Comparative Effectiveness Research for Prostate Cancer Radiation Therapy

Current Status and Future Directions

Xinglei Shen; Nicholas G Zaorsky; Mark V Mishra; Kathleen A Foley; Terry Hyslop; Sarah Hegarty; Laura T Pizzi; Adam P Dicker; Timothy N Showalter


Future Oncol. 2012;8(1):37-54. 

In This Article

CER for Prostate Cancer RT

PC treatment is a high-impact topic for CER within oncology. PC is the most commonly diagnosed cancer among men in the USA with 217,000 new cases in 2010 and is the second-leading cause of cancer mortality, with 28,660 deaths in 2008.[23,24] The estimated direct cost of PC treatment in 2010 was US$11.9 billion and is projected to increase 38% by 2020.[25] PC patients have multiple options at each stage of disease, including active surveillance (AS), androgen deprivation therapy, surgery with radical prostatectomy (RP) and multiple forms of RT. In most instances, there are no data to demonstrate clear differences in efficacy or in quality-of-life (QoL) outcomes[26] among the treatment options. Treatment decisions are often based on subjective considerations, such as the preferences or biases of patients and physicians. For example, the long duration of external-beam RT (EBRT) for PC has been cited as a primary reason for patients deciding against RT for definitive treatment.[27] Current data demonstrate both overtreatment of low-risk disease and under-treatment of high-risk disease, with wide difference in patterns of care across the USA,[28] suggesting that decisions are often based on physician and patient biases rather than evidence of comparative effectiveness. A better understanding of the effectiveness of each therapeutic alternative in a usual-care setting may improve decision-making.

Comparing the pros and cons of each modality for an individual patient lies at the heart of CER. This article reviews recent efforts in CER regarding RT for PC. We focus on comparison between RT modalities as well as comparison of outcomes between RT and RP. We also discuss examples of cost analyses in RT. Using example CER studies, we highlight the complementary role of CER to traditional efficacy research, the limitations of the current tools for CER and areas of development for the future of CER.

Comparisons Between RT & Other Modalities

RT versus RP While RT and RP are considered to be equivalent in terms of biochemical control and long-term survival, this hypothesis has never been formally tested in a randomized setting. Multiple attempts at randomized comparisons have been abandoned owing to poor accrual, including the Surgical Prostatectomy Versus Interstitial Radiation Intervention Trial (SPIRIT), Medical Research Council (MRC) PR06 and SWOG 8890 trials. Relatively small older studies using suboptimal RT doses reported inferior outcomes with RT as compared with RP.[29,30] In the contemporary era of dose-escalated RT, single and multi-institutional studies have generally reported similar outcomes for all risk groups.[31–33]

Population-based CER studies of outcomes following RT or RP have reported improved PC-specific survival with RP, but their results must be interpreted with caution owing to limitations in adjusting for confounding variables, both measured and unmeasured. Using the Cancer of the Prostate Strategic Urology Research Endeavor (CaPSURE) database, Cooperberg et al. reported that RT was associated with more than double the risk of PC-specific mortality compared with RP (hazard ratio [HR]: 2.21; 95% CI: 1.50–3.24).[34] Concerns were raised regarding the study in subsequent letters to the editor that highlighted deficiencies in adjustment for confounders, which would lead to overestimation of the effectiveness of RP compared with RT.[35,36] A study of a cohort of older men in the Surveillance, Epidemiology and End Results (SEER)-Medicare linked database, which included patients with localized PC diagnosed in 1992 and treated with RT or RP showed similar findings: an estimated 10-year PC-specific survival of 93.8 versus 98.1% (p < 0.0001) in favor of RP.[37] However, a subsequent comparative study using the SEER database did not identify a difference in survival for men with low-to-intermediate-risk PC treated with RP versus brachytherapy (BT).[38]

