Abstract
Introduction
American College of Surgeons’ Oncology Group Z0030 found no survival difference between patients with early-stage non-small-cell lung cancer who had mediastinal nodal dissection or systematic sampling. However, a meta-analysis of 1,980 patients in five randomized controlled trials from 1989–2007 associated better survival with nodal dissection. We tested the survival impact of the extent of nodal dissection in curative-intent resections for early-stage non-small cell lung cancer in a population-based observational cohort.
Methods
Resections for clinical T1 or T2, N0 or non-hilar N1, M0 non-small-cell lung cancer, within four contiguous United States Hospital Referral Regions from 2009–2019 were categorized into mediastinal nodal dissection, systematic sampling, and neither, based on lymph node stations examined. We compared demographic and clinical characteristics, perioperative complication rates and survival after assessing statistical interactions and confounding.
Results
Of 1,942 eligible patients, 18% had nodal dissection, 6% had systematic sampling, 75% had intraoperative nodal evaluation that met neither standard. In teaching hospitals, nodal dissection was associated with a lower hazard of death than ‘neither’ resections (0.57 [95% confidence interval 0.41, 0.79]) but not systematic sampling (0.74 [0.40, 1.37]) after adjusting for multiple comparisons. There was no significant difference in hazard ratios at non-teaching institutions. Perioperative complication rates were not signficantly worse after mediastinal nodal dissection or systematic sampling, compared to neither.
Conclusions
In teaching institutions, mediastinal nodal dissection was associated with superior survival over less comprehensive pathologic nodal staging. There was no survival difference in non-teaching institutions, a finding which warrants further investigation.
Keywords: ACOSOG Z0030, mediastinal lymphadenectomy, survival, thoracic oncology, quality of surgical care
Introduction
Accurate pathologic nodal staging is associated with improved survival through multiple putative mechanisms- improved risk-categorization, increased detection of candidates for, and deployment of, beneficial adjuvant therapy, and, possibly, resection of oligo-metastatic disease.1 The American College of Surgeons Oncology Group (ACOSOG) Z0030 trial showed that a thorough systematic sampling procedure provided equivalent survival to mediastinal lymph node dissection in patients with early-stage non-small-cell lung cancer (NSCLC).2,3 This, despite detection of unexpected mediastinal nodal metastasis in 4% of the mediastinal nodal dissection cohort.2
The rigor of this large randomized controlled trial notwithstanding, the findings have been criticized by multiple observers, even including the study investigators.2,4,5 A recent meta-analysis of five randomized controlled trials suggested that mediastinal nodal dissection might yet provide better survival and recommended further evaluation of this question through pragmatic multicenter trials.5 A review of new developments in early-stage NSCLC in 2018 suggested that the question of adequacy of systematic sampling was still open.6
Because most lung cancer resections in the United States are performed in community-level hospitals, unlike most institutions in ACOSOG Z0030, we pragmatically re-examined the survival effects of the extent of nodal dissection in a large, contemporary, population-based surgical resection dataset developed from predominantly community-level institutions in a region with some of the highest lung cancer incidence and mortality rates in the United States. We hypothesized that the outcomes of mediastinal lymphadenectomy and systematic sampling would be equivalent, but that Z0030-quality nodal evaluation would be infrequently achieved, to the detriment of patient outcomes. To test these hypotheses, we compared the survival and perioperative complication rates of patients who received mediastinal lymph node dissection, systematic sampling and nodal evaluation not attaining ACOSOG specifications. We also examined the patient characteristics and pattern of failure when neither mediastinal lymphadenectomy nor systematic sampling was achieved.
Materials and methods
Cohort
The Mid-South Quality of Surgical Resection (MS-QSR) cohort includes all curative-intent resections for NSCLC from January 2009 to February 2019 at all 12 eligible institutions within four contiguous Dartmouth Hospital Referral Regions in North Mississippi, Eastern Arkansas and Western Tennessee. Eligible institutions had five or more annual resections for NSCLC and came from seven different healthcare systems (Table A.1).7–9 The MS-QSR cohort is derived from an ongoing National Institutes of Health-funded population-level Dissemination and Implementation Science project to improve surgical resection and pathology examination practices. The database is designed to have greater details about surgical resection practices than are currently available in national databases such as the Surveillance, Epidemiology and End-Results database, or the National Cancer Data Base. It includes greater details about anatomic lymph node evaluation than can be found in such national datasets. Contribution of data is approved by the Institutional Review Boards of all participating institutions, with a waiver of the informed consent requirement.
Patient selection.
Similar to Z0030 selection criteria, we restricted analysis to patients with clinical (c) T1 or T2 (by American Joint Committee on Cancer 8th edition criteria, rather than the 5th edition as with Z0030),10 N0 or non-hilar N1 (metastasis in stations 11–14), M0 who underwent primary anatomic resection for their first NSCLC. We excluded all patients with metastasis to lymph node stations that would have been examined at the time of systematic sampling, as in Z0030 (Table A.2).2,11 We also excluded patients who received neo-adjuvant therapy, had a previously treated lung cancer, re-resections, and wedge resections.
Definition of extent of nodal dissection.
Using the nodal stations specified for inclusion in Z0030, we defined mediastinal nodal dissection as inclusion of stations 2R, 4R, 7, 8, 9 and 10R for right-side resections, stations 5, 6, 7, 8, 9, and 10L for left-sided resections; systematic sampling as inclusion of a minimum of 2R, 4R, 7, and 10R on the right and 5,6,7 and 10L on the left, but excluding stations required for mediastinal nodal dissection (Table A.2).2,11 As a control group, we evaluated patients whose lymph node examination did not meet either of the ACOSOG criteria. Using a conservative approach, to account for the possibility of pre-resection removal of lymph nodes, we counted nodes removed at mediastinoscopy as part of the resected nodes and stations. The alternative approach in which these nodes were not counted did not change the results of our analysis (data not shown). Because endobronchial ultrasound-guided biopsy does not remove whole nodes for pathologic evaluation, we did not include nodes evaluated solely by this technique in the definition of lymphadenectomy. Node stations were identified from the final pathology report, rather than the surgeon’s report because of prior evidence of sharp discordance between surgeons’ reports of the nodal dissection procedure performed and findings from objective review of pathology reports.12,13
Survival.
