Helping Patients Make Better Personal Health Decisions: Title and subTitle BreakThe Promise of Patient-Centered Outcomes Research

The new Patient-Centered Outcomes Research Institute has proposed a working definition of patient-centered outcomes research¹ that addresses 4 patient questions: “Given my personal characteristics, conditions, and preferences, what should I expect will happen to me?” “What are my options and what are the benefits and harms of those options?” “What can I do to improve those outcomes?” and “How can the health care system improve my chances of achieving the outcomes I prefer?”

In this issue of JAMA, Alemozaffar and colleagues² present their findings from a study that primarily addresses the first question for men regarding treatment for clinically localized prostate cancer. Does the study answer this question, and what are its implications for the health care system?

The study by Alemozaffar et al² focuses on prediction of functional erections 2 years after treatment with radical prostatectomy, external beam radiotherapy, or brachytherapy. Erectile dysfunction is a common adverse effect many men care about.³ Men who received these therapies were enrolled and underwent follow-up for 2 years, with periodic administration of a validated patient-reported outcome measure focused on sexual function. Baseline patient and tumor characteristics were then used to predict the preservation of functional erections. Prediction of erectile function was more accurate for patients who received radiotherapy than for those who underwent prostatectomy. For radiotherapy, the models correctly discriminated 87% to 90% of the time between men with preserved erectile function and those with impaired erectile function, compared with 77% discrimination for men treated with surgery. The reasons are not known, but the authors suggest that the greater unpredictability might reflect variation in surgeon proficiency.

This study has many strengths. The sample size was reasonably large, and follow-up was relatively complete. Validated self-reported sexual outcome measures were used. Patients were recruited from 9 hospitals, enhancing generalizability. Importantly, the predictive models were independently validated in a separate population, and the findings were reasonably consistent.

However, the study also has several limitations. The 9 hospitals were all linked with academic medical centers, and thus outcomes may be different in lower-volume community hospitals. A better study sample would have been population based, rather than hospital based, although population-based sampling presents many challenges. Selecting 2-year follow-up may bias comparisons across treatments to the extent that the time course for loss of erectile function differs.³ Longer follow-up would be desirable. But that creates another problem: the longer the time between treatment and outcome measurement, the higher the risk that treatment will change such that outcomes may no longer represent what current patients can expect. For example, robotically assisted laparoscopic radical prostatectomy is now ubiquitous, compared to 2003-2006 when the study cohort was assembled, and the outcomes of prostate operations may have gotten better (or worse) as a result.⁴ This problem is greatest for outcomes such as mortality, which may require years or even decades to define with precision, and treatments are often much different over such long time courses.

An important limitation of the study was that no outcomes were provided for men who chose a watchful waiting or active surveillance approach.⁵ Better “natural history” data are needed for many conditions, a fundamental knowledge gap that patient-centered outcomes research must address.

An issue highlighted by the authors is the observational nature of the study, with the possibility that unmeasured confounders may account for some of the differences observed in outcomes across treatments. A randomized trial would minimize the effect of confounding but would ask a fundamentally different question than patients would ask. Rather than asking, “What will happen to me after I choose treatment A or treatment B based on my preferences?” the trial research question would ask, “What will happen if I cannot decide between treatments A and B and am assigned one at random?” The answers to these questions are complementary but different, and both study designs are needed.

How might a man who has undergone therapy for clinically localized prostate cancer use this information? Consider a 60-year-old African American man with a prostate-specific antigen level of 6 ng/mL, a body mass index of 30, and a baseline sexual health-related quality of life score of 83, indicating some sexual dysfunction, as well as a tumor Gleason score sum of 6 (3 + 3). This patient could learn that men like him would have about a 1 in 3 chance of maintaining erectile function after radical prostatectomy if a nerve-sparing operation could be performed but a 1 in 7 chance without it. Given that with these tumor characteristics he would likely not receive neoadjuvant hormone therapy, he could expect a 6 in 10 chance of maintaining erectile function after external radiotherapy and an 8 in 10 chance after brachytherapy. To achieve these outcomes, however, there is a chance that this patient might require some assistance, most likely from a 5-phosphodiesterase inhibitor.

Given the limitations of the study by Alemozaffar et al,² could patients be confused or misled by these data? First, this study is observational, and patients should use the findings cautiously to help choose among treatments. Even though confounding by indication should be less of a problem for adverse effects than therapeutic effects, confounding by unmeasured factors may still threaten the validity of comparisons of adverse outcomes across treatments. Nevertheless, and particularly in the absence of randomized trials comparing the sexual outcomes of surgery and radiotherapy, well-conducted prospective observational studies like this one may indeed provide some useful information for men still weighing their options.

Moreover, the study addressed only 1 outcome, erectile function. Hopefully, these investigators will develop similar predictive models for incontinence. For cancer survival, results from the recently presented Prostate Cancer Intervention Versus Observation Trial (PIVOT) need to be considered and would suggest that conservative management, rather than immediate treatment, might be the best way for a patient like the one presented to preserve erectile function and continence while maximizing survival.⁶ However, if this patient seeks immediate intervention, he might be misled if he were offered treatment by a clinician who could not deliver the anticipated outcomes. The patient also would need to understand the uncertainty inherent in the point estimates, especially for smaller subgroups. Nevertheless, for most scenarios, the take-away message is that if the patient has chosen surgery, he will more than likely lose erectile function, whereas if he has chosen radiotherapy, he has a better than even chance of preserving it, at least for 2 years. Even if the risk of erectile dysfunction continues to increase over time after radiotherapy, it is better to have an undesirable outcome later rather than sooner.

What are the implications of studies like this one for the health care system? In this case, optimizing outcome prediction requires detailed knowledge of baseline sexual function. Routinely collecting objective measures of subjective phenomena like sexual function from patients will need to become part of usual care rather than just research.

More importantly, better ways of ensuring that patients are informed about their choices and invited to participate in making decisions need to be identified and disseminated. Outcomes data for all important options and outcomes need to be integrated in a useful synthesis for the broad spectrum of men facing this decision. A meta-analysis of randomized trials of patient decision aids to support a shared decision-making process for preference-sensitive conditions demonstrated better decision quality when these tools are used.⁷ For the most fateful decisions, coaches or navigators may be helpful and efficient as well.^{8 - 9}

The promise of patient-centered outcomes research will be realized not only when high-quality outcomes data are available for all common medical problems but also when patients are routinely informed and invited to participate in their health care decisions. To achieve this promise, patients must increasingly be encouraged to adopt the position of “no decision about me, without me.”¹⁰