The Challenge of Communication1
The workshop next focused on the challenges involved with effectively communicating risk information, Dr. Lipkus presented an overview of the issues.2 Clinical barriers (for example, limited patient involvement in discussions, limited time) make it difficult for physicians to effectively communicate about risks and benefits with their patients. Physicians tend to underestimate the amount of information that patients want, control discussions and discourage patient involvement, overestimate how much patients know, overestimate the efficacy with which they accomplish important communication tasks (how well they have communicated information to their patients), and have limited time.
Furthermore, physicians and patients understand risk differently. He referenced a study in which patients identified “possible side effects” as the most important piece of information to consider when making a decision about a drug (Berry et al. 1997). Physicians, however, ranked side effects tenth among 15 items, and the number one consideration for physicians was “interaction with other drugs for long-term use.”
While most physicians agree that conveying risk in a quantitative format is important, very few are confident in their ability to do so. In
a study of primary care physicians in Massachusetts, 93 percent agreed that conveying risk numerically is important, and 63 percent felt that quantitative and qualitative communications are equally important, but only 36 percent felt that they could convey numerical risk information effectively (Gramling et al. 2004). One of the most difficult challenges in risk communication is conveying probabilistic information. The difficulty stems in part from the fact that most patients are interested in what their own chances of benefit and risk are, not population-level probabilities.
If physicians are not confident in their ability to communicate risks numerically, what can be done to help them? More generally, how can risk information about medication be communicated effectively? The magnitude of this challenge is evidenced by the fact that even after several decades of effort and a large body of evidence, there is still a lack of consensus concerning the most appropriate method by which to communicate medical risk.
Both individual-level (information directed toward the individual patient) and population-level (information about the population of which the individual is a member) risk information can be communicated in any of several formats—numerically, verbally, visually, or through the use of narrative. Numerical formats for presenting risk include percentages (e.g., 10 percent greater risk), frequencies (e.g., 1 out of 10 people is expected to have side effects), classical probabilities (e.g., 0.10 chance), and “numbers needed to treat” (e.g., need to treat 100 people to get one person to benefit). The advantages to using numbers include the following: they are precise, they add an aura of “scientific credibility,” they are easy to compare and convert to varying metrics, they can be used in existing or new algorithms, and they are verifiable. Numerical usage also has disadvantages, such as the discrepancy between actual (or objective) and perceived risks that results when numerical risk information is used, even just moments after the information has been provided. Dr. Lipkus stated that studies have shown that innumeracy is problematic across all educational levels, even among the college educated.
The problem of innumeracy raises the question, Why can’t we just verbally communicate the risk? Verbal terms tap into gut-level reactions, they seem to be easy, and they convey uncertainty on multiple levels. Yet verbal communication is vague, terminology is difficult to standardize across contexts and between people, and interpretation is highly variable.
If not verbal, how about visual communication? Visual aids can range from bar charts and line graphs to risk ladders and stick figures. The advantages of visual displays are that they can summarize lots of data; help the patient see patterns that would otherwise go undetected; help the patient perform some mathematical operations, such as comparisons,
automatically; and attract and hold the patient’s attention because the data are displayed in concrete, visual terms. However visual aids have their drawbacks too: data patterns may discourage people from paying attention to details; some visuals, such as risk ladders, are poorly understood unless they are explained; creating visuals usually requires technical programming; and we don’t really know how visual aids affect risk perception.
Dr. Lipkus concluded his talk by posing a final challenge: What should the outcomes of risk communication be? He argued that we spend a great deal of time discussing how to communicate risk, but we don’t spend much time discussing what the outcomes should be. For example, does risk communication lead to higher or lower rates of adherence? Does risk communication lead to more or less conflict or mistrust? Does it unnecessarily increase anxiety or other negative emotions?
PRESCRIPTION DRUG FACTS BOX
The 1938 Food, Drug and Cosmetics Act, the basic law that established the FDA’s actions, reads: “Information in drug labels should appear only in such medical terms as are not likely to be understood by the ordinary individual.” Although the law has been amended since then, Dr. Woloshin argued that sometimes it still feels as though we are still trying to live by its spirit, particularly when it comes to direct-to-consumer advertisement.