A common criticism of these observational studies has been that the typical RT doses used in the timeframe of these studies would be considered suboptimal by current standards. An inherent source of bias in these analyses is that up to 30% of patients with high-risk pathologic features would go on to receive adjuvant RT and are excluded from analysis, while these patients would be included in the radiation-alone group.[39] Furthermore, the demographics of those who receive RT versus RP are different, as RT tends to be used in older patients with higher-stage disease, and more comorbid and chronic illnesses.[40,41] And, in addition to measured confounders, such as age, comorbidities and tumor stage, there may be additional, unmeasured confounders associated with treatment selection that may limit the ability of observational studies to estimate treatment effects for RP versus RT. In other words, most methodological flaws in these analyses relate to nonrandom treatment allocation and the potential for selection biases; for instance, higher stage predicts both for use of radiation and for poorer survival. CER studies using observational data often utilize statistical methods to adjust for selection biases, such as traditional regression, propensity score and instrumental variable analyses.[42,43] Regardless of the statistical methods used to adjust for confounders, comparative studies of RP versus RT should be interpreted with attention to selection biases such as tumor stage, age and comorbidities.

Prospective, controlled trials in PC have also compared the toxicity and QoL after initial treatment with RT versus RP. In a prospective study of QoL and toxicity after definitive therapy for PC, RP had a higher incidence of sexual dysfunction and urinary incontinence, but a lower incidence of urinary irritation or obstruction, bowel or rectal dysfunction, and hormonal dysfunction, when compared with EBRT or BT.[44] SPIRIT was designed to compare patients treated with BT or RP, but was closed prematurely owing to poor patient accrual. An analysis of QoL measures for patients enrolled in SPIRIT suggested improved QoL in the urinary, sexual and overall satisfaction domains for BT compared with RP.[30] In summary, there are limited data from prospective trials that compare QoL and toxicity outcomes after RP versus RT.

A number of noncontrolled, observational studies have also attempted to use population-based data to investigate differences in toxicity after RT or RP. In a report of 5-year outcomes for men with localized PC enrolled in the Prostate Cancer Outcomes Study, RT was associated with higher rates of bowel urgency and painful hemorrhoids, and lower rates of erectile dysfunction and urinary incontinence, compared with RP.[45] In their systematic review of the effectiveness and harms of treatments for localized PC, Wilt et al. observed that adverse event severity varied between studies and that comparison of the harms of PC treatments was limited owing to the lack of sufficient high-quality data.[46]

RT versus AS AS has emerged as a viable treatment option for low-risk PC,[47] and is considered a preferred management option for selected men in the National Comprehensive Cancer Network guidelines.[48] There is evidence to suggest that AS, when performed according to accepted guidelines, is not associated with inferior outcomes compared with aggressive treatment.[47,49] However, concerns have been raised regarding current diagnostic tools available to select patients appropriate for AS, and there is a need for improved biomarkers and imaging techniques to better distinguish indolent from aggressive cancers.[50] Although AS is appealing because it avoids overtreatment and may delay, or avoid altogether, the complications of aggressive treatment, exposure to the toxicity and complications of PC treatment, less than 10% of eligible men choose AS over aggressive treatments.[40] Decisions in this setting are strongly influenced by the opinions of physicians and patient perceptions of treatment-related toxicities from aggressive treatments.[51] In a decision analysis of risks and QoL benefits of AS compared with aggressive treatment (RP, RT or BT) for low-risk PC, published in JAMA in 2010, AS was associated with greater quality-adjusted life expectancy than other treatment options. Although the gains in expected quality-adjusted life expectancy in this situation were sensitive to individual preferences for AS versus aggressive treatment, this study supports AS as a reasonable alternative to other treatment modalities, including RT, for management of low-risk PC.[49] In summary, for low-risk PC, AS is a reasonable option that is currently underutilized, and there is an opportunity to improve QoL and reduce overall PC treatment burden by developing data that will allow patients to understand the benefits of AS in certain situations.

