Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Setting Standards for Access to Radiotherapy 1 Mackillop WJ, 1Chen Z, 3Pearcey R, 1King W, 1Kerba M, 2 Shortt S, 1Feldman-Stewart D, 4Brouwers M, 1Brundage M, 5Coldman A 1 Cancer Research Institute, and 2Centre for Health Services Research and Policy, Queen’s University, Kingston; 3Cross Cancer Research Institute, Edmonton; 4Program in Evidence-Based Care, CCO/ McMaster; 5BC Cancer Agency, Vancouver In collaboration with Neil Berman, George Browman, Timothy Caulfield, Steve Hall, Mike Milosevic, Tom Pickles, Brent Schacter, and Simon Sutcliffe Taming of the Queue III March 2006 Objectives of this talk To report on our progress in setting “evidence-based benchmarks” for waiting times for RT and discuss next steps in this process To demonstrate that waiting times alone are an inadequate measure of access to care To demonstrate that setting benchmarks and monitoring compliance with them, is not useful unless it leads to effective action to reduce waiting times Radiotherapy (RT) • RT = Cancer treatment using ionizing radiation • It has diverse roles, e.g. – It may be used alone to cure some localized cancers (head and neck cancers, cervix) – It may used after surgery to reduce the risk of cancer recurrence (breast, rectum) – It may be used to control symptoms in people with incurable cancers • Overall, RT is useful in about 50% of patients. – Most common indications for RT are supported by level 1 evidence Objectives of this talk To report on our progress in setting “evidence-based benchmarks” for waiting times for RT and discuss next steps in this process To demonstrate that waiting times alone are an inadequate measure of access to care To remind ourselves that setting benchmarks and monitoring compliance with them is worthless unless it leads to effective corrective action Guidelines or Benchmarks? • Evidence-based guidelines: recommendations for care based on a systematic review of the literature • Benchmarks: benchmarking is “measuring products or services against those provided by companies recognized as industry leaders*” • Evidence-based benchmarks!!!!???? *from Benchmarking for Best Practices: Winning through Innovative Adaptation, Bogan and English 1994 Defining maximum acceptable WT’s for RT: What kinds of information are available? – Expert opinion • Surveys of individual opinion, existing guidelines/standards published by expert groups, clinical trial protocols – Prevailing waiting times (benchmarks) – Radiobiological Models based on: • a) the relationship between the volume of tumor and risk of recurrence, • b) the observed rates of growth of human cancers – Direct observations of associations between waiting times and risk of recurrence in clinical practice – Patients’ opinions Defining maximum acceptable WT’s for RT: What kinds of information are available? – Expert opinion • Surveys of opinions of individual experts, clinical trial protocols, existing guidelines/standards published by expert groups, – Prevailing waiting times (benchmarks) – Radiobiological Models based on: • a) the relationship between the volume of tumor and risk of recurrence, • b) the observed rates of growth of human cancers – Direct observations of associations between waiting times and risk of recurrence in clinical practice – Patients’ opinions Radiation Oncologists’ opinions about the maximum acceptable interval between referral and start of RT for a T2,N0 Ca larynx Maximum acceptable waiting times from surgery to post-operative RT in protocols of current and recent clinical trials. Prostate Number of Trials 6 4 2 0 6 4 2 0 Rectum Head & Neck 6 4 2 0 6 4 2 0 Cervix Breast 6 4 2 0 NSCLC 0 5 10 Waiting Time (weeks) 15 6 4 2 0 Guidelines of the Canadian Association for Radiation Oncology (CARO) • Created by an informal consensus-based process; grounded