© 1999 by Oxford University Press
Journal of the National Cancer Institute Monographs, No. 25, 85-87,
1999
© 1999 Oxford University Press
Matching Strength of Message to Strength of Evidence: a Discussion
Affiliation of author: Division of Cancer Prevention, National Cancer Institutes, Bethesda, MD.
Correspondence to: B. S. Kramer, M.D., M.P.H., National Institutes of Health, Bldg. 31, Rm. 10A46, MSC 2580, Bethesda, MD 20892.
 |
INTRODUCTION |
|---|
 Top Introduction An Additional "Axis" of...References |
|---|
The four discussions in this session present different and sometimes contrasting aspects of cancer screening from the perspective of behavioral science. As a nonbehavioral scientist, I have tried to compare the behavioral aspects from the perspective of the public health professional in the field of cancer screening and to see the implications of a given behavioral strategy. I picked up, in essence, three general approaches or themes from the three papers on general population screening (Table 1
). The genetic
screening paper adds complexity that I will address at the end.
|
One (1) of the papers that we discussed today was primarily aimed at removing barriers to screening. For example, there was a statement in the paper that "the doctor should simply recommend screening." Some examples were given for mammography and prostate cancer screening, and McCaul and Tulloch in their presentation said some physicians urge prostate cancer screening to patients as the standard of care.
In their paper, O'Connor et al. (2) refer to this approach as "supplier induced demand." In essence, that is what happens to some patients during a routine checkup. The supplier virtually demands that the patient get a prostate-specific antigen test. We are seeing much of that in the medical practice today. The pharmaceutical or industrial suppliers are going straight to the public and advertising, or doctors are creating the demand for the types of things they do. The paper by Rothman and Kiviniemi (3) dealt with a conceptualized approach. Some of the means suggested to supply context were the provision of testimonials, imagery, and fear. More about that in a bit. The third paper O'Connor et al. (2) was a personal values-based approach or "informed choice" strategy, as they term it.
From my perspective, each of the three approaches has advantages and disadvantages. I listed
attendant potential pros and cons in Table 2. The first approach is
removing barriers. One of its advantages is that it is quite direct. You go for what you would like
the patient's behavior to be. On the con side, the strategy can be directive. I think being
direct has advantages but being directive has downsides, such as oversimplification of complex
decisions or unproven interventions.
|
The contextual approach has a great deal of intuitive appeal. The word that comes to mind in terms of pros is "poignant." I could see how it would be extremely poignant and very compelling for someone to hear the context in which he or she is making the decision. In point of fact, we get a lot of contextual information these days. Congressional hearings are full of contextual approaches in which you will often, if not universally, have a panel of people who testify to very personal and moving experiences while a given committee is contemplating legislation on national policy. The recent Internal Revenue Service (IRS) hearings are an example. A panel of people gave stories of personal sufferings in dealings with the IRS. That is extremely poignant because most people will resonate with the suffering and fear. The problem is one of numerator versus denominator. You do not know how large the numerator of suffering people is. The denominator is many, in this case more than 100 million people. The question is: How reflective is the contextual story (i.e., numerator) of the general? This also has happened in hearings that have dealt with a variety of medical issues. You have someone who is personally affected. You have seen on national television people lamenting what happened to them within the medical system. And although it is poignant, it is hard to know whether it is oversimplified or if the problem is general. It is hard to know whether it is an anecdote and, sometimes, it is hard to verify its accuracy in the first place. In other words, accuracy of inference may suffer at the altar of poignancy.
Finally, informed choice has the pros of empowerment in engaging the patient in an informed decision. It does have its downsides as well. It can be time consuming for the health professional. It can be quite confusing to the person trying to choose among options. And it can affect compliance. If you, as a health provider or health professional, think you know exactly what the right thing is to do, and you complicate the message with any issues of pros and cons, the person may not comply and you might consider that a defeat of your purpose. Resources required for presenting complex messages are another problem.
|
AN ADDITIONAL "AXIS" OF COMPLEXITY |
|---|
|
TopIntroductionAn Additional "Axis" of...References |
|---|
I have some additional thoughts in terms of research directions in behavioral approaches to cancer screening. The papers in this session deal almost exclusively with what may happen to an individual, whether medically or socially, in the screening setting. However, I think that there is yet another aspect to address in health communications to the individual (or to the public)—that is, not just what the range of outcomes is and the likelihood that these outcomes could occur, but the strength of the evidence that informs the discussion in the first place. The Physician Data Query (PDQ) of the National Cancer Institute provides a means to do this. Formal levels of evidence are attached to each discussion. This represents a completely different conceptual axis because it does not represent the actual health outcomes. It is the strength of data that support the statement about outcomes—good or bad. There can be strong, intermediate, or weak levels of evidence.