Comparisons Among RT Modalities

A number of different techniques are available to deliver RT for PC, and the treatment choices vary based on risk group. RT may be divided simply into EBRT and BT. EBRT may be delivered either using x-rays (photons) with 3D conformal RT (3D-CRT) planning or intensity-modulated RT (IMRT) planning, or using particle therapy such as in proton-beam therapy (PBT). Since the introduction of IMRT and the accompanying CMS Healthcare Common Procedure Coding System level 1 billing code for its use in 2002, the utilization of IMRT has increased dramatically.[52] Interest in PBT has also accelerated in recent years.[53] Despite the rapid adoption of these new technologies, little efficacy data have been generated to substantiate the clinical benefits and value of IMRT or PBT rather than 3D-CRT.[46] In this section, we will review current CER comparisons among radiation modalities.

3D-CRT versus IMRT The primary rationale for IMRT is to allow for dose-escalated RT by minimizing the volume of rectum exposed to high RT dose. Multiple RCTs have established the role of dose escalation to improve biochemical control, although not overall survival, in patients with localized PC.[54–58] Higher doses are associated with lower rates of positive post-treatment biopsies, which are correlated with improvements in subsequent biochemical control.[59,60] However, pivotal studies that established the efficacy of dose escalation were performed using 3D-CRT techniques. Since higher RT doses may also result in higher rates of severe urinary, rectal and sexual complications,[61] IMRT has been embraced in the clinic with the intent to reduce rectal and urinary toxicity. Despite clear dosimetric advantages in treatment planning,[62,63] the value of IMRT has not been validated in a RCT. With the rapid diffusion of IMRT into clinical practice, it is unlikely that any randomized trial will ever be attempted to specifically test the benefit of IMRT for PC. It is unlikely that patients or physicians would agree to participate in a RCT that randomizes patients to IMRT versus 3DCRT, since IMRT is so widely available and has some inherent dosimetric advantages. However, the added expense of IMRT creates a need for careful evaluation of this technique. Therefore, CER using real-world observational data will be necessary for providing data that are critical for decision-making.

Using the SEER-Medicare-linked database, investigators examined rates of new-onset urinary, erectile and bowel complications after treatment 3D-CRT or IMRT for patients with PC diagnosed from 2002 to 2004. In this population-based CER study, they found a reduction in the cumulative incidence of aggregate bowel complications from 22.5 to 18.8% (HR: 0.86; 95% CI: 0.79–0.93), and nonsignificant reduction in urinary complications from 11.2 to 10.4% (HR: 0.93; 95% CI: 0.83–1.04), with IMRT versus 3D-CRT.[64] In a separate analysis of the SEER-Medicare database, however, the rates of grade 3 gastrointestinal toxicities were lower, but IMRT was again associated with a modest decrease in toxicity rates. For patients with T1–T2, localized PC, diagnosed from 1992 to 2005, rates of grade 3 gastrointestinal toxicities were 4.8 and 3.3% for patients treated with 3D-CRT and IMRT, respectively.[65] Both studies have the typical limitations of observational cohort studies, including measured and unmeasured sources of bias,[42] and there are additional limitations regarding the reliability of claims data for measuring RT technical details[66] and the complications of cancer treatment.[67] There continues to be a need for high-quality evidence to evaluate the comparative effectiveness of IMRT versus 3D-CRT.[46]

There are complementary sources of evidence that provide an attenuated level of support for IMRT. A European study of QoL, as measured by the SF-36, EORTC-QLQ-30 and EORTC-QLQ-PR25 survey instruments, of patients treated with either IMRT to 76 Gy or 3D-CRT to 70 Gy found less deterioration in QoL between 1 and 6 months with IMRT compared with 3D-CRT, despite use of a higher dose.[68] IMRT also appears to permit dose escalation for PC with minimal increases in toxicity, based on prospective, nonrandomized data. The Memorial Sloan Kettering Cancer Center has pioneered extreme dose escalation using IMRT up to 86.4 Gy, a dose not feasible with 3D-CRT.[69] Whether these efforts towards extreme dose escalation will translate into improved survival outcomes, when evaluated in a larger cohort, remains unanswered.

