Fewer than 5% of all patients with cancer in the United States (US) participate in clinical trials,1 and this proportion is probably even lower in New Zealand.2 Almost 30% of trials funded by the US National Cancer Institute fail to enrol even a single patient3 and many trials do little better, as illustrated by a study described in a recent article in the New York Times: ...A new trial [of intensive chemotherapy for colorectal cancer] in the United States has been temporarily suspended so that researchers can find a way to recruit patients. After nearly a year only one patient had enrolled, because people were reluctant to chance winding up in the control group, according to one of the investigators.4 Overall, only about half of all cancer trials manage to enrol enough patients to generate meaningful results, and only one in five studies results in a published report.5 Difficulty with recruitment is also reflected in the steadily increasing costs of cancer studies, which now stand at about US$60,000 per subject.6 The cost of studies and difficulty recruiting patients has led some pharmaceutical companies to scale back their research and development programmes.7 A recent Health Committee Inquiry into improving New Zealands environment to support innovation through clinical trials concluded that several steps should be taken to ensure that New Zealand did not lose its advantage as a good place to carry out clinical trials . Whilst this report made several recommendations it did not address the issue of low participation rates in clinical trials.8 The causes of poor enrolment in cancer trials have been much studied.9-12 Probably the main contributing factor is patients natural reluctance to wind up in the control group , as noted in the above quote. In addition, doctors are often hesitant to inform patients about randomised controlled trials (RCTs) because of the hassle factor involved - especially when half the patients counselled will not be offered the new treatment. Moreover, doctors often cite their dislike of discussions of uncertainty with patients in explaining why they failed to enrol patients in trials.13 Several suggestions have been made for improving enrolment in cancer trials, including paying doctors higher fees for recruiting patients,11 but such measures seem unlikely to increase participation in cancer research to any substantial extent given the dynamics described above. Improving enrolment via post-randomisation consent Another approach gaining increased international attention is to obtain consent for experimental treatment after randomisation, rather than before (as with standard consent protocols).14-17 Post-randomisation consent (PRC) designs, also known as randomised consent or pre-randomisation , were first formally described by Marvin Zelen in the 1970s,18,19 though they had been used previously. PRC designs have been much discussed20-22 and various modifications developed.15, 23, 24 The recruitment process for New Zealand cancer research proposed herein is based on Zelens original single-consent design, under which subjects randomised to be offered the experimental treatment are approached for consent, while subjects randomised to continue to receive current best (or standard) treatment are not approached for consent, since nothing has changed for them. Subjects randomised to be offered the experimental treatment but who decline to receive it would also continue to receive current best practice. Figures 1 and 2 depict the sequence of events under standard RCTs and under the proposed PRC approach. For analysis purposes, the outcomes of patients who decline offered experimental treatment (offered-declined) are usually included with the offered-accepted patients and collectively compared against the non-offered patients (intention-to-treat analysis). Alternatively, the offered-declined group can be amalgamated with the non-offered patients and collectively compared to the offered-accepted patients (treatment received). The former approach has the advantage of preserving randomisation, but estimates of treatment effects will usually be biased towards the null (no effect) because of treatment dilution (i.e., some patients counted as receiving the treatment will not actually receive it). The treatment-received approach, on the other hand, breaks the randomisation and potentially magnifies any systematic bias, although careful data analysis can to some extent mitigate this problem. PRC designs were used in some cancer studies in the 1970s and 1980s. A review by Altman et al25 found 11 such trials, the five largest of which were conducted by co-operative groups: three by the National Surgical Adjuvant Breast and Bowel Project, one by the Eastern Co-operative Oncology Group in the US, and one by the Danish Breast Cancer Co-operative. The PRC design appears to have been used less frequently in cancer research since that time, probably due to the ethical considerations described below. PRC designs continue to be used in a wide range of other clinical contexts, however, as described in two reviews published in 2006, one of 58 PRC studies21 and one of 50 studies.20More recently, PRC was used in a New Zealand study examining the effectiveness of a behavioural treatment for Mori patients who present with self-harm.26 PRC designs are also still widely used in cluster randomised trials, where the units of randomisation consist of medical practices, hospitals, clinics, or communities.27, 28 In such trials, consent is obtained only from patients residing within the units randomised to receive treatment.29 Advantages of the proposed approach The main advantage of the PRC approach in cancer trials is that study recruitment would likely increase substantially, primarily because the offer group would be able to obtain the experimental treatment for sure (i.e., no chance of being relegated to the control group). Also, doctors may be more inclined to take the time to enrol patients if they have already been selected to receive the study treatment. Likely recruitment rates under PRC designs can be estimated by examining previous PRC studies to determine the rate of acceptance by