Advertisement

Time to treatment as a quality metric in lung cancer: Staging studies, time to treatment, and patient survival

      Abstract

      Purpose

      Prompt staging and treatment are crucial for non-small cell lung cancer (NSCLC). We determined if predictors of treatment delay after diagnosis were associated with prognosis.

      Materials and methods

      Medicare claims from 28,732 patients diagnosed with NSCLC in 2004–2007 were used to establish the diagnosis-to-treatment interval (ideally ⩽35 days) and identify staging studies during that interval. Factors associated with delay were identified with multivariate logistic regression, and associations between delay and survival by stage were tested with Cox proportional hazard regression.

      Results

      Median diagnosis-to-treatment interval was 27 days. Receipt of PET was associated with delays (57.4% of patients with PET delayed [n = 6646/11,583] versus 22.8% of those without [n = 3908/17,149]; adjusted OR = 4.48, 95% CI 4.23–4.74, p< 0.001). Median diagnosis-to-PET interval was 15 days; PET-to-clinic, 5 days; and clinic-to-treatment, 12 days. Diagnosis-to-treatment intervals <35 days were associated with improved survival for patients with localized disease and those with distant disease surviving ⩾1 year but not for patients with distant disease surviving <1 year.

      Conclusion

      Delays between diagnosing and treating NSCLC are common and associated with use of PET for staging. Reducing time to treatment may improve survival for patients with manageable disease at diagnosis.

      Keywords

      Non-small cell lung cancer (NSCLC) grows rapidly, and delays in initiating treatment can result in disease progression and death [
      • Everitt S.
      • Herschtal A.
      • Callahan J.
      • Plumridge N.
      • Ball D.
      • Kron T.
      • et al.
      High rates of tumor growth and disease progression detected on serial pretreatment fluorodeoxyglucose-positron emission tomography/computed tomography scans in radical radiotherapy candidates with nonsmall cell lung cancer.
      ,
      • Mohammed N.
      • Kestin L.L.
      • Grills I.S.
      • Battu M.
      • Fitch D.L.
      • Wong C.Y.
      • et al.
      Rapid disease progression with delay in treatment of non-small-cell lung cancer.
      ,
      • Everitt S.
      • Plumridge N.
      • Herschtal A.
      • Bressel M.
      • Ball D.
      • Callahan J.
      • et al.
      The impact of time between staging PET/CT and definitive chemo-radiation on target volumes and survival in patients with non-small cell lung cancer.
      ]. In one study, delays of >8 weeks from initial diagnosis to treatment led to disease progression in 31% of patients and new metastases in 13% [
      • Mohammed N.
      • Kestin L.L.
      • Grills I.S.
      • Battu M.
      • Fitch D.L.
      • Wong C.Y.
      • et al.
      Rapid disease progression with delay in treatment of non-small-cell lung cancer.
      ]. In addition, studies of PET early in treatment for locoregionally confined NSCLC have shown that these changes in PET are correlated with overall survival in this setting [
      • van Loon J.
      • Offermann C.
      • Ollers M.
      • van Elmpt W.
      • Vegt E.
      • Rahmy A.
      • et al.
      Early CT and FDG-metabolic tumour volume changes show a significant correlation with survival in stage I-III small cell lung cancer: a hypothesis generating study.
      ]. Thus consensus panels have recommended that treatment be initiated in a timely manner, defined as within 35 days of pulmonary consultation [
      • Hermens R.P.
      • Ouwens M.M.
      • Vonk-Okhuijsen S.Y.
      • van der Wel Y.
      • Tjan-Heijnen V.C.
      • van den Broek L.D.
      • et al.
      Development of quality indicators for diagnosis and treatment of patients with non-small cell lung cancer: a first step toward implementing a multidisciplinary, evidence-based guideline.
      ,
      • Lennes I.T.
      • Lynch T.J.
      Quality indicators in cancer care: development and implementation for improved health outcomes in non-small-cell lung cancer.
      ,
      • Ouwens M.
      • Hermens R.
      • Hulscher M.
      • Vonk-Okhuijsen S.
      • Tjan-Heijnen V.
      • Termeer R.
      • et al.
      Development of indicators for patient-centred cancer care.
      ,
      • Ouwens M.M.
      • Hermens R.R.
      • Termeer R.A.
      • Vonk-Okhuijsen S.Y.
      • Tjan-Heijnen V.C.
      • Verhagen A.F.
      • et al.
      Quality of integrated care for patients with nonsmall cell lung cancer: variations and determinants of care.
      ]. However, the prevalence, impact, and factors contributing to such delays remain unknown.
      Accordingly, the purpose of this population-based study was threefold. First, we determined the prevalence of treatment delay in a large cohort of patients aged ⩾66 years with NSCLC diagnosed in 2004–2007. Second, we assessed patient, disease, and physician supply components that contributed to treatment delay, and determined the effect of delay on survival in specific stage groups. Finally, we derived discrete benchmarks for timeliness of staging studies that, if implemented, could significantly reduce such delays.

      Materials and methods

      This study was granted exempt status by The University of Texas MD Anderson Cancer Center’s institutional review board. Patients were selected from the Surveillance, Epidemiology, and End Results (SEER)-Medicare and Texas Cancer Registry (TCR)-Medicare databases, which collectively report data on incident malignancies diagnosed in patients residing in 17 geographic catchments representing approximately 34% of the US population. The patient population consisted of 28,732 patients and is further detailed in the Supplementary methods and in Supplementary Table S1.