Survival updates to the MS-QSR are provided by the tumor registries across all participating institutions at the same time every 6-months. In the current analysis, we measured survival from the date of surgical resection until the most recent survival update, censored on February 28, 2019. Cause of death was not available in this dataset at the time of this analysis.
Statistical analysis.
We used descriptive statistics to summarize patient demographics, clinical characteristics, and thoroughness of staging measures across the 3 nodal evaluation groups. We tested for statistical differences across the three groups with Chi-squared (or Fisher’s exact tests if expected cell counts were small) and Kruskal-Wallis tests. Survival was visually explored with Kaplan-Meier plots and statistical differences tested with log-rank tests. The association between survival and node evaluation was further examined with Cox Proportional Hazards models. Potential effect-modifiers and confounders considered included patient age, race, insurance, number of comorbidities, aggregate clinical stage, histology, resection technique, invasive staging, preoperative use of positron emission tomographic-computed tomographic (PET-CT) scanning, extent of resection, surgeon board certification, surgeon years of practice, and institutional characteristics including rurality, bed-size, and teaching status.
Institutional rurality was determined based on the Rural-Urban Continuum Code, a classification of counties into nine subgroups of metropolitan and non-metropolitan areas based on their population size. Codes 1–3 indicate a metropolitan area and 4–9 indicate a rural area.14 Effect modification, a difference in the association between type of sampling and survival across the levels of another covariate, checked at the bivariate level with an interaction term, was deemed significant at the alpha=0.05 level. Confounding, which occurs when the presence of a covariate alters the overall assocation between type of sampling and survival, was assessed among those covariates that were not effect modifiers with a percent change in the crude and adjusted hazard ratio >10% considered a confounder. As a secondary approach to adjust for confounding, we used a stabilized inverse probability weight within the Cox model which was calculated adjusting for all potential confounders. Although the intra-correlation among surgeons and institutions was small, we performed a sensitivity analysis repeating the primary analysis using a robust sandwich covariance estimator for the Cox Proportional Hazards model. We opted against using this approach in the primary analysis because the results were similar, but our sample sizes became smaller and variance estimates unreliable. We also performed an additional analysis to account for potential confounding by differences in the evaluation and management of patients with cN2 disease by analyzing survival in patients with cN0 disease only.
We performed pairwise comparisons among the types of nodal evaluation using Tukey’s multiple comparison adjustment.15 Hazard ratios with 95% confidence intervals (CI) are provided for each analysis. The proportional hazards assumption, assessed visually through log(−log) survival curves, was met. All analyses were performed at the alpha=0.05 significance level using SAS 9.4 (2013, SAS Institute Inc., Cary NC).
Results
Cohort characteristics.
From January 1, 2009 to February 28, 2019, there were 3,615 resections in the MS-QSR cohort, of which 1,258 did not meet the stage inclusion criteria. Further excluded were 51 resections for diagnoses other than NSCLC, 57 neoadjuvantly treated patients, 78 patients treated for a previous lung cancer, 20 re-resections, and 209 wedge resections, leaving an analytical cohort of 1,942 eligible resections (Figure 1). Of these, 358 (18%) had mediastinal nodal dissection, 120 (6%) systematic sampling, and 1,464 (75%) had nodal evaluation that met neither standard (Figure 2).
Figure 1.
Selection criteria leading to the analytic cohort of patients with clinical T1 or T2, N0 or non-hilar N1, M0 (by American Joint Commission on Cancer 8th edition) non-small-cell lung cancer who underwent primary curative resection, stratified by institutional teaching status, and then as having had mediastinal lymph node dissection, systematic sampling, or neither nodal evaluation standard. Patients with metastasis to lymph node stations which would have been evaluated at the time of the American College of Surgeons Z0030 trial-defined systematic sampling (stations 2R, 4R, 7 and 10 for right-side resections and 5,6,7 and 10L for left side resections) were excluded in keeping with Z0030 trial selection criteria. For this analysis, mediastinal lymph node dissection was identified by inclusion of lymph nodes from stations 2R, 4R, 7, 8, 9 and 10 for right-side resections and nodes from 5, 6, 7, 8, 9 and 10L for left side resections. The Mid-South Quality of Surgical Resection Cohort is a population-based dataset including 12 hospitals in 4 contiguous Dartmouth Hospital Referral regions in East Arkansas, North Mississippi, West Tennessee.
Figure 2.
Proportion of patients in the population-based Mid-South Quality of Surgical Resection cohort who received curative-intent surgery with mediastinal lymph node dissection or systematic sampling, as defined by the American College of Surgeons Oncology Group, or neither sampling method.
These resections were performed at 12 institutions of which 7 (providing care for 812 patients [42% of cohort]) were non-teaching; 3 (providing care for 353 patients [18%]) were rurally located. The resections were performed by 52 surgeons of which 4 (providing care for 112 patients [6%]) were board-certified general surgeons, 40 (providing care for 999 patients [51%]) were board-certified cardiovascular surgeons without general thoracic surgery focus, and 8 (providing care for 831 [43%]) were board-certified with a dedicated general thoracic surgery practice. Surgeons had a median of 28 years of practice (range: 3–51 years; interquartile range [IQR]: 16–40 years).