patients randomised to be offered experimental treatments. In the only controlled study examining this question, a Canadian multicentre trial of platelet infusion in premature infants with deficient platelet counts,30 reported that three centres using standard (pre-randomisation) consent designs had recruitment rates of 19%, 26% and 52%. By contrast, a centre using PRC reported 86% recruitment. Comparable acceptance rates were observed by Altman et al in their 1995 review of PRC cancer studies,25 with a median acceptance rate of 88% and inter-quartile range (IRQ) of 84-89%. Similar acceptance rates were seen in two more recent reviews, one of 58 PRC studies, in which the median acceptance rate (out of 39 studies reporting this statistic) was 88% with an inter-quartile range (IQR) of 81-93%,21 and one of 50 studies, in which the median reported acceptance rate was 85% with an IQR of 61-93%.20 Comparable acceptance rates might reasonably be expected in the context of new treatments for patients with cancer in New Zealand, especially when the cancer is at an advanced stage (regional or distant spread). The limited range of effective treatments for most such patients could make uptake of offered new treatment very high - indeed the norm. Coupled with the high incidence of cancer in New Zealand (see Table 1), such robust acceptance rates could translate into hundreds or even thousands of patients being able to try new cancer treatments each year who otherwise would not have access to such treatments. High acceptance rates will also have positive implications with respect to power requirements for comparative effectiveness studies. For example, in a Cox proportional hazards model with survival as the outcome, assuming one-sided statistical significance for \u03b1 of 0.05 and 1-\u03b2 (power) of 0.8, one needs 58 deaths to detect a hazard ratio of 0.5 and 560 deaths to detect a ratio of 0.8.31 Table 1 depicts the most recent statistics on annual incidence and number of deaths for the 10 most common types of cancer in New Zealand. Table 1. New Zealand cancer mortality statistics (ranked by numbers of deaths for the top 10 cancers) Cancer type New registrations (2009) Deaths (2008) Trachea/lung Colorectal Prostate Breast Pancreas Melanoma Stomach Bladder Kidney Uterine 2008 2837 3369 2781 472 2212 372 361 482 436 1634 1280 670 624 373 317 283 200 165 94 Source: New Zealand Ministry of Health (http://www.health.govt.nz/nz-health-statistics/health-statistics-and-data-sets/cancer-data-and-stats). Figure 1. Design of conventional randomised controlled trials (RCTs) Figure 2. Design of proposed post-randomisation consent RCT RCT = randomised controlled trial TR = treatment received ITT = Intention-to-treat Clinical input will be required to determine if and when it is necessary to define standard treatment, to be employed nationally, as opposed to accepting a variety of standard treatments , which can collectively be compared to the new treatment. Depending on the acceptance rate and the proportion of patients within each type who are eligible for a particular experimental treatment (e.g., based on tumour stage or receptor subtype), several types of cancer appear to be associated with a sufficient number of deaths per year to enable detection of a hazard ratio of 0.8, especially if recruitment extends over more than 1 year. Recruitment will likely be greater in the setting of the more common types of cancer and in later stages of cancer and perhaps this is where the focus should be. In some cases, data from New Zealand studies may often need to be combined with data from parallel international studies, especially when narrow inclusion criteria limit enrolment. Ideally such studies would also be of the PRC type and would use similar standard-treatment controls. Bias control\u2014Adamson et al found that 44 studies (out of 58) specified the rationale for using the PRC design: The single largest justification for using the design was to avoid bias (n = 23). Trialists justified using Zelens method: to avoid the Hawthorne effect, to obtain \u2018effectiveness estimates as opposed to efficacy effects, to avoid contamination of the control group or to prevent resentful demoralisation. Few trials (N = 7) explicitly used the approach to increase participant recruitment.21 These bias-avoiding features must be balanced against the concern that patients who reject offered experimental treatment may differ in some significant way from patients who accept the treatment. The effect of any such resultant bias will increase as acceptance rates for the experimental treatment fall, but acceptance rates in the 85+% range, as seems likely based on evidence from previous studies, would probably be sufficient to warrant confidence in the validity of the findings. This is especially true when effect sizes are large and when comprehensive and accurate information on potential confounders (i.e., baseline variables that are significantly correlated with untreated prognosis, such as age, ethnicity, co-morbidities, and stage of cancer), is available on all patients. Additional attention would need to be paid to comparability of acceptors and non-acceptors if the acceptance rate falls substantially below 85%. Fortunately, New Zealand has a world-class system of linked health data, made possible via a national system of health identifiers. Over 98% of New Zealand residents have been assigned a unique health identification number, permitting linkage of individual patient data across multiple databases (e.g., demographic, hospitalisation, prescription drugs, mortality). Critically, New Zealand also has a comprehensive and well-functioning national cancer registry, into which all patients with newly diagnosed cancer (except minor skin cancers) are entered. Information from this registry can also be linked with the larger national databases. A limiting factor in these databases is the lack of information on clinical outcomes other than mortality (e.g., quality of life), although work is underway to expand the breadth of variables collected in cancer databases. Because comprehensive, routinely collected information would be available on all patients\u2014offered-accepted, offered-declined, and non-offered\u2014it would generally be possible to stratify, match or adjust on relevant predictor variables across groups. This sort of cross-group matching is commonly performed via the use of propensity or prognostic scores.32 The ability of PRC designs to take advantage of information on potential confounders was acknowledged by MacLehose et al:13 [The PRC design] allows the possible biases associated with patient selection for the experimental therapy to be investigated. The characteristics of patients allocated to the control treatment can be compared with those allocated to the new intervention who receive the control treatment after refusing the new treatment. If a comparison is made between the \u2018as-treated groups, the role of self-selection [i.e., confounding] can be examined.