      Defining diagnosis, staging, and treatment interventions

      The distinction between diagnosis, staging, and treatment interventions is further described in the Supplementary methods and in Supplementary Table S2. The diagnosis date was extracted from the cancer registry, either the Texas Cancer Registry or SEER, depending on the specific database. Staging studies were defined as positron emission tomography (PET), brain imaging (magnetic resonance imaging [MRI] or head computed tomography [CT]), mediastinal evaluation (staging mediastinoscopy or staging bronchoscopy), or bone scan performed at any time between the date of diagnosis and the date of treatment.

      Guidelines for timeliness of care

      Timely care, or “adherence,” was defined as a diagnosis-to-treatment interval of ⩽35 days, and treatment delay was defined as a diagnosis-to-treatment interval of >35 days. This definition was derived from a proposed quality measure that states that therapy should be started within 35 calendar days from the patient’s first visit to the pulmonologist. This quality measure was evaluated in prior studies of relatively small cohorts [
      • Hermens R.P.
      • Ouwens M.M.
      • Vonk-Okhuijsen S.Y.
      • van der Wel Y.
      • Tjan-Heijnen V.C.
      • van den Broek L.D.
      • et al.
      Development of quality indicators for diagnosis and treatment of patients with non-small cell lung cancer: a first step toward implementing a multidisciplinary, evidence-based guideline.
      ,
      • Ouwens M.
      • Hermens R.
      • Hulscher M.
      • Vonk-Okhuijsen S.
      • Tjan-Heijnen V.
      • Termeer R.
      • et al.
      Development of indicators for patient-centred cancer care.
      ,
      • Ouwens M.M.
      • Hermens R.R.
      • Termeer R.A.
      • Vonk-Okhuijsen S.Y.
      • Tjan-Heijnen V.C.
      • Verhagen A.F.
      • et al.
      Quality of integrated care for patients with nonsmall cell lung cancer: variations and determinants of care.
      ] and proposed as a relevant metric for evaluating care quality in the United States [
      • Lennes I.T.
      • Lynch T.J.
      Quality indicators in cancer care: development and implementation for improved health outcomes in non-small-cell lung cancer.
      ]. However, because only 44.7% of patients in our cohort were seen by a pulmonologist within 3 months before diagnosis, we chose to evaluate time from diagnosis to treatment, rather than time from pulmonary consultation to treatment.

