Serious and Actionable Risks, Plus Disclosure: Investigating an Alternative Approach for Presenting Risk Information in Prescription Drug Television Advertisements

Abstract

Background:

Broadcast direct-to-consumer (DTC) prescription drug ads that present product claims are required to also present the product’s major risks. Debate exists regarding how much information should be included in these major risk statements. Some argue that such statements expose people to unnecessary amounts of information, while others argue that they leave out important information.

Objectives:

Examine the impact of type of risk statement (unedited versus serious and actionable risks only) and a disclosure indicating that not all risks are presented on consumers’ ability to remember the important risks and benefits of a drug following exposure to a DTC television advertisement (ad). Risk and benefit perceptions, ad-prompted actions, recognition of the disclosure statement, and evaluations of both the disclosure and risk statement were also examined.

Methods:

A web-based experiment was conducted in which US adults who self-reported as having depression (N = 500), insomnia (N = 500), or high cholesterol (N = 500) were randomly assigned to view one of four versions of the television ad, and then complete a questionnaire.

Results:

The type of risk statement had a significant effect on risk recall and recognition, benefit recognition, perceived risk severity (depression condition only), and perceived benefit magnitude (high cholesterol condition only). Disclosure recognition (using bias-corrected scores) ranged from 63% to 70% across the three illness samples.

Conclusions:

The revised risk statement improved overall processing of the television ad, as evidenced by improved risk recall and recognition and improved benefit recognition. Further, the presence of the disclosure did not adversely affect consumers’ processing of drug risk and benefit information. Therefore, limiting the risks presented in DTC television ads and including a disclosure alerting consumers that not all risks are presented may be an effective strategy for communicating product risks.

INTRODUCTION^*

Prescription drug advertising regulations (21 CFR 202.1) require that broadcast (television, telephone, or radio) advertisements (ads) that present product claims also present the product’s major side effects and contraindications in either the audio or audio and visual parts of the ad. This requirement ensures that consumers are provided with a fair balance of information about product benefits and risks which both serve to inform consumer judgment and decision making about the advertised product. The statement of major side effects and contraindications is often referred to as the major statement or major risk statement. Typically, the amount of side effect or risk information in drug labeling is greater than the amount of benefit information, a characteristic which is similarly reflected in the ad. Including such information requires the use of time and space in the ad, which means that a requirement for a major statement necessarily constrains the inclusion of other types of content in an ad. Such risk information may affect consumer perceptions and decision-making, although available research does not yet offer unequivocal conclusions in that vein, as Frosch and colleagues have noted.^1,2 The major statement is mandated, yet there are various views as to the optimal form for such a statement. Many stakeholders have contributed to the debate regarding the design of major statements, including legislators, physicians, consumers, the pharmaceutical industry, and researchers.

The Case For and Against Abbreviation of Major Risk Statement Information

Arguments in favor of maximizing the amount of information to be included in a major risk statement can draw support from both social science research on informed decision-making and available work on consumer preferences. Numerous scholars have argued that direct-to-consumer (DTC) television ads historically have not included adequate risk information and that some leave out important risk information necessary for informed decision-making.^1,3 Moreover, consumer research also suggests that consumers would prefer ads to contain more information about drug risks than they currently do. Consider, for example, studies finding that a majority of consumers agree that ads do not provide enough information about risks.^3–5 Such consumer preferences are particularly important to note in light of the premise that the effectiveness of communication increases when messages reflect consumer preferences for format.^6–8 In addition, consumers often do not seek information beyond DTC prescription drug ads. Friedman and Gould,³ for example, reported that only a small minority (roughly 17%) of study respondents said they would seek out additional information after seeing a DTC ad. Although it can be debated whether this number represents a positive or negative outcome, it indicates that by not seeking additional information or visiting a health care provider, DTC ads are the main, and perhaps only, source of drug information for some individuals. As a result, DTC television ads arguably offer a prime opportunity for risk information presentation, so limiting major risk statements might represent a missed opportunity.

Is more necessarily better in the case of risk information, however? Some researchers have noted that consumer preferences may not necessarily relate to their actual ability to recall or understand risk information (e.g., Aikin, O’Donoghue, Swasy, and Sullivan⁹; Barnes et al.¹⁰; Hamstra et al.¹¹), which means researchers should look to evidence beyond expressed preferences. Moreover, as noted earlier, typical broadcast ads have distinct time limits that require trade-offs. Given the typical length of a DTC commercial at 30 to 120 seconds, the inclusion of complex or lengthy risk information in that time frame may reduce consumer resources to process important parts of an ad. Extensive major risk statements conceivably could expose people to an overwhelming amount of information, which could lead to an overall reduction in comprehension due to cognitive overload or strategic avoidance. Even if consumers engage and process a thorough presentation of risk information, they may not be able to interpret and use the information because of constraints on working memory or information processing style.^12,13 In this way, the additional risk information would be physically present but functionally absent.

Earlier studies have reported mixed results regarding the potential influence of risk information amount and type on consumer recall and recognition of pertinent facts. For example, research by Morris and colleagues in the 1980s failed to find a consistent pattern of effects. One study¹⁴ found no difference in recall and knowledge when four versus two risks were presented, yet two follow-up studies found that ads with four¹⁵ or six¹⁶ risks actually produced greater risk recall than when two or three risks were presented, respectively. In another study, Hoek et al.¹⁷ tested the recall of information in a standard DTC ad versus one with limited information. The unedited version mentioned two side effects and two contraindications and recommended that patients consult their doctor to determine if the medication is suitable for them. The limited information, or edited version, contained the same two side effects but did not contain the precaution or contraindications. Contrary to predictions, the study failed to find any significant difference in the recall of the drug’s indication, risks, or benefits between conditions. In a more recent study, Kavadas et al.¹⁸ found that highly involved consumers recalled fewer ad claims when exposed to a high ratio of risk-to-benefit information (e.g., 6 risk and 2 benefit claims) than those exposed to an ad with a low risk-to-benefit ratio (e.g., 6 benefit and 2 risk claims). This outcome suggests that listing a large number of additional risks could hinder memory for ad concepts beside risks.

Although valuable in their contribution to the literature, the studies summarized here are limited in terms of the small number of risks presented across conditions. Anecdotally, recent DTC television ads are seen as including a very long list of risks.¹⁹ Whether consumers can adequately process a large number of risks remains an open question and one that is critical to address given the potential implications for informed consumer judgment and decision-making.

Limitations to working memory may also affect decision-making.^20–24 People may be better able to make decisions based on an abridged and well-curated set of risks than under the circumstances of more complete information presentation. As just one example, Kahn and Kupor²¹ demonstrated that a medical product may be perceived as less threatening when nonserious risks (congestion, fatigue) are presented alongside serious risks (seizures) compared with presentation of serious risks only. Inclusion of more nonserious risks promoted less accurate judgments about risk in this case. Similarly, Mayer and Moreno²³ argue that a cognitive load-reducing technique (what they call “weeding”) that eliminates interesting but extraneous material could help prime consumers to engage only in essential processing of the critical information presented. If risk information were restricted to only the most critical drug facts, then consumers conceivably would have more working memory available for information processing and improved decision-making.

