Abstract
The format in which risk is communicated influences how patients make health care decisions. The same statistical information may be summarized differently according to relative risk reduction, absolute risk reduction, number needed to treat, or odds ratio. A total of 76 men participated in focus groups exploring their understanding of information about prostate cancer treatments when framed across these different formats. Using thematic analysis, it was identified that the study participant best understood information when outcomes were framed as an absolute risk reduction and in a positive frame. Patient education materials about prostate cancer treatment options should be reported as an increase in probability of survivorship rather than decrease in risk of mortality and incorporate impact of treatment on patient-centered quality-of-life outcomes.
Introduction
Increased access to health information has led to the opportunity to adopt a shared approach to health care decision making between patients and their health carers (Woolf, 1997). To enable this, patients should be effectively informed about the benefits and limitations of potential medical procedures and treatments (Coulter, Entwistle, & Gilbert, 1999). The integration of patient values, along with the best available empirical evidence and the clinical expertise of clinicians, facilitates shared decision making between patient and clinician, with the patient having a comprehensive understanding of the potential consequences of the chosen intervention (Calman, 2002; Sackett, Straus, Richardson, Rosenberg, & Haynes, 2000).
Patient education materials, including decision aids, improve patient knowledge and decision-making ability (O’Connor et al., 1999). The manner in which medical information is presented to patients can influence how they interpret the information and approach subsequent medical decisions (Edwards, Elwyn, Covey, Matthews, & Pill, 2001; Edwards, Unigwe, Elwyn, & Hood, 2003). Relative risk (RR), absolute risk (AR), and number needed to treat (NNT) are numerical modes commonly used to report research results (Sarfati, Howden-Chapman, Woodward, & Salmond, 1998). Each of these summary statistics may be used to report the same data, with all appearing to report a different result. For example, a hypothetical trial may report a survival rate of 91.8% in the active intervention group, compared to 88.5% in the control group (O’Connell, Gebski, & Keech, 2004). The results of such a study may be reported as a RR of 0.70 (i.e., the risk of mortality in the intervention group compared to the control group), a relative risk reduction (RRR) of 30% (i.e., the risk of mortality is reduced by 30% in patients randomized to receive the intervention compared to those receiving the control), an absolute risk reduction (ARR) of 3.3% (i.e., there is 3.3% less deaths occurring in the group allocated to the intervention compared to the control group), or a NNT of 30 (i.e., 30 patients need to be treated by the intervention to prevent one additional adverse event [i.e., death] from occurring; O’Connell et al., 2004). The odds ratio (OR) may also be used as an approximation of RR, with the exception that odds rather than risk is being calculated (Davies, Crombie, & Tavakoli, 1998).
An ability to interpret the difference between RR, RRR, ARR, and NNT is essential, since such data may often be manipulated to amplify a relatively small difference between interventions (Flaherty, 2004). Presenting patients with meaningful health information to inform decision making is critical, given that patients have a more accurate perception of risk if probabilistic information is presented as specific quantitative data rather than a confusing written discourse (Man-Son-Hing et al., 2002; Marteau et al., 2000). In such a case, patients are more likely to understand absolute frequencies since they experience information expressed in this nature on a daily basis (e.g., 1 out of 10 people will be affected by disease x; Gigerenzer, 1996). Patients may understand information presented as an ARR since it provides information on the magnitude of the risk, along with the baseline risk (Feinstein, 1992). Conversely, RRR may present big reduction in risk, but does not offer any indication of the absolute risk (Nuovo, Melnikow, & Chang, 2002). Information presented as an NNT may also be misinterpreted by patients as an absolute frequency (Cook & Sackett, 1995).
Framing numerical information as either benefit or harm influences the manner in which patients interpret health information (Gurm & Litaker, 2000). Patients have demonstrated a preference for undertaking treatment options that are presented in the positive frame (e.g., 90% survival) rather than a negative frame (10% chance of mortality; Gurm & Litaker, 2000). The manner in which such information is framed may alter patient preferences and decision making with respect to treatment.