      Statistical methods

      Logistic regression was conducted to examine the potential effect of the signal factor on the likelihood of initiating treatment(s) in 35 days. Unadjusted odds ratios (single covariate) were estimated along with the Wald statistics test for each category in comparison to the reference of the factor (Table 1). The average and median time from diagnosis to treatment were reported. Due to the non-normality nature of the time from diagnosis to treatment, the Wilcoxon rank sum test was used to evaluate differences in time between categories of each factor (Table 1). All p-values were 2-sided, and a threshold of 0.05 was used to determine significance.
      Table 1Delay in treatment by baseline characteristics and staging studies.
      CharacteristicsNo. of patientsDelay (>35 days)Length of delay
      %%OR95% CIp-ValueMeanMedianQ1Q3p-Value
      Wilcoxon rank sum test.
      Age at diagnosis, years
       66–69647022.5233.230.3251043
       70–74843429.3536.71.16(1.09, 1.25)<.00132.1271147<.001
       75–79760826.4837.31.20(1.12, 1.28)<.00132.6271148<.001
       80–84457715.9340.21.35(1.25, 1.46)<.00134.3281250<.001
       85+16435.7238.51.26(1.12, 1.41)<.00134.0281149<.001
      Gender
       Male1466451.0436.332.2261147
       Female1406848.9637.21.04(0.99, 1.09)0.10932.42711480.742
      Ethnicity
       White non-Hispanic23,98183.4636.632.2271147
       Black non-Hispanic20277.0538.41.08(0.99, 1.19)0.09533.12710500.255
       Hispanic15315.3336.91.02(0.91, 1.13)0.79032.32710480.805
       Other11934.1537.01.02(0.90, 1.15)0.78032.32710470.714
      Edu. (% <12-yr quartiles)
       Highest quartile680223.6737.532.7271148
       3rd quartile681223.7137.10.98(0.92, 1.05)0.99033.12712480.582
       2nd quartile684823.8337.10.98(0.92, 1.05)0.47832.32711470.217
       Lowest quartile (highest edu.)684323.8235.20.91(0.85, 0.97)0.00631.3261045<.001
       Unknown14274.9737.41.00(0.88, 1.12)0.96431.52610460.249
      Income(quartiles)
       Lowest quartile701824.4336.532.2261147
       2nd quartile703324.4836.71.01(0.94, 1.08)0.11032.62711480.343
       3rd quartile703724.4937.81.06(0.99, 1.13)0.06432.72711480.362
       Highest quartile704524.5235.80.97(0.91, 1.04)0.87231.62610460.120
       Unknown5992.0838.21.08(0.91, 1.28)0.26633.52913460.076
      SEER registry region
       Texas554819.3134.631.0261044
       West/Hawaii914731.8439.21.22(1.14, 1.31)<.00133.8281249<.001
       Northeast552519.2338.81.20(1.11, 1.30)<.00132.8279490.030
       Midwest325211.3235.91.06(0.97, 1.16)0.19832.02611460.127
       South526018.3133.00.93(0.86, 1.01)0.08930.62410440.589
      Charlson score
       010,53336.6635.031.5261245
       1969033.7336.61.07(1.01, 1.14)0.01632.12711470.378
       2+810628.2139.61.22(1.15, 1.30)<.00133.9281050<.001
       UNK4031.4026.80.68(0.54, 0.85)<.00125.519738<.001
      Treatment received
       Surg12,36043.0237.030.426048
       RT725625.2532.50.82(0.77, 0.87)<.00130.421944<.001
       Chemo554119.2942.61.26(1.18, 1.35)<.00138.4312050<.001
       ChemoRT357512.4435.60.94(0.87, 1.02)0.12033.3271444<.001
      Stage
       Distant11,81041.1030.729.322941
       Localized796027.7041.61.61(1.52, 1.71)<.00134.0291051<.001
       Regional896231.1940.41.53(1.44, 1.62)<.00134.7291450<.001
      Diagnostic facility
       Freestanding22,53578.4337.432.9271148
       Hospital-based619721.5734.20.87(0.82, 0.92)<.00130.125844<.001
      Radiation oncologist density
       1st (lowest) quartile735325.5936.032.3271246
       2nd quartile700524.3837.81.08(1.01, 1.15)0.07732.32710480.520
       3rd quartile722425.1435.30.97(0.91, 1.04)0.61231.6269470.002
       4th (highes t) quartile715024.8937.71.08(1.01, 1.15)0.01433.02711490.356
      Surgeon density
       1st (lowest) quartile719625.0537.432.6271247
       2nd quartile732125.4835.90.94(0.88, 1.00)0.21431.82610460.034
       3rd quartile705524.5537.00.98(0.92, 1.05)0.93132.12610480.039
       4th (highest) quartile716024.9236.50.96(0.90, 1.03)0.43132.62711480.548
      PET scan(Y/N)
       No17,14959.6922.822.915334
       Yes11,58340.3157.44.56(4.33, 4.80)<.00146.2402662<.001
      MRI Brain(Y/N)
       No23,60782.1633.730.124844
       Yes512517.8450.82.03(1.91, 2.16)<.00142.2362257<.001
      Bronchoscopy(Y/N)
       No23,96483.4133.029.824943
       Yes476816.5955.42.52(2.37, 2.69)<.00144.8402163<.001
      CT Head(Y/N)
       No24,94686.8234.730.825945
       Yes378613.1850.01.88(1.75, 2.01)<.00142.4352158<.001
      Mediastinoscopy(Y/N)
       No27,31395.0635.031.3261045
       Yes14194.9469.74.27(3.80, 4.79)<.00151.9483170<.001
      Bone scan
       No23,79482.8134.530.625845
       Yes493817.1947.41.71(1.61, 1.82)<.00140.7342056<.001
      Abbreviations: SEER, Surveillance Epidemiology and End Results; PET, positron emission tomography; MRI, magnetic resonance imaging; CT, computed tomography.
      Wilcoxon rank sum test.
      To assess the consistency of the effect of a particular staging study across other factors (treatment year, stage, initial treatment, and total number of staging studies [PET, mediastinoscopy/bronchoscopy, brain MRI/CT, and bone scan]), we used an analysis of variance, using first a parametric test treating delay as a continuous variable and comparing the mean length of delay between patients who did vs. did not receive PET, and next a nonparametric test comparing the median delays among patients who did vs. did not receive PET.
      Initial prognostic parameters for the statistical model were selected based on the clinical judgment of the authors and prior data supporting these variables as significant in impacting outcomes in lung cancer. Then, multivariate logistic regression was used to evaluate associations of staging studies and other covariates with treatment delay. Stepwise selection was used to select variables with p values ⩽0.1 for entry and ⩽0.05 for remaining in the model. Due to the large study sample, both backward and forward stepwise selection result in the same set of predictors on multivariate analysis. Bootstrap validation addressed concerns of a substantial decrease in the predictive ability of the model through data-driven model building procedures (such as stepwise selection). Brier score was calculated for validation, and an overfitting corrected R-squared value was used to address the possibility of overfitting. The apparent model fit was assessed with the Hosmer–Lemeshow goodness-of-fit test, Pearson’s correlation tests, and AUC.

      Assessing the effect of delay on survival outcomes

      To determine the correlation of delay with survival, we used the Kaplan–Meier method and log-rank tests to determine how overall survival varied with stage and adherence. This approach is detailed in Supplementary methods. Hazard ratios (HRs) and 95% confidence intervals were estimated with the Cox regression model, with time dependent covariates. Non-proportionality was detected graphically, and time-dependent effects of independent variables were added to the model when violation of the proportional hazards was detected. Separate models were built for localized, regional, and distant disease. The models were then adjusted for multiple covariates, as outlined in the Supplementary methods.

      Assessing approaches to improve adherence to timeliness of care

      We used the results from the adjusted and unadjusted analyses above to determine clinically relevant benchmarks for three distinct intervals: Interval 1, time from diagnosis to PET; Interval 2, time from PET to post-PET clinic visit with a physician; and Interval 3, time from post-PET clinic visit to treatment. We characterized the time distribution of each interval and then altered the intervals to determine the effect on delay if the upper bound of the interquartile range for each interval was lowered to a clinically achievable, prespecified threshold.