(p.67) These anti-bias advantages may be offset to some degree by the fact that provision of experimental treatments within studies is often accompanied by enhanced levels of clinical scrutiny (e.g., attention, monitoring), which could itself result in improved outcomes. If such were to occur, estimates of treatment effects could be biased upwards. Although a similar bias can occur in standard RCTs, in the standard design control patients are managed within the same in-study framework as the treated patients (including blinding where possible, though usually not in cancer studies), which could potentially result in more comparable levels of non-treatment-specific therapeutic care. Efficient research platform\u2014In addition to enabling many more patients to enrol in cancer studies, the proposed PRC approach, combined with the availability of New Zealands health databases, would enable cancer trials to be conducted in an extremely efficient and cost-effective manner.33 Indeed, most required data collection would occur automatically (e.g., age, type and stage of cancer, co-morbidities, pharmaceutical usage, mortality). The need for new data collection would not be completely obviated under the PRC approach, however, because safety monitoring and reporting would be required as part of detailed treatment protocols, which would be developed in consultation with appropriate clinical and ethical advisory groups. Moreover, it may be necessary to upgrade cancer-related clinical infrastructure, including hiring additional research nurses and data managers, in order to ensure accurate data collection. Improvement of some features of routine data collection, such as documentation of chemotherapy and radiation therapy, would also likely be required, and efforts in this direction are already underway through initiatives supported by the Ministry of Health and Cancer Control New Zealand. In addition, provision for rapid documentation of adverse effects will be needed, including the ability to stop trials promptly if required. Such upgrades of system capabilities would in themselves be likely to lead to improved quality and cost-effectiveness of cancer care. A further feature contributing to the value of the proposed research platform is that the resulting data would reflect long-term, real-world effectiveness, which is not the case in standard RCTs. These efficiencies could make New Zealand a potential leader in smart and cost-effective trial design for testing new treatments for cancer. If so, not only would thousands of New Zealand cancer patients gain early access to investigational treatments that would otherwise be unobtainable; in addition, the New Zealand research infrastructure could be given a major financial boost. The global pharmaceutical budget for cancer drug research and development is in the billions of US dollars annually34 and a proportion of this funding could potentially be captured. A potential challenge arising from creation of the envisioned enhanced-efficiency cancer research platform is that international pharmaceutical manufacturers might expand their presence and influence in New Zealand, with possible detrimental effect on Pharmacs ability to determine the public subsidy of cancer drugs based on beneficial impact and cost-effectiveness. This could come about through increased patient expectations or political lobbying, for example. This risk could be minimised, however, if Pharmac and the Ministry of Health were to participate in negotiating the terms under which studies are conducted (e.g., user fees, future discounts, pay-for-outcomes policies). It is worth noting that additional factors contribute to low participation rates in cancer research in New Zealand and elsewhere, aside from concerns about winding up in the control group . These include poor trial design, dominance of industry in trial design, and not enough time spent explaining trials to patients. Adoption of a PRC framework could mitigate some of these factors. For example, it was noted above that doctors would likely spend more time explaining trials if the patient had already been offered access to the experimental treatment. On the other hand, as also noted above, the role of industry in funding trials would likely increase, with associated benefits and risks. Disadvantages of the PRC approach The principal disadvantage of the PRC approach is that randomising patients without their consent can seem ethically dubious or inappropriate. Because randomisation is usually conducted as an integral part of standard RCTs, the fact that randomisation occurs without advance consent under a PRC process can appear to constitute enrolling patients in studies without their knowledge or consent. This concern has led to the exclusion of PRC designs as unacceptable to some bodies developing research guidelines.35 In the New Zealand setting, however, comprehensive data from patients listed on the Cancer Registry are already routinely collected for use in research on an aggregated and anonymous basis, without individual consent, as sanctioned by the Cancer Registry Act 1993. Data from thousands of patients have been used in these studies without their individual knowledge or consent for clear scientific, ethical and practical reasons. Nor is there any provision for