• © 2010 by American Society of Clinical Oncology

Simple Rules Can Improve Prognostic Accuracy

A new decision aid, a clever mnemonic, or simple rules of thumb can add precision, clarity, and focus to clinical dialogue. Computer-based decision tools that provide reliable estimates of risk reduction with adjuvant therapy help to anchor difficult conversations in solid statistical estimates of benefit.^1,2 Adjvuvant! Online³ was quickly adopted by oncologists in the United States and abroad because it is simple, clear, and readily available. Colorful bar graphs provide visual estimates of risk reduction with various adjuvant therapies; these estimates are dramatically contrasted with the natural history without intervention. The program enables the clinician to show both relapse and mortality figures and to help a patient understand how source data was used to generate the estimate. It can strengthen the collaborative nature of the physician-patient relationship by modeling how we routinely sort through statistics and can provide a useful platform for deliberation in search of the best treatment for an individual patient. Numeric estimates and diagrams assist clinicians in their efforts to find the language to convey important information and diminish the anxiety experienced by both the physician and patient when discussing uncertain outcomes.⁴ Similarly, simple communication skills, tasks, and protocols improve clinical performance and may contribute to better outcomes overall. More than 10 years ago, Baile and Buckman⁵ introduced a catchy mnemonic known as SPIKES to provide a cognitive road map for discussions of difficult news. Six easy steps were presented in a straightforward manner that was designed to minimize the physician's anxiety by directing the physician's focus to what is best for the patient. Both of these tools, Adjuvant! Online and SPIKES, were designed to respond to real clinical concerns and have continued to prove useful after years of scrutiny and critique.

There is less agreement and perhaps greater variation in practice with respect to the communication of prognostic information and estimates of life expectancy. Philosophic arguments for and against complete disclosure and patient surveys shed some light on this challenging issue, and there are helpful suggestions to improve communication.^6–8 Clearly, new research and insight are needed to help oncologists work through these difficult conversations and optimize both the quantity and delivery of such vital information. The fact remains that many patients with incurable cancer live for months or years, but it is also true that most eventually die of their disease. Without realistic expectations and time frames, some patients may underestimate or overestimate their life expectancy and make poor choices that they may live to regret or that result in a complicated bereavement for their loved ones.⁹

The desire to protect patients by limiting prognostic information is not driven by available evidence. On the contrary, it is a gut reaction. The literature suggests that there is a large variability in the frequency of prognostic discussions in the metastatic setting.¹⁰ We know little about how the information is conveyed and how it is received and processed by patients and families. An interesting study of parents of dying children showed that they remained more hopeful if fully informed, even though the information itself was devastating.¹¹ Perhaps this reflects the degree of support provided by multidisciplinary teams and the importance of recognizing the varying and multiple needs of family members who support the patient. In the world of adult oncology, most patients welcome prognostic estimates when the information is good, but when the news is less favorable, many prefer to know as little as possible. Decision aids such as computer-generated bar graphs, pie charts, and diagrams, which are so helpful to patients who are deciding between surgical treatments or contemplating choices for adjuvant therapy, may be threatening when life expectancy is short and available treatments cannot offer significant extensions. Physicians may use verbal acrobatics to highlight interventions and symptom management and steer clear of discussions of survival or disability. Qualifiers and disclaimers abound; we rush to reassure and often miss opportunities to explore the patient's understanding and validate their emotions.

To assist clinicians and to standardize practices, Australian researchers published consensus and evidence-based guidelines for communicating prognoses.¹² The starting point for these guidelines is respect for the individual's right to know crucial health information. Included are clear recommendations to use mixed framing (ie, the chances of improvement or cure should be mentioned first, then chance of relapse) and a range of time estimates rather than specific end points. The guidelines emphasize the importance of hope-giving aspects of information, such as the existence of extraordinary survivors and the value of personalizing the discussion to include the effect of the cancer on the individual patient's quality of life. As oncologists struggle to strike a balance and protect patients, respect patients' autonomy, and honor their commitment to tell the truth, they may find some handy tools to improve their prognostic accuracy buried in survival curves.