Dr. Woloshin showed a series of direct-to-consumer drug advertisements, demonstrating their lack of accurate factual information. Some ads assert efficacy rather than provide data (e.g., “works for me”); others contain data about popularity but, again, nothing about efficacy (e.g., “more than one million people have begun using Rezulin to help manage diabetes”); some contain data irrelevant to the assertions made; and still others contain incomplete information (e.g., informing that the drug cuts a risk “by nearly half,” raising the question, half of what?).
Dr. Woloshin commented on the brief summaries of harm information that the FDA requires to be included in all advertisements. He remarked that while the FDA has recently issued new guidance to industry about providing these summaries in a user-friendly format, the fundamental problem about the lack of efficacy data remains. If consumers are going to make benefit–risk decisions, they need to have access to both benefit and risk data. Only if they know what the benefit is, will they be able to make informed decisions about whether the risks are worth that benefit.
Dr. Woloshin discussed a possible solution for providing user-friendly benefit–risk data in advertisements: the Prescription Drug Facts Box. Modeled after the FDA’s Nutrition Facts Box, this box would contain the data from the brief summary but in a simple tabular format. He showed a
prototype of the Prescription Drug Facts Box with two parts: the first part containing descriptive information (answering questions such as, What is the drug for?), the second part containing a data table including both risk and benefit information (likelihood of intended outcome if the drug is taken versus not taken; likelihood of risk if the drug is taken versus not taken).
Dr. Woloshin described two studies that he and his colleagues conducted in an effort to determine whether the Prescription Drug Facts Box is effective—that is, if it helps people understand and judge the benefits and harms of drugs. In one study, researchers concluded that not only were the boxes easy to read and were preferred by participants (compared to ads without the boxes), they also helped participants more accurately interpret the drug’s benefits (Woloshin et al. 2004). In the other study, researchers found that participants across a range of educational backgrounds did quite well in interpreting extracting, manipulating, and applying both benefit and risk data.
Questions were raised about the discrepancy between Dr. Woloshin’s encouraging results with respect to patients’ abilities to analyze, digest, and make fairly sophisticated decisions about benefit–risk information and Dr. Lipkus’ less optimistic perspective. Dr. Woloshin responded that the explanation probably is in how the information is provided. He and his colleagues chose methods and representations, including a simple tabular format, that had been shown to be understandable and accessible even to less well-educated people. He said that the boxes have been through countless iterations and were based on large numbers of focus groups and cognitive interviews. Dr. Lipkus agreed that the tabular format made it easy for people to find the information they need. He also noted that presenting numbers in two ways, as the box does, could provide multiple senses of meaning. He remarked that some of the studies that he reported utilized more complex information.
Dr. Fendrick remarked that his work has shown that informed patients become more anxious and less likely to undergo screening and that many patients want their physicians to make the decisions. He also expressed concern about whether efficacy data drawn from registration trials is suitable for labeling, given differences between efficacy and effectiveness.
PHYSICIAN USE OF THE PRESCRIPTION DRUG FACTS BOX
Dr. Schwartz argued that the Prescription Drug Facts Box would also help physicians make better prescribing decisions by providing a fast, efficient way to access information. She remarked on the limitations of several ways that physicians currently access drug information:
Information provided by drug company representatives is often selective and incomplete.
Many clinical trials are not published, many are not published in journals that physicians regularly read, and not all peer-reviewed journal articles report all the benefits and risk measured during a clinical trial.
Brief summaries that accompany advertisements in medical journals, newspapers, and magazines typically do not include efficacy data; when they do, the data tend to be incomplete or exaggerated.
FDA approval labels, or package inserts, have much more complete information than these other sources, but there is some question as to how often clinicians read those inserts.
FDA drug approval documents, including medical, chemical, statistical, and other reviews of drug company applications, are freely available on the Internet;3 however the large quantity of critical data, lack of structure, and difficult reading make accessing the information overwhelming.