All teaching institutions were urban, had significantly larger number of beds than non-teaching hospitals (median 793 vs 359; p<0.001), and had resections performed by cardiovascular (29%) or dedicated general thoracic surgeons (71%) only. Surgeons in teaching hospitals had been in practice longer (median practice 38 v 29 years; p<0.001). Of the 812 resections performed in non-teaching hospitals, 43% were in rural and 57% in urban hospitals; 83% were performed by cardiovascular surgeons, 14% by general surgeons, and only 4% were performed by general thoracic surgeons.
Patient, surgeon and institutional characteristics associated with type of nodal dissection.
Patient-level characteristics including sex, histology, PET-CT, pathologic stage, change in nodal staging, extent and technique of resection varied across the 3 nodal evaluation groups (Table 1). Mediastinal lymphadenectomy and systematic sampling recipients were more likely to be female; the procedures were predominantly robotically-assisted (68/52/21%, respectively; p<0.001), whilst the ‘neither’ resections were mostly open resections (29/34/63%). Mediastinal lymphadenectomy and ‘neither’ resections predominated at teaching hospitals (81/47/53; p<0.001) and were often performed at larger hospitals (median bed size 793/359/617; p<0.001). Dedicated general thoracic surgeons performed 78% of the mediastinal nodal dissections and 43% of the systematic sampling procedures; cardiovascular surgeons performed 62% of the resections that did not meet the Z0030 standard.
Table 1.
Demographic, clinical and surgery characteristics across types of nodal evaluation.
| Characteristics* | Medistinal nodal dissection N (%) | Systematic sampling N (%) | Neither N (%) |
|---|---|---|---|
| 358 | 120 | 1464 | |
| Sex (p-value: 0.0023) | |||
| Male | 165 (46) | 55 (46) | 808 (55) |
| Female | 193 (54) | 65 (54) | 656 (45) |
| Age, years (p-value: 0.2361) | 69 (62–74, 31–89) | 68 (63–73, 22–86) | 68 (61–73, 27–93) |
| Race (p-value: 0.3775) | |||
| Caucasian | 286 (80) | 103 (86) | 1178 (80) |
| African American | 64 (18) | 16(13) | 268 (18) |
| Other/Not reported | 8 (2) | 1 (1) | 18 (1) |
| Insurance (p-value: 0.3636) | |||
| Medicare | 174 (49) | 57 (48) | 724 (49) |
| Medicaid | 52 (15) | 21 (18) | 203 (14) |
| Commercial | 128 (36) | 38 (32) | 489 (33) |
| Self-Insured/None | 4 (1) | 4 (3) | 48 (3) |
| Number of comorbidities (p-value: 0.9926) | 2 (2–3, 0–6) | 2 (2–3, 0–7) | 2 (2–3, 0–7) |
| Histology (p-value: 0.0006) | |||
| Adenocarcinoma | 236 (66) | 64 (53) | 779 (53) |
| Squamous | 97 (27) | 43 (36) | 526 (36) |
| Other | 25 (7) | 13 (11) | 159 (11) |
| PET-CT (p-value: 0.0406) | |||
| No | 64 (18) | 10 (8) | 246 (17) |
| Yes | 294 (82) | 110 (92) | 1218 (83) |
| Invasive staging (p-value: 0.3009) | |||
| No | 313 (87) | 102 (85) | 1232 (84) |
| Yes | 45 (13) | 18 (15) | 232 (16) |
| Clinical T category¶ (p-value: 0.1864) | |||
| T1a | 26 (7) | 8 (7) | 75 (5) |
| T1b | 146 (41) | 48 (40) | 544 (37) |
| T1c | 96 (27) | 23 (19) | 414 (28) |
| T2a | 69 (19) | 28 (23) | 317 (22) |
| T2b | 21 (6) | 13 (11) | 114 (8) |
| Clinical N category¶ (p-value: 0.2129) | |||
| N0 | 347 (97) | 112 (93) | 1407 (96) |
| Non-hilar N1 | 11 (3) | 8 (7) | 57 (4) |
| Pathologic T category¶ (p-value: 0.0021) | |||
| pT1a | 37 (10) | 6 (5) | 68 (5) |
| pT1b | 104 (29) | 30 (25) | 395 (27) |
| pT1c | 61 (17) | 29 (24) | 290 (20) |
| pT2a | 116 (32) | 32 (27) | 443 (30) |
| pT2b | 14 (4) | 9 (8) | 102 (7) |
| pT3 | 23 (6) | 13 (11) | 139 (9) |
| pT4 | 3 (1) | 1 (1) | 27 (2) |
| Pathologic N category¶ (p-value: 0.0002) | |||
| pNx | 0 (0) | 0 (0) | 44 (3) |
| pN0 | 333 (93) | 114 (95) | 1279 (87) |
| pN1 | 20 (6) | 6 (5) | 122 (8) |
| pN2 | 5 (1) | 0 (0) | 19 (1) |
| Clinical to pathologic nodal stage (p-value: 0.0450) | |||
| Down-stage | 10 (3) | 5 (4) | 76 (5) |
| No change | 324 (91) | 112 (93) | 1268 (87) |
| Up-stage | 24 (7) | 3 (3) | 120 (8) |
| Extent of resection (p-value: 0.0229) | |||
| Pneumonectomy | 4 (1) | 2 (2) | 42 (3) |
| Bilobectomy | 9 (3) | 5 (4) | 70 (5) |
| Lobectomy | 334 (93) | 107 (89) | 1259 (86) |
| Segmentectomy | 11 (3) | 6 (5) | 93 (6) |
| Resection technique (p-value: <0.001) | |||
| Open | 103 (29) | 41 (34) | 916 (63) |
| Robotically-assisted | 245 (68) | 62 (52) | 302 (21) |
| Video-assisted | 10 (3) | 17 (14) | 246 (17) |
| Margin status (p-value: 0.0002) | |||
| Positive | 12 (3) | 3 (3) | 43 (3) |
| Negative | 341 (95) | 117 (98) | 1394 (95) |
| Not Reported | 5 (1) | 0 (0) | 27 (2) |
P-value testing differences across three columns using Chi-squared (Fisher’s exact if small sample size) tests or Kruskal-Wallis test, not significant unless otherwise specified, PET-CT=positron emission tomography-computerized tomography;
American Joint Committee on Cancer 8th edition staging system.