An Alternative Approach: Present Serious and Actionable Risks, Plus a Disclosure

The notion of an abbreviated risk statement has invited concerns and critiques. Limiting the amount of presented risk information requires in many cases omitting some specific risks or attributes of risks that, in turn, could reduce consumers’ opportunity to learn various drug facts and limit informed decisions about whether to seek out those treatments (e.g., Davis²⁵). One approach to enhance consumer comprehension while simultaneously reducing the amount of risk information presented is to focus only on the risks that are serious and actionable. Examples of serious and actionable risks include suicidal thoughts or behavior and allergic reactions, whereas unpleasant taste and morning drowsiness, while actionable, might be considered insufficiently serious to warrant inclusion in the face of more serious risks. Limiting the risks to those that are serious and actionable serves a dual purpose. First, it may reduce the cognitive burden that may result from having to process a large amount of information. Second, serious and actionable risks may be perceived as more relevant, thus improving attention and comprehension of the risk statement. Such an approach addresses concerns raised by some that DTC ads often do not sufficiently distinguish between clinically important and unimportant problems.^26–28 Relatedly, certain signaling techniques might also facilitate the success of a relatively brief major risk statement, for example, by providing a cue to identify the risks as serious or important at the start of the risk statement^29,30 (although also see O’Donoghue, Sullivan, Aikin, and Betts³¹ for negative findings on this topic). Adding such a cue necessarily adds some length to the major statement, but it may also convey important information about the degree of risk entailed by the risks that follow.

Another complementary option to counter concerns is to disclose to the consumer that the presented risk information has been restricted. Advertisers could include a disclosure statement indicating that the risks presented are not a full list and directing consumers to seek additional information, for example, by consulting one’s doctor or consulting patient labeling, if available. Such a strategy may help address the concern that simply omitting risk information might lead consumers to believe there are no other risks beyond those mentioned in the ad. Whether consumers would notice such disclosure is not yet settled, however. Earlier research cited by Morgan and Stoltman³² found low levels of recall for warnings and disclosures for a variety of products.^33–36 Special steps should be taken, then, to ensure that such a disclosure would be noticed and processed. For example, presenting information using both visual and audio modalities has been shown to result in better processing outcomes than single-modality presentations (see, for example, Aikin, O’Donoghue, Squire, Sullivan, and Betts³⁷ and Brasel and Gips³⁸). Still, investigation within the specific context of prescription drug television advertising is needed.

Assuming consumers attend to the disclosure statement, research is also needed to examine whether the presence of this statement aids or hinders overall consumer understanding of statements in the ad. No studies to date have explored the impact of this specific type of disclosure, but initial research on the use of other contextual statements in television ads, including prescription drug ads, suggests there may be little impact on consumers’ information processing, recall, and understanding. For example, a study that investigated the impact of adding a toll-free statement to DTC television ads found that although the statement was well recognized, processing of other ad elements was not significantly affected.³⁷ Similarly, another study presented a disclosure (in text form in the lower portion of the screen for a limited time), and although participants reported feeling competent in their ability to recall the disclosure statement, the actual recognition of it was significantly lower compared with other ad content.³² More broadly, at least some previous research has found that government-mandated disclosures can have unintended effects on consumer behavior (e.g., Green and Armstrong³⁹) and understanding.⁴⁰ Further research is needed to explore whether including a disclosure about risks specifically has a detrimental, beneficial, or neutral effect on consumer recall, perceptions of a product’s risks and benefits, and other ad-prompted actions.

Overview of Study and Research Questions

The goal of this study was to experimentally test the effects of two actions on consumer perceptions: (1) limiting the risks in the major statement to those that fall into the categories of serious or severe, such as boxed warnings, and those that are actionable, such as those a patient would be able to recognize and do something to mitigate or avoid, and (2) including a disclosure to alert consumers that other product risks are not included in the ad. The project team also sought to replicate the study across three unique contexts and with three illness populations: consumers who self-reported as having depression, insomnia, or high cholesterol. These medical conditions were chosen because they represented medical conditions with low, medium, or high prevalence, they affected both men and women, they represented a mix of chronic and asymptotic versus symptomatic conditions, and they were all advertised to consumers. Of note, because the ads for each medical condition vary in ways not accounted for by our experimental manipulations, we do not aggregate findings across medical conditions; rather, we replicated the design in this manner to learn about the generalizability of our findings across ads targeted at different medical conditions. Outcomes of interest comprised memory of the disclosure as well as the risk and benefit statements, perceived product risk and benefit, information seeking and other ad-prompted actions, and perceived clarity of the risk and disclosure statements. Based on the issues and literature summarized above, key research questions are presented below.

• Research Question 1. Does limiting the risks presented in DTC prescription drug television ads (and including a signal about the severity of the risks to follow) improve retention of the drug’s serious and actionable risks and benefits?

• Research Question 2. Does limiting the risks presented in DTC prescription drug television ads (and signaling the risks as serious) affect consumers’ perception of the drug’s risks and benefits?

• Research Question 3. Does limiting the risks presented in DTC prescription drug television ads (and signaling the risks as serious) affect intentions to seek more information about the drug?

• Research Question 4. Do people have more positive evaluations of the revised versus unedited risk statement?

• Research Question 5a. What percentage of people notice an advertisement disclosure regarding limits of included risk information?

• Research Question 5b. Do people have positive evaluations of the disclosure?

• Research Question 6. Does including a disclosure alerting consumers that there are other risks not disclosed in the ad improve retention of the drug’s serious and actionable risks and benefits?

• Research Question 7. Does including a disclosure alerting consumers that there are other risks not disclosed in the ad affect consumers’ perception of the drug’s risks and benefits?

• Research Question 8. Does including a disclosure alerting consumers that there are other risks not disclosed in the ad affect consumers’ intentions to seek more information about the drug?

• Research Question 9. Does including a disclosure moderate the relationship between limiting risks and recall and recognition of serious and actionable risks, risk perceptions, and ad-prompted actions?

METHODS

Experimental Design and Stimulus

The present research examined two independent variables in a 2 × 2 between-subjects design: the type of risk statement (unedited versus serious and actionable risks only [also referred to in this article as “revised”]) and the presence versus absence of a disclosure indicating that not all risks are presented. Within each illness population, respondents were randomly assigned to view one of four versions of an approximately 1-minute television ad for a prescription drug used for the treatment of depression, insomnia, or high cholesterol. Base ads comprised an existing ad from the marketplace for each of the three drugs (depression—Abilify [aripiprazole]; insomnia—Lunesta [eszopiclone]; and high cholesterol—Crestor [rosuvastatin calcium]) and served as the “unedited risk statement—disclosure absent” condition. Three additional conditions were created by modifying the major risk statements from these base ads: 1) unedited risk statement—disclosure present; 2) revised risk statement—disclosure absent; and 3) revised risk statement—disclosure present.

To manipulate the type of risk statement, the project team replaced the audio track during the major risk statement and recorded a new voiceover for the revised risk statement that only included those risks from the unedited risk statement that were serious and actionable. Thus, both the unedited and revised risk statement included serious and actionable risks, but the revised risk statement comprised only serious and actionable risks. The specific number of risks in the revised risk statement was not predetermined; instead, the unedited risk statement was refined until it included only serious and actionable risks. The revised statements began by alerting consumers that the drug can cause “serious reactions” or, in the case of the depression ad, “severe, life-threatening reactions,” whereas the first sentence of the unedited risk statement noted that this drug is not appropriate for some people (e.g., “Crestor is not right for everyone”). Across the three illness populations, the reduction in the number of risks varied (depression: unedited statement—17 versus revised statement—8; insomnia: unedited statement—11 versus revised statement—6; high cholesterol: unedited statement—6 versus revised statement—4).

