If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Address correspondence to: Jan M. Eberth, M.S.P.H., Ph.D., South Carolina Cancer Prevention and Control Program, Department of Epidemiology and Biostatistics, University of South Carolina, 915 Greene Street, Room 234, Columbia, SC 29208.
South Carolina Cancer Prevention and Control Program, University of South Carolina, Columbia, South CarolinaDepartment of Epidemiology and Biostatistics, University of South Carolina, Columbia, South CarolinaDepartment of Communication, University of Texas at San Antonio, San Antonio, Texas
Department of Epidemiology, Human Genetics and Environmental Sciences, University of Texas School of Public Health, Houston, TexasDan L. Duncan Cancer Center, Baylor College of Medicine, Houston, Texas
This study aimed to determine the temporal association of print media coverage and Internet search activity with adverse events reports associated with the human papillomavirus vaccine Gardasil (HPV4) and the meningitis vaccine Menactra (MNQ) among United States adolescents.
We used moderated linear regression to test the relationships between print media reports in top circulating newspapers, Internet search activity, and reports to the Vaccine Adverse Event Reporting System (VAERS) for HPV4 and MNQ during the first 2.5 years after Food and Drug Administration approval.
Compared with MNQ, HPV4 had more coverage in the print media and Internet search activity, which corresponded with the frequency of VAERS reports. In February 2007, we observed a spike in print media for HPV4. Although media coverage waned, Internet search activity remained stable and predicted the rise in HPV4-associated VAERS reports.
We demonstrate that media coverage and Internet search activity, in particular, may promote increased adverse event reporting. Public health officials who have long recognized the importance of proactive engagement with news media must now consider strategies for meaningful participation in Internet discussions.
The abundance of media coverage and public interest in human papillomavirus vaccination in the months after the Food and Drug Administration's approval of Gardasil was strongly related to the increase in human papillomavirus vaccine–associated adverse events reported to the Vaccine Adverse Event Reporting System.
In June 2006, the United States (U.S.) Food and Drug Administration (FDA) approved Gardasil (Merck & Co., Inc., Whitehouse Station, NJ), a quadrivalent recombinant vaccine that prevents certain strains of human papillomavirus (HPV) that can cause genital warts and are linked with the development of cancers, including cervical cancer [
]. Although Gardasil (HPV4) was lauded as a medical breakthrough, the social and political climate surrounding vaccination was tumultuous because of questionable lobbying practices and conflicts of interest, anecdotal reports of side effects, and public debate about parental rights and adolescent sexual practices.
The HPV vaccine differs from other childhood vaccinations in ways that the public found salient [
]. First, the FDA initially approved the vaccine for use only in females aged 9–26 years, so vaccine recommendations were gender specific, unlike other vaccines on the market. This targeting was purposeful in an effort to reduce the future burden of cervical cancer in women; however, it was also perceived as biased, because males are also able to be infected and transmit the virus. Second, legislation was introduced in several states to mandate HPV vaccine education and/or vaccination in middle school–aged girls [
], a period when the likelihood of HPV exposure is low (and thus, vaccination is most effective for prevention) and other adolescent vaccines are administered. The first such legislation was introduced by Texas Governor Rick Perry in February 2007. Governor Perry issued an executive order [
] that school girls over the age of 11 years be vaccinated against HPV—a mandate that received substantial national news coverage although it was quickly overturned by the Texas legislature after parental outcry and political descent on the issue of parental rights. Similar bills under consideration in most states were tabled, withdrawn, or vetoed soon thereafter. Third, the sexually transmitted nature of HPV garnered a great deal of media attention, differentiating it from diseases that can be transmitted by casual contact in the classroom (e.g., measles, mumps, whooping cough) [
] and distracted viewers from the primary messages about cervical cancer and prevention of sexually transmitted infections being presented by health professionals.
