All product-specific prescription drug advertisements appearing in these magazines from January 1989 through December 1998 were photocopied. We excluded disease education messages that did not mention a specific drug and advertisements for medical devices and veterinary drugs. Advertisements for the same brand often differed in nonsubstantive ways and were coded as a single case.*
Two judges independently coded each advertisement. To maximize reliability, almost all codes involved a determination of the presence or absence of a word or phrase in the advertisement. The mean kappa across all classifications reported here was .93 (range=0.90-1.0).
Classifications
Medical conditions. We classified each drug brand into 1 of 14 medical condition categories Table 1. In most cases, this was a straightforward process; however, when a drug was promoted for more than one indication it was classified according to the indication for which it was most extensively advertised.
Inducements. The advertisements were coded for the presence or absence of an offer of each of the following inducements: patient support services (eg, assistance in helping the reader find a local support group for smoking cessation), additional information about the drug or condition in print or audiotape/video form, and monetary incentives (rebate, discount coupon, money-back guarantee, or free product trial).
Advertising appeals. Each advertisement wascoded for the presence or absence of a number of inductively derived claims about the drug’s effectiveness, social-psychological benefits, ease of use, and safety. We looked at whether each of 42 adjectives, adjectival phrases, or adverbs was used to describe the drug’s nature or impact. We initially coded for the presence or absence of each of these terms or phrases and then collapsed across related claims to create more general product attribute variables as shown in Table 2. For instance, to create the attribute category “Innovative,” we coded for the terms “advancement,” “breakthrough,” “a first,” the “only” drug of kind, “innovative,” “novel,” and “new” and then recoded each advertisement for whether at least one of these descriptors was used to depict the drug.
Statistical Analyses
We employed 2 units of analysis. In most instances, the unit of analysis was the advertisement, with descriptive statistics serving as the primary analytic tool. Use of the advertisement as the unit of analysis, however, is inappropriate when inferential tests such as the chi square are used. Doing so would violate the assumption of independence of observations, because most advertisements for a brand represent modifications of previously placed advertisements. We thus relied on the brand as the unit of analysis when using inferential statistical procedures. Specifically, each brand was treated as a single case by giving each advertisement for the brand a weight of 1/n, when n represents the number of advertisements for that brand. For instance, if 5 advertisements were found for a particular brand, each advertisement was given a weight of 0.2 in these analyses. This approach allows for fractional values on the coded variables. For example, if half the advertisements for a brand provided a monetary incentive, the value for the brand on that variable would be 0.5. Thus, each brand contributed equally to the analyses regardless of the extent to which it was advertised. This is appropriate, because the objective of content analysis is to describe message content, not the effects of that content. In 2 analyses, trends in the frequency with which advertisements and brands were introduced during the decade were examined using the Statistical Package for Social Sciences curve fit procedure.20
Results
After aggregating essentially identical advertisements, 320 remained for analysis, covering 101 brands Table 1. Sixty-eight percent of the brands were oral medications; 17% were topicals or transdermals; 7% required injections; 5% were inhalants; 2% were implants; and one brand (less than 1%) was a suppository.
Trends in DTC Advertising
The first set of research questions pertained to trends in DTC advertising. As shown in Table 1, the most common brands advertised were for dermatologic conditions, HIV/AIDS, cardiovascular disease, and obstetrical/gynecological conditions. Treatments for allergies, gastrointestinal conditions, musculoskeletal ailments, psychiatric/neurologic disorders, and urological conditions were also well represented. Brands for cancer, diabetes, infectious/non-HIV diseases, respiratory conditions, and tobacco addiction were less common.
We examined trends in the breadth of DTC advertising in 2 ways. First, to examine changes in the introduction of new brands, each of the 101 brands was classified under the calendar year in which it first appeared in the sampled publications. These data are reported in the first data series in the Figure 1, which shows a tendency for new brand introductions to increase during the 10-year period. Although brand introductions leveled off in 1996, the overall trend is best modeled as a linear relationship (adjusted R2=.80; B=2.03; P=.0003). Second, we classified each of the 320 advertisements under the calendar year in which it first appeared. These data are displayed in the second data series in the Figure 1. The number of new advertisement introductions grew from only 3 in 1989 to 76 in 1998. The best-fitting trend line for this increase is provided by a linear model (adjusted R2=.91; B=7.45; P <.0001).