In this issue of Journal of Clinical Oncology, Kiely et al¹³ report the results of a comprehensive review of published survival estimates for patients who have participated in clinical trials of first-line chemotherapy for metastatic breast cancer (MBC). Using the register maintained by the Cochrane Breast Cancer Group, they identified 27 suitable trials that they supplemented with another nine trials identified through Medline. The 36 selected trials included 78 survival curves and 13,083 patients. The median follow-up for the 36 trials ranged from 10 months to 102 months with a median of 29 months. Most trials had two treatment groups, and the majority of the women enrolled were postmenopausal with visceral metastases and an Eastern Cooperative Oncology Group performance status of 0 to 1. Approximately half of the patients had tumors that expressed estrogen receptors, and in six trials, trastuzumab was administered along with chemotherapy to patients with HER2+ tumors. The authors analyzed the survival distribution and more specifically the median (ie, the middle value in a group ranked from smallest to largest) and the interquartile range, which extends from the 25th percentile to the 75th percentile. Typically, 50% of observations are within the range; 25% are higher and 25% are lower. Kiely et al found that the distribution of the survival times fell within a narrow range. They estimated the mean value of the interquartile range in months for four identified scenarios: the best case, worst case, and lower and upper typical cases. They found that the median and mean times for each percentile (or scenario) were almost identical.

The authors then used the data to generate simple rules of thumb that provide reliable quantitative estimates of selected scenarios. The median × 0.25 accurately estimated the 90th percentile (worst case) in 73% of curves, the median × 0.5 accurately estimated the 75th percentile (lower typical) in 95% of curves, the median × 2 accurately estimated the 25th percentile (upper typical) in 95% of curves, and the median × 3 accurately estimated the 10th percentile (best case) in 96% of curves. The accuracy of using simple multiples of the median survival was independent of the duration of median survival. For this group of patients who participated in trials of first-line chemotherapy for MBC, the median overall survival was typically about 3 × the median progression-free survival (PFS). The authors conclude that these observations can help estimate and explain life expectancy to women with MBC who are beginning chemotherapy. It remains to be seen whether using simple multiples of the median to calculate estimates of scenarios for survival can be applied to other advanced cancers; this is being actively explored by the same investigators.¹²

These simple calculations can assist oncologists in the communication of prognostic estimates more clearly and precisely. If the recommendation to treat was determined on the basis of a randomized clinical trial that identified a PFS in the treatment arm of 8 months versus 4 months in the control group, there are a number of numeric estimates that may help a patient understand the possible benefits of treatment as well as the range of possible outcomes, which are quite broad for this population of women with MBC. For a patient with an estimated PFS of 8 months, the worst case scenario (10% of cases) is 2 months and the best (also 10%) is 24 months. The majority of patients will have a PFS between 4 and 16 months. Patients may interpret the data optimistically and imagine that they will be among the exceptional survivors or may interpret the data more realistically and understand the magnitude of possible benefits. Numeric estimates can steer prognostic discussions toward a realistic framing of personal benefit and help a patient to make a wise decision about future treatment. Considerations of toxicity as well as planning for the resources and time required to receive treatment should also be included as a topic of discussion.

Can these data-driven mathematic estimates help us in practice? Numeric estimates and accompanying explanatory phrases can help personalize prognostic discussions, but it may be necessary to adapt the estimates from trials to best fit the individual scenario. Not every patient meets the criteria for enrollment onto trials, so it is important to account for the fact that they may have a worse prognosis. Similarly, we need to remember the biologic heterogeneity of this sample of patients and to consider that other physiologic parameters may influence predictions of life expectancy and response to subsequent treatment. Chemotherapy, beyond first or even second line, can provide benefits to many women with MBC, especially to those who benefitted from prior regimens. We have sufficient evidence to respond to patients' needs for prognostic information. Data supports frank disclosure that is accompanied by supportive and skilled emotional guidance, not just for the patient, but for his or her loved ones. Estimating the best and worst case scenarios with words and concrete time frames may take time and practice. Our hope is that it will also facilitate coping and planning for an uncertain future.