3D-CRT/IMRT versus PBT Although PBT has been used to treat PC for decades, interest in PBT has increased in recent years. In theory, PBT has a physical advantage over photon-based treatments due to the rapid drop-off in dose beyond the proton Bragg peak, which can provide better sparing of normal tissues in some situations.[70] Whereas dose-escalation studies from RCTs that used 3D-CRT reported grade 3 gastrointestinal and urinary toxicity rates in the range of 1.0–6.9% for the high-dose arms,[71–73] the late toxicity rates observed in the PROG 95–09 study, which employed 50.4 Gy of conventional photon RT plus a proton boost of 19.8 GyE or 28.8 GyE, were 2% in both arms.[74] In a prospective trial of PBT to a cumulative dose of 82 GyE, Mendenhall et al. reported a 1.9% rate of grade 3 urinary toxicity and a <0.5% rate of grade 3 gastrointestinal toxicities.[75] The early experience at Loma Linda University (CA, USA) for treatment of 1961 patients with PC also reported a low 1.2% rate of ≥grade 3 late toxicity.[76] The estimated rate of late-onset ≥grade 3 at 18 months on the American College of Radiology (ACR) 0312 trial, a Phase II study of proton therapy to 82 GyE for localized PC, was 6%.[77] Although these limited prospective data are available, there have been no RCTs to compare survival or toxicity after PBT versus photon therapy with 3D-CRT or IMRT.

In the absence of direct evidence from RCTs, observational studies have been conducted to compare the effectiveness of PBT to 3D-CRT and IMRT. An evaluation of the SEER-Medicare database, which included patients diagnosed with T1–T2 PC, showed that receipt of PBT was associated with an 8.5% risk of severe gastrointestinal toxicity at 4 years, compared with 4.8% for 3D-CRT (HR: 2.12; 95% CI: 1.45–3.13) and 3.3% for IMRT (HR: 3.32; 95% CI 2.12–5.20).[65] However, caution must be used with interpreting these results because only 335 patients in the study received PBT. Absence of dose information and other undocumented confounding variables may further bias the results of this study. Finally, the majority of PBT cases likely originated from a single-proton facility, since five out of six proton facilities in operation during the timeframe of the study existed outside of SEER areas, and the findings may not be generalizable to this modality as a whole. QoL comparisons between PBT and photon therapy are also lacking. Coen et al. reported QoL data for the ACR 0312 study, and in general, declines in the urinary, sexual and bowel domains are comparable to reports from patients undergoing treatment with photons.[77] Additional studies are needed to characterize the effectiveness and potential harmful effects of PBT, particularly in light of the concerning findings from the above analysis of the SEER-Medicare database.[78]

EBRT versus BT For those patients who are eligible to receive it, BT presents an attractive alternative to EBRT or RP. Typical indications for BT in PC are limited to patients with low-risk localized PC and select patients with intermediate-risk PC.[79] Early retrospective evidence suggested that, when compared with EBRT or RP, BT provides equivalent biochemical control rates in patients with low- and intermediate-risk PC, but inferior biochemical control rates among patients with high-risk localized PC.[80] Subsequent reports from high-volume BT centers have described biochemical control rates that are similar to or superior than those reported for EBRT for all risk groups of localized PC.[32,81–83] Case-matched, retrospective comparisons of PBT and BT have also suggested equivalent biochemical control rates.[84] A prospective cohort study conducted by a team from Seattle (WA, USA) has suggested a biochemical control benefit to the combination of BT with supplemental EBRT for the treatment of patients with high-risk PC.[85] These studies provide data to support the efficacy of BT as at least equal to EBRT techniques.

However, it has also been demonstrated that outcomes after BT are highly operator-dependent[86,87] and the published data from expert BT centers that deliver the highest-quality implants may overestimate the effectiveness of BT in many real-world settings. Observational data capturing BT versus other treatment options are limited, however, and details of BT quality indicators are not available in large registries such as the SEER database. Our group recently compared PC-specific mortality rates after BT, BT combined with EBRT and EBRT alone for patients in the SEER database with high-risk PC. The results of our analysis indicate that that the use of BT was associated with lower rates of PC-specific mortality.[88,89] These results should be interpreted with caution as the effectiveness of BT may be overestimated owing to selection biases. However, these findings highlight the need for well-designed, prospective studies that evaluate the efficacy and/or effectiveness of BT in men with high-risk PC. This work highlights the complementary interaction between observational data and prospective studies. Just as efficacy studies (e.g., RCTs) may be followed by observational CER studies designed to verify effectiveness under usual-care conditions, CER studies using observational data may also generate hypotheses that may be best evaluated in RCTs.