opting out of contributing ones data to the cancer registry or to other healthcare databases. This statutory provision reflects a decision on the part of New Zealands elected representatives that the moral good of autonomy (e.g., to withhold ones clinical data from the databases) is outweighed, in this context, by the greater public good emanating from health research in terms of prevention and healthcare effectiveness. A similar judgement might be applied to the proposed PRC approach. It is important to acknowledge, however, that studies using New Zealand Cancer Registry data are observational, whilst PRC studies are interventional. The proposed PRC policy detailed here would therefore represent an extension of the doctrine of presumed consent to intervention studies of comparative effectiveness of new versus standard treatments. This extension would provide a further reason for obtaining explicit societal consent, as discussed below. Such societal consent would not obviate the need for each individual proposed PRC study meeting the normal requirements of independent ethical review. A related psychological-ethical challenge to instituting PRC unique to New Zealand is that proposals to modify consent procedures in cancer research are likely to remind some observers of the National Womens Hospital (NWH) cervical cancer saga.36, 37 This event involved a group of women with early forms of cervical cancer (or pre-cancer) being inappropriately just observed, rather than actively treated according to prevailing standards of care at the time. Consent was not obtained from patients who received non-standard treatment (i.e., observation only). In the PRC approach proposed herein, however, control patients wouldreceive standard (often state-of-the-art) treatment and care, and patients randomised to receive non-standard experimental care would be asked to give consent. Furthermore, as with all RCTs, trials would be proposed only when genuine clinical equipoise existed about the benefits of the experimental treatment compared to standard, best-practice treatments. A more pragmatic challenge to use of the PRC approach is that implementation requires the availability of comprehensive data on baseline, treatment, and outcome variables on all patients, including those not offered experimental treatment or who decline offered treatment. Such data will generally be available only where data are routinely collected on all patients with a given condition as part of a large (e.g., regional or national) cohort, registry or database.15 New Zealand is well positioned to meet this requirement in the cancer arena. A final methodological limitation of the PRC approach is that patient blinding cannot be achieved. This limitation is probably not a major drawback because much cancer research is already conducted in an open-label manner. Moreover, blinding is less necessary when mortality (an objective measure) is the major or only outcome, as is often the case with studies of advanced cancer. Societal consent How acceptable would a PRC approach for cancer trials be to New Zealanders? The only way to start answering this question is to ask them. In the first instance, a group of patient representatives, ethicists, and clinicians could examine the issues, with wider citizen deliberation as the next step. The latter could start in a small way by convening focus groups and in-depth interviews with stakeholders, including patients and their advocates, caregivers, and researchers. If the idea were supported at these levels, more detailed mechanisms such as formal citizen juries (used previously in New Zealand38) and formal public consultation could be considered. These activities could perhaps be supported by the National Health Committee, National Ethics Advisory Committee (NEAC), or Cancer Society of New Zealand, amongst other possibilities. The envisioned societal consent process would in effect embed PRC in cancer studies into the national culture, fulfilling the World Health Organizations Good Clinical Practice Principle 7, which states that: Freely given informed consent should be obtained from every subject prior to research participation in accordance with national culture(s) and requirements. Pursuant to societal consent, presumed consent for pre-randomisation consent would be considered in accordance with New Zealands culture and requirements. As noted above, patients randomised to be offered the new treatment would be informed and asked for consent. One topic requiring discussion during public consultation is what to tell patients who have been randomised to not be offered the experimental treatment. If such patients ask if this has occurred, of course they would be told the truth, but what if they dont ask? Some patients will prefer to receive this (arguably useless) information and others not. Presumably doctors could decide whether to volunteer the information if not asked, but guidance on this point from patients and the public would be useful. Conclusion In the most recent available review of PRC studies and designs, Schellings et al39 struck quite a positive note, concluding that: [B]ased on well-defined indications and requirements, prerandomization [PRC] designs have an essential contribution to evidence-based medicine. . . . [M]ethodologically, the prerandomization design seems preferable when: an attractive experimental treatment is involved; the reference is the standard treatment; a sham procedure or placebo cannot be used; and possible contamination caused by outcome measurements may be prevented. It could readily be argued that the cancer research setting in New Zealand meets all of these criteria. At the least, we believe that patients, doctors, researchers, and members of the public should be provided with an opportunity to learn about and to discuss the idea of a PRC approach to cancer research in New Zealand. Clear support and active leadership from these constituencies is probably essential for PRC to be sustainably adopted in this country. Perhaps it will turn out that PRC is an idea whose time has come - again. If so, it could be a defining moment for cancer services in New Zealand.

