively resected rectal cancer: findings from the Intergroup 0114 study, J. Clin. Oncol. 22 (2004) 166–174. [11] J.E. Rosen, J.G. Hancock, A.W. Kim, F.C. Detterbeck, D.J. Boffa, Predictors of mortality after surgical management of lung cancer in the National Cancer Database, Ann. Thorac. Surg. 98 (2014) 1953–1960. [12] P.J. Speicher, B.R. Englum, A.M. Ganapathi, et al., Traveling to a high-volume center is associated with improved survival for patients with esophageal cancer, Ann. Surg. 265 (2017). [13] C.J. Balentine, A.D. Naik, C.N. Robinson, et al., Association of high-volume hospitals with greater likelihood of discharge to home following colorectal surgery, JAMA Surg. 149 (2014) 244–251. [14] K. Bilimoria, A. Stewart, D. Winchester, C. Ko, The National Cancer Data Base: a powerful initiative to improve cancer care in the United States, Ann. Surg. Oncol. 15 (2008) 683–690. [15] J.M. Stahl, C.D. Corso, V. Verma, et al., Trends in stereotactic body radiation therapy for stage I small cell lung cancer, Lung Cancer 103 (2017) 11–16. [16] V. Verma, M.T. McMillan, S. Grover, C.B. Simone 2nd, Stereotactic body radiation therapy and the influence of chemotherapy on overall survival for large (≥5 centimeter) non-small cell lung cancer, Int. J. Radiat. Oncol. Biol. Phys. 97 (2017) 146–154. [17] M.T. McMillan, E. Ojerholm, V. Verma, et al., Radiation treatment time and overall survival in locally advanced non-small cell lung cancer, Int. J. Radiat. Oncol. Biol. Phys. 98 (2017) 1142–1152. [18] V. Verma, C.B. Simone 2nd, C. Lin, Human papillomavirus and nasopharyngeal cancer, Head Neck 40 (4 (April)) (2018) 696–706, http://dx.doi.org/10.1002/hed. 24978. [19] V. Verma, J.M. Ryckman, C.B. Simone 2nd, C. Lin, Patterns of care and outcomes with the addition of chemotherapy to radiation therapy for stage I nasopharyngeal cancer, Acta Oncol. 57 (2 (Feb.)) (2018) 257–261, http://dx.doi.org/10.1080/ 0284186X.2017.1351039. [20] V. Verma, C.A. Ahern, C.G. Berlind, et al., National cancer database report on pneumonectomy versus lung-sparing surgery for malignant pleural mesothelioma, J. Thorac. Oncol. 12 (2017) 1704–1714. [21] W. Haque, V. Verma, E.B. Butler, B.S. Teh, Definitive chemoradiation at high volume facilities is associated with improved survival in glioblastoma, J. Neurooncol.
5. Conclusions This contemporary NCDB dataset, the largest of its kind to date, illustrates the influence of facility volume on postoperative outcomes and survival for MPM. Patients treated at HVFs may experience more favorable postoperative outcomes than those treated at LVFs, particularly for length of postoperative hospitalizations, 30-day readmission rates, and 90-day mortality rates. Although these data must be corroborated by other investigations, they have implications for postoperative management, patient counseling, referring providers, and cost-effectiveness. Declaration There are no acknowledgements. There was no funding for this study. This study has not been presented or published in part or full form elsewhere. CAA, CGB, and WDL report personal fees from Oncora Medical during the conduct of the study and stock ownership in Oncora Medical. All other authors declare that conflicts of interest do not exist. All other authors declare no conflicts of interest. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.lungcan.2018.03.019. References [1] National Comprehensive Cancer Network, Malignant Pleural Mesothelioma. Version 2, (2017) https://www.nccn.org/professionals/physician_gls/pdf/mpm.pdf . (Accessed 30 October 2017). [2] D.J. Sugarbaker, R.M. Flores, M.T. Jaklitsch, et al., Resection margins, extrapleural nodal status, and cell type determine postoperative long-term survival in trimodality therapy of malignant pleural mesothelioma: results in 183 patients, J. Thorac. Cardiovasc. Surg. 117 (1999) 54–65. [3] V.W. Rusch, K. Rosenzweig, E. Venkatraman, et al., A phase II trial of surgical resection and adjuvant high-dose hemithoracic radiation for malignant pleural mesothelioma, J. Thorac. Cardiovasc. Surg. 122 (2001) 788–795. [4] P. Bovolato, C. Casadio, A. Bille, et al., Does surgery improve survival of patients with malignant pleural mesothelioma? A multicenter retrospective analysis of 1365
12
Lung Cancer 120 (2018) 7–13
V. Verma et al.
Stat. 10 (2011) 150–161. [26] S.L.T. Normand, M.B. Landrum, E. Guadagnoli, et al., Validating recommendations for coronary angiography following an acute myocardial infarction in the elderly: a matched analysis using propensity scores, J. Clin. Epidemiol. 54 (2001) 387–398. [27] H.F. Batirel, M. Metintas, H.B. Caglar, et al., Adoption of pleurectomy and decortication for malignant mesothelioma leads to similar survival as extrapleural pneumonectomy, J. Thorac. Cardiovasc. Surg. 151 (2016) 478–484. [28] J.S. Friedberg, C.B. Simone 2nd, M.J. Culligan, et al., Extended pleurectomy-decortication-based treatment for advanced stage epithelial mesothelioma yielding a median survival of nearly three years, Ann. Thorac. Surg. 103 (2017) 912–919. [29] C.B. Simone 2nd, K.A. Cengel, Photodynamic therapy for lung cancer and malignant pleural mesothelioma, Semin. Oncol. 41 (2014) 820–830.
135 (1 (Oct.)) (2017) 173–181, http://dx.doi.org/10.1007/s11060-017-2563-0. [22] V. Verma, P.K. Allen, C.B. Simone 2nd, H.A. Gay, S.H. Lin, Impact of treatment at high-volume facilities on survival in patients receiving chemoradiotherapy for nasopharyngeal cancer, JAMA Otolaryngol. Head Neck Surg. (Nov) (2017), http://dx. doi.org/10.1001/jamaoto.2017.1874 [Epub ahead of print] No abstract available. PMID: 29098291. [23] R Core Team, R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria, 2017 URL https://www.Rproject.org/. [24] D.E. Ho, I. Kosuke, G. King, E.A. Stuart, MatchIt: nonparametric preprocessing for parametric causal inference, J. Stat. Softw. 42 (2011) 1–28. [25] P.C. Austin, Optimal caliper widths for propensity-score matching when estimating differences in means and differences in proportions in observational studies, Pharm.
13