As the discourse unfolded, reports of HPV4-associated adverse events escalated on the Vaccine Adverse Events Reporting System (VAERS) maintained by the U.S. Centers for Disease Control and Prevention and the FDA. Slade et al [
] found that the VAERS reporting rate for HPV4 was triple the rate for all other vaccines combined from June 1, 2006 to December 31, 2008, citing a possible “Weber effect”—in which reporting of adverse events spikes after the licensure of a new product. In his work from the 1980s, Weber [
] described how the adverse event reporting rate can be influenced by factors such as exposure to a new treatment, new dosage forms, and exposure to journal articles or regulatory agency warnings. With the widespread availability of health-related information in the print news and online media, some have suggested that biased reporting can also result from “increased reporting after publicity about a particular known or alleged type of adverse event” [
] found that HPV-associated knowledge levels increased as media coverage increased, and found that knowledge remained high even as media coverage levels dwindled over time. This suggests that once gained, knowledge stays; conversely, perhaps information propagates on other sources (e.g., Internet). Ball et al [
] also mentioned the role of media outlets in molding the public's perceptions of vaccine risk. To our knowledge, however, we know of no studies that have quantified the association of print media with the frequency of adverse event reports to VAERS.
The current study aimed to determine the temporal association of print media and HPV4-associated VAERS reports among adolescents in the U.S. We also examined the mediating effect of Internet search activity. We conducted the same analyses for Menactra (Sanofi Pasteur, Inc., Swiftwater, PA) (MNQ), a meningococcal vaccine licensed for a similar age group just 18 months before HPV4, as a comparison case. We hypothesized that (1) the increased media coverage surrounding HPV4 in the initial period after FDA approval would be associated with an increased number of HPV-associated VAERS reports; and (2) the number of HPV-associated VAERS reports would remain steady despite waning print media coverage over time because of continued public interest in HPV vaccination, as evidenced by Internet search activity.
Data sources and procedures
Multiple data sources were included in our analysis, including the monthly total of adverse event reports from VAERS, the monthly total of print media reports from top-circulating U.S. newspapers, and the monthly frequency of Internet search activity as measured by Google Insights for Search beta. The public availability and de-identified nature of the data received from these sources did not necessitate institutional review.
The Vaccine Adverse Events Reporting System is a post-marketing, passive vaccine safety surveillance system established in the 1990s to collect and analyze data on reported adverse events after the administration of U.S.-licensed vaccines [
]. Anyone can report an adverse event, including health care providers, vaccine manufacturers, and the public. Medical events are encouraged to be reported, even if the reporter is unsure whether there is a causal link between the vaccination and the event, and selected reports will be followed up by VAERS staff. Over 85% of all the reports received are mild and include fever, injection site reactions, and irritability [
]. Publicly available data from 2005 to 2008 were retrieved from the VAERS Web site (http://vaers.hhs.gov/data/data/) in early 2009. We included VAERS reports from June 2006 to December 2008 for HPV4 and from January 2005 to July 2007 for MNQ. (The abbreviations HPV4 and MNQ are the respective terms used by VAERS for Gardasil and Menactra.) The dates we chose correspond to approximately 2.5 years after initial FDA approval of each vaccine. The VAERS data were only extracted for adolescents aged 11–18 years, because this range falls within the Advisory Committee on Immunization Practices' recommended age of vaccination for both vaccines (i.e., HPV4: 11–26 years; and MNQ: 11–18 years) [
]. Vaccine Adverse Events Reporting System reports that had both MNQ and HPV listed in the sample report were excluded (n = 1,160).
We searched LexisNexis, the largest database of public record news articles, for print media reports from top circulating U.S. newspapers (top 20) or top largest U.S. cities (top 10). Five newspapers unavailable through LexisNexis were not included. A total of 16 newspapers were analyzed (Table 1).