To manipulate the presence versus absence of the disclosure statement, for the disclosure-present conditions, the following statement was spoken and presented on screen at the end of the major risk statement: “This is not a full list of risks and side effects. Talk to your doctor and read the patient labeling for more information.” This dual presentation of the risk disclosure was intended to increase retention of the disclosure.⁴¹ For those not exposed to this disclosure, the major statement ended following announcement of the final risk.

For all revised ads, the size and font of new on-screen text matched the size and font of other on-screen text included in the unedited ad, and the new voiceover matched (as closely as possible) the voice, speed, and cadence of the narrator in the unedited ad. No other changes were made to the ads. All on-screen text included in the unedited ad was also included in the revised ad, so long as the on-screen text did not refer to a risk that was excluded from the revised risk statement. Before beginning of the experimental study, the project team conducted a set of cognitive interviews with consumers to ensure the stimuli were seen as realistic and the disclosure was noticeable.

Procedure

Respondents were recruited from an online panel vendor. The sample consisted of US adults aged 18 or older who self-reported as having been diagnosed with depression (N = 500), insomnia (N = 500), or high cholesterol (N = 500). Health care professionals and those that worked for a pharmaceutical company, advertising agency, or market research company were excluded from the sample. The survey was administered online in May 2016. The recruitment procedures were not intended to draw a probability-based sample from the full population of adults meeting the eligibility criteria. A total of 17,103 panelists were invited to participate in the study. Of those panelists, 5,119 responded, 2,552 completed the screener, based on their responses to the screener, 2,178 were eligible, and 2,092 consented. A small number of panelists (n = 19) who opted to participate reported that they could not view the ads when shown and thus could not complete the study. A total of 1,500 respondents ultimately completed the study.

Within each illness population, respondents who were eligible and consented were randomly assigned (through the use of a computer algorithm) to one of the four experimental arms and directed to the appropriate stimuli and questionnaire. The randomization procedure was checked and confirmed in the study’s pretest and main study. The questionnaire was designed using cognitive interviews⁴² and a pretest. The same questionnaire was given to all respondents, except that the drug name, medical condition, and specific benefits and risks changed based on the medical condition featured in the ad to which they were assigned. Respondents were blinded to their specific experimental arm assignment throughout the study. At the end of the study, respondents were debriefed and told that the ad they saw was modified for the purpose of this study. This study was reviewed and granted an exemption from the relevant institutional review boards.

Measurement of Key Variables

This study focused on five categories of outcomes: (1) attention to the disclosure statement; (2) retention of the risks and benefits; (3) risk and benefit perceptions; (4) measures of ad-prompted actions, such as interest in reading the patient labeling and intentions to seek additional information; and (5) evaluations of the disclosure and risk statements. Each outcome was measured via self-reported survey items.

Attention to the disclosure statement was assessed by presenting the disclosure statement and asking respondents if they remembered hearing or seeing the statement in the ad. A measure was created to assess “correct disclosure recognition” based on whether the participant recognized seeing the disclosure and whether the disclosure was present based on their experimental condition. Respondent error is a common measurement problem that must be addressed when computing advertising recognition scores. As such, in the text, both the raw recognition scores and a bias-corrected score that adjusts for yea- and nay-saying response tendencies among study respondents are reported. This bias-corrected recognition score is based on a recognition-scoring approach used in research addressing measurement issues in advertising recognition testing^43,44 and is grounded in signal detection theory.^{45,46 †} Scores on this bias-corrected recognition scale were calculated using the following formula:

$$A^{\prime }=.5+\frac{\left(H-FA\right)\left(1+H-FA\right)}{4H\left(1-FA\right)}.$$

In the present study, respondents were randomized to experimental study arms differentiated by whether the disclosure was either present or absent. Therefore, H is the aggregate hit rate for the disclosure at the study level (proportion of respondents in the disclosure-present condition who correctly reported seeing the disclosure), and FA is the overall false-alarm rate for the disclosure (proportion of respondents in the disclosure-absent condition who incorrectly reported seeing the disclosure). The point value of A′ can then be interpreted as the proportion of correct recognition (or by multiplying it by 100, the percentage correct), controlling for response bias and differences in motivation.

For the measure of risk recall, respondents had the opportunity to list the risks they could remember from the ad in an open-ended text box. After open-ended response codes were established, two team members independently coded 10% of the raw participant responses for each illness population. A third staff member then reviewed coding for accuracy and evidence of intercoder reliability between coders using Krippendorff’s alpha.⁴⁷ In all cases, alpha was .79 or higher, indicating acceptable intercoder reliability, and the remaining responses were divided between the two coders. A measure of recall of serious and actionable risks was created by summing the number of correct serious and actionable risks respondents listed in response to this open-ended question (ranging from 0–8 in depression, 0–6 in insomnia, and 0–3 in high cholesterol).^‡

A separate item assessed respondents’ recognition of the drug’s risks and benefits. Respondents were shown a combination of statements about the potential risks and benefits of taking the advertised drug (some statements were true and some were foils) and were asked to check off which statements were mentioned as a benefit or risk of taking the product. A separate benefit and risk recognition item was created based on those responses. Benefit recognition scores could range from 0 to 4 for all three illness populations (two benefits were in the ad and two were foils). Risk recognition scores could range from 0 to 7 for insomnia and depression conditions (four risks were in the ad and three were foils) and from 0 to 4 for high cholesterol (two risks were in the ad and two were foils). The risk recall, risk recognition, and benefit recognition measure were then converted into proportions to reflect the proportion of key risks recalled, the proportion of risks correctly recognized, and the proportion of benefits correctly recognized.

Two different items measured respondents’ perceived risk of the drug. People reported on a 5-point scale (not at all serious to very serious) how serious they thought the risks and side effects of the drug were as a whole and how serious they thought the side effects would be if the drug caused a person to have a side effect. The average of these two items constituted a measure of perceived risk severity (Cronbach’s alpha = 0.77 for the depression sample, 0.75 for insomnia, and 0.72 for high cholesterol). Respondents reported efficacy magnitude using the following item, “Based on the information in the [Drug Name] ad, if [Drug Name] did help a person’s [disease condition], how much would it help?” Responses on the 5-point scale ranged from would help a little to would help a lot. To measure risk and benefit trade-off, respondents rated their perceptions of the balance between risks and benefits by responding to the question, “Thinking overall about the risks and benefits of [Drug Name], would you say it has …” with response options ranging from 1 (many more risks than benefits) to 5 (many more benefits than risks).

Intentions to seek out additional information about the advertised drug were measured by asking respondents how likely it was that they would engage in four searching behaviors (using a 5-point scale ranging from very unlikely to very likely): ask a doctor about the advertised drug, read the patient labeling for more information about the advertised drug, look for more information about the advertised drug, and look for more information about the medical condition of interest. These items were averaged to create an index of information seeking (Cronbach’s alpha = 0.88 for the depression sample, 0.88 for insomnia, and 0.91 for high cholesterol). To assess whether respondents would choose to read the patient labeling, respondents were asked the following, “Would you like to receive the patient labeling for [Drug Name] at the end of this survey to learn more about the complete list of risks and side effects?” Respondents who said yes were directed to a link to the approved patient labeling at the end of the study.

A 7-item, 5-point, Likert scale (strongly disagree to strongly agree) was used to measure perceptions of the disclosure statement (noticeable, believable, distracting [reverse-coded], important, clear, too long [reverse-coded], helpful). The seven items were averaged to form a scale, where higher scores indicate more positive evaluations of the disclosure statement (Cronbach’s alpha = 0.84 for the depression sample, 0.85 for insomnia, and 0.85 for high cholesterol).