Fewer than 5% of all patients with cancer in the United States (US) participate in clinical trials,1 and this proportion is probably even lower in New Zealand.2 Almost 30% of trials funded by the US National Cancer Institute fail to enrol even a single patient3 and many trials do little better, as illustrated by a study described in a recent article in the New York Times: ...A new trial [of intensive chemotherapy for colorectal cancer] in the United States has been temporarily suspended so that researchers can find a way to recruit patients. After nearly a year only one patient had enrolled, because people were reluctant to chance winding up in the control group, according to one of the investigators.4 Overall, only about half of all cancer trials manage to enrol enough patients to generate meaningful results, and only one in five studies results in a published report.5 Difficulty with recruitment is also reflected in the steadily increasing costs of cancer studies, which now stand at about US$60,000 per subject.6 The cost of studies and difficulty recruiting patients has led some pharmaceutical companies to scale back their research and development programmes.7 A recent Health Committee Inquiry into improving New Zealands environment to support innovation through clinical trials concluded that several steps should be taken to ensure that New Zealand did not lose its advantage as a good place to carry out clinical trials . Whilst this report made several recommendations it did not address the issue of low participation rates in clinical trials.8 The causes of poor enrolment in cancer trials have been much studied.9-12 Probably the main contributing factor is patients natural reluctance to wind up in the control group , as noted in the above quote. In addition, doctors are often hesitant to inform patients about randomised controlled trials (RCTs) because of the hassle factor involved - especially when half the patients counselled will not be offered the new treatment. Moreover, doctors often cite their dislike of discussions of uncertainty with patients in explaining why they failed to enrol patients in trials.13 Several suggestions have been made for improving enrolment in cancer trials, including paying doctors higher fees for recruiting patients,11 but such measures seem unlikely to increase participation in cancer research to any substantial extent given the dynamics described above. Improving enrolment via post-randomisation consent Another approach gaining increased international attention is to obtain consent for experimental treatment after randomisation, rather than before (as with standard consent protocols).14-17 Post-randomisation consent (PRC) designs, also known as randomised consent or pre-randomisation , were first formally described by Marvin Zelen in the 1970s,18,19 though they had been used previously. PRC designs have been much discussed20-22 and various modifications developed.15, 23, 24 The recruitment process for New Zealand cancer research proposed herein is based on Zelens original single-consent design, under which subjects randomised to be offered the experimental treatment are approached for consent, while subjects randomised to continue to receive current best (or standard) treatment are not approached for consent, since nothing has changed for them. Subjects randomised to be offered the experimental treatment but who decline to receive it would also continue to receive current best practice. Figures 1 and 2 depict the sequence of events under standard RCTs and under the proposed PRC approach. For analysis purposes, the outcomes of patients who decline offered experimental treatment (offered-declined) are usually included with the offered-accepted patients and collectively compared against the non-offered patients (intention-to-treat analysis). Alternatively, the offered-declined group can be amalgamated with the non-offered patients and collectively compared to the offered-accepted patients (treatment received). The former approach has the advantage of preserving randomisation, but estimates of treatment effects will usually be biased towards the null (no effect) because of treatment dilution (i.e., some patients counted as receiving the treatment will not actually receive it). The treatment-received approach, on the other hand, breaks the randomisation and potentially magnifies any systematic bias, although careful data analysis can to some extent mitigate this problem. PRC designs were used in some cancer studies in the 1970s and 1980s. A review by Altman et al25 found 11 such trials, the five largest of which were conducted by co-operative groups: three by the National Surgical Adjuvant Breast and Bowel Project, one by the Eastern Co-operative Oncology Group in the US, and one by the Danish Breast Cancer Co-operative. The PRC design appears to have been used less frequently in cancer research since that time, probably due to the ethical considerations described below. PRC designs continue to be used in a wide range of other clinical contexts, however, as described in two reviews published in 2006, one of 58 PRC studies21 and one of 50 studies.20More recently, PRC was used in a New Zealand study examining the effectiveness of a behavioural treatment for Mori patients who present with self-harm.26 PRC designs are also still widely used in cluster randomised trials, where the units of randomisation consist of medical practices, hospitals, clinics, or communities.27, 28 In such trials, consent is obtained only from patients residing within the units randomised to receive treatment.29 Advantages of the proposed approach The main advantage of the PRC approach in cancer trials is that study recruitment would likely increase substantially, primarily because the offer group would be able to obtain the experimental treatment for sure (i.e., no chance of being relegated to the control group). Also, doctors may be more inclined to take the time to enrol patients if they have already been selected to receive the study treatment. Likely recruitment rates under PRC designs can be estimated by examining previous PRC studies to determine the rate of acceptance