Prostate cancer remains a controversial subject, due to the varying levels of evidence available on screening and treatment of prostate cancer (Ilic, Green, O’Connor, & Wilt, 2010). Despite a high level of public advocacy, the evidence suggests that there is limited benefit in the reduction of mortality through screening for prostate cancer (Ilic et al., 2010). Similarly, the relative benefits and limitations of treatments such as radical prostatectomy, watchful waiting, hormone therapy, and chemotherapy remain uncertain due to a lack of high-quality randomized controlled trials (RCTs; Hegarty et al., 2010; Kumar et al., 2006; Shelley et al., 2006).
Various decision aids to assist men in making a decision about prostate cancer screening and treatment choices have been developed in a bid to reduce decisional conflict and anxiety, enhancing patient knowledge, and promoting shared medical decision making (O’Connor et al., 1999; Volk et al., 2007). The manner in which such information is presented to patients may influence their decision making. For example, a RCT investigating how patients interpret information about taking statins concluded that presenting the benefits of treatment as an RRR increased the likelihood of patients accepting treatment with the statins compared to absolute summary statistics (Carling et al., 2009). We know of no study that has investigated this with respect to prostate cancer.
The overall aim of this qualitative study was to explore how men interpreted information about different prostate cancer treatments. The primary objective of the study was to identify which summary statistic men best understood and how the framing of the summary statistic influences their perception of the information. A secondary aim of the study was to identify whether men diagnosed with prostate cancer have a differing perception from men not diagnosed with prostate cancer.
Method
Theoretical Framework
The qualitative approach of thematic analysis was used to provide an inductive analysis of how men make sense of risk information on prostate cancer treatment to inform decision making and resolve decisional conflict (Liamputtong, 2010; Rice & Ezzy, 1999).
Participant Selection
Participants were recruited through purposive sampling techniques including public advertisements in newspapers and local prostate cancer support groups. Eligible participants included adult men (aged 40 years and older), who could speak and comprehend English fluently and provide informed consent. Recruited participants were then categorized into two groups: (a) men diagnosed with prostate cancer (PC) and (b) men who did not have prostate cancer (NPC). All men diagnosed with prostate cancer were at least 1 year posttreatment for their prostate cancer. Potential participants were given a verbal explanation of the study prior to attending a focus group. Each participant was given an explanatory statement, which further outlined the study, and provided written consent prior to the commencement of the respective focus group.
Setting
Separate focus groups were held for the two groups (NPC or PC), with participants given a preference to attend focus groups at community halls across Melbourne or the Monash Institute of Health Services Research (place of work for the researchers). Participants were offered an honorarium (AU$50) to cover the costs associated with their participation. This study was approved by the Monash University Standing Committee on Ethics in Research Involving Humans.
Data Collection
The use of focus groups was chosen as an appropriate methodology to facilitate greater discussion on prostate cancer with men. Prostate cancer, and topics around male reproductive health, can be sensitive for men to discuss, as men often feel a sense of embarrassment when discussing reproductive health issues (Ilic, Risbridger, & Green, 2005; O’Shaughnessy & Laws, 2009). Previous literature has demonstrated that the use of focus groups permits participants from similar backgrounds (e.g., diagnosed with prostate cancer) to feel greater comfort in discussing any perceived sensitive issues within a group environment rather than participating in a one-off individual meeting (Bryan et al., 2008; Rice & Ezzy, 1999). By sharing a common interest, participants are able to listen to the discussion generated within the group context and feel greater confidence in presenting their own opinions in this environment (Grbich, 1999).
A total of 11 focus groups were conducted with 76 participants. Standard questions were developed from a review of the literature. All focus groups were led solely by the same experienced female facilitator (KM, who has a PhD in qualitative research and is a registered psychologist). Questions were asked based on a semistructured interview schedule to ensure that all questions/discussion points were raised across all focus groups. Focus groups were 1½ to 2 hours in duration. All focus groups were audio-taped by a digital audio-recorder, with the digital files transcribed verbatim at the conclusion of the focus groups by an independent transcription service. Copies of the focus group transcripts were offered to participants for an opportunity to provide feedback—although no took up the proposal. The focus group discussions were conducted until the data reached a point of theoretical saturation (Patton, 2002; Silverman, 2004; Strauss & Corbin, 1998). Theoretical saturation of data was assessed at the conclusion of each focus group. Theoretical saturation was determined when the final focus group (from each of the respective PC and NPC groups) did not generate any further novel discussion points (Braun & Clarke, 2006; Guest, Bunce, & Johnson, 2006).