To assess respondents’ perceptions of the amount of risk information included in the ads, respondents were asked to report (on a 5-point scale ranging from strongly disagree to strongly agree) the extent to which they agreed with the statement, “The ad did not give enough information about the possible risks and side effects of using [Drug Name].” This item is similar to items used in previous research.^3–5 A 3-item, 5-point, Likert scale (strongly disagree to strongly agree) was used to measure the perceived clarity of the risk statement. Respondents were asked to rate their level of agreement that the risk statement was clear, that the risk statement was informative, and that the ad clearly communicated the risks and side effects of the drug. The three items were averaged, and higher scores indicate greater perceived clarity of the risk statement (Cronbach’s alpha = 0.79 for the depression sample, 0.79 for insomnia, and 0.83 for high cholesterol).

In addition to the dependent variables, study data included key demographic and background variables, including age, race and ethnicity, gender, household income, and items related to self-reported attention to the ad (ranging from 1 = a little attention to 5 = a lot of attention), whether the respondent was currently taking a medication for their illness, and attitude toward the ad (good-bad, pleasant-unpleasant, favorable-unfavorable, convincing-unconvincing, entertaining-boring, interesting-uninteresting, honest-dishonest, simple-complicated, important-unimportant, unique-ordinary, Cronbach’s alpha = 0.89 for the depression sample, 0.91 for insomnia, and 0.90 for high cholesterol).

ANALYSES

We report both descriptive statistics and comparisons of dependent variable means and percentages with either analysis of variance (ANOVA) or chi-square statistics. For the majority of analyses, 2×2 ANOVA tests were conducted to examine the main effects of the type of risk statement (unedited versus revised) and the presence versus absence of a disclosure statement, and the interaction between the two. When the interaction between type of risk statement and presence versus absence of the disclosure was statistically significant at P < .05, planned comparisons were conducted to identify any significant differences among four comparisons of interest using a Bonferroni-adjusted significance threshold of .0125.

After running descriptive statistics for the items, the associations between the dependent variables and the key demographic and background variables were examined to determine whether subsequent analyses should control for those variables. To justify a variable’s inclusion in the model, the potential covariate needed to have at least a moderately strong association with the dependent variable (r, V, or square-root of R² ≥ .30).^48,49 Based on those bivariate analyses, attitude toward the ad was included as a covariate for four outcomes (benefit magnitude, risk-benefit trade-off, information-seeking intentions, and perceptions of the risk statement) across all three illness populations, and self-reported attention was included as a covariate for the insomnia condition for analyses related to risk recall, evaluations of the disclosure, and perceptions of the risk statement. For significant findings, Cohen’s f, Cramer’s V, and Cohen’s d are reported as measures of effect size.⁵⁰ When covariates are included, all reported statistics reflect the adjusted statistics (i.e., statistics that have been adjusted for the effects of the covariate) (with the exception of the effect size for Cohen’s d). For those analyses that included covariates, we conducted sensitivity analyses to see if the pattern of results changed based on inclusion of the covariate. The pattern of significant findings remained the same for all but one analysis. For insomnia, the finding that risk statement significantly predicted the proportion of risks recalled was no longer significant (P = .086). All analyses were conducted using SAS Version 9.4 or SPSS Version 23.0.

RESULTS

Demographic Characteristics

Table 1 contains the demographic characteristics of the three illness populations. For those in the depression sample, the mean age was 53, 56.2% were female, and 75.4% were white and non-Hispanic. Of the 79.0% that were taking a prescription drug for their depression, only 3.4% were taking the advertised drug. For those in the insomnia sample, the mean age was 55, 53.7% were female, and 69.0% were white and non-Hispanic. Of the 63.4% that were taking a prescription drug for their insomnia, 4.8% were taking the advertised drug. For those in the high cholesterol sample, the mean age was 63, 48.2% were female, and 63.6% were white and non-Hispanic. Of the 84.0% that were taking a prescription drug for their high cholesterol, 10.8% were taking the advertised drug.

Table 1.

Distribution of Demographic Characteristics, by Illness Population

Category	Depression (N = 500)	Insomnia (N = 500)	High Cholesterol (N = 500)
Gender
Male	219 (43.8%)	232 (46.3%)	259 (51.8%)
Female	281 (56.2%)	268 (53.7%)	241 (48.2%)
Age
Mean in years (SD)	53.0 (14.0)	55.4 (12.4)	62.9 (11.2)
18–29	32 (6.4%)	18 (3.6%)	9 (1.8%)
30–44	115 (23.0%)	81 (16.2%)	26 (5.2%)
45–59	148 (29.6%)	184 (36.8%)	106 (21.2%)
60+	205 (41.0%)	217 (43.4%)	359 (71.8%)
Race/Ethnicity
White (non-Hispanic)	376 (75.4%)	345 (69.0%)	318 (63.6%)
Black (non-Hispanic)	37 (7.4%)	51 (10.2%)	61 (12.2%)
Hispanic	66 (13.2%)	77 (15.4%)	71 (14.2%)
Other (non-Hispanic)	10 (2.0%)	19 (3.8%)	41 (8.2%)
Multiracial (non-Hispanic)	10 (2.0%)	8 (1.6%)	9 (1.8%)
Education
High school or less	118 (23.6%)	147 (29.4%)	101 (20.2%)
Some college	147 (29.4%)	145 (29.0%)	148 (29.6%)
Bachelor’s degree	124 (24.8%)	115 (23.0%)	131 (26.2%)
Master’s degree or higher	111 (22.2%)	93 (18.6%)	120 (24.0%)
Income
Less than $30,000	93 (18.7%)	104 (20.8%)	56 (11.3%)
$30,001–$75,000	210 (42.3%)	211 (42.3%)	191 (38.6%)
$75,001–$150,000	162 (32.6%)	140 (28.1%)	194 (39.2%)
$150,001 or more	32 (6.4%)	44 (8.8%)	54 (10.9%)
Currently taking prescription for medical condition
Yes	395 (79.0%)	317 (63.4%)	420 (84.0%)
Yes—Taking advertised drug	17 (3.4%)	24 (4.8%)	54 (10.8%)
Yes—Not taking advertised drug	378 (75.6%)	293 (58.6%)	366 (73.2%)
Not taking a prescription for medical condition	105 (21.0%)	183 (36.6%)	80 (16.0%)

Note: Percentages may not always add up to 100.0% because of rounding. These data generally reflect patterns among adults in the U.S. with depression, insomnia, and high cholesterol. Our sample includes more women than men and women are more likely than men to report recent depression⁵⁴ and insomnia^55–56. Similarly, insomnia and high cholesterol is more prevalent in older adults in the U.S., and our average age was 55.4 and 62.9, respectively^55,57. Finally, for all three medical conditions, our sample included more than 60% of the sample that was currently taking a prescription medication for their condition, which is in line with patterns found in the U.S.^54,55,57.

Disclosure Recognition and Evaluations

Depression.

When respondents were exposed to the disclosure, 72.3% of them reported seeing or hearing the statement in the ad. The bias-corrected recognition was slightly lower at 63.0%. Significantly more people in the disclosure-present condition reported seeing or hearing the disclosure than people in the disclosure-absent condition, χ² (1, N = 500) = 11.05, P < .001, V = .15. Respondents in the disclosure-present conditions were asked follow-up questions regarding their evaluations about the disclosure statement. On average, respondents held positive evaluations of the disclosure statement (M = 3.99, SD = 0.69), and these evaluations were not affected by type of risk statement, P > .05.