on expert opinion; not explicitly evidence-based. • first adopted 1993, last reviewed and approved by members, Sept.2000 • Waiting time from referral to consultation with a radiation oncologist should not exceed 10 working days. • Waiting time for radiotherapy should not exceed 10 working days – – For single modality treatment with RT, WT for RT is defined as the time from consultation to start of RT, or from date that the RO makes the decision to treat, whichever is later. For multi-modality treatments, WT for RT is defined as the interval between the target RT start date and the actual RT start date. • Consistent with guidelines from the UK, NZ, and several European countries which are also based entirely on expert opinion • http://www.caro-acro.ca/ Defining acceptable WT’s for RT: What kinds of information are available? – Expert opinion • Surveys of individual opinion, existing guidelines/standards published by expert groups, clinical trial protocols – Prevailing waiting times (benchmarks) – Radiobiological Models based on: • a) the relationship between the volume of tumor and risk of recurrence, • b) the observed rates of growth of human cancers – Direct observations of associations between waiting times and risk of recurrence in clinical practice – Patients’ opinions WT’s from referral to start of RT for a T2, N0, Ca larynx at Cancer Centres in Canada (n=27) and the US (n=65) Defining acceptable WT’s for RT: What kinds of information are available? – Expert opinion • Surveys of individual opinion, existing guidelines/standards published by expert groups, clinical trial protocols – Prevailing waiting times (benchmarks) – Radiobiological Models based on: • the relationship between the volume of tumor and risk of recurrence • the observed rates of growth of human cancers – Direct observations of associations between waiting times and risk of recurrence in clinical practice – Patients’ opinions Biological Background Information • The probability that RT will eradicate a cancer is inversely related to the volume of the tumor – Experimental studies in animals – Direct observation in human subjects • Untreated cancers usually grow exponentially – Experimental studies in animals – Direct observations in human subjects Tumour Doubling Time by Primary Cancer Site 260 104 52 26 30 22 21 18 8 12 13 6 4 4 1 Tumor Doubling Time (weeks) 520 Based on the results of 128 studies including ~10,000 cases Prostate Breast Colorectal Gastric Lung Hepatic Pancreas Head/Neck Work in progress: with thanks to Gavin Li and Zheng Chen Distribution of Tumour Doubling Times 0.4 0.3 0.2 Breast median 30 wks Percent of Total 0.1 0.0 0.4 Lung 0.3 median 18 wks 0.2 0.1 0.0 0.4 0.3 Head & Neck median 4 wks 0.2 0.1 0.0 7 30 90 360 Tumour Doubling Time (days) % Local Control Rate Predicted Decrease in Local Control with Increasing WT 100 for RT for Cancer of the Tonsil 80 T1 T2 60 40 20 0 T3 T4 mean Td = 82 days mean Td = 58 days 20 40 Waiting Time (days) Mackillop, Bates, O’Sullivan, Withers et al, Radiation Research, 60 Defining acceptable WT’s for RT: What kinds of information are available? – Expert opinion • Surveys of individual opinion, existing guidelines/standards published by expert groups, clinical trial protocols – Prevailing waiting times (benchmarks) – Radiobiological Models based on: • a) the relationship between the volume of tumor and risk of recurrence, • b) the observed rates of growth of human cancers – Direct observations of associations between waiting times and risk of recurrence in clinical practice – Patients’ opinions A systematic review of the literature relating WT’s to outcomes of RT: Study Questions • Is delay in RT associated with an increased risk of local recurrence? • Is there a threshold below which delay is free of risk? • Does the impact of delay on the risk of recurrence differ from