Pattern of failure.
Mediastinal lymph node dissection provided a median (IQR) of 8 (6–11), systematic sampling provided 7 (4.5–10), and resections meeting neither criteria provided 3 (1–5) mediastinal lymph nodes; and 5 (5–6), 4 (4–4), and 2 (1–3) mediastinal nodal stations, respectively (p<0.001; Table A.3). Resections meeting neither quality criteria primarily missed stations 6 (81%) and 8 (78%) in left-sided resections and stations 2R (92%), 8 (75%), 9 (66%) and 4R (57%) in right-sided resections. However, 3% had no nodes (pathologic NX), 14% had no mediastinal nodes, and 42% had no hilar nodes (Table A.3).
Survival associations.
With a median follow up time among censored patients of 3.27 years and median survival time of 7.55 years, 631 patients died during follow up: 64 (18%) had mediastinal nodal dissection, 28 (23%) had systematic sampling, and 539 (37%) had neither. Institutional teaching status was the only effect-modifying covariate, resection technique the only confounder (Figure 3). We therefore present hazard ratios for each type of institution while adjusting for technique. At teaching institutions, recipients of mediastinal nodal dissection had a significantly lower risk of death than recipients of neither sampling (hazard ratio 0.57 [0.41, 0.79]; p<0.001; adjusted p=0.002); the difference between mediastinal lymphadenectomy and systematic sampling was not statistically significant. There were no significant differences between types of nodal evaluation in non-teaching institutions (Table 2). Fully adjusted models including all covariates provided similar results (Tables A.4).
Figure 3.
Schematic illustrating the hazard ratio of patients with clinical T1 or T2, N0 or non-hilar N1, M0 non-small-cell lung cancer who underwent primary surgical resection, stratified by the extent of nodal dissection based on lymph node stations in the pathology report: mediastinal lymph node dissection (stations 2R, 4R, 7, 8, 9 and 10R present in right-side resections; 5, 6, 7, 8, 9 and 10L present in left-side resections), systematic sampling (stations 2R, 4R, 7, and 10 in right side resections; 5, 6, 7, and 10L in left-side resections) or neither. Upon stratification by the sole effect-modifier, institutional teaching status, significant survival differences were found in resections performed in teaching institutions but not in non-teaching institutions.
Table 2.
Stratified hazard ratios by institution teaching status.
| Type of sampling* | Stratified Hazard Ratios† (95% Confidence Interval) | Unadjusted P-value | Adjusted P-value‡ |
|---|---|---|---|
| Among non-teaching institutions, after adjusting for resection technique | |||
| Mediastinal nodal dissection vs Neither | 1.03 (0.63, 1.69) | 0.897 | 0.9908 |
| Systematic sampling vs Neither | 1.48 (0.87, 2.52) | 0.1447 | 0.3111 |
| Mediastinal nodal dissection vs systematic sampling | 0.7 (0.35, 1.4) | 0.3094 | 0.5664 |
| Among teaching institutions, after adjusting for resection technique | |||
| Mediastinal nodal dissection vs Neither | 0.57 (0.41, 0.79) | 0.0007 | 0.0021 |
| Systematic sampling vs Neither | 0.77 (0.44, 1.35) | 0.3654 | 0.637 |
| Mediastinal nodal dissection vs systematic sampling | 0.74 (0.4, 1.37) | 0.3386 | 0.6042 |
Neither = resections that did not examine all specific lymph node stations required for mediastinal nodal dissection or systematic sampling by American College of Surgeons Oncology Group definition.
Teaching status proved to be an effect modifier and therefore provided the stratified hazard ratios, while adjusting for the only confounder, resection tehcnique.
Adjusted for pairwise comparisons using Tukey’s method.
Sensitivity analyses.
Additional analyses with a robust sandwich covariance estimator produced the same results, as did a stabilized inverse probability approach to balance the groups and the sources of potential bias across baseline covariates (Table A.5). Analyses limited to patients with cN0 provided similar results (Table A.6). At the crude level, mediastinal nodal dissection was associated with significantly lower risk of mortality compared to ‘neither’ (0.58 [0.45, 0.76]); institutional teaching status was an effect-modifier and resection technique a confounder. Among teaching institutions, only mediastinal lymphadenectomy had lower risks of mortality compared to neither (0.60 [0.43, 0.83]) after adjusting for surgeon specialty (data not shown). There were no significant survival differences in non-teaching hospitals.
Healthcare utilization and perioperative complications.
Resections with systematic sampling had a median surgery time of 160 minutes (IQR: 113–209) compared to 109 minutes (83–160) for mediastinal nodal dissection and 141 minutes (103–186) for resections meeting neither standard (p<0.001; Table 3). The ‘neither’ cohort had longer duration of chest tube drainage (p<0.001), Intensive Care Unit admission (p=0.0056), and hospital length of stay (p<0.001) but the lowest rate of respiratory failure (p<0.001). A detailed summary of perioperative outcomes stratified by teaching status is provided in Table A.7.
Table 3.