Insomnia.

Among respondents who were exposed to the disclosure statement, 81.4% of them reported seeing or hearing the statement in the ad. The bias-corrected recognition was slightly lower at 70.0%. Significantly more people in the disclosure-present condition reported noticing the disclosure statement than in the disclosure-absent condition, χ² (1, N = 500) = 29.96, P < .001, V = .24. On average, respondents held positive evaluations of the disclosure statement (M = 4.01, SD = 0.70), and these evaluations were not affected by type of risk statement, P > .05.

High cholesterol.

When respondents were exposed to the disclosure statement, 77.1% of them reported seeing or hearing the statement in the ad. The bias-corrected recognition was slightly lower at 69.0%. Significantly more people in the disclosure-present condition reported noticing the disclosure statement than in the disclosure-absent condition, χ² (1, N = 500) = 28.00, P < .001, V = .24. On average, respondents held positive evaluations of the disclosure statement (M = 3.91, SD = 0.67), and these evaluations were not affected by type of risk statement, P > .05.

Recall and Recognition of Product Risks

Depression.

There was a statistically significant main effect of risk statement on risk recall (F[1, 478] = 13.47, P < .001, f = .17) and risk recognition (F[1, 496] = 14.10, P < .001, f = .17). Respondents who saw the revised risk statement correctly recalled and accurately recognized more serious and actionable risks than respondents who saw the unedited risk statement (see Table 2).

Table 2.

Means (SE) of Dependent Variables by Risk Statement and Disclosure Presence for Each Illness Population

		Type of Risk Statement			Disclosure Presence
Illness Population and Dependent Variable	Overall Sample M (SE)	Unedited Risk Statement M (SE)	Revised Risk Statement M (SE)	P value*	Disclosure Absent M (SE)	Disclosure Present M (SE)	P value*
Depression		n = 232	n = 268		n = 265	n = 265
Proportion of key risks recalled	0.14 (0.01)	0.12 (0.01)	0.16 (0.01)	<.001	0.13 (0.01)	0.15 (0.01)	.089
Proportion of risks correctly recognized	0.70 (0.01)	0.68 (0.01)	0.73 (0.01)	<.001	0.71 (0.01)	0.70 (0.01)	.706
Proportion of benefits recognized	0.89 (0.01)	0.86 (0.01)	0.91 (0.01)	.006	0.90 (0.01)	0.87 (0.01)	.176
Risk severity overall	4.00 (0.04)	3.88 (0.06)	4.11 (0.06)	.006	3.96 (0.06)	4.03 (0.06)	.337
Benefit magnitudea	3.75 (0.04)	3.71 (0.06)	3.79 (0.06)	.283	3.73 (0.06)	3.77 (0.06)	.628
Risk-benefit trade-offa	2.90 (0.05)	2.82 (0.07)	2.98 (0.07)	.112	2.93 (0.07)	2.87 (0.07)	.546
Info-seek intentionsa	2.92 (0.05)	2.95 (0.07)	2.90 (0.07)	.617	2.87 (0.07)	2.98 (0.07)	.265
Not enough information about risksa	2.56 (0.05)	2.53 (0.08)	2.59 (0.07)	.536	2.50 (0.07)	2.61 (0.07)	.266
Clarity of risk statementa	3.79 (0.05)	3.85 (0.05)	3.94 (0.04)	.155	3.96 (0.05)	3.83 (0.05)	.062
Insomnia		n = 252	n = 248		n = 258	n = 242
Proportion of key risks recalledb	0.22 (0.01)	0.20 (0.01)	0.24 (0.01)	.024	0.21 (0.01)	0.23 (0.01)	.386
Proportion of risks correctly recognized	0.73 (0.01)	0.70 (0.01)	0.75 (0.01)	.008	0.73 (0.01)	0.73 (0.01)	.835
Proportion of benefits recognized	0.77 (0.01)	0.74 (0.01)	0.81 (0.01)	<.001	0.74 (0.01)	0.81 (0.01)	<.001
Risk severity overall	3.87 (0.04)	3.83 (0.06)	3.91 (0.06)	.350	3.88 (0.06)	3.86 (0.06)	.800
Benefit magnitudea	4.12 (0.04)	4.13 (0.06)	4.11 (0.06)	.747	4.13 (0.06)	4.11 (0.06)	.786
Risk-benefit trade-offa	3.07 (0.05)	3.11 (0.07)	3.04 (0.07)	.437	3.11 (0.07)	3.04 (0.07)	.465
Info-seek intentionsa	3.26 (0.05)	3.34 (0.07)	3.18 (0.07)	.087	3.29 (0.07)	3.24 (0.07)	.604
Not enough information about risksa,b	2.44 (0.05)	2.30 (0.07)	2.59 (0.07)	.005	2.50 (0.07)	2.39 (0.07)	.287
Clarity of risk statementa,b	3.93 (0.04)	4.11 (0.05)	3.90 (0.05)	.001	3.97 (0.04)	4.04 (0.05)	.246
High Cholesterol		n = 263	n = 237		n = 234	n = 266
Proportion of key risks recalled	0.30 (0.01)	0.28 (0.02)	0.33 (0.02)	.035	0.27 (0.02)	0.33 (0.02)	.012
Proportion of risks correctly recognized	0.73 (0.01)	0.72 (0.01)	0.73 (0.02)	.805	0.73 (0.02)	0.72 (0.01)	.465
Proportion of benefits recognized	0.90 (0.01)	0.90 (0.01)	0.90 (0.01)	.781	0.90 (0.01)	0.90 (0.01)	.949
Risk severity overall	3.63 (0.04)	3.58 (0.06)	3.68 (0.06)	.194	3.62 (0.06)	3.64 (0.06)	.843
Benefit magnitudea	4.29 (0.04)	4.22 (0.05)	4.37 (0.05)	.038	4.28 (0.05)	4.31 (0.05)	.617
Risk-benefit trade-off1	3.44 (0.05)	3.47 (0.06)	3.41 (0.07)	.518	3.44 (0.07)	3.44 (0.06)	.965
Info-seek intentionsa, †	3.15 (0.05)	3.16 (0.07)	3.14 (0.07)	.851	3.07 (0.07)	3.23 (0.07)	.092
Not enough information about risksa	2.80 (0.05)	2.75 (0.07)	2.84 (0.07)	.358	2.82 (0.07)	2.78 (0.07)	.660
Clarity of risk statementa	3.69 (0.04)	3.78 (0.05)	3.72 (0.05)	.519	3.71 (0.05)	3.79 (0.04)	.208

Note: All items ranged from 1 to 5.

^aAttitude toward the ad was included as a covariate, so adjusted statistics are reported.

^bAttention to the ad was included as a covariate, so adjusted statistics are reported.

^†A statistically significant interaction between risk statement and disclosure presence was found.

^*P values are from 2×2 ANOVA tests.

Bold items are statistically significant.

Insomnia.

This study found a statistically significant main effect of risk statement on risk recall (F[1, 482] = 5.10, P = .024, f = .10) and risk recognition (F[1, 496] = 7.08, P = .008, f = .12). Respondents who saw the revised risk statement correctly recalled and accurately recognized more serious and actionable risks than respondents who saw the unedited risk statement (see Table 2).

High cholesterol.