one clinical situation to another? • Does delay affect other outcomes of RT? Methods • Search MEDLINE and CANCERLIT databases: any language; Jan 1975 to June 2005 • Manual search of Abstracts from CARO and ASTRO • Secondary search of reference lists Flowchart for study identification Reviewed 72 potentially relevant studies were reviewed in detail Entry Criteria 36 studies met basic entry criteria1 Quality Criteria 20 studies met quality criteria2 Entry criteria and quality criteria Entry criteria: All patients were treated with RT; WT from diagnosis or surgery to start of RT was reported; Local recurrence rate was reported. Quality criteria: There were no RCT’s: Quality assessment was therefore designed for observational studies. We asked 2 questions about each study: Were potential confounding factors adequately described? patient-related, disease-related, and treatment-related factors known to be associated with risk of recurrence Were the potential confounding factors either balanced between WT groups or controlled for in a multivariate analysis? If the answer to both questions was yes, the study was classified as meeting quality criteria. Characteristics of 36 studies that met the basic entry criteria • Year: 75-89, 1; 90-94, 9; 95-00, 17; 01-05, 9 • Origin: Canada, 7; Europe 9; US: 15; other, 5. • Cancer site: Breast, 22; Head & Neck, 12; Sarcoma, 2 • Sample Size: <100, 6; 100-299, 10; 300-49, 7; 500-1000, 10; >1000, 3 • Waiting time measurement: – continuous variable, 8; categorical variable 28 Statistical Methods • Data transformation – Results of studies which described WT as a categorical variable were re-analyzed to estimate relative risk per month of delay • Median waiting times in each category were attributed to all cases in that category • Meta-analysis – Fixed effect model – Weighting factor (1/SE2) Breast Cancer (met quality criteria) Study RR (95%CI) Weight% Postop RT Without Chemo Froud,2000 0.91 (0.69, 1.22) 6.75% Vujovic,1998 1.06 (0.80, 1.39) 7.37% Whelan,1996 1.46 (0.88, 2.42) 2.20% Clarke,1985 2.07 (1.19, 3.59) 1.83% 1.11 (0.94, 1.33) Subtotal Postop RT With Chemo Hebert-C,2004 1.18 (1.00, 1.39) 20.32% Benk,2004 1.13 (0.96, 1.33) 20.93% Yock,2004 0.95 (0.81, 1.12) 20.34% Wallgren*,1996 1.11 (0.62, 1.98) 1.67% Wallgren,1996 1.12 (0.91, 1.37) 13.44% Recht,1996 1.37 (0.92, 2.03) 3.55% Slotman,1994 2.62 (1.45, 4.72) 1.61% Subtotal All Breast Studies 1.12 (1.03, 1.22) 1.12 (1.04, 1.21) 0.2 0.5 1.0 2.0 Relative Risk 5.0 10.0 Breast Cancer (All studies) Study RR (95%CI) Weight% Postop RT Without Chemo 3.29% 0.91 (0.69, 1.22) Froud,2000 3.59% 1.06 (0.80, 1.39) Vujovic,1998 3.51% 1.08 (0.82, 1.42) Bahena,1998 1.07% 1.46 (0.88, 2.42) Whelan,1996 4.92% 1.16 (0.92, 1.47) Fourquet,1995 1.08% 0.42 (0.26, 0.70) Nixon,1994 0.89% 2.07 (1.19, 3.59) Clarke,1985 Subtotal 1.06 (0.94, 1.20) Postop RT With Chemo 9.90% 1.18 (1.00, 1.39) Hebert-C,2004 10.20% 1.13 (0.96, 1.33) Benk,2004 9.91% 0.95 (0.81, 1.12) Yock,2004 20.42% 0.99 (0.88, 1.11) Metz,2000 0.81% 1.11 (0.62, 1.98) Wallgren*,1996 6.55% 1.12 (0.91, 1.37) Wallgren,1996 1.88% 0.84 (0.57, 1.23) Meek,1996 1.73% 1.37 (0.92, 2.03) Recht,1996 10.94% McCormick,1996 1.25 (1.07, 1.47) 2.41% 0.92 (0.66, 1.28) Leonard,1995 0.93% 1.59 (0.93, 2.74) Hartsell,1995 0.78% 2.62 (1.45, 4.72) Slotman,1994 0.53% 1.10 (0.54, 2.25) Nguyen,1993 1.70% 1.64 (1.10, 2.45) Buchholz,1993 2.95% 1.38 (1.02, 1.87) Recht,1991 Subtotal 1.11 (1.05, 1.18) 1.10 (1.04, 1.16) All Breast Studies 0.2 0.5 1.0 2.0 Relative Risk 5.0 10.0 Head & Neck Cancer (met quality criteria) Study RR (95%CI) Weight% Leon,2003 1.08 (0.94, 1.23) 50.80% Fortin,2002 2.03 (1.24, 3.33) 3.84% Brouha,2000 1.03 (0.68, 1.56) 5.49% O'Sullivan,1998 1.52 (1.07, 2.17) 7.47% Lee,1994 0.63 (0.23, 1.70) 0.93% Definitive RT Subtotal 1.15 (1.02, 1.29) Postop RT Marshak,2004 1.63 (0.67, 4.01) 1.16% Suwinski,2003 