      Results

      Of 28,732 patients, 27.7% had local, 31.2% regional, and 41.1% distant disease. Other patient characteristics are listed in Table 1. The incidence of PET according to SEER stage was 38.9% for those with localized disease (n = 3069/7960), 46.4% for regional (n = 4158/8962), and 36.9% for distant (n = 4356/11,810). The median time from diagnosis to treatment was 27 days, and 36.7% of patients (n = 10,554) experienced delay between diagnosis and treatment. Both staging studies and other study covariates were associated with time from diagnosis to treatment and with delay (Table 1). PET was particularly associated with delay, as 42.6% of patients undergoing PET were treated within 35 days of diagnosis, versus 77.2% of patients who did not (p< 0.001). The association of PET with delay was consistent regardless of treatment year, disease stage, number of other staging studies, and treatment received (p< 0.001) for each year, stage, treatment, and staging study in both parametric and non-parametric tests (Supplementary Fig. S1).
      In adjusted analysis for the outcome of treatment delay, receipt of PET demonstrated the largest effect size (odds ratio [OR] 4.48, 95% confidence interval [CI] 4.23–4.74, p< 0.001). Each additional staging study was also associated with increased odds of delay (OR range 1.34–2.35, p< 0.001 for all staging studies) (Table 2). Other factors associated with increased delay, including chemoradiation, higher comorbidity score, advanced age, and race are detailed in Table 2.
      Table 2Multivariate logistic regression analysis to identify factors associated with treatment delay.
      FactorOR95% CIp-Value
      PET Scan
       Yes vs. No4.484.234.74<.0001
      Bronchoscopy/Mediastinoscopy
       Yes vs. No2.352.192.52<.0001
      CT/MRI Head
       Yes vs. No1.581.481.68<.0001
      Bone Scan
       Yes vs. No1.341.241.44<.0001
      Treatment
       Chemotherapy vs. Surgery0.980.901.060.6169
       Chemoradiation vs. Surgery0.610.550.67<.0001
       Radiation Therapy vs. Surgery0.720.670.78<.0001
       Length of stay (days)1.051.041.05<.0001
      Stage
       Localized vs. Distant1.991.842.15<.0001
       Regional vs. Distant1.521.421.63<.0001
      Charlson comorbidity score
       1 vs. 01.020.961.090.5552
       2+ vs. 01.131.051.210.0006
       UNK vs. 00.940.741.210.6479
      Location
       Hospital-based vs. Freestanding1.101.021.180.0104
      Area
       Midwest vs. Texas0.840.750.940.0033
       Northeast vs. Texas1.060.961.180.2633
       South vs. Texas0.750.680.82<.0001
       West/Hawaii vs. Texas1.161.061.270.0008
      Radiation oncologist density
       2nd quartile vs. Lowest quartile1.020.931.110.7130
       3rd quartile vs. Lowest quartile0.980.891.070.6137
       Highest quartile vs. Lowest quartile1.101.001.220.0555
      Surgeon density
       2nd quartile vs. Lowest quartile0.880.800.960.0026
       3rd quartile vs. Lowest quartile0.920.841.010.0659
       Highest quartile vs. Lowest quartile0.990.891.100.8831
      Age, years
       70–74 vs. 66–691.151.071.240.0002
       75–79 vs. 66–691.231.141.33<.0001
       80–84 vs. 66–691.451.321.58<.0001
       85 + vs. 66–691.451.271.64<.0001
      Sex
       Female vs. male1.061.011.120.0314
      Ethnicity
       Black non-Hispanic vs. White non-Hispanic1.181.061.310.0032
       Hispanic vs. White non-Hispanic1.000.891.130.9445
       Other vs. White non-Hispanic0.910.791.040.1654
      Education
       Lowest quartile vs. Highest quartile(lowest edu)0.800.740.87<.0001
       2nd quartile vs. Highest quartile0.910.840.990.0222
       3rd quartile vs. Highest quartile0.960.881.040.2812
       Unknown vs. Highest quartile0.950.831.090.4591
      Abbreviations: CI, confidence interval; UNK, unknown.
      The AUC of the fitted model was 0.759. Both Hosmer Lemeshow (p = 0.10) and Pearson’s correlation (p = 0.29) tests were conducted for model performance assessments, and showed no systematic patterns in the residuals across predictors. With 500 replicated samples (test sets), the estimated AUC was 0.759 (95% CI 0.7587–0.7592) and the Brier Score was 0.1869 (95% CI 0.1870–0.1837). The overfitting corrected R-square is 0.1839, which is close to 0.1846 in the final model. This small difference between values suggests only a minimal overfitting issue in the final model and the estimations were robust.
      The overall median follow-up time for survival was 16.8 months (36.9 months for those with localized disease, 21.8 months for regional, and 8.1 months for distant). Supplementary Table S3 illustrates the impact of delay on survival for patients with localized, regional, or distant disease, and Fig. 1 illustrates Kaplan–Meier survival for each stage (log-rank p< 0.001). In patients with localized disease, adherence was associated with improved survival (hazard ratio [HR] = 0.86, 95% CI 0.80–0.91, p< 0.001). No association was found between treatment delay and survival for patients with regional disease, although a trend was evident toward reduced survival with increased adherence (HR = 1.05, 95% CI 0.99–1.11, p= 0.054). In patients with distant disease, adherence was a time-dependent-variable—that is, the HR was not proportional across time between adherence and non-adherence. Specifically, for patients who died within 1 year, adherence was associated with worse survival (HR = 1.35, 95% CI 1.28–1.42, p< 0.001). However, for those patients with distant disease who survived for at least 1 year, adherence was associated with improved survival (HR = 0.86, 95% CI 0.74–0.99, p= 0.042) compared with non-adherent patients.
      Figure thumbnail gr1
      Fig. 1Kaplan–Meier survival plot based on stage subgroup and adherence to delay guidelines. Adherence was associated with improved survival for patients with localized disease, reduced survival for those with distant disease, but not for patients with regional disease. However, incorporating time-dependent effects into the model revealed that for patients with distant disease, the influence of adherence differed by survival time (<1 year vs. ⩾1 year), suggesting that shorter treatment times may reflect the need for immediate palliation of aggressive disease. Ad, adherence; nAd, non-adherence.
      Median intervals from diagnosis to PET (interval 1), PET to physician clinic visit (interval 2), and physician clinic visit to treatment (interval 3) were 15, 5, and 12 days (Supplementary Fig. S2 and Table S4). Fig. 2 illustrates results of a “threshold” analysis for selecting appropriate thresholds that may improve the timeliness of care. We varied the upper bound of the interquartile range of interval 3 while fixing the upper bound of the interquartile range (Q3) of interval 1 at 7 days (e.g., 75% of patients underwent PET within 7 days of diagnosis) and interval 2 at 3 days. In doing so, the proportion of patients receiving treatment within 35 days improved by 70–82%, depending on the Q3 of interval 3 (Fig. 2, scenario A). When the upper bound of Q3 is fixed at 7 days for interval 1, 3 days for interval 2, and 10 days for interval 3, the prevalence of delay in patients undergoing PET was only 20%. In contrast, if intervals 1 and 2 remained at their observed values and only the Q3 of interval 3 was decreased, then compliance with timely care ranged from only 45–60% in patients undergoing PET (Fig. 2, scenario B), indicating that all three periods had substantial roles in determining delay.
      Figure thumbnail gr2
      Fig. 2The influence of varying the intervals from diagnosis to treatment on delay for patients undergoing PET, with adherence defined as ⩽35 days, and using PET as a reference point. Interval 1 is from diagnosis to PET; Interval 2, from PET to first follow-up clinic visit; and Interval 3, from first follow-up clinic visit to treatment start. Scenario A demonstrates the result of shortening Interval 3 when the upper bound of the interquartile range (Q3) for intervals 1 and 2 are fixed at 7 and 3 days, respectively. By doing so, adherence rates approach 80% when Q3 of Interval 3 = 10 days (i.e., 75% of patients were treated within 10 days of the first follow-up clinic visit). Scenario B shows the effect of maintaining Intervals 1 and 2 at their observed median values (15 and 5 days) and altering Interval 3. In this scenario, even when Q3 of Interval 3 is reduced to 6 days, adherence is at a much lower rate, 60%.