There was a statistically significant main effect of risk statement on risk recall, F(1, 475) = 4.45, P = .035, f = .10, such that respondents who saw the revised risk statement correctly recalled more serious and actionable risks than respondents who saw the unedited risk statement. There was no significant effect of risk statement on risk recognition, P > .05. Somewhat unexpectedly, there was a statistically significant main effect of the disclosure statement on risk recall, F(1, 475) = 6.33, P = .012, f = .11. Respondents who saw the disclosure statement correctly recalled more serious and actionable risks than respondents who did not see the disclosure statement (see Table 2).

Recognition of Product Benefits

Depression.

There was a statistically significant main effect of risk statement on benefit recognition, F(1, 496) = 7.72, P = .006, f = .12. Respondents who saw the revised risk statement correctly recognized more benefits than respondents who saw the unedited risk statement (see Table 2).

Insomnia.

There was a statistically significant main effect of risk statement and disclosure presence on benefit recognition (see Table 2). Respondents who saw the revised risk statement correctly recognized more benefits than respondents who saw the unedited risk statement, F(1, 495) = 15.55, P < .001, f = .17. When the disclosure was present, respondents correctly recognized more benefits than respondents who did not see a disclosure, F(1, 495) = 13.15, P < .001, f = .16.

High cholesterol.

Type of risk statement or disclosure presence had no effect on benefit recognition, P > .05.

Perceived Product Risks and Benefits

Depression.

There was a statistically significant main effect of risk statement on overall risk severity, F(1, 496) = 7.57, P = .006, f = .12 (see Table 2). Respondents who saw the revised risk statement rated the risks and side effects as more severe than respondents who saw the unedited risk statement. This study found no evidence of a statistically significant effect of type of risk statement or disclosure presence on benefit magnitude or risk-benefit trade-off, P > .05.

Insomnia.

There was no evidence of a statistically significant effect of type of risk statement or disclosure presence on overall risk severity, benefit magnitude, or risk-benefit trade-off, P > .05.

High cholesterol.

There was a statistically significant main effect of risk statement on benefit magnitude, F(1, 493) = 4.33, P = .038, f = .09 (see Table 2). Respondents who saw the revised risk statement rated the product as more effective than respondents who saw the unedited risk statement. This study found no evidence of a statistically significant effect of type of risk statement or disclosure presence on overall risk severity or risk-benefit trade-off, P > .05.

Information Seeking and Interest in Patient Labeling

Depression.

Type of risk statement or disclosure presence did not have a statistically significant effect on information-seeking intentions, P > .05. Overall, 30.2% of respondents said they wanted to receive the patient labeling for the advertised drug at the end of the study; there were no differences in patient labeling interest by disclosure presence or type of risk statement, P > .05.

Insomnia.

Type of risk statement or disclosure presence did not have a statistically significant effect on information-seeking intentions or interest in receiving patient labeling at the end of the study, P > .05. Overall, 34.8% of respondents said they wanted to receive the patient labeling for the advertised drug at the end of the study; there were no differences in patient labeling interest by disclosure presence or type of risk statement, P > .05.

High cholesterol.

This study found a statistically significant interaction effect between risk statement and disclosure statement, F(1, 492) = 12.76, P < .001, f = .15. Among respondents who saw the disclosure, those who saw the unedited risk statement (M = 3.42) reported higher intentions to seek information about the drug than respondents who saw the revised risk statement (M = 3.05), P = .006, Cohen’s d = .27. In addition, among respondents who saw the unedited risk statement, those who saw the disclosure statement (M = 3.42) reported higher intentions to seek information about the drug than respondents who did not see the disclosure statement (M = 2.91), P < .001, Cohen’s d = .36. Overall, 27.9% of respondents said they wanted to receive the patient labeling for the advertised drug at the end of the study; there were no differences in patient labeling interest by disclosure presence or type of risk statement, P > .05.

Clarity of Risk Statement

Depression.

Overall, respondents disagreed that the ad did not give enough information about the possible risks and side effects of using the advertised drug (M = 2.56, SE = 0.05). Only 23.0% of respondents agreed or strongly agreed that the ad did not give enough information about the risks and side effects. Perceptions on the amount of risks presented did not differ by type of risk statement or disclosure presence, P > .05. Respondents also thought the ad presented the risks clearly (M = 3.90, SE = 0.03). Type of risk statement or disclosure presence did not have a statistically significant effect on the perceived clarity of the risks, P > .05 (see Table 2).

Insomnia.

Overall, respondents disagreed that the ad did not give enough information about the possible risks and side effects of using the advertised drug (M = 2.45, SE = 0.05). Only 20.6% of respondents agreed or strongly agreed that the ad did not give enough information about the risks and side effects. Type of risk statement had a statistically significant effect on perceptions of the amount of risk information included in the ad, F(1, 491) = 7.92, P = .005, f = .12. Respondents who saw the unedited risk statement had greater disagreement that the ad did not contain enough information about risks (see Table 2).

Respondents also thought the ad presented the risks clearly (M = 4.00, SE = 0.03). Once again, type of risk statement had a statistically significant effect on perceived clarity of the risks, F(1, 490) = 10.84, P = .001, f = .13. Respondents who saw the unedited risk statement had greater perceived clarity scores than respondents who saw the revised risk statement (see Table 2).

High cholesterol.

Overall, respondents disagreed that the ad did not give enough information about the possible risks and side effects of using the advertised drug (M = 2.80, SE = 0.05). Thirty percent (30.0%) of respondents agreed or strongly agreed that the ad did not give enough information about the risks and side effects. Perceptions on the amount of risks presented did not differ by type of risk statement or disclosure presence, P > .05. Respondents also had somewhat positive perceptions of how clearly the ad presented the risks (M = 3.75, SE = 0.03). Type of risk statement or disclosure presence did not have a statistically significant effect on the perceived clarity of the risks, P > .05 (see Table 2).

DISCUSSION

The length of major risk statements in DTC prescription drug television ads has proven controversial; some believe these statements include too much information, and some believe they include too little information. This study demonstrates the effectiveness of a limited risks plus disclosure strategy that takes into consideration these varied perspectives. Relative to the unedited major statement, presentation of serious and actionable drug risks plus a disclosure indicating that not all risks are presented can facilitate retention of the serious and actionable risks as well as product benefits and allow for high levels of recognition for the disclosure statement. Moreover, participants perceive the revised risk statements to be similar in clarity to risk statements currently in the marketplace. Although some variation in outcomes between the three medical conditions was observed, the general pattern of effects observed is compelling. Arguably, this alternative approach is a viable one, and one that policy makers could consider. Although consumers may not remember all presented risks in any case, the types of risks that are presented can be restricted to those serious and actionable risks deemed most relevant to them and, by extension, can help them remember those more relevant risks as well as the product’s benefits. At the same time, a disclosure may effectively communicate that not all risks are presented and provide information about where additional information can be obtained.

It is also noteworthy that consumer perceptions and intentions regarding the product are similar regardless of whether they view the unedited or revised risk statement. One might expect that restricting the list of risks to those that are serious (and actionable) would cause consumers to perceive the risks as more severe, but as the results for two out of the three medical conditions show, this tended not to be the case. Serious risks are presented in both the unedited and revised risk statements, so perhaps inclusion of such risks in a major statement is paramount. Or reactions may depend in part on the overall number of risks. For those in the depression sample, participants exposed to the revised risk statement did perceive the product risks to be more severe, and the manipulations for this medical condition involved the largest reduction in number of risks presented. Consistent with findings from Kahn and Kupor,⁵¹ participants in this condition may have underestimated the severity of risks when such risks were presented alongside numerous nonserious risks. As yet another possibility, participants exposed to the revised risk statement in the depression sample may have perceived the risks as greater due to the framing statement regarding “severe, life-threatening reactions” (whereas the framing statement for the high cholesterol and insomnia samples referenced “serious reactions” only). Future research can address these questions; for now, we take note that risk perceptions did not vary for those in either the insomnia or high cholesterol samples.