Each focus group explored the following constructs:
Participants’ understanding of various numerical methods of communicating risk (i.e., RRR, NNT etc.)
Proposed methods of best framing information (i.e., positive or negative frames)
Participants were presented with four summary statistic examples each depicting the same quantitative information about a prostate cancer treatment but framed differently. The four examples were framed as NNT, OR, RRR, and ARR (Table 1). Each scenario was individually presented in both text and graphical formats to participants during the focus group discussion.
Examples Used to Frame Information About Prostate Cancer Treatment
The information presented via these four examples was based on results from a RCT comparing watchful waiting to radical prostatectomy (Bill-Axelson et al., 2005). None of the participants were told which example related to which intervention (i.e., watchful waiting or radical prostatectomy) during the focus group. This approach was adopted as many men, particularly those in the PC group, have preexisting attitudes toward the value of existing prostate cancer treatments including watchful waiting, radical prostatectomy, brachytherapy, and complementary therapy, which are not necessarily informed by evidence. The scenarios were all framed with respect to mortality, rather than an increase in survival, across all examples. Participants were presented with the scenarios in a sequential manner, beginning with NNT, OR, RRR, and ARR.
Data Analysis
The de-identified transcripts were analyzed using thematic analysis, which enables the identification, coding, and categorization of the primary patterns within the data (Braun & Clarke, 2006; Patton, 2002; Silverman, 2004). Each transcript was analyzed individually by two investigators (DI and KM), independently coding and categorizing participants’ responses from the transcripts with respect to the four presented examples. Both investigators individually categorized the data for each example (which included several iterations from the initial coding theme), which outlined how participants’ perceived and understood the example summary statistic provided (Rice & Ezzy, 1999). The two investigators then met to discuss common themes from their analysis, before a final iteration of the results was developed and agreed on collectively by all investigators. Thematic coding of the data was assisted with the NVivo software program (QSR International, 2008). Verbatim quotes from focus group participants that best represent the key findings for each theme were highlighted, for subsequent reporting purposed (Liamputtong, 2010).
Results
A total of 76 men participated in the project, including 42 with prostate cancer and 34 without. Their demographic details are presented in Table 2. Significantly more men in the PC group were either retired, or not in the workforce, and older compared to men in the NPC group. Focus groups were homogeneous with respect to prostate cancer diagnosis, as it was initially hypothesized that PC men would have different perceptions of the prostate cancer treatment information compared to the NPC men. During the data analysis it was identified that both NPC and PC men shared common perceptions. The data presented below are a combined response from both PC and NPC men.
Demographic Data of Study Participants
Note. Two participants in the NPC group did not document their age. One participant in the PC group, and two in the NPC group, did not document their education. One participant in the PC group, and two in the NPC group, did not document their employment status. Three participants in the NPC group did not document their marital status.
Significant difference at p < .01.
Number Needed to Treat
Participants reported interpreting information presented as an NNT to be confusing. Some participants interpreted the NNT as 16 out of 17 people dying after being treated with this treatment. Other participants interpreted its meaning as a 17% chance of survival. Participants reported that the statement also made the treatment sound ineffective, thus discouraging its use.
Doesn’t give you much confidence in Treatment A. If you have to treat 17 people before 1 is saved there’s not much possibility that it will work. (PC) Well I would like to know what it really means. It says here one in every 17 for treatment A over treatment B. Does that mean treatment B is useless? (NPC)
When used as a way of comparing two treatments, NNT created the impression that participants were being influenced to prefer another treatment. There was a tendency to attempt to translate the figures into percentages, since percentages were perceived to be easier to understand.
Odds Ratio
Participants also experienced difficulty when interpreting information presented as an OR. Approximately half of the participants from both groups considered this statement to be unclear. The remaining participants felt they could make a decision between two treatments based on it as they had a better understanding of information presented as an OR rather than NNT. Overall, participants believed that the statement made Treatment A sound much better than Treatment B, so much so that they would not consider Treatment B. However, others pointed out that although it told you that your odds of dying were doubled with Treatment B, it did not give sufficient information to work out what the likelihood of this was.