The Wall Street Journal, Chicago Tribune, Post/Rocky Mountain News, Boston Globe, San Diego Union-Tribune, and Arizona Republic were not index in LexisNexis and could not be queried, despite being in the top 20 circulating newspapers or top 10 in population size.
used to extract print media reports
Newspaper circulation rank
Rank by population size (top 10 only)
Largest reported circulation as of March 31, 2006
USA Today (Arlington, VA)
New York Times (New York, NY)
Los Angeles Times (Los Angeles, CA)
Washington Post (Washington, DC)
Daily News (New York, NY)
Inquirer (Philadelphia, PA)
Chronicle (Houston, TX)
New York Post (New York, NY)
News/Free Press (Detroit, MI)
Morning News (Dallas, TX)
Star Tribune (Minneapolis, MN)
Star-Ledger (Newark, NJ)
Journal-Constitution (Atlanta, GA)
Newsday (Long Island, NY)
Chronicle (San Francisco, CA)
Express-News (San Antonio, TX)
a The Wall Street Journal, Chicago Tribune, Post/Rocky Mountain News, Boston Globe, San Diego Union-Tribune, and Arizona Republic were not index in LexisNexis and could not be queried, despite being in the top 20 circulating newspapers or top 10 in population size.
We limited our document search to the time periods June 2006 to December 2008 for print media reports related to HPV4 and January 2005 to July 2007 for those related to MNQ, so that they would also correspond to the 2.5 years after initial FDA approval of each vaccine. The following search terms were used to identify print media reports related to HVP4: “HPV” AND “vaccine OR vaccination OR immunization”; “Human papillomavirus” AND “vaccine OR vaccination OR immunization”; and “Gardasil.” For those related to MNQ, we used the following search terms: “Menactra”; “Meningococcal” AND “vaccine OR vaccination OR immunization”; and “Meningitis” AND “vaccine OR vaccination OR immunization.” We did not specifically search for the terms “HPV4” or “MNQ” because these are abbreviations used by VAERS, not the media or public. We excluded reports related solely to Merck or Sanofi Pasteur's financial status (e.g., profits) and duplicate reports from the monthly count. We did not exclude reports that fit the search criteria but were only tangentially related to HPV4 or MNQ (e.g., articles mentioning Cervarix or Prevnar). However, we noted few reports that met this criterion (e.g., Cervarix and Prevnar were approved outside our study time frame).
Monthly Internet search activity was measured using Google Insights for Search (now Google Trends), a Web tool published by Google, Inc., that shows the popularity of specific search terms entered into Google Search. The tool analyzes a sample of Google Web searches to calculate how many searches have been conducted on the topic of interest relative to the total number of searches conducted. Repeated queries on a particular topic from a single user over a short time span are not included, to minimize potential bias. Google Insights data are graphically displayed over time and space. Specific news headlines on that particular topic can also be plotted, to give the viewer a better sense of what may be driving the popularity of search terms over time. We searched for the terms “Gardasil” and “Menactra” and extracted the monthly total search popularity (normalized across geographic regions and scaled by Google from 0 to 100) for analysis. More details on the methods used to derive the data are available online (http://support.google.com/trends; Site Help Section).
Descriptive statistics were used to summarize news media reports, VAERS reports, and Internet search activity. t-tests measured mean differences between key variables by vaccine type. A moderated regression model was used to assess whether the spike in print media reports observed in Month 9 after FDA approval of HPV4 (which coincided temporally with Governor Perry's HPV4 mandate) was significantly different from the preceding and subsequent months [
]. This model also controlled for continuous time, to ensure a clean estimate of our categorical construction of time (Months 1–8, Month 9, and Months 10–31). We ran two additional moderated regression models to examine the differences in VAERS reports between HPV4 and MNQ over time and by Internet search activity. In these models, we also controlled for continuous time and the number of monthly print media reports. Significant interaction effects were deconstructed to identify the specific differences between HPV4 and MNQ over time and by Internet search activity. Mediations tests were conducted using Sobel statistics [
]. JMP 9.0 for Mac (SAS Statistical Software, Cary, NC) was used to conduct all of these described statistical tests.
Summary of Vaccine Adverse Event Reporting System and print media reports
In the initial 2.5 years post-FDA licensure, 1,220 MNQ- and 5,040 HPV4-associated adverse events reported to VAERS met our inclusion and exclusion criteria. The monthly mean number of HPV4-associated VAERS reports was 159.19 (standard deviation [SD], 92.55), which was significantly greater (t = 6.96; p < .001) than MNQ-associated VAERS reports (mean, 18.77, SD, 28.34). In the 2.5 years after FDA licensure, we identified 582 and 249 unique news reports published in top circulating U.S. newspapers related to the HPV4 and MNQ vaccines, respectively. The mean monthly number of print media reports was significantly greater for HPV4 (mean, 18.77; SD, 28.3) compared with MNQ (mean, 8.03; SD, 3.26), (t = 2.09; p = .04). This t-test was inclusive of Month 9, when print media coverage spiked (147 reports in Month 9; z-score, 4.52).