In addition to improving physician decision making, Dr. Schwartz argued that the process of creating a structured table of rates of outcomes for all the treatment groups can reveal whether and which patient outcomes are unavailable or missing from the label. She used a prototype Prescription Drug Facts Box for Zometa to illustrate. By sifting through the drug approval documents, she and colleagues discovered a statistically significant dose-related difference in mortality that had been noted but without any strong warning bells. Nor was the difference included in the current label. Dr. Schwartz argued that the mortality finding warrants a stronger statement.
Dr. Schwartz remarked that one of the challenges of presenting side effect information is doing it so that people can sort through the multiplicity of side effects, which requires establishing arbitrary rules for deciding which side effects to include. Her team decided, for Zometa, to make separate rules for life-threatening and symptom side effects. Specifically, life-threatening side effects would be included if the p-value was 0.5 or less with respect to the difference between the drug and the comparison. A large p-value was chosen to prevent missing potentially life-threatening harms. Symptomatic side effects, by contrast, would be included only if the p-value was less than 0.2, limiting unnecessary concerns. There were concerns, however, that manipulating statistical precision in this manner could ultimately harm the integrity of the process.
Dr. Galson remarked that FDA drug labels are in fact undergoing total transformation; that there is going to be a standardized format, which the
FDA has been working on for some 10 years; and that changing the format of the labels is in practice very difficult. Additionally, the FDA is working with the National Library of Medicine to move all labels onto the latter’s website, in a machine-readable format, giving people the opportunity to examine the data (through hyperlinks to full prescribing information) and make their own facts boxes. He opined that while these changes are certainly expected to improve communication, we still have a long way to go. For example, various types of graphics might be better than words in explaining the different components of benefit–risk ratios.4
LESSONS LEARNED FROM THE EPA
Dr. Goldman noted that there were two types of safety communication: routine and crisis. She relayed lessons learned about both from her experience with the Environmental Protection Agency.
She described how the EPA’s Pesticide Consumer Labeling Initiative5 found that while the language on labels seemed to be complete, it was often unintelligible and useless. In response, the EPA changed its standard labeling language; conducted more comprehensive research to find out in more detail what kind of information people wanted and how they could best find that information; took a top-down approach to reforming the internal process of managing and developing labels; and ran a campaign to encourage consumers to read labels.
Dr. Goldman showed the label from a prescription drug and remarked on two features of the “Information for Patient” section. First, she noted the difference between information and facts, arguing that the labels included only the latter and that not even her physician could understand them. Second, she observed that there is no indication of what judgment the FDA made about the risks of this drug. She argued that the label demands too much of patients—not only must they find a way to understand these facts, but they also have to make their own judgment about what the risks may be. She asked why we are afraid to tell consumers what the FDA’s judgment is, in plain language.
She described a 1994 crisis situation, when an FDA market basket survey detected an illegal pesticide, chloropyrifos-ethyl, in oat cereal and discussed how the EPA and FDA cooperated on risk assessment and communication strategies and successfully resolved the situation.
Dr. Goldman suggested that we study the information needs and preferences of consumers and learn how to communicate in their language, and that we develop procedures for making decisions quickly and
See, for example, Edwards et al. 2002.
collaboratively in crisis situations and then informing the public about those decisions.
Dr. Goldman’s suggestion that more judgment be put in the materials that the FDA gives to consumers prompted a comment that this was very different from Mr. Hutt’s argument (see Section 4) that patients be given the facts only and be allowed to make their own judgments. It was noted that this is an excellent demonstration of the fact that very reasonable people have diametrically opposed views about what the FDA should do on any given matter.
A comment was made about how much of the workshop discussion focused on academics, industry and consumers, but that the entire FDA regulation process relies on the prescribing physician as a critical “learned intermediary.” The complete absence of participation by such groups in the FDA regulatory process is striking. While, in the past, efforts have been made to build relationships with medical professional organizations, budgetary constraints have eliminated that component from the agency. Little has been done to reestablish those relationships and have had varying degrees of success in trying to partner with them. Dr. Goldman noted that this is why she mentioned how difficult it is for physicians to understand some drug labels. She agreed that they are “out of the loop.” Dr. Leiden agreed and noted that “opinion leaders” practice in highly controlled academic medical centers where drugs are used very differently than they are in the “real world.” We need to consider the knowledge and input of practitioners who are regularly seeing patients and prescribing drugs.