Likewise, very little impact was observed on intentions regarding the product based on exposure to the unedited versus revised risk statements. For those in the depression and insomnia samples, no effects on intention were observed. For those in the high cholesterol sample, presence of the disclosure prompted interest in seeking additional information, but the unedited versus revised risk statements did not result in increased information seeking. In addition to asking about various intentions regarding the advertised product, participants were also asked whether they would like to receive the relevant patient labeling. Inclusion of this behavioral measure served as a strength of this study given that much research relies on intention measures alone.⁵² Nonetheless, no effects were observed for this outcome, suggesting that the proposed revisions to major risk statements are unlikely to affect intentions and behaviors.

At least some recommendations can be made for implementing a limited risks plus disclosure strategy; or when necessary, further research needs can be noted. First, the disclosure statement used in this research was presented in both audio and visual formats, or with dual modality, and without competing text information during the disclosure presentation. This approach was adopted based on observations in pilot tests that audio communication alone resulted in low levels of disclosure recognition, as well as indications in the broader literature that disclosures and warnings often go unnoticed.^32–36 With dual-modality presentation, 63% to 70% of participants (depending on condition and bias corrected) correctly indicated that they saw or heard the disclosure statement. These observations are consistent with extensive literature pertaining to dual- versus single-modality presentations and impact on information processing.^37,38 Therefore, policy makers and drug ad sponsors should consider dual modality to be a critical element in the effectiveness of this approach. Without dual modality, far fewer consumers may notice and remember the disclosure statement.

Additionally, an opening sentence that framed the risks prefaced each modified major statement. Whereas the opening statement for the insomnia and high cholesterol ads both referred to serious risks, the opening statement for the depression ad referred to severe, life-threatening risks. These opening statements reflect the seriousness of the risks as determined in consultation with professional reviewers at FDA. We did not specifically examine the impact of these framing statements apart from the risks, however; thus, it is not possible to determine whether differences between medical conditions could have been influenced by the framing of the risks that followed. Thus, we advise that researchers pursue empirical investigations into the impact of such opening statements.

Additionally, it is acknowledged that identification of risks as serious and actionable is a somewhat subjective pursuit. For this study’s purposes, expert reviewers at FDA were consulted to identify such risks, and independent tests were also conducted to determine likely consumer perceptions (not reported here). We acknowledge both the challenges of 1) determining which risks should be deemed serious and actionable and 2) securing regulatory agency (FDA) agreement with such determinations. Further research on risk perceptions would be useful in this regard, particularly in regard to specific risks.

One consequence of revising major statements to include only serious and actionable risks involves a greater reduction in the number of risks presented for some products than for others. In the present research, the revised risk statement for the depression ad resulted in a 53% reduction in risks presented; for insomnia, 46%; and for high cholesterol, 33%. Another point to consider is that although some drug products may not have both serious and actionable risks, all DTC broadcast advertisements are required to present a fair balance of risk information when presenting information relating to the drug’s effectiveness. DTC broadcast ads that provide information about a drug’s effectiveness would be expected to contain at least some risk information, even if the risks are not serious and actionable. The lack of uniform applicability associated with these issues warrant additional consideration from a regulatory and practical perspective.

The limits of the present research inspire several directions for future research. Of note, no statistical comparisons can be made across medical conditions in the present research. Including multiple medical conditions allowed for replication of the study across different illness populations. However, such an approach also required the use of different ads focused on these different medical conditions. Thus, determining whether differences in responses between medical conditions result because of the experimental manipulations or merely because of variations in the ads is not possible. Interested researchers may be able to find ways to overcome this limitation in future work, such as by using fictitious ads where only the intended experimental manipulations and indication vary. The study used a non-probability based convenience sample, as we did not intend to generate nationally or locally representative results or precise estimates of population parameters from this study. Instead, the strength of the experimental design used in this study lies in its internal validity, on which meaningful estimates of differences caused by carefully manipulated variables can be produced and generalized. The potential bias of using an online convenience sample cannot be quantified, although there is evidence to suggest that this approach has comparable reliability and validity to more traditional survey methods.⁵³

Future research should also explore longer-term effects on outcomes of interest. The present research captured only short-term effects. Participants responded to the series of dependent measures immediately following exposure to the stimuli. It would be useful to additionally collect information at varied time points following exposure to the stimuli. It may be that serious and actionable risks are more resistant to decay or interference effects relative to other risks. Likewise, consumer perceptions of overall or specific risks may be found to vary over time, as may intentions regarding the advertised product. Further, given each study had nine dependent variables, concerns related to type I error should be taken into consideration. We decided, a priori, to statistically control for pairwise comparisons within an analysis. Observed effect sizes, along with exact p values are provided to allow the reader to evaluate the importance and statistical significance of results.

Future research should also investigate risk comprehension. The present research emphasized memory for risks across the experimental conditions. In some sense, remembering information may be indicative of comprehending the information. Nonetheless, recognition and comprehension are not identical constructs, so future research should investigate impact on comprehension more fully.

Researchers may also wish to consider alternative strategies for limiting risks in the major statement. In this study, serious and actionable risks were considered to be of greatest relevance to consumers. Accordingly, we restricted the major statements in the test conditions to include only serious (which includes severe) and actionable risks (plus the disclosure). However, other approaches may be suitable for identifying the most important risks, and restricting the major statements to these other risks may reveal similar outcomes. The present research does not attempt to separate out the impact of type of risks versus number of risks, but it seems likely that the mere number of risks had some impact. Indeed, such an expectation would be well grounded in the literature on working memory and was considered in the review of the literature. Consumers are limited in the amount of information that they can process at one time, so limiting the risks they are presented with should help them process pertinent information regardless of whether such risks are serious or actionable or of some other type. Thus, consideration of other strategies for limiting risks in major statements, as well as empirical testing of such strategies, is a worthwhile pursuit.

DTC television ads are an important part of the overall universe of information about prescription drug treatments and, in some cases, provide the only information consumers are exposed to about a product before they interact with a health care professional. Therefore, it is important that such ads provide clear, useful, and non-misleading information to the intended audience. Currently, DTC television ads may expose consumers to large amounts of information, which may restrict their ability to process the information most relevant to them. Regulatory authorities and sponsors may facilitate consumers’ ability to process important product information by considering including only risks in the major statement that are serious and actionable and including a disclosure to indicate that not all risks are presented.

Appendix A:

Table 1A.