1.22 (1.01, 1.48) 25.55% Kajanti,1991 1.56 (1.01, 2.43) 4.77% Subtotal 1.28 (1.08, 1.52) All Head&Neck Studies 1.19 (1.08, 1.31) 0.2 0.5 1.0 2.0 Relative Risk 5.0 10.0 Head & Neck Cancer (All studies) Study RR (95%CI) Weight% Leon,2003 1.08 (0.94, 1.23) 45.64% Fortin,2002 2.03 (1.24, 3.33) 3.45% Brouha,2000 1.03 (0.68, 1.56) 4.93% O'Sullivan,1998 1.52 (1.07, 2.17) 6.71% Barton,1997 0.74 (0.47, 1.17) 3.97% Lee,1994 0.63 (0.23, 1.70) 0.84% Definitive RT Subtotal 1.12 (1.00, 1.25) Postop RT Marshak,2004 1.63 (0.67, 4.01) 1.04% Suwinski,2003 1.22 (1.01, 1.48) 22.95% Dixit,1998 2.76 (1.39, 5.46) 1.79% Ampil,1993 1.60 (0.88, 2.93) 2.29% Kajanti,1991 1.56 (1.01, 2.43) 4.29% Schiff,1990 1.73 (0.92, 3.24) 2.11% Subtotal 1.38 (1.18, 1.61) All Head&Neck Studies 1.20 (1.09, 1.31) 0.2 0.5 1.0 2.0 Relative Risk 5.0 10.0 0.2 0.3 0.4 0.5 0.6 Standard Error of Log RR 0.1 0.0 Funnel plot with pseudo 95% confidence limit All studies 0.13 -1.0 -0.5 0.0 Log Relative Risk 0.5 1.0 Summary High Quality Studies RR/m 95%CI All Studies 1.15 (1.09, 1.22) All Breast Studies 1.12 (1.04, 1.21) All Head&Neck Studies 1.19 (1.08, 1.31) All Studies RR/m 95%CI All Studies 1.13 (1.09, 1.18) All Breast Studies 1.10 (1.04, 1.16) All Head&Neck Studies 1.20 (1.09, 1.31) Is delay in RT associated with an increased risk of local recurrence? • Yes • Longer WT was significantly associated with a higher risk of local recurrence in breast and H &N cancer. • There was no direct evidence that delay is associated with increased recurrence rates in other cancer sites, but beware of the problem of beta error “absence of evidence is not evidence of absence” Is there a threshold below which delay is free of risk? • No • There is no empirical evidence that there is a threshold below which delay has no adverse effect • There is no theoretical reason to suspect that such a threshold exists. • Similar stochastic risks are common with environmental hazards, hence the ALARA principle i.e. keep exposures As Low As Reasonably Achievable. Does the impact of delay on the risk of recurrence differ from one clinical situation to another? • Unanswerable! • No significant differences in effect size in the different situations studied. Study power was inadequate to address this question. • But, the RR/month delay in post-op RT was greater for head and neck cancer than for breast cancer. • The RR/month delay was greater for post-operative RT than for definitive RT in the context of head and neck cancer. Does delay affect other outcomes of RT? Data not presented. Analysis is incomplete • There is no good empirical evidence that delay in RT is associated with an increase in the risk of distant metastasis. There are good theoretical reasons to believe that any such effect would be small. • There is some empirical evidence that delay in RT decreases the probability of long term survival in situations in which RT is used with curative intent, including head and neck cancer and cervical cancer. • There is almost no information about the effect of delay in RT on symptoms or quality or life. Defining acceptable waiting times • What kinds of information are available? – Expert opinion • Surveys of individual opinion, existing guidelines/standards published by expert groups, clinical trial protocols – Prevailing waiting times (benchmarks) – Models based on the observed relationship between the volume of tumor and risk of recurrence, and the observed rates of growth of human cancers – Direct observations of associations between waiting times and risk of recurrence in clinical practice – Patients’ opinions Patients Attitudes to Waiting for RT • When offered the choice between delayed RT close to home or immediate RT at distant centre, many patients choose to accept significant delays rather than relocate. – Palda, VA., et al (1997). Journal of Clinical Oncology, 