      Discussion

      In this population-based analysis of the effect of disease-staging studies on delays in beginning treatment for newly diagnosed NSCLC, our pertinent findings are as follows. First, almost 40% of patients had substantial treatment delays after the diagnosis of lung cancer. Second, delays in treatment were correlated with changes in survival, with this association being stage dependent. Finally, the cause of delays was multifactorial and related to several staging studies obtained after diagnosis in addition to patient and treatment factor, though with PET being the most substantial cause of delay. To this end, by adhering to a regimen of 7-3-10 days from diagnosis to PET scan, PET scan to first follow-up, and follow-up to treatment in 75% of patients, at least 80% of patients could receive timely treatment within 35 days of diagnosis.
      Several studies have assessed the correlation between time to treatment and tumor progression, and in more aggressive malignancies it has been found that even relatively short delays can be consequential [
      • Everitt S.
      • Plumridge N.
      • Herschtal A.
      • Bressel M.
      • Ball D.
      • Callahan J.
      • et al.
      The impact of time between staging PET/CT and definitive chemo-radiation on target volumes and survival in patients with non-small cell lung cancer.
      ,
      • van Loon J.
      • Offermann C.
      • Ollers M.
      • van Elmpt W.
      • Vegt E.
      • Rahmy A.
      • et al.
      Early CT and FDG-metabolic tumour volume changes show a significant correlation with survival in stage I-III small cell lung cancer: a hypothesis generating study.
      ,
      • Seoane J.
      • Takkouche B.
      • Varela-Centelles P.
      • Tomas I.
      • Seoane-Romero J.M.
      Impact of delay in diagnosis on survival to head and neck carcinomas: a systematic review with meta-analysis.
      ,
      • Lawrence Y.R.
      • Blumenthal D.T.
      • Matceyevsky D.
      • Kanner A.A.
      • Bokstein F.
      • Corn B.W.
      Delayed initiation of radiotherapy for glioblastoma: how important is it to push to the front (or the back) of the line?.
      ,
      • Goy J.
      • Hall S.F.
      • Feldman-Stewart D.
      • Groome P.A.
      Diagnostic delay and disease stage in head and neck cancer: a systematic review.
      ,
      • Gomez I.
      • Seoane J.
      • Varela-Centelles P.
      • Diz P.
      • Takkouche B.
      Is diagnostic delay related to advanced-stage oral cancer? A meta-analysis.
      ,
      • Froud P.J.
      • Mates D.
      • Jackson J.S.
      • Phillips N.
      • Andersen S.
      • Jackson S.M.
      • et al.
      Effect of time interval between breast-conserving surgery and radiation therapy on ipsilateral breast recurrence.
      ,
      • Butof R.
      • Baumann M.
      Time in radiation oncology - keep it short!.
      ,
      • Muijs C.T.
      • Pruim J.
      • Beukema J.C.
      • Berveling M.J.
      • Plukker J.T.
      • Langendijk J.A.
      Oesophageal tumour progression between the diagnostic (1)(8)F-FDG-PET and the (1)(8)F-FDG-PET for radiotherapy treatment planning.
      ]. For example, as alluded to above, investigators from Australia found that in 83 patients with NSCLC in which definitive chemoradiation was planned, tumor volumes almost doubled (from 105 cc to 198 cc) in the 23 days between a diagnostic PET and planning PET scan, leaving almost 30% of patients no longer eligible for radical treatment [
      • Everitt S.
      • Plumridge N.
      • Herschtal A.
      • Bressel M.
      • Ball D.
      • Callahan J.
      • et al.
      The impact of time between staging PET/CT and definitive chemo-radiation on target volumes and survival in patients with non-small cell lung cancer.
      ]. In an analogous study in esophagus cancer, in 42 patients with consecutive PET/CT scans prior to treatment for esophagus cancer at a median time difference of 22 days apart, TNM progression was found in 27% of patients, and 18% of patients had newly detected mediastinal lymph nodes [
      • Muijs C.T.
      • Pruim J.
      • Beukema J.C.
      • Berveling M.J.
      • Plukker J.T.
      • Langendijk J.A.
      Oesophageal tumour progression between the diagnostic (1)(8)F-FDG-PET and the (1)(8)F-FDG-PET for radiotherapy treatment planning.
      ]. Indeed, a review by investigators from Germany outlined four categories of delay, as well as measures that could be taken to counteract these factors causing delay (described in parentheses): delayed detection of tumor until symptoms arise (screening programs), diagnostic delay (improved workflow efficiency and increased population awareness), prolonged waiting/preparation time (resource optimization, technical resources that prevent waiting lists), and overall treatment time (accelerated treatment regimens and concurrent vs. sequential multimodality approaches) [