What does it really mean? Does that mean, you know, that Treatment A that 1 person in every 1000 survives but under Treatment B 1 person in every 2000 survives. Like, you know, you sort of want to know the odds as well. (NPC)
Relative Risk Reduction
Relative risk reduction was used as an alternative format for describing relative risk. Participants again expressed confusion about the wording of RRR, as some interpreted the phrase “risk reduction” as a negative outcome, whereas others viewed it in a positive sense.
Is that saying four of us are going to survive out of this 10 if we have Treatment A, if we have Treatment B we’re all going to die? (NPC)
However, RRR was considered a more informative presentation than NNT or OR, since it was presented in percentage terms, which participants felt more comfortable and familiar with. It was also seen as providing a strong argument for one treatment over another. Despite a better perceived ability to interpret the data, RRR was still not necessarily interpreted correctly with many participants interpreting it as stating that Treatment B was more effective than Treatment A. While understanding the concept of RRR, some participants could not envisage how much their actual risk of dying from prostate cancer was reduced by potentially receiving Treatment A, since they did not have an absolute risk to compare it to.
Well this statement is actually cloudy enough that you don’t actually realize that it’s not actually giving you a proper survival rate, because you don’t know what the survival rate is, you need to know B before you can know that 40% of what. (PC)
Absolute Risk Reduction
Absolute risk reduction was used as an alternative format for describing absolute risk to the participants. Information presented in the ARR format was strongly endorsed by the participants across all focus groups. Participants stated that information framed as an ARR permitted a comparison of survival rates between the two treatments. Participants commented that it gave “real information” and enabled clear comparisons to be made. Participants strongly preferred information to be framed as an ARR, over OR, RRR, and NNT.
That statement (ARR) is a lot clearer because . . . it gives you of the total, whereas all these other ones were very hard to understand of the total. They just compared to each other but you didn’t get the total picture. (PC)
Context
All participants reported a preference for outcomes to be framed positively. The use of the word “dying” was considered by participants as being very negative. Participants across all groups stated that it was more preferable to use “surviving” or “survival” when referring to treatment, since those words promoted a positive outcome.
Well if it was me, I’m going to look at what my chances of survival are. (NPC)
Participants from the PC groups also commented that it would be very frightening to be given information framed in a negative manner after being diagnosed with prostate cancer. Participants in the PC groups stated a preference for RRR to be framed as a relative risk increase (i.e., “your chance of survival from prostate cancer is increased by 60% if you receive Treatment A compared to Treatment B”).
You want to be told about survival. They said “your survival chances” that’s . . . but as soon as you see “your risk of dying” you think “oh, I’m going to die.” I mean, you don’t want to hear that. (PC)
Participants also stated that knowing the survival statistics of treatments was not sufficient; they needed to know the implications of the treatment (i.e., side effects, quality of life, etc.). Statistics help in making a decision but not in isolation. Survival rate was not the only factor that participants considered when deciding between treatments.
Well I think that it (information) should show the benefits and the risks associated with the whole box and dice, so I think that’s needed when looking at the stats. (NPC)
Discussion
Although participants found the interpretation of summary statistics about prostate cancer treatment effects, the study participants were in agreement that summary statistics presented as an ARR were easier to understand than information presented as an OR, RRR, or NNT. Additionally, participants demonstrated a preference for information to be presented in a positive frame. Rather than presenting information as an ARR (i.e., the reduction in the risk of prostate cancer–specific mortality), information should be presented as an absolute risk increase (i.e., the increase in “risk,” or probability, of survival posttreatment). Information regarding adverse events and quality of life indicators was also raised as key issues by participants that would influence decision-making behavior. These themes were common across all focus groups despite participants differing with respect to prostate cancer diagnosis, age, and employment status.
A 2011 systematic review by Akl identified that information presented as natural frequencies were more favored and understood by study participants over information presented as probabilities (Akl et al., 2011). The review also identified moderate quality evidence indicating that participants understood RRR better than NNT (standardized mean difference [SMD] = 0.73; 95% confidence interval [CI] = 0.43, 1.04), with little or no difference in understanding between RRR and ARR (SMD = 0.02; 95% CI = −0.39, 0.43; Akl et al., 2011). It also identified strong quality evidence indicating that ARR may be better understood than NNT (SMD = 0.42; 95% CI = 0.12, 0.71; Akl et al., 2011). Our findings also support conclusions from the Akl review that patients perceive interventions to be more effective and have better understanding when exposed to ARR compared to NNT (Akl et al., 2011).