15, 3192-3200 • But, if patients are informed that delay is associated with any decrease in the effectiveness of RT, they will choose treatment at a distant centre to avoid even short delays – Lehman, et la (2004). Radiotherapy and Oncology, 70, 283-289. That’s the evidence! How should we interpret it? • The paradigm of “Evidence-based medicine” • The paradigm of “Risk assessment and risk management” The Paradigm of “Evidence-based Medicine” • Context: widely used for evaluating the effectiveness of clinical interventions and establishing CPGs • Rules of Evidence: Sackett’s rules: The RCT is the gold standard; observational data is of secondary value; the results of animal studies are inadmissible • Action Guide: the default position is not to recommend treatment unless there is clear evidence of its effectiveness (because unproven therapy may do more harm than good) • Who interprets the data and makes the call? Expert health professionals+ clinical epidemiologists The Paradigm of “Risk Assessment” • Context: evaluation of the risks of environmental hazards and establishment of regulations for human exposures • Rules of Evidence: Hill’s criteria; observational studies in humans and experimental studies in animals are both admissible. • Action Guide:, The evidentiary threshold for action is lower because inaction may lead to harm and there is no health risk associated with limiting exposures to environmental hazards • Who interprets the data and makes the call? Experts AND those potentially affected Our next step will be to engage stakeholders in evaluation of the risks Risk Assessment Framework from the Presidents Commission Report In the meantime, we recommended adoption of the CARO/WTA guidelines • • • • • • They are clear and simple They are known to be achievable They are auditable They are widely accepted by Canadian ROs They are consistent with standards elsewhere They are consistent with the available evidence and the principle that WTs for RT should be As Short As Reasonably Achievable Were these guidelines accepted? CARO/WTA • Waiting time from referral to consultation with a radiation oncologist should not exceed 10 working days. • Waiting time for radiotherapy should not exceed 10 working days – For single modality treatment with RT, WT for RT is defined as the time from consultation to start of RT, or from date that the RO makes the decision to treat, whichever is later. – For multi-modality treatments, WT for RT is defined as the interval between the target RT start date and the actual RT start date. Objectives To report on our progress in setting “evidence-based benchmarks” for waiting times for RT and discuss next steps in this process To demonstrate that waiting times alone are an inadequate measure of access to care To remind ourselves that setting benchmarks and monitoring compliance with them is worthless unless it leads to effective corrective action The Concept of Accessibility in Health Care – Accessibility embraces all factors that influence the level of use of a service, given a particular level of need…. “describes the degree of fit between the system and the patients” Aday, Penchansky Components of Accessibility • Availability* : total system capacity/ total needs – Determined by total resources, efficiency, flexibility • Spatial Accessibility* – Distance and travel times • Accommodation* (Convenience) – hours of operation; lodges; transportation services • Affordability* – direct and indirect costs; ability and willingness to pay • Awareness of services, and the indications for their use, among doctors and patients * terms used by Penchansky 1981 Limitations of Waiting Time as a Measure of Access to RT Waiting times are sensitive to supply-side problems associated with availability. BUT Waiting times are entirely insensitive to demand-side problems associated with