Perioperative complications among the pathologic nodal evaluation groups.
| Complications | Mediastinal lymph node dissection N (%) | Systematic sampling N (%) | Neither N (%) | P-value* |
|---|---|---|---|---|
| N | 358 | 120 | 1464 | |
| Duration of surgery, minutes† | 109 (83–160), (16–335) | 159.5 (113–209), (39–388) | 141 (103–186), (16–641) | <0.001 |
| Number with any post-operative complications, N (%) | 182 (51) | 50 (42) | 708 (48) | 0.2197 |
| Number of post-operative complications† | 1 (0–1), (0–4) | 0 (0–1), (0–4) | 0 (0–1), (0–6) | 0.2711 |
| Cardiac arrhythmias, N (%) | 53 (15) | 14 (12) | 222 (15) | 0.5845 |
| Atelectasis, N (%) | 74 (21) | 19 (16) | 307 (21) | 0.4085 |
| Pneumonia, N (%) | 10 (3) | 2 (2) | 64 (4) | 0.1634 |
| Rate of reoperation, N (%) | 0 (0) | 1 (1) | 8 (1) | 0.2711 |
| Rate of blood transfusion, N (%) | 4 (1) | 2 (2) | 39 (3) | 0.1941 |
| Empyema, N (%) | 0 (0) | 0 (0) | 3 (0) | 1.000 |
| Chylothorax, N (%) | 1 (0) | 0 (0) | 3 (0) | 0.6774 |
| Bronchopleural fistula, N (%) | 0 (0) | 0 (0) | 6 (0) | 0.7305 |
| Air leak > 7 day s (in summary), N (%) | 20 (6) | 9 (8) | 118 (8) | 0.2842 |
| Respiratory failure, N (%) | 110 (31) | 30 (25) | 283 (19) | <0.001 |
| Myocardial infarction, N (%) | 0 (0) | 0 (0) | 10 (1) | 0.3189 |
| Adult respiratory distress syndrome, N (%) | 1 (0) | 0 (0) | 20 (1) | 0.1182 |
| Recurrent laryngeal nerve injury, N (%) | 0 (0) | 0 (0) | 0 (0) | NA |
| Increased lymphatic drainage, N (%) | 0 (0) | 0 (0) | 1 (0) | 1.000 |
| ICU readmittance, N (%) | 11 (3) | 4 (3) | 68 (5) | 0.3546 |
| Hospital readmittance within 60 day, N (%) | 34 (10) | 19 (17) | 186 (13) | 0.1173 |
| Duration of chest tube drainage, days† | 3 (1–5), (0–33) | 3 (2–6), (1–38) | 4 (3–7), (0–370) | <0.001 |
| ICU length of stay, days† | 1 (1–2), (0–33) | 1 (1–2), (0–24) | 1 (1–2), (0–43) | 0.0056 |
| Hospital length of stay, days† | 4 (3–7), (1–373) | 5 (3–7), (2–43) | 6 (4–9), (0–171) | <0.001 |
| 30-day mortality, N (%) | 12 (3) | 9 (8) | 63 (4) | 0.1538 |
| 60-day mortality, N (%) | 19 (5) | 10 (8) | 90 (6) | 0.4880 |
| 90-day mortality, N (%) | 23 (6) | 12 (10) | 119 (8) | 0.3878 |
| 120-day mortality, N (%) | 32 (9) | 15 (13) | 137 (9) | 0.4907 |
P-value testing differences across three columns using Chi-squared (Fisher’s exact if expected cell counts were less than 5) tests or Kruskal-Wallis test.
Median (interquartile range), (range).
Discussion
In this population-based evaluation of the pattern and outcomes of pathologic nodal staging practice in a high lung cancer incidence and mortality region of the United States, we found that although mediastinal lymphadenectomy and systematic sampling were not associated with greater perioperative morbidity or mortality, they were relatively infrequently performed in resections for early-stage NSCLC: 75% did not meet the ACOSOG definition of systematic sampling. Stations 6, 7 and 8 were unexamined in 81%, 69% and 78%, respectively, in left-side resections and 2R, 8, 9, 4R and 7 were most often ignored (in 92%, 75%, 66%, 57%, and 51%, respectively) in right-side resections. Moreover, 12% of the whole analytic cohort had extremes of poor nodal evaluation, including 2% without any nodal evaluation and 10% without any mediastinal nodal evaluation.
It is unclear from these data why prevailing practice falls so far short of the existing evidence. Although the final results of ACOSOG Z0030 were not published until 20112 and our cohort spans the years 2009 to 2019, the initial safety results were published in 2006.16 In the MS-QSR, there was no difference in 30-day postoperative mortality between the mediastinal nodal dissection, systematic sampling and ‘neither’ cohorts (3% v 8% v 4%). Although the rates seem higher than the 0.76% for mediastinal nodal dissection and 2% for systematic sampling reported in Z0030,16 this probably reflects inherent biases such as the ‘healthy-volunteer effect’ in clinical trial populations. Postoperative survival statistics in the population-based MS-QSR cohort are also probably more reflective of ‘real-world statistics’ than reports from non-population-based datasets such as the Society of Thoracic Surgeons database, and the National Cancer Database. We speculate that erroneous belief that Z0030 was a ‘negative trial’ might have inadvertently devalued thorough pathologic nodal evaluation among community surgeons.