by patients randomised to be offered experimental treatments. In the only controlled study examining this question, a Canadian multicentre trial of platelet infusion in premature infants with deficient platelet counts,30 reported that three centres using standard (pre-randomisation) consent designs had recruitment rates of 19%, 26% and 52%. By contrast, a centre using PRC reported 86% recruitment. Comparable acceptance rates were observed by Altman et al in their 1995 review of PRC cancer studies,25 with a median acceptance rate of 88% and inter-quartile range (IRQ) of 84-89%. Similar acceptance rates were seen in two more recent reviews, one of 58 PRC studies, in which the median acceptance rate (out of 39 studies reporting this statistic) was 88% with an inter-quartile range (IQR) of 81-93%,21 and one of 50 studies, in which the median reported acceptance rate was 85% with an IQR of 61-93%.20 Comparable acceptance rates might reasonably be expected in the context of new treatments for patients with cancer in New Zealand, especially when the cancer is at an advanced stage (regional or distant spread). The limited range of effective treatments for most such patients could make uptake of offered new treatment very high - indeed the norm. Coupled with the high incidence of cancer in New Zealand (see Table 1), such robust acceptance rates could translate into hundreds or even thousands of patients being able to try new cancer treatments each year who otherwise would not have access to such treatments. High acceptance rates will also have positive implications with respect to power requirements for comparative effectiveness studies. For example, in a Cox proportional hazards model with survival as the outcome, assuming one-sided statistical significance for \u03b1 of 0.05 and 1-\u03b2 (power) of 0.8, one needs 58 deaths to detect a hazard ratio of 0.5 and 560 deaths to detect a ratio of 0.8.31 Table 1 depicts the most recent statistics on annual incidence and number of deaths for the 10 most common types of cancer in New Zealand. Table 1. New Zealand cancer mortality statistics (ranked by numbers of deaths for the top 10 cancers) Cancer type New registrations (2009) Deaths (2008) Trachea/lung Colorectal Prostate Breast Pancreas Melanoma Stomach Bladder Kidney Uterine 2008 2837 3369 2781 472 2212 372 361 482 436 1634 1280 670 624 373 317 283 200 165 94 Source: New Zealand Ministry of Health (http://www.health.govt.nz/nz-health-statistics/health-statistics-and-data-sets/cancer-data-and-stats). Figure 1. Design of conventional randomised controlled trials (RCTs) Figure 2. Design of proposed post-randomisation consent RCT RCT = randomised controlled trial TR = treatment received ITT = Intention-to-treat Clinical input will be required to determine if and when it is necessary to define standard treatment, to be employed nationally, as opposed to accepting a variety of standard treatments , which can collectively be compared to the new treatment. Depending on the acceptance rate and the proportion of patients within each type who are eligible for a particular experimental treatment (e.g., based on tumour stage or receptor subtype), several types of cancer appear to be associated with a sufficient number of deaths per year to enable detection of a hazard ratio of 0.8, especially if recruitment extends over more than 1 year. Recruitment will likely be greater in the setting of the more common types of cancer and in later stages of cancer and perhaps this is where the focus should be. In some cases, data from New Zealand studies may often need to be combined with data from parallel international studies, especially when narrow inclusion criteria limit enrolment. Ideally such studies would also be of the PRC type and would use similar standard-treatment controls. Bias control\u2014Adamson et al found that 44 studies (out of 58) specified the rationale for using the PRC design: The single largest justification for using the design was to avoid bias (n = 23). Trialists justified using Zelens method: to avoid the Hawthorne effect, to obtain \u2018effectiveness estimates as opposed to efficacy effects, to avoid contamination of the control group or to prevent resentful demoralisation. Few trials (N = 7) explicitly used the approach to increase participant recruitment.21 These bias-avoiding features must be balanced against the concern that patients who reject offered experimental treatment may differ in some significant way from patients who accept the treatment. The effect of any such resultant bias will increase as acceptance rates for the experimental treatment fall, but acceptance rates in the 85+% range, as seems likely based on evidence from previous studies, would probably be sufficient to warrant confidence in the validity of the findings. This is especially true when effect sizes are large and when comprehensive and accurate information on potential confounders (i.e., baseline variables that are significantly correlated with untreated prognosis, such as age, ethnicity, co-morbidities, and stage of cancer), is available on all patients. Additional attention would need to be paid to comparability of acceptors and non-acceptors if the acceptance rate falls substantially below 85%. Fortunately, New Zealand has a world-class system of linked health data, made possible via a national system of health identifiers. Over 98% of New Zealand residents have been assigned a unique health identification number, permitting linkage of individual patient data across multiple databases (e.g., demographic, hospitalisation, prescription drugs, mortality). Critically, New Zealand also has a comprehensive and well-functioning national cancer registry, into which all patients with newly diagnosed cancer (except minor skin cancers) are entered. Information from this registry can also be linked with the larger national databases. A limiting factor in these databases is the lack of information on clinical outcomes other than mortality (e.g., quality of life), although work is underway to expand the breadth of variables collected in cancer databases. Because comprehensive, routinely collected information would be available on all patients\u2014offered-accepted, offered-declined, and non-offered\u2014it would generally be possible to stratify, match or adjust on relevant predictor variables across groups. This sort of cross-group matching is commonly performed via the use of propensity or prognostic scores.32 The ability of PRC designs to take advantage of information on potential confounders was acknowledged by MacLehose et al:13 [The PRC design] allows the possible biases associated with patient selection for the experimental therapy to be investigated. The characteristics of patients allocated to the control treatment can be compared with those allocated to the new intervention who receive the control treatment after refusing the new treatment. If a comparison is made between the \u2018as-treated groups, the role of self-selection [i.e., confounding] can be examined.