      • Butof R.
      • Baumann M.
      Time in radiation oncology - keep it short!.
      ]. The current report expands on many prior studies that have shown a relationship between treatment delay and tumor progression by: (1) including a large, population database of greater than 28,000 patients, (2) addressing the effect of multiple patient, disease, and staging/treatment factors on treatment delays, (3) providing specific benchmarks of interval timeframes that will improve adherence to timely treatment, and (4) correlating treatment time to survival in NSCLC.
      Indeed, the influence of treatment delays on prognosis in NSCLC has varied by study, with some showing progression of disease or reduced survival with longer delays [
      • Everitt S.
      • Herschtal A.
      • Callahan J.
      • Plumridge N.
      • Ball D.
      • Kron T.
      • et al.
      High rates of tumor growth and disease progression detected on serial pretreatment fluorodeoxyglucose-positron emission tomography/computed tomography scans in radical radiotherapy candidates with nonsmall cell lung cancer.
      ,
      • Radzikowska E.
      • Roszkowski-Sliz K.
      • Glaz P.
      The impact of timeliness of care on survival in non-small cell lung cancer patients.
      ] and others showing no difference or even improvements in survival with longer times to treatment [
      • Diaconescu R.
      • Lafond C.
      • Whittom R.
      Treatment delays in non-small cell lung cancer and their prognostic implications.
      ,
      • Yilmaz A.
      • Damadoglu E.
      • Salturk C.
      • Okur E.
      • Tuncer L.Y.
      • Halezeroglu S.
      Delays in the diagnosis and treatment of primary lung cancer: are longer delays associated with advanced pathological stage?.
      ]. It has been hypothesized that any relationship between time to treatment and survival outcomes is influenced by the fact that urgent treatment correlated with a negative prognosis because of high symptom burden. In the current study, we found that adherence was associated with improved survival in localized disease and indolent metastatic disease in which patients survive ⩾1 year, but reduced survival in patients with distant disease surviving <1 year. Our working hypothesis based on these population-based findings is that for patients who have “manageable” disease, shortening times to treatment can enhance survival by reducing rates of progression. However, for patients with malignancies that are quickly fatal (distant metastases in patients surviving <1 year), quick treatment times are also a surrogate for disease severity. Regional disease did not consistently fall into one of these two categories, and thus offsetting influences probably explained the lack of clear correlation between treatment times and survival in this context.
      As a final aim, we proposed discrete benchmarks that could improve adherence to timely treatment in this clinical context. We believe that the “7-3-10” benchmark proposed in this analysis should be feasible in many clinical practices. The availability of PET and the speed of financial clearance have improved over the past decade, beginning when Medicare first expanded coverage for diagnosis and staging/restaging of NSCLC in July 2001 [

      National Coverage Determination (NCD) for Positron Emission Tomography (PET) Scans (220.6). Center for Medicare and Medicaid Services website, 2013.

      ]. Therefore, a 1-week interval between diagnosis and PET, followed by a prescheduled post-PET clinic visit, should then serve to streamline other potential studies, such as cardiopulmonary function tests for a presurgical workup for localized disease, thereby increasing the practicality of starting treatment within 10 days of the post-PET visit. We acknowledge that for some patients, this timeframe is not feasible because of factors such as more complex disease requiring extensive multidisciplinary discussion or logistics from the patient’s perspective, and therefore these proposed benchmarks acknowledge that 25% of patients cannot adhere to the 7-3-10 rule. However, the strength of this recommendation is that it presents novel guidance to treating physicians that is realistic, achievable, and can substantially influence times to treatment in this context.
      Given the delays in treatment associated with PET, this study underscores the importance of this study in the staging workup. Notably, although 37% of patients with distant disease underwent PET, this modality is not recommended as for initial staging of metastatic disease in the updated National Comprehensive Cancer Network guidelines [

      “Non-Small Cell Lung Cancer” Guidelines, National Comprehensive Cancer Network, www.nccn.org, 2013.