Institutional teaching status had a significant impact on the association between the type of nodal evaluation and survival. Recipients of mediastinal nodal dissection in teaching hospitals had significantly better survival than those whose surgery failed to meet systematic sampling criteria. Although, similar to Z0030, there was no significant difference in survival between mediastinal nodal dissection and systematic sampling, mediastinal lymphadenectomy was associated with a lower hazard for death in teaching hospitals (0.74 [CI 0.40–1.37]), raising the possibility that this difference might be significant if sustained in a larger systematic sampling cohort. Surprisingly, we found no difference in survival between the systematic sampling and ‘neither’ cohorts. Furthermore, there was no difference in outcomes associated with any of the types of nodal evaluation in patients who had surgery in non-teaching hospitals. These findings suggest a need for a more nuanced interpretation of the Z0030 results. There is precedence for this suggestion: post-hoc re-examination of the mediastinal nodal dissection arm of Z0030 from a pathologic evaluation quality perspective suggested that the thoroughness of pathologic evaluation, defined by the number of lymph nodes examined, had significant survival impact.17
Most operations in teaching hospitals were performed by general thoracic surgeons, whereas most resections in non-teaching hospitals were performed by cardiovascular or general surgeons. Indeed, all resections by general surgeons were performed in non-teaching hospitals. General thoracic surgeons have superior outcomes to cardiovascular surgeons;18 and general surgeons’ lung resection outcomes are significantly worse than those of board-certified cardiovascular surgeons.19,20 Proficiency in pathologic nodal staging possibly drives some of these survival differences. Disparities in proficiency are suggested by the paradoxical differences between node evaluation cohorts in the duration of surgery and short-term markers of morbidity such as duration of chest tube drainage and Intensive Care Unit admission.
ACOSOG Z0030 will never be re-enacted, given the enormous logistic difficulty of executing such a trial. The published results were unequivocal- mediastinal lymph node dissection did not provide additional survival benefit beyond a thoroughly-conducted systematic nodal sampling procedure.2,3 Nevertheless, the debate about the equivalence of these two procedures continues.4–6 Z0030 was a very ‘explanatory’ (or ‘non-pragmatic’) trial, on the ‘Pragmatic Explanatory Continuum Indicator Summary’ scale.21 The external validity of such highly explanatory trials to real-world populations and clinical practice is increasingly questioned. Designed to demonstrate, once and for all, the benefit of mediastinal nodal dissection, multiple aspects of the trial such as patient-, surgeon- and institution-level eligibility, inflexibility in intervention delivery and adherence, and the exclusion of 13.9% of randomized patients from the primary outcomes analysis, paradoxically open up the possibility that the trial results might not fully apply in ‘real-world’ experience.2,21
Limitations.
Our effort to re-enact the Z0030 trial using observational data has obvious inherent limitations, including greater susceptibility to bias than a randomized clinical trial. For example, it is likely that differences in surgeon (and lung cancer care delivery team) proficiency drove some of the differences we found. We eliminated patients with metastasis to nodal stations that would have been detected at systematic sampling to pattern after the Z0030 approach in which eligibility was determined intraoperatively after confirmation of the absence of hilar and mediastinal node metastasis after systematic sampling. Similar to Z0030, we have not accounted for the use of adjuvant chemotherapy, which may have a modest impact on long-term survival.
We used the 8th edition of the American Joint Committee on Cancer staging system, rather than the 5th edition in use during the Z0030 patient accrual window. By thus eliminating patients with tumors greater than 5cm, we have selected a lower-risk population, which could bias our analysis towards the null. Our findings remained consistent when the cohort was further restricted to the ostensibly even lower-risk subset with cN0 disease. Our criteria for categorizing mediastinal nodal dissection and systematic sampling on the basis of minimum lymph node stations identified, although pragmatic, are not as clearly demarcated as in the Z0030 trial. Our systematic sampling cohort sometimes included more than the stipulated stations in Z0030, which would also tend to bias toward the null. Our definition of mediastinal nodal dissection, based on identification of lymph nodes from specific anatomic stations in the pathology report rather than the surgeon’s description of the lymphadenectomy procedure, may not truly reflect the stringency of definition applied in Z0030. This may be suggested by the retrieval of a median of 11 to 12 mediastinal lymph nodes in the nodal dissection arm of Z0030, compared to 7 in our cohort.11 Furthermore, quality measures based on lymph node counts can be confounded by the manner of handling fragmented nodes. Discordance between surgeons and pathologists in identifying the lymphadenectomy procedure performed may be caused by variation in thoroughness of pathologic evaluation or accuracy of pathology reporting, in addition to surgical practice. We have previously shown the existence of such discordance between surgeons and pathologists.12,13 Although our findings are not based on lymph node counts, but rather the anatomic provenance of examined lymph nodes which is not subject to the fragmentation problem, they remain susceptible to variation in pathology practices.22 Finally, we are currently unable to evaluate the cause of death in this cohort, although this was also not provided in Z0030.
Strengths of our study.
This is the first population-based examination of ACOSOG Z0030 findings. The MS-QSR database is a unique, population-based surgical resection dataset. It includes information from 12 hospitals, seven different healthcare systems across 3 states (Arkansas, Mississippi and Tennessee), with a catchment area including contiguous counties in 6 states (including counties in Southwestern Kentucky, Western Alabama and Southeastern Missouri). Supported by the National Cancer Institute, the database provides ‘real-world data’ to facilitate close contemporary examination of surgical lung cancer care at the heart of the United States lung cancer mortality belt. The MS-QSR contains more anatomic lymph node details than can be found in any national-level datasets, including the Surveillance, Epidemiology and End Results database, the National Cancer Data Base and the General Thoracic Surgery database of the Society of Thoracic Surgeons, the latter two of which are not population-based. The lymph node station-specific analysis we report in this paper cannot be re-enacted in any of these datasets.
Furthermore, we include a more representative range of institutions and surgeons than the Z0030, which was limited to general thoracic surgeons and mostly academic institutions. We included another control population- those (the vast majority) who did not attain the systematic sampling control of Z0030. This is important because of widespread de-facto mis-interpretation of Z0030 as having indicated that thorough pathologic nodal staging is unnecessary. Although we corroborate Z0030 survival findings in a ‘real-world’ population and confirm that systematic nodal dissection is comparatively safe in these settings, we also demonstrate the significant under-utilization of systematic nodal dissection in these settings. Revelation of institutional teaching status as an effect-modifier is important. In future analyses, we plan to further investigate the failure to identify a survival benefit for systematic nodal dissection in nonacademic institutions, including the generalizability of these findings in other datasets.