Depression Condition Risk Statement Manipulations and Operationalization for Risk Recall and Recognition

Condition	RS Language	Length of RS1	# of Risks	Key Risks in Evaluating Recall	Key risks in Evaluating Recognition
Unedited RS - Disclosure Absent	Abilify is not for everyone. [Abilify is not for everyone.] Call your doctor if your depression worsens or you have unusual changes in behavior or thoughts of suicide. Antidepressants can increase these in children, teens, and young adults. [ABILIFY is not approved for depression in children under 18.] Elderly dementia patients taking Abilify have an increased risk of death or stroke. [ABILIFY is not approved for dementia.] Call your doctor if you have high fever, stiff muscles, and confusion to address a possible life threatening condition, or if you have uncontrollable muscle movements as these could become permanent. [www.AddAbilify.com 1-855-69-ABILIFY] High blood sugar has been reported with Abilify and medicines like it, and in extreme cases can lead to coma or death. [Your blood sugar should be monitored if you have or are at risk for diabetes.] Other risks include increased cholesterol, weight gain, decreases in white blood cells, which can be serious, dizziness on standing, seizures, trouble swallowing, and impaired judgment or motor skills. [Your weight should be monitored.] [Use caution before driving or operating heavy machinery.]	00:40 (WC:127)	17	Increased risk of suicidal thoughts or behavior Sudden changes in mood or behavior Increased risk of stroke High fever Stiff muscles Confusion Uncontrollable muscle movements Death	Sudden changes in mood or behavior Increased risk of suicidal thoughts or behavior Increased risk of stroke Uncontrollable muscle movement Three Foil Risks Nausea or vomiting Increased risk of developing certain forms of cancer Liver problems such as yellowing of skin or eyes
Unedited RS - Disclosure Present	Same as above plus: This is not a full list of risks and side effects. Talk to your doctor and read the patient labeling for more information. [This is not a full list of risks and side effects. Talk to your doctor and read the patient labeling for more information]	00:54 (WC:150)	17	Same as above	Same as above
Revised RS - Disclosure Absent	Abilify can cause severe, life-threatening reactions. [Abilify is not for everyone]. These include suicidal thoughts or behavior, sudden changes in behavior or mood, stroke, and uncontrollable muscle movements. [www.AddAbilify.com 1-855-69-ABILIFY] Get medical help right away if you experience any of these symptoms or high fever, stiff muscles, or confusion. Abilify can cause more risks in some people, which can be severe and life-threatening. Talk to your doctor about all your medical conditions.	00:26 (WC: 65)	8	Same as above	Same as above
Revised RS - Disclosure Present	Same as above plus: This is not a full list of risks and side effects. Talk to your doctor and read the patient labeling for more information. [This is not a full list of risks and side effects. Talk to your doctor and read the patient labeling for more information]	00:31 (WC: 88)	8	Same as above	Same as above

Note. Text in italics indicates when the narration is coupled with a matching on-screen text. On-screen text appears in brackets.

¹The length of the risk statement includes that duration in seconds and the word count (wc).

Insomnia Condition Risk Statement Manipulations and Operationalization for Risk Recall and Recognition

Condition	RS Language	Length of RS1	# of Risks	Key Risks in Evaluating Recall	Key risks in Evaluating Recognition
Unedited RS - Disclosure Absent	Do not take Lunesta if you are allergic to anything in it. [Do not take LUNESTA if you are allergic to anything in it.] When taking Lunesta, don’t drive or operate machinery until you feel fully awake. Walking, eating, driving, or engaging in other activities while asleep without remembering it the next day have been reported. [Take LUNESTA before bed, and only if you have 7 to 8 hours to sleep before becoming active.] Lunesta should not be taken together with alcohol. Abnormal behaviors may include aggressiveness, agitation, hallucinations, or confusion. In depressed patients, worsening of depression including risk of suicide may occur. Alcohol may increase these risks. Allergic reactions such as tongue or throat swelling occur rarely and may be fatal. [LUNESTA has some risk of dependency. It’s non-narcotic.] Side effects may include unpleasant taste, headache, dizziness, and morning drowsiness.	00:34 (WC:102)	11	Sleep walking and other activities Changes in behavior Depression Allergic reaction Dependency Death	Sleep walking and other activities Changes in behavior or mood such as aggressiveness Can increase the risk of suicide Can cause potentially fatal allergic reactions Three Foil Risks Nausea or vomiting Can increase the risk of developing certain forms of cancer Can cause liver problems such as yellowing of skin or eyes
Unedited RS - Disclosure Present	Same as above plus: This is not a full list of risks and side effects. Talk to your doctor and read the patient labeling for more information. [This is not a full list of risks and side effects. Talk to your doctor and read the patient labeling for more information]	00:42 (WC:125)	11	Same as above	Same as above
Revised RS - Disclosure Absent	Lunesta can cause serious reactions. These include sleep walking and other activities, changes in behavior or mood, and life-threatening allergic reactions. Get medical help right away if you experience symptoms of an allergic reaction. [Do not take LUNESTA if you are allergic to anything in it.] Lunesta can increase the risk of suicide in people who are depressed. [Take LUNESTA before bed, and only if you have 7 to 8 hours to sleep before becoming active.] Only take Lunesta if you have 7 to 8 hours to sleep. Do not drive or operate heavy machinery until you are fully awake. [LUNESTA has some risk of dependency. It’s non-narcotic.]	00:26 (WC:70)	6	Same as above	Same as above
Revised RS - Disclosure Present	Same as above plus: This is not a full list of risks and side effects. Talk to your doctor and read the patient labeling for more information. [This is not a full list of risks and side effects. Talk to your doctor and read the patient labeling for more information]	00:35 (WC:93)	6	Same as above	Same as above

Note. Text in italics indicates when the narration is coupled with a matching on-screen text. On-screen text appears in brackets.

¹The length of the risk statement includes that duration in seconds and the word count (wc).

High Cholesterol Condition Risk Statement Manipulations and Operationalization for Risk Recall and Recognition

Condition	RS Language	Length of RS1	# of Risks	Key Risks in Evaluating Recall1	Key risks in Evaluating Recognition
Unedited RS - Disclosure Absent	Crestor is not right for everyone, like people with liver disease, or women who are nursing, pregnant, or may become pregnant. [See our ad in Weight Watchers Magazine] Tell your doctor about other medicines you’re taking. [Call your doctor if you have muscle pain or weakness, even after stopping CRESTOR] Call your doctor right away if you have muscle pain or weakness, feel unusually tired, have loss of appetite, upper belly pain, dark urine, or yellowing of skin or eyes. [800-CRESTOR CRESTOR.COM]. These could be signs of rare but serious side effects.	00:21 (WC:69)	6	Muscle pain or weakness Liver problems Do not take if nursing or pregnant	Muscle problems such as pain or weakness Liver problems such as yellowing of skin or eyes Two Foil Risks Nausea or vomiting Can increase the risk of developing certain forms of cancer
Unedited RS - Disclosure Present	Same as above plus: This is not a full list of risks and side effects. Talk to your doctor and read the patient labeling for more information. [This is not a full list of risks and side effects. Talk to your doctor and read the patient labeling for more information]	00:28 (WC:92)	6	Same as above	Same as above
Revised RS - Disclosure Absent	Crestor can cause serious reactions. These include muscle and liver problems. [See our ad in Weight Watchers Magazine] Crestor can cause more risks in some people. [Call your doctor if you have muscle pain or weakness, even after stopping CRESTOR.] Do not take Crestor if you have liver disease, are nursing, pregnant, or may become pregnant, or if you are allergic to anything in Crestor.	00:17 (WC:44)	4	Same as above	Same as above
Revised RS - Disclosure Present	Same as above plus: This is not a full list of risks and side effects. Talk to your doctor and read the patient labeling for more information. [This is not a full list of risks and side effects. Talk to your doctor and read the patient labeling for more information]	00:21 (WC:67)	4	Same as above	Same as above

Note. RS = Risk statement. Text in italics indicates when the narration is coupled with a matching on-screen text. On-screen text appears in brackets.

¹The length of the risk statement includes that duration in seconds and the word count (wc).

¹Allergic reactions were not included as a key risk because it did not overlap with the unedited risk statement conditions.