Because the toxicity profiles differ between BT and EBRT, a thorough comparison of toxicities may help patients to make informed decisions regarding treatment. In the prospective QoL study reported by Sanda and colleagues, BT had a lower incidence of rectal toxicity, a similarly low incidence of urinary incontinence, a higher risk of urinary irritation or obstruction, and higher risk of impaired vitality, compared with EBRT.[44] In an analysis of the SEER-Medicare database, Kim et al. found that the lowest rates of late gastrointestinal toxicity for men treated with RT were observed among those who were treated with BT.[65] These observations are useful when discussing treatment options with patients.

Cost–effectiveness & Cost–benefit Analyses

Despite a relative paucity of data to confirm improved survival or reduced toxicity, advanced technologies, such as IMRT and PBT, have been adopted rapidly by radiation oncologists and patients. From 2002 to 2005, the use of IMRT among Medicare beneficiaries for PC increased from 29 to 82%.[52] Based on calculated total Medicare reimbursements for each treatment modality, the authors estimated the additional cost of widespread IMRT utilization, rather than 3D-CRT, for PC treatment from 2002 to 2005 was US$891 million in the USA, with costs rising sequentially each year. While such sensational figures suggest uncontrolled spending on IMRT, these numbers must also be viewed in the context of changes in reimbursement rates. As the price of IMRT decreased each year, the differential cost between IMRT and 3D-CRT similarly narrowed from $15,000 in 2002 to $11,000 in 2005.[52] Since costs associated with medical interventions can change very quickly as a result of CMS policy changes, interpretation of cost trends must be interpreted carefully and extrapolation to later years may not be reliable in such analyses.

Although PCORI specifically does not allow consideration of cost in CER, the rapidly rising costs of care in oncology have stimulated interest in cost–effectiveness and comparative effectiveness analyses. Analyses of the cost–effectiveness of interventions often include evaluating the incremental cost per QALY.[90] In the radiation oncology literature, Konski has performed several cost–effectiveness analyses to evaluate the cost per QALY associated with RT approaches for PC. In brief, the methodology used in these studies consists of determining; based on interviews of a representative sample of men with PC, the relative utility (i.e., global QoL) attached to each health state (i.e., free of cancer, cancer recurrence and toxicity).[91] A model is then developed, using data from previously published studies, to compare the benefits and harms of the treatment options. Combining the value of each health state and the likelihood of each health state allows the investigator to derive the difference in QALYs.

Konski et al. have estimated that IMRT results in an incremental cost of $40,101 per additional QALY gained, compared with 3D-CRT,[92] and that PBT results in an incremental cost of $63,578 per additional QALY gained, compared with IMRT, for the base-case of a 70-year-old man with intermediate-risk PC.[93] Such models heavily rely on the quality of the input data regarding patient-reported preferences for health states, efficacy of treatments, rates of toxicity and costs of treatment. Despite these limitations, these models can be useful in CER. One recent example is the Markov model, reported by Elliott et al., which found that adjuvant RT was associated with slightly higher PSA recurrence-free, metastasis-free and overall survival, but lower QALYs, when compared with observation plus salvage RT over a 10-year time horizon. This model is a useful tool for informing current decision-making while awaiting the results of RCTs conducted to compare these two treatment approaches.[94]

It is not clear what the role cost considerations will play in CER for PC RT in the USA in the future. Although PCORI is prohibited from funding research with cost-centered end points such as cost per QALY,[202] the rising costs of RT for PC[52] and federal budget concerns in a depressed economy may result in future emphasis on cost–effectiveness analyses of PC RT interventions. The challenge for investigators interested in studying the cost–effectiveness of PC RT approaches, such as PBT or IMRT, in the USA will be to design studies that provide outcome measures that reflect both the importance of an individual's desire to choose among treatment options and the needs of payers, healthcare systems and policymakers to understand the comparative effectiveness and value of treatment options.