There is good evidence that the nodal staging quality gap exists across the Americas, Australia, Asia and Europe.23–28 Coupled with evidence in support of corrective interventions such as timely, confidential feedback on institutional performance,29 use of pre-labeled lymph node specimen collection kits,30,31 and improved gross dissection of lung resection specimens,32,33 our findings should be of great interest to the relevant guideline-making surgery organizations, healthcare policymakers and institutional administrators. Reflecting the potential policy-level relevance, the American College of Surgeons Commission on Cancer 2020 phase-in standard for pulmonary resection now states: ‘The surgical pathology report following any curative intent pulmonary resection for primary lung malignancy must contain lymph nodes from at least one (named and/or numbered) hilar station and at least 3 distinct (named and/or numbered) mediastinal stations.’34
Conclusion.
In this pragmatic population-based cohort, majority of early-stage NSCLC resections did not apply the minimum standards of pathologic nodal evaluation defined by the ACOSOG. In teaching institutions, survival of patients who underwent mediastinal lymphadenectomy was superior to that of patients who had lesser dissections. It is not clear why there was no difference in the survival of patients who received surgery in non-teaching facilities. The overwhelming impression from our analysis however, is the great need to improve the quality of pathologic nodal staging during curative-intent resection of early-stage NSCLC.
Supplementary Material
Acknowledgements.
We thank members of the Thoracic Oncology Research Group over the years, as well as the surgeons, hospital administrators, cardiovascular surgery operating room team members, pathologists, pathology assistants and Tumor Registry data managers at all Mid-South Quality of Surgical Resection Cohort institutions over the years.
Research Support: This work was supported by the National Institutes of Health [grant numbers 1R01CA172253 and 2R01CA172253]; the Mid-South Quality of Surgical Resection database was initially approved by the Baptist Memorial Healthcare Corporation Institutional Review Board on March 8, 2013 (IRB #13-06).
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflict of Interest Statement: Patents for surgical specimen collection kit; stocks in Eli Lilly, Gilead Sciences Inc, and Pfizer; paid research consultant for the American Cancer Society, the Association of Community Cancer Centers, and AstraZeneca; founder Oncobox Device, Inc (Osarogiagbon). Paid research consultant for the Association of Community Cancer Centers (Smeltzer).
References.
- 1.Smeltzer MP, Faris NR, Ray MA, Osarogiagbon RU. Association of Pathologic Nodal Staging Quality With Survival Among Patients With Non-Small Cell Lung Cancer After Resection With Curative Intent. JAMA Oncol. 2018. January 1;4(1):80–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Darling GE, Allen MS, Decker PA, et al. Randomized trial of mediastinal lymph node sampling versus complete lymphadenectomy during pulmonary resection in the patient with N0 or N1 (less than hilar) non-small cell carcinoma: results of the American College of Surgery Oncology Group Z0030 trial. J Thorac Cardiovasc Surg 2011;141:662–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Murthy SC. Less is more…(more or less…). J Thorac Cardiovasc Surg 2011;141:670–72. [DOI] [PubMed] [Google Scholar]
- 4.Osarogiagbon RU, Darling GE. Towards optimal pathologic staging of resectable non-small cell lung cancer. Transl Lung Cancer Res. 2013. October;2(5):364–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Mokhles S, Macbeth F, Treasure T, et al. Systematic lymphadenectomy versus sampling of ipsilateral mediastinal lymph-nodes during lobectomy for non-small-cell lung cancer: a systematic review of randomized trials and a meta-analysis. Eur J Cardiothorac Surg. 2017. June 1;51(6):1149–1156. doi: 10.1093/ejcts/ezw439. Erratum in: Eur J Cardiothorac Surg. 2018 Oct 1;54(4):795. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Osarogiagbon RU, Veronesi G, Fang W, et al. Early-Stage NSCLC: Advances in Thoracic Oncology 2018. J Thorac Oncol. 2019. June;14(6):968–978. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Faris NR, Smeltzer MP, Lu F, et al. Evolution in the Surgical Care of Patients With Non-Small Cell Lung Cancer in the Mid-South Quality of Surgical Resection Cohort. Semin Thorac Cardiovasc Surg. 2017. Spring;29(1):91–101. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Smeltzer MP, Faris NR, Ray MA, et al. Survival Before and After Direct Surgical Quality Feedback in a Population-Based Lung Cancer Cohort. Ann Thorac Surg. 2019. May;107(5):1487–1493. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Osarogiagbon RU, Lee YS, Faris NR, Ray MA, Ojeabulu PO, Smeltzer MP. Invasive mediastinal staging for resected non-small cell lung cancer in a population-based cohort. J Thorac Cardiovasc Surg. 2019;158:1220–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Goldstraw P, Chansky K, Crowley J, Rami-Porta R, et al. The IASLC Lung Cancer Staging Project: Proposals for Revision of the TNM Stage Groupings in the Forthcoming (Eighth) Edition of the TNM Classification for Lung Cancer. J Thorac Oncol. 2016. January;11(1):39–51. [DOI] [PubMed] [Google Scholar]
- 11.Darling GE, Allen MS, Decker PA, et al. Number of lymph nodes harvested from a mediastinal lymphadenectomy. CHEST 2011;139:1124–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Osarogiagbon RU, Allen JW, Farooq A, Wu JT. Objective review of mediastinal lymph node examination in a lung cancer resection cohort. J Thorac Oncol. 2012. February;7(2):390–6. [DOI] [PubMed] [Google Scholar]
- 13.Detterbeck FC. The fable of Babel and building a foundation for quality. J Thorac Oncol. 2012. February;7(2):267–8. [DOI] [PubMed] [Google Scholar]
- 14. Available at: https://www.ers.usda.gov/data-products/rural-urban-commuting-area-codes.aspx.