Table 3 collapses the five-level rating system of PDQ into three
general levels. The strongest level of evidence comes from a randomized, controlled prospective
trial, in which the only thing that has been changed is the actual intervention. The observed
outcomes, therefore, reflect directly a cause-and-effect relationship between intervention and
outcomes. Intermediate-strength evidence can come from a variety of study designs. The example
in Table 3 is evidence from observational studies. These designs can
suggest causation, but confidence is not nearly as strong. Rather than causation, they
show association between intervention and outcome. The weakest level of evidence
would be opinion, logic, and clinical judgment. In the area of screening, there often tends to be
more opinion rather than other forms of evidence. The important bottom line of this second
conceptual axis is to match the strength of opinion to the strength of the evidence, which I think
is a somewhat new dimension that is often lost in crafting health-related messages for the public.
|
In summary, the first dimension of communication on cancer screening is the medical, social, and economic outcomes; both pros and cons; and the consequences of intervening, or not, medically. The other dimension is the actual strength of the evidence. There are even higher orders of complexity (Table 4
). There may be multiple risks with which
you have to deal. As discussed earlier in this conference, there is very little research going on in
how to deal with communicating multiple risks. There are also multiple potential screening
interventions, so one not only has to weigh risks versus benefits of a single screening test but also
has to consider, for example, "Should I get a mammogram or the heart scan that's
also advertised a lot," or "Should I get a barium enema or a
sigmoidoscopy," and it goes on. (It is virtually impossible to get all medical screening
tests advertised to the public.) One way to approach such a situation of multiple choices and
trade-offs is to examine the number needed to treat. There's little information on its use as
a communication tool, but research could be helpful. When a health professional is explaining a
medical test or treatment, it may be, for example, helpful to say, "We would have to treat
10 people with your blood pressure level to avert one stroke," or "We would have
to screen 2000 women your age for breast cancer annually for 10 years to increase life by 1
year," thus providing context among medical interventions.
|
There is also the complexity of communicating consequences of multiple tests, not just each one in isolation. The public is deluged by health messages. Apart from the strength of evidence, and totally apart from the actual outcomes, there are many screening tests of varying accuracy. Multiple professional societies issue numerous screening recommendations in isolation. What could be the consequences of following each and every recommendation? It has been estimated that about 125 screening recommendations are made to the U.S. public by various professional and advocacy groups. If each recommended test were independent and had a specificity of 0.95 (considerably better than most such tests), the likelihood that a perfectly healthy person would be declared healthy on all screening tests is 0.0016 (0.95125). Even at the extremely unrealistic specificity of 0.99, the likelihood would be 0.28. That is a behavioral research challenge—not just approaching one screening tool at a time or one intervention at a time, but putting it into the full context of other recommendations and health strategies.
Can we help the person considering the messages with additional information such as level of evidence or comparisons to other strategies in terms of "number needed to treat"? Research on how to frame such messages seems important to me as a nonbehavioral scientist.
Finally, a comment on the presentation by Croyle and Lerman (4) on the more restricted area of genetic screening. We have gotten into a brave new world in which risk is not homogenous across the nation. As we learn about genetic tests, we are going to learn that there are subpopulations and founder populations who are at much higher risks. Then there are going to be the far larger number of people who have polymorphisms who are at moderately elevated risk. Almost everyone will have, in some sense, a "pre-existing condition," predisposing them to harms from specific exposures. If an individual is exposed to tobacco smoke or an industrial carcinogen, risk depends on how well he or she metabolizes and excretes carcinogens and procarcinogens or repairs molecular damage. The message may have to be far more complex and tailored than any that we have imagined before. We have immense and growing challenges in the field of risk communication.
|
REFERENCES |
|---|
|
TopIntroductionAn Additional "Axis" of...References |
|---|
1 McCaul KD, Tulloch HE. Cancer screening decisions. Monogr Natl Cancer Inst 1999; 25:52-8.
2 O'Connor AM, Fiset V, DeGrasse C, Graham ID, Evans W, Stacey D, et al. Decision aids for patients considering options affecting cancer outcomes: evidence of efficacy and policy implications. Monogr Natl Cancer Inst 1999;25:67-80.
3 Rothman AJ, Kiviniemi MT. Treating people with information: an analysis and review of approaches to communicating health risk information. Monogr Natl Cancer Inst 1999;25:44-51.
4 Croyle R, Lerman C. Risk communication in genetic testing for cancer susceptibility. Monogr Natl Cancer Inst 1999;25:59-66.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||