awareness, spatial accessibility, affordability, and accommodation Therefore, the absence of a waiting list does not imply that access is optimal. Indicators of Access to RT • Waiting times for RT • Rates of use of RT Measures of Use of RT • Use of RT in initial treatment of cancer – *defined as the proportion of incident cases treated within one year of diagnosis • Use of palliative RT among patients dying of cancer – **defined as the proportion of cases that die of cancer who receive palliative RT in the last two years of life *Mackillop et al, JCO, 1997, **Huang at al JCO, 2002 a) RT within 1 Year of Diagnosis Northern Ontario Ottaw a Thunder Bay 0 25 0 Kingston 500 Miles Quintile 17.5% - 24.3% 25.1% - 26.8% 27.2% - 29.1% 29.2% - 30.7% 30.8% - 35.2% data not available Toronto Sudbury Windsor London Ham ilton 0 100 200 Miles Southern Ontario Radiotherapy Centre Zhang-Salomons 2005 b) Palliative RT in the Last 2 Years of Life Northern Ontario Ottaw a Thunder Bay 250 0 Kingston 500 Miles Quintile 10.9% - 23.5% 23.5% - 24.5% 24.5% - 26.5% 26.5% - 29.6% 29.6% - 35.5% data not available Toronto Sudbury Windsor London Ham ilton 0 100 200 Miles Southern Ontario Radiotherapy Centre Zhang-Salomons 2005 Factors associated with the use of palliative RT in Ontario* Characteristic OR 95% CI Low 1.00 – Medium 1.09 1.04-1.15 High 1.17 1.11-1.24 Diagnosing Hospital Not associated with an RT center 1.00 – Associated with an RT center 1.35 1.30-1.40 Place of residence Counties with no cancer center 1.00 – Counties with cancer center 1.24 1.21-1.27 Toronto 0.88 0.84-0.92 Ottawa 1.00 – Hamilton 1.20 1.14-1.26 Median household income Region *controlling for primary site and age Huang et al, JCO, 2002 RT Use in Older Patients RT Rate or Functional State * 100% Adjuvant RT for resected Breast Ca 80% Palliative RT within yr of death from Prostate Ca 60% 40% Radical RT for Oropharynx Ca 20% Estimated Functional Status 0% 45-54 55-64 65-74 >75 Age Groups *Relative to 45-54 years Tyldesley et al, 2000 The value of Rates of RT Utilization as a Measure of Accessibility • May identify demand-side problems relating to spatial accessibility and awareness of services, that would otherwise remain invisible • May provide a quantitative measure of the shortfall between supply of service and the need for service. • Public policy about access to care that is informed only by WT’s is blind to half of the problem. • Waiting times and utilization rates both require to be “benchmarked” and monitored to ensure adequate access to RT. Objectives To report on our progress in setting “evidence-based benchmarks” for waiting times for RT and discuss next steps in this process To demonstrate that waiting times alone are an inadequate measure of access to care To remind ourselves that setting benchmarks and monitoring compliance with them is worthless unless it leads to effective corrective action Median Wait Time from Referral to Start of RT: CCO 10 8 Breast GI GU Lung 6 4 2 19 98 19 99 20 00 20 01 20 02 20 03 20 04 19 97 19 96 19 95 19 94 19 93 0 19 92 Median Wait (weeks) 12 Calendar Year Conclusion • Our plans for setting WT standards and monitoring access to care will be worthless unless they are linked to effective strategies to deal with problems that they identify. 5-year Survival (%) Cause-specific Survival in Ontario and the U.S.(SEER), by Income Group (1987-92) 60 All sites except prostate 55 50 Ontario U.S. 45 1 2 Highest Quintile 3 4 5 Lowest Quintile Median Household Income Boyd et al, J Clin Oncol, 1999 Initial Rate of RT for Lung Cancer in Ontario by County Estimated Appropriate Rate=44.6% RT Rates 50% 40% Estimated Shortfall=10.6% 30% 20% Overall Observed Rate=34% 10% 1 5 9 13 17 21 25 29 33 37 41 45 Counties ordered by RT rates Bars (-) represent the 90% CI Tyldesley et al, IJROBP 2001; Barbera et al, Medical Care, 2003