- 15.Kramer CY. Extension of multiple range tests to group means with unequal numbers of replications. Biometrics. 1956;12(3):307–310 [Google Scholar]
- 16.Allen MS, Darling GE, Pechet TTV, et al. Morbidity and mortality of major pulmonary resections in patients with early-stage lung cancer: initial results of the randomized, prospective ACOSOG Z0030 trial. Ann Thorac Surg 2006;81:1013–20. [DOI] [PubMed] [Google Scholar]
- 17.Osarogiagbon RU, Decker PA, Ballman K, Wigle D, Allen MS, Darling GE. Survival Implications of Variation in the Thoroughness of Pathologic Lymph Node Examination in American College of Surgeons Oncology Group Z0030 (Alliance). Ann Thorac Surg. 2016. August;102(2):363–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Farjah F, Flum DR, Varghese TK, Symons RG, Wood DE. Surgeon specialty and long-term survival after pulmonary resection for lung cancer. Ann Thorac Surg 2009;87:995–1006. [DOI] [PubMed] [Google Scholar]
- 19.Silvestri GA, Handy J, Lackland D, Corley E, Reed CE. Specialists achieve better outcomes than generalists for lung cancer surgery. CHEST 1998;114:675–680. [DOI] [PubMed] [Google Scholar]
- 20.Schipper PH, Diggs BS, Ungerleider RM, Welke KF. The influence of surgeon specialty on outcomes in general thoracic surgery: a national sample 1996 to 2005. Ann Thorac Surg 2009;88:1566–1573. [DOI] [PubMed] [Google Scholar]
- 21.Loudon K, Treweek S, Sullivan F, Donnan P, Thorpe KE, Zwarenstein M. The PRECIS-2 tool: designing trials that are fit for purpose. BMJ. 2015. May 8;350:h2147. [DOI] [PubMed] [Google Scholar]
- 22.Butnor KJ, Asamura H, Travis WD. Node Doubt: Rigorous Surgical Nodal Procurement Combined With Thorough Pathologic Evaluation Improves Non-Small Cell Lung Carcinoma Staging Accuracy. Ann Thorac Surg. 2016;102(2):353–356. [DOI] [PubMed] [Google Scholar]
- 23.Varlotto JM, Recht A, Nikolov M, Flickinger JC, DeCamp MM. Extent of lymphadenectomy and outcome for patients with stage I nonsmall cell lung cancer. Cancer 2009;115:851–8. [DOI] [PubMed] [Google Scholar]
- 24.Verhagen AF, Schoenmakers MC, Barendregt W, et al. Completeness of lung cancer surgery: Is mediastinal dissection common practice? Eur J Cardiothorac Surg. 2012;41(4):834–838. [DOI] [PubMed] [Google Scholar]
- 25.Van Schil PE. Lymph node dissection for lung cancer: Does Europe behave better? J Thorac Oncol 2013. March;8(3):e29. [DOI] [PubMed] [Google Scholar]
- 26.Osarogiagbon RU, Smeltzer MP, Faris N, Rami-Porta R, Goldstraw P, Asamura H. Comment on the Proposals for the Revision of the N Descriptors in the Forthcoming Eighth Edition of the TNM Classification for Lung Cancer. J Thorac Oncol. 2016. October;11(10):1612–4. [DOI] [PubMed] [Google Scholar]
- 27.Heineman DJ, Ten Berge MG, Daniels JM, Versteegh MI, Marang-van de Mheen PJ, Wouters MW, Schreurs WH. The Quality of Staging Non-Small Cell Lung Cancer in the Netherlands: Data From the Dutch Lung Surgery Audit. Ann Thorac Surg. 2016. November; 102(5):1622–1629. [DOI] [PubMed] [Google Scholar]
- 28.Liang W, He J, Shen Y, et al. Impact of Examined Lymph Node Count on Precise Staging and Long-Term Survival of Resected Non-Small-Cell Lung Cancer: A Population Study of the US SEER Database and a Chinese Multi-Institutional Registry. J Clin Oncol. 2017. April 10;35(11):1162–1170. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Smeltzer MP, Faris NR, Ray MA, et al. Survival Before and After Direct Surgical Quality Feedback in a Population-Based Lung Cancer Cohort. Ann Thorac Surg. 2019. May;107(5):1487–1493. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Osarogiagbon RU, Sareen S, Eke R, et al. Audit of lymphadenectomy in lung cancer resections using a specimen collection kit and checklist. Ann Thorac Surg. 2015. February;99(2):421–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Osarogiagbon RU, Miller LE, Ramirez RA, et al. Use of a surgical specimen-collection kit to improve mediastinal lymph-node examination of resectable lung cancer. J Thorac Oncol. 2012. August;7(8):1276–82. [DOI] [PubMed] [Google Scholar]
- 32.Osarogiagbon RU, Eke R, Sareen S, et al. The impact of a novel lung gross dissection protocol on intrapulmonary lymph node retrieval from lung cancer resection specimens. Ann Diagn Pathol. 2014. August;18(4):220–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Ray MA, Faris NR, Smeltzer MP, et al. Effectiveness of Implemented Interventions on Pathologic Nodal Staging of Non-Small Cell Lung Cancer. Ann Thorac Surg. 2018. July;106(1):228–234. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Commission on Cancer. Optimal Resources for Cancer Care. Patient care: Expectations and Protocols. 5.8 Pulmonary resection. Page 65 Accessed on 04/18/20; available at: https://www.facs.org/-/media/files/quality-programs/cancer/coc/optimal_resources_for_cancer_care_2020_standards.ashx
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.