Methods for Comparative Effectiveness Analyses

Although they have traditionally provided the highest level of evidence for evaluation of the efficacy of medical interventions, RCTs have been criticized for being inefficient, costly, not representative of usual-care setting, and not inclusive of all treatment options.[95] CER is intended to provide the evidence that decision makers need;[95,96] therefore, CER research tools and methods should be aimed at providing useful, relevant information on treatment alternatives in real-world settings.

CER approaches applicable to radiation oncology, including pragmatic clinical trials, adaptive trials and observational studies, were recently described[12] and will be reviewed briefly here. To evaluate the effectiveness of a medical intervention in a usual-care setting, pragmatic clinical trials relax the usually strict entry and treatment criteria common in efficacy trials, and compare the effectiveness of treatments under usual care conditions.[97] This trial design has been used successfully to determine effectiveness in multiple other areas of research; for instance, diabetes management[98] and depression.[99] Pragmatic clinical trials, however, often need to enroll large numbers of patients to meet prespecified end points, and can, therefore, be very expensive and may take just as long as traditional RCTs to conduct. Adaptive trials offer flexibility, as protocols may be adjusted at interim points based on new evidence from the trial itself or other sources of evidence.[12,100] Adaptive trials are viewed as useful tools for evaluating technological advances, since protocols can be modified to include new technologies without closing the trial.[12]

Pragmatic clinical trials and adaptive trials are particularly appealing for PC, since many variables can change during the study period that limit the relevance of the study findings to clinical practice. One illustrative example is the limited implementation of the results of three RCTs of adjuvant RT versus observation for patients with adverse pathological features after RP. Although adjuvant RT was shown in these RCTs to improve PSA relapse-free survival,[101–103] distant metastasis-free survival and overall survival,[5] less than 20% of qualifying patients in the USA actually receive adjuvant RT.[104–106] Why, then, is there the limited utilization of adjuvant RT in the face of supportive evidence from three RCTs? In the time since these trials were designed, the advent of PSA testing transformed the definition of adjuvant RT and created the possibility of closely monitoring patients with serial PSA testing after RP. Moreover, new evidence in support of salvage RT (compared with observation) became available. Therefore, although the trial compared adjuvant RT to observation, the current decision instead involves choosing between adjuvant RT and close observation with delayed salvage RT at the time of a biochemical recurrence. In current practice, salvage RT approaches are often preferred over adjuvant RT in order to reduce overtreatment.[7] In retrospect, an adaptive clinical trial of adjuvant RT could have been designed instead that would have permitted protocol revision to incorporate enrollment, and a policy for the use of salvage RT in the observation arm, based on PSA testing.

Observational studies are important tools for CER. The power of observational databases lies in their broad generalizability and large number of cases. There are a variety of data sources for observational studies in CER, each type with its own set of strengths and limitations ( Table 1 ). However, since these data originate, for the most part, from nonrandomized studies, careful attention must be paid to minimize the effects of confounding variables. In contrast to randomized studies, assignment to a given treatment in observational cohorts are biased by selection factors that may also influence outcome.[8] For instance, patients with higher stage cancer are more likely to receive RT and to die from PC independent of treatment modality. Patients who are treated with RT tend to be older, have higher-stage disease and more comorbidities, all of which portend worse survival. Evaluation of adjuvant RT is particularly thwarted by confounding by indication, since patients who have more advanced tumors are more likely both to receive adjuvant RT after surgery and to have a worse overall prognosis, regardless of treatment received.[11]

One technique to adjust for selection biases is to use propensity score matching, which is designed to account for measured covariates that are associated with increased likelihood of receiving the treatment.[107,108] However, propensity score analysis is not capable of accounting for confounders that are unmeasured or unavailable in the database, such as comorbidities or smoking status. For example, in the SEER database, only very limited PSA data are available, despite its prognostic significance, so PSA is an unmeasured variable that would be expected to affect treatment selection. Another technique to adjust for selection bias is instrumental variable analysis (IVA). IVA uses a variable ('instrument'), which is correlated to the treatment selection but not with the outcome, to control for both known and unknown confounders.[43,109,110] When an appropriate instrumental variable is available, IVA has been shown to be useful in PC for the estimation of treatment effect.[43] These statistical tools, and others, for CER improve the reliability of observational studies, and familiarity with these methods is essential for effective analysis of observational data for CER studies pertaining to prostate RT.