      ]. Therefore, this modality could be excluded in patients with evidence of metastatic disease on CT and MRI. Also, the current analysis included patients treated in 2004–2007, and improvements in the ease of obtaining PET scans over the past 5–10 years are not reflected in this study. Other than reducing times to follow-up as noted above, measures such as rapid assessment clinics and navigators to facilitate test ordering would be useful for reducing time to treatment.
      Other than the constraints in any retrospective, population-based study, our analysis had some limitations. First, the quality guidelines we used were adapted from the Netherlands, which has a different health care system than the United States (e.g., it includes obligatory health insurance). Those guidelines combined evidence-based consensus publications with appraisal and discussion by expert panels, using the Rand-modified Delphi method [
      • Campbell S.M.
      • Cantrill J.A.
      • Roberts D.
      Prescribing indicators for UK general practice: Delphi consultation study.
      ,
      • Cantrill J.A.
      • Sibbald B.
      • Buetow S.
      Indicators of the appropriateness of long-term prescribing in general practice in the United Kingdom: consensus development, face and content validity, feasibility, and reliability.
      ]. As noted previously, the quality indicator used the first visit to the pulmonologist as the reference date. However, because a significant percentage of patients in our cohort did not visit a pulmonologist, we chose to use date of diagnosis instead. We also expected that this date would provide a more conservative estimate of adherence in a percentage of patients, given the proportion who undergo biopsy only after seeing a pulmonologist.
      Indeed, these guidelines have not been validated or rigorously tested in the United States as a quality measure. We therefore acknowledge that a time to treatment of 35 days should not be considered a rigid threshold for “appropriate” care, but rather is presented as one potential measure endorsed by investigators in the United States. In addition, it should be emphasized that the 7-3-10 benchmark is supported by patients receiving PET scans only in the United States and during a specific time period, and that they are likely to vary depending on the year that delay is measured and in countries outside of the United States. We certainly agree with Butof et al., advocate for “global” solutions that can improve delay within any healthcare system such as adequate allocation of resources, sufficient staffing, and the consideration of accelerated fractionation regimens when supported by available data [
      • Butof R.
      • Baumann M.
      Time in radiation oncology - keep it short!.
      ]. The proposition of a benchmark should be viewed as a rough guideline for improving care in the context of the patient population in our study and as a methodology to identify critical points of delay. In a specific hospital setting, the solution should be tailored to the barriers preventing expedited treatment in that system and using the framework that Butof et al. [
      • Butof R.
      • Baumann M.
      Time in radiation oncology - keep it short!.
      ] provide.
      Finally, the database itself has limitations in accounting for complexities in individual cases. That is, the number of diagnostic tests ordered (and thus time to treatment) is often a function of clinical indication, with more tests ordered in more difficult cases. Although we acknowledge this limitation (inherent in SEER studies of this nature), our findings strongly support that, even when controlling for other patient, disease, and treatment factors, PET remains the strongest factor associated with delay. Further, we are not suggesting that delay is greater in a proportion of patients with NSCLC simply because patients are not being seen for follow-up appointments and treatment in an efficient manner. Rather, our findings suggest that if the infrastructure of an individual clinical practice could be changed to reduce times to PET, follow-up, and treatment in the manner described, regardless of the manner in which this goal is achieved, treatment would be more timely, which could in turn improve patient care.

      Conflict of interest statement

      The authors have no conflicts of interest to declare.

      Funding sources and acknowledgments

      Supported in part by Cancer Center Support (Core) Grant CA016672 to The University of Texas MD Anderson Cancer Center and the Center for Radiation Oncology Research. Drs. Gomez, Smith, and Giordano are supported by the Cancer Prevention and Research Institute of Texas’s Comparative Effectiveness Research in Texas ( RP101207 ). Dr. Giordano is also supported by American Cancer Society 117488-RSGI-09-149-01-CPHPS . The authors sincerely thank Christine Wogan, M.S., E.L.S., of MD Anderson’s Division of Radiation Oncology, for her valuable input in reviewing and editing this manuscript.

      Appendix A. Supplementary data

      Figure thumbnail fx1
      Supplementary Fig. 1This file contains supplementary Fig. 1.
      Figure thumbnail fx2
      Supplementary Fig. 2This file contains supplementary Fig. 2.