(p.67) These anti-bias advantages may be offset to some degree by the fact that provision of experimental treatments within studies is often accompanied by enhanced levels of clinical scrutiny (e.g., attention, monitoring), which could itself result in improved outcomes. If such were to occur, estimates of treatment effects could be biased upwards. Although a similar bias can occur in standard RCTs, in the standard design control patients are managed within the same in-study framework as the treated patients (including blinding where possible, though usually not in cancer studies), which could potentially result in more comparable levels of non-treatment-specific therapeutic care. Efficient research platform\u2014In addition to enabling many more patients to enrol in cancer studies, the proposed PRC approach, combined with the availability of New Zealands health databases, would enable cancer trials to be conducted in an extremely efficient and cost-effective manner.33 Indeed, most required data collection would occur automatically (e.g., age, type and stage of cancer, co-morbidities, pharmaceutical usage, mortality). The need for new data collection would not be completely obviated under the PRC approach, however, because safety monitoring and reporting would be required as part of detailed treatment protocols, which would be developed in consultation with appropriate clinical and ethical advisory groups. Moreover, it may be necessary to upgrade cancer-related clinical infrastructure, including hiring additional research nurses and data managers, in order to ensure accurate data collection. Improvement of some features of routine data collection, such as documentation of chemotherapy and radiation therapy, would also likely be required, and efforts in this direction are already underway through initiatives supported by the Ministry of Health and Cancer Control New Zealand. In addition, provision for rapid documentation of adverse effects will be needed, including the ability to stop trials promptly if required. Such upgrades of system capabilities would in themselves be likely to lead to improved quality and cost-effectiveness of cancer care. A further feature contributing to the value of the proposed research platform is that the resulting data would reflect long-term, real-world effectiveness, which is not the case in standard RCTs. These efficiencies could make New Zealand a potential leader in smart and cost-effective trial design for testing new treatments for cancer. If so, not only would thousands of New Zealand cancer patients gain early access to investigational treatments that would otherwise be unobtainable; in addition, the New Zealand research infrastructure could be given a major financial boost. The global pharmaceutical budget for cancer drug research and development is in the billions of US dollars annually34 and a proportion of this funding could potentially be captured. A potential challenge arising from creation of the envisioned enhanced-efficiency cancer research platform is that international pharmaceutical manufacturers might expand their presence and influence in New Zealand, with possible detrimental effect on Pharmacs ability to determine the public subsidy of cancer drugs based on beneficial impact and cost-effectiveness. This could come about through increased patient expectations or political lobbying, for example. This risk could be minimised, however, if Pharmac and the Ministry of Health were to participate in negotiating the terms under which studies are conducted (e.g., user fees, future discounts, pay-for-outcomes policies). It is worth noting that additional factors contribute to low participation rates in cancer research in New Zealand and elsewhere, aside from concerns about winding up in the control group . These include poor trial design, dominance of industry in trial design, and not enough time spent explaining trials to patients. Adoption of a PRC framework could mitigate some of these factors. For example, it was noted above that doctors would likely spend more time explaining trials if the patient had already been offered access to the experimental treatment. On the other hand, as also noted above, the role of industry in funding trials would likely increase, with associated benefits and risks. Disadvantages of the PRC approach The principal disadvantage of the PRC approach is that randomising patients without their consent can seem ethically dubious or inappropriate. Because randomisation is usually conducted as an integral part of standard RCTs, the fact that randomisation occurs without advance consent under a PRC process can appear to constitute enrolling patients in studies without their knowledge or consent. This concern has led to the exclusion of PRC designs as unacceptable to some bodies developing research guidelines.35 In the New Zealand setting, however, comprehensive data from patients listed on the Cancer Registry are already routinely collected for use in research on an aggregated and anonymous basis, without individual consent, as sanctioned by the Cancer Registry Act 1993. Data from thousands of patients have been used in these studies without their individual knowledge or consent for clear scientific, ethical and practical reasons. Nor is there any provision for