Our study shares common findings with another study that explored the effects of framing information with respect to breast cancer (Fortin, Hirota, Bond, O’Connor, & Col, 2001). This study concluded that participants preferred information framed as an ARR, rather than RRR or NNT, as information framed as an ARR provided them with a “baseline” risk to which the benefit of the intervention could be compared (Fortin et al., 2001). Presenting information as only as an ARR, or an RRR, does not provide information about the underlying risk of the outcome (Sarfati et al., 1998). Information reported as an RRR may subsequently exaggerate the absolute impact of the intervention when the baseline risk is very low, and conversely diminish the impact when the risk is high (Gigerenzer, 2003; Sarfati et al., 1998). Commonly, changes in risk are better understood if ARR or RRR are presented with baseline risk (Christensen, Brosen, Brixen, Andersen, & Kristiansen, 2003; Sheridan, Pignone, & Lewis, 2003).
Evidence to date has identified that information framed in a positive format is associated with a greater uptake of the intervention by patients (Edwards et al., 2001; Edwards, et al., 2003; Moxey, O’Connell, McGettigan, & Henry, 2003). Findings from our study support this conclusion. Presenting information in a positive rather than negative context may decrease a patient’s level of anxiety while increasing their confidence in the treatment outcome.
Preferences did not differ with respect to whether the men had been diagnosed with prostate cancer (and therefore potentially knew more about the disease and treatments) and men never tested for prostate cancer. This may be explained in part by an initial lack of understanding when presented with a cancer diagnosis. Many men diagnosed with prostate cancer report feeling great decisional conflict and anxiety when presented with choices about potential treatment (Feldman-Stewart, Brundage, Nickel, & MacKillop, 2001). Subsequently, many newly diagnosed prostate cancer patients will defer their decision making on treatment to their treating clinician (Feldman-Stewart et al., 2001). Similarly, the stage of diagnosis may limit the potential treatment options available (Feldman-Stewart et al., 2001).
Strengths and Limitations
Our study implemented a purposive sampling technique, which would have led to recruiting men with a greater interest in this topic than otherwise found in the general male population. Prostate cancer diagnosis and treatment are subject to various cultural and social determinants, which were not explored in this study. Although our study did not undertake methodological triangulation in the form of using different data collection techniques, it did incorporate data from two different populations and used two researchers to undertake the data analysis. None of the researchers occupied dual roles during the research (i.e., clinician and research). Although the researchers informally examined their own influence on the formulation of the research question, data collection, and interpretation through iterative discussions, a formal evaluation of these influences was not examined. Similarly, no form of methodological triangulation of the data was conducted. The use of focus groups also provides internal validity for the results, as variety of viewpoints on the various summary statistics assisted in data interpretation (Grbich, 1999).
Information presented in our study was taken from an RCT that compared radical prostatectomy (surgery) with watchful waiting. Watchful waiting is a conservative method of treatment, in which no actual intervention is implemented; rather, the patient is monitored for increases in prostate-specific antigen over a course of time. This is considered a legitimate treatment since a localized, nonaggressive cancer may be present but never be the cause of death in a patient. In such a case, it is recommended that patients are monitored via watchful waiting rather than undertake a radical prostatectomy and experience adverse events (erectile dysfunction and incontinence) as a result of the surgery. This study did not present showing a comparison of watchful waiting to surgery. Such a comparison would have resulted in information being presented as a relative risk increase, absolute risk increase, and number needed to harm—which might have yielded differing views by participants.
Conclusion
Patient education materials about prostate cancer treatment options should be reported as an increase in probability of survivorship, rather than decrease in risk of mortality, and incorporate impact of treatment on patient-centered quality-of-life outcomes. Further research is required to gain a better understanding the impact of social, cultural, and language components on risk communication of prostate cancer treatments to men.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This project was funded by ANZ Trustees, Medical Research & Technology in Victoria, Victorian Community Foundation, James & Vera Lawson Trust.