      References

        • Everitt S.
        • Herschtal A.
        • Callahan J.
        • Plumridge N.
        • Ball D.
        • Kron T.
        • et al.
        High rates of tumor growth and disease progression detected on serial pretreatment fluorodeoxyglucose-positron emission tomography/computed tomography scans in radical radiotherapy candidates with nonsmall cell lung cancer.
        Cancer. 2010; 116: 5030-5037
        • Mohammed N.
        • Kestin L.L.
        • Grills I.S.
        • Battu M.
        • Fitch D.L.
        • Wong C.Y.
        • et al.
        Rapid disease progression with delay in treatment of non-small-cell lung cancer.
        Int J Radiat Oncol Biol Phys. 2011; 79: 466-472
        • Everitt S.
        • Plumridge N.
        • Herschtal A.
        • Bressel M.
        • Ball D.
        • Callahan J.
        • et al.
        The impact of time between staging PET/CT and definitive chemo-radiation on target volumes and survival in patients with non-small cell lung cancer.
        Radiother Oncol. 2013; 106: 288-291
        • van Loon J.
        • Offermann C.
        • Ollers M.
        • van Elmpt W.
        • Vegt E.
        • Rahmy A.
        • et al.
        Early CT and FDG-metabolic tumour volume changes show a significant correlation with survival in stage I-III small cell lung cancer: a hypothesis generating study.
        Radiother Oncol. 2011; 99: 172-175
        • Hermens R.P.
        • Ouwens M.M.
        • Vonk-Okhuijsen S.Y.
        • van der Wel Y.
        • Tjan-Heijnen V.C.
        • van den Broek L.D.
        • et al.
        Development of quality indicators for diagnosis and treatment of patients with non-small cell lung cancer: a first step toward implementing a multidisciplinary, evidence-based guideline.
        Lung Cancer. 2006; 54: 117-124
        • Lennes I.T.
        • Lynch T.J.
        Quality indicators in cancer care: development and implementation for improved health outcomes in non-small-cell lung cancer.
        Clin Lung Cancer. 2009; 10: 341-346
        • Ouwens M.
        • Hermens R.
        • Hulscher M.
        • Vonk-Okhuijsen S.
        • Tjan-Heijnen V.
        • Termeer R.
        • et al.
        Development of indicators for patient-centred cancer care.
        Suppor Care Cancer. 2010; 18: 121-130
        • Ouwens M.M.
        • Hermens R.R.
        • Termeer R.A.
        • Vonk-Okhuijsen S.Y.
        • Tjan-Heijnen V.C.
        • Verhagen A.F.
        • et al.
        Quality of integrated care for patients with nonsmall cell lung cancer: variations and determinants of care.
        Cancer. 2007; 110: 1782-1790
        • Seoane J.
        • Takkouche B.
        • Varela-Centelles P.
        • Tomas I.
        • Seoane-Romero J.M.
        Impact of delay in diagnosis on survival to head and neck carcinomas: a systematic review with meta-analysis.
        Clin Otolaryngol. 2012; 37: 99-106
        • Lawrence Y.R.
        • Blumenthal D.T.
        • Matceyevsky D.
        • Kanner A.A.
        • Bokstein F.
        • Corn B.W.
        Delayed initiation of radiotherapy for glioblastoma: how important is it to push to the front (or the back) of the line?.
        J Neurooncol. 2011; 105: 1-7
        • Goy J.
        • Hall S.F.
        • Feldman-Stewart D.
        • Groome P.A.
        Diagnostic delay and disease stage in head and neck cancer: a systematic review.
        Laryngoscope. 2009; 119: 889-898
        • Gomez I.
        • Seoane J.
        • Varela-Centelles P.
        • Diz P.
        • Takkouche B.
        Is diagnostic delay related to advanced-stage oral cancer? A meta-analysis.
        Eur J Oral Sci. 2009; 117: 541-546
        • Froud P.J.
        • Mates D.
        • Jackson J.S.
        • Phillips N.
        • Andersen S.
        • Jackson S.M.
        • et al.
        Effect of time interval between breast-conserving surgery and radiation therapy on ipsilateral breast recurrence.
        Int J Radiat Oncol Biol Phys. 2000; 46: 363-372
        • Butof R.
        • Baumann M.
        Time in radiation oncology - keep it short!.
        Radiother Oncol. 2013; 106: 271-275
        • Muijs C.T.
        • Pruim J.
        • Beukema J.C.
        • Berveling M.J.
        • Plukker J.T.
        • Langendijk J.A.
        Oesophageal tumour progression between the diagnostic (1)(8)F-FDG-PET and the (1)(8)F-FDG-PET for radiotherapy treatment planning.
        Radiother Oncol. 2013; 106: 283-287
        • Radzikowska E.
        • Roszkowski-Sliz K.
        • Glaz P.
        The impact of timeliness of care on survival in non-small cell lung cancer patients.
        Pneumonol Alergol Pol. 2012; 80: 422-429
        • Diaconescu R.
        • Lafond C.
        • Whittom R.
        Treatment delays in non-small cell lung cancer and their prognostic implications.
        J Thorac Oncol. 2011; 6: 1254-1259
        • Yilmaz A.
        • Damadoglu E.
        • Salturk C.
        • Okur E.
        • Tuncer L.Y.
        • Halezeroglu S.
        Delays in the diagnosis and treatment of primary lung cancer: are longer delays associated with advanced pathological stage?.
        Upsala J Med Sci. 2008; 113: 287-296
      1. National Coverage Determination (NCD) for Positron Emission Tomography (PET) Scans (220.6). Center for Medicare and Medicaid Services website, 2013.

      2. “Non-Small Cell Lung Cancer” Guidelines, National Comprehensive Cancer Network, www.nccn.org, 2013.

        • Campbell S.M.
        • Cantrill J.A.
        • Roberts D.
        Prescribing indicators for UK general practice: Delphi consultation study.
        BMJ. 2000; 321: 425-428
        • Cantrill J.A.
        • Sibbald B.
        • Buetow S.
        Indicators of the appropriateness of long-term prescribing in general practice in the United Kingdom: consensus development, face and content validity, feasibility, and reliability.
        Qual Health Care. 1998; 7: 130-135