Print media reports varied across time and by vaccine type
The results prompted us to test whether this specific month was individually different from the 8 preceding and 22 subsequent months. To do this, we ran a moderated regression model (Figure 1) in which the number of print media reports was regressed on the interaction between vaccine type and spike level. The model was significant (F[6, 55] = 28.50; p < .0001; R2 = .76) and the interaction term improved the fit of the model (F[2, 55] = 36.07; p < .0001; ΔR2 =.32). A stratified analysis revealed no significant differences in print media reports over time periods for MNQ, whereas HPV4 had a significant, positive slope before the spike (t = 6.96; p < .0001; ß = .58) and a significant, negative slope afterward (t = −7.59; p < .0001; ß = −.92) (Table 2). This confirmed that a substantial change in print media reports occurred in Month 9 for HPV4 but not MNQ.
Table 2Moderated regression
p > |t|
Model 1: Print media reports by interaction of vaccine type and spike level (Months 1–8 and 10–31 vs. Month 9)
Spike in print media drives Vaccine Adverse Event Reporting System reporting for Gardasil
We next posited that this print media spike in Month 9 would act as a catalyst to increase monthly VAERS reports for HPV4 but not for MNQ. The catalytic event was coded as a dichotomous variable (Months 1–8 vs. Months 9–31) and a new interaction term was created from the product of vaccine type and the catalytic event. The model was significant (F[5, 56] = 68.62; p < .0001; R2 = .86) and the interaction term added significantly to the fit of the model (F[1, 56] = 62.58; p < .0001, ΔR2 = .19). In Months 1–8, the number of VAERS reports per month was not significantly different between HPV4 and MNQ. However, in Months 9–31, HPV4 was associated with more VAERS reports than MNQ (t = 7.91; p < .0001; ß = .45) (Table 1). This significant interaction supports our hypothesis that the spike in print media in Month 9 may have been the catalyst needed to increase HPV4-associated VAERS reporting over the following months (Figure 2).
Internet search activity varied across time and by vaccine type
Although this catalytic event was associated with an increase in the number of VAERS reports, it did not necessarily explain why the number of VAERS reports increased even in the post-spike period when the number of print media reports decreased to levels matching MNQ (Figure 1). We posited that the relationship between the catalytic event and increase in VAERS reports was actually mediated by a third variable: Internet search activity. Before testing this hypothesis, we examined the correlation between print media reports and Internet search activity, and evaluated whether there was a significant rise in Internet search activity in Month 9 for HPV4 but not MNQ. Gardasil generated significantly more average activity (mean, 62.29; SD, 18.38) than MNQ (mean, 7. 35; SD, 3.29) (t = 16.38; p < .0001). In a similar fashion as print media reports, Internet search activity increased in Months 9–31 for HPV4 (ß = .23) but remained stable for MNQ (F = 22.67; p < .0001) (Figure 1).
Internet search activity as a mediator between print media and Vaccine Adverse Event Reporting System reporting
Considering the similar pattern, we examined the effect of Internet search activity on VAERS reporting and whether Internet search activity was a mediating step between the catalytic event and the increase in VAERS reports (Figure 3). We included an interaction term of Internet search activity by vaccine type to evaluate whether it accounted for some or all of the variance explained by the interaction of the catalytic event and vaccine type. Running a moderated regression model controlling for continuous time and monthly print media reports, and including these two interaction terms, our model was statistically significant (F[8, 53] = 72.95; p < .0001; R2 = .91). Not only was the new interaction term of Internet search activity by vaccine type significant (t = 1.98; ß = .24; F[1, 53] = 3.99; p = .05, ΔR2 = .01), it pushed the catalytic event by vaccine type interaction out of statistical significance (from p < .0