opting out of contributing ones data to the cancer registry or to other healthcare databases. This statutory provision reflects a decision on the part of New Zealands elected representatives that the moral good of autonomy (e.g., to withhold ones clinical data from the databases) is outweighed, in this context, by the greater public good emanating from health research in terms of prevention and healthcare effectiveness. A similar judgement might be applied to the proposed PRC approach. It is important to acknowledge, however, that studies using New Zealand Cancer Registry data are observational, whilst PRC studies are interventional. The proposed PRC policy detailed here would therefore represent an extension of the doctrine of presumed consent to intervention studies of comparative effectiveness of new versus standard treatments. This extension would provide a further reason for obtaining explicit societal consent, as discussed below. Such societal consent would not obviate the need for each individual proposed PRC study meeting the normal requirements of independent ethical review. A related psychological-ethical challenge to instituting PRC unique to New Zealand is that proposals to modify consent procedures in cancer research are likely to remind some observers of the National Womens Hospital (NWH) cervical cancer saga.36, 37 This event involved a group of women with early forms of cervical cancer (or pre-cancer) being inappropriately just observed, rather than actively treated according to prevailing standards of care at the time. Consent was not obtained from patients who received non-standard treatment (i.e., observation only). In the PRC approach proposed herein, however, control patients wouldreceive standard (often state-of-the-art) treatment and care, and patients randomised to receive non-standard experimental care would be asked to give consent. Furthermore, as with all RCTs, trials would be proposed only when genuine clinical equipoise existed about the benefits of the experimental treatment compared to standard, best-practice treatments. A more pragmatic challenge to use of the PRC approach is that implementation requires the availability of comprehensive data on baseline, treatment, and outcome variables on all patients, including those not offered experimental treatment or who decline offered treatment. Such data will generally be available only where data are routinely collected on all patients with a given condition as part of a large (e.g., regional or national) cohort, registry or database.15 New Zealand is well positioned to meet this requirement in the cancer arena. A final methodological limitation of the PRC approach is that patient blinding cannot be achieved. This limitation is probably not a major drawback because much cancer research is already conducted in an open-label manner. Moreover, blinding is less necessary when mortality (an objective measure) is the major or only outcome, as is often the case with studies of advanced cancer. Societal consent How acceptable would a PRC approach for cancer trials be to New Zealanders? The only way to start answering this question is to ask them. In the first instance, a group of patient representatives, ethicists, and clinicians could examine the issues, with wider citizen deliberation as the next step. The latter could start in a small way by convening focus groups and in-depth interviews with stakeholders, including patients and their advocates, caregivers, and researchers. If the idea were supported at these levels, more detailed mechanisms such as formal citizen juries (used previously in New Zealand38) and formal public consultation could be considered. These activities could perhaps be supported by the National Health Committee, National Ethics Advisory Committee (NEAC), or Cancer Society of New Zealand, amongst other possibilities. The envisioned societal consent process would in effect embed PRC in cancer studies into the national culture, fulfilling the World Health Organizations Good Clinical Practice Principle 7, which states that: Freely given informed consent should be obtained from every subject prior to research participation in accordance with national culture(s) and requirements. Pursuant to societal consent, presumed consent for pre-randomisation consent would be considered in accordance with New Zealands culture and requirements. As noted above, patients randomised to be offered the new treatment would be informed and asked for consent. One topic requiring discussion during public consultation is what to tell patients who have been randomised to not be offered the experimental treatment. If such patients ask if this has occurred, of course they would be told the truth, but what if they dont ask? Some patients will prefer to receive this (arguably useless) information and others not. Presumably doctors could decide whether to volunteer the information if not asked, but guidance on this point from patients and the public would be useful. Conclusion In the most recent available review of PRC studies and designs, Schellings et al39 struck quite a positive note, concluding that: [B]ased on well-defined indications and requirements, prerandomization [PRC] designs have an essential contribution to evidence-based medicine. . . . [M]ethodologically, the prerandomization design seems preferable when: an attractive experimental treatment is involved; the reference is the standard treatment; a sham procedure or placebo cannot be used; and possible contamination caused by outcome measurements may be prevented. It could readily be argued that the cancer research setting in New Zealand meets all of these criteria. At the least, we believe that patients, doctors, researchers, and members of the public should be provided with an opportunity to learn about and to discuss the idea of a PRC approach to cancer research in New Zealand. Clear support and active leadership from these constituencies is probably essential for PRC to be sustainably adopted in this country. Perhaps it will turn out that PRC is an idea whose time has come - again. If so, it could be a defining moment for cancer services in New Zealand.