CER to Inform an Individual Patient's Decisions

Central to patient-centered outcomes research is the concept of CER that informs individual decisions, which brings the potential to improve the health and safety of each patient by providing the patient and physician an individualized estimate of the effect of a medical intervention.[111] To provide evidence that is relevant to individual-level decision-making, as opposed to population-level data aimed at policy makers, one of the areas of focus for current CER is to compare medical interventions while accounting for the clinical heterogeneity of a given population.[112] Heterogeneity among patients arises from differences in patient comorbidities, care facilities or context, providers and biological differences that can result in different outcomes after a medical intervention.[9] By incorporating such potential sources of heterogeneity, high-quality CER in PC can better inform the decisions of individual patients. In this section, we review a few potential data considerations that may influence patient outcomes and should be addressed in future CER efforts to evaluate PC RT.

PC staging and risk stratification is important to categorize the disease severity, estimate prognosis, recommend treatment, and to aid healthcare providers and researchers in the exchange of information about patients.[113] As new factors are identified that refine risk stratification, it will be important to include these new factors in CER to better understand clinical heterogeneity in observational studies and to evaluate treatment effect based on all factors. At present, nearly all risk-factor stratification systems incorporate the American Joint Commission on Cancer (AJCC) 7th edition[114] T-stage, Gleason score and PSA value to determine appropriate treatment. However, not all of these risk-stratification factors are routinely available in large databases such as SEER-Medicare or linked registry-claims data.

One example of the potential for evaluation of treatment effect based on an expanded list of prognostic factors is the role of androgen deprivation therapy. Multiple trials have evaluated the addition of androgen deprivation therapy to RT for patients with high-risk PC.[115–118] However, these trials are older and primarily included patients with advanced T-stage. By contrast, contemporary PCs are more likely to be defined as high-risk by virtue of Gleason score.[119] CER that incorporates multiple prognostic factors and treatment-related factors, in addition to AJCC staging, can help to estimate the effectiveness of androgen deprivation therapy among contemporary high-risk patients. CER studies that incorporate data on evolving trends can help to re-evaluate existing prognostic models in PC as well. Factors, including the percentage of positive cores, prostatectomy Gleason score, seminal vesicle invasion, absolute pre-RT PSA and pre-RT PSA doubling time, are emerging as important determinants of outcome following RT. These factors should be integrated into future staging systems.[120,121] For instance, Radiation Therapy Oncology Group 94–13, which was designed in the mid-1990s, used the Roach formula:[122]

to select a population predicted to have a 15% risk of pelvic lymph node metastases.[117] Today, however, PSA is more often used to diagnose PC and, as the percentage of patients with lower T-stage (i.e., T1–T2a) increases, the Roach formula overestimates the risk of lymph node metastasis.[123] In summary, CER that incorporates a wide range of prognostic and treatment-related data can be a useful tool for re-evaluating established medical interventions in PC.

In the future, newer-generation prognostic factors and biomarkers may become important for medical decisions in PC, and CER tools should be refined to include these considerations. For example, more advanced calculation of PSA kinetics[124] and interpretation of Gleason score have been recommended[125] for prognostication. For PC RT, new biomarkers may be evaluated in CER to predict the probability of treatment success with regard to recurrence, survival outcomes or toxicity. Such promising biomarkers include IL-16,[126] Bcl-2 and Bax,[127] the bcl-2:bax ratio,[128] p53, p21/waf1,[128] COX-2,[129] androgen receptor density,[130] androgen receptor coregulatory proteins[131] and the p450 system.[132] These factors, and additional ones yet to be identified, should be integrated into CER in PC in order to provide individualized information for patient decisions.

Finally, it is important to consider the preferences and desires of individual patients when designing CER studies and interpreting their results, since what a patient ought to do based on CER-generated evidence is not always the same as what a patient may want to do.[133] This is particularly relevant for PC treatment decisions, where patient preferences and perceptions are often as integral to decision-making as the clinical situation.[134] Although patient-centered outcomes research that includes individualized CER evidence, and includes the most comprehensive list of clinical and biological considerations, may be helpful for determining which treatment is best for a patient, it will remain essential to consider patient preferences and beliefs.

Developing Better Tools for CER in Prostate Cancer RT

To improve the effectiveness and efficiency of CER for PC and PC RT, a high priority must be placed on: developing, funding and conducting trials designed to compare effectiveness, such as pragmatic clinical trials and adaptive trials; creating high-quality PC registries designed to incorporate a wide-array of data types and sources; expanding the scope and quality of existing databases for observational studies; and improving the medical informatics system for PC CER.[12,135] Pragmatic clinical trials and adaptive trials can provide estimates of the effectiveness of PC treatments, while potentially adjusting for some limitations created by the prolonged time from study design to results, which is typical of PC trials. As the diagnostic and contextual landscape shifts over time in PC, such as occurred with the advent of PSA testing, these newer CER trial types may be designed to allow adjustments that will permit evaluation of treatment effect despite these changes. Efforts to create high-quality PC registries and to improve existing databases will also improve the potential for meaningful CER for PC RT. Medical informatics will be essential for such improvements in the quality of data for CER, as well as for improving the implementation of CER findings (Figure 3).[135] By connecting various sources of data in a useful manner, and by improving access to and analysis of these data, medical informatics can enhance the capacity to conduct CER.[135]

Figure 3.

The areas of focus of medical informatics applied to the information-based needs of comparative effectiveness research.
CER: Comparative effectiveness research; NLP: Natural language processing.
Reproduced with permission from [135].

Meaningful CER in PC RT has some unique challenges that should be addressed. Radiation details in databases for observational studies, such as the SEER-Medicare database,[66] are currently limited, and future efforts should focus on enrichment of radiation technical details in databases and registries designed for CER. Information should include RT doses and number of fractions, dates of radiation delivery, immobilization devices used, image-guidance methods, type of RT (i.e., 3D-CRT and IMRT), quality assurance methods and acute toxicities observed. Ideally, such databases would also be designed to include imaging and dose-volume data, since dose-volume considerations are important for estimating clinical heterogeneity in studies designed to evaluate radiation-induced normal-tissue toxicity.[16] Furthermore, it will be essential to incorporate data from image-guidance approaches, since the availability of online imaging for PC RT has become important. Currently, patients who receive IMRT, 3D-CRT or PBT undergo daily imaging to confirm positioning prior to treatment. Image-guidance permits reduction of treatment margins and a mechanism to verify positioning, so the imaging data should be included in RT registries. Similarly, biomarkers should be incorporated in databases and registries for PC CER, since these factors may influence PC control and normal-tissue toxicity.[16,136]

Although the Institute of Medicine identified treatment options for localized prostate as the highest-priority prostate topic for CER in the USA,[1] it will be important to focus on several additional issues pertinent to PC RT. At each point in a patient's treatment and follow-up, there are decisions regarding PC RT that warrant evaluation using CER methods. The availability of multiple RT modalities, including emerging techniques, such as stereotactic body RT and PBT, have created new needs for high-quality CER to provide individualized data for patient decisions. Methods of immobilization, treatment planning, quality assurance, image-guidance and delivery should also be evaluated in CER studies. For practicing radiation oncologists, as well as for patients, these additional aspects of PC RT should be explored to optimize treatment and tailor radiation techniques for individual patients and clinical settings. Interestingly, cooperative groups, such as the Radiation Therapy Oncology Group, who have traditionally been most interested in conducting RCTs to evaluate the efficacy of cancer treatments in controlled settings, may be particularly well suited to lead efforts for CER in PC RT, since the infrastructures already exist for central collection of imaging and dose-volume information through digital data submission and of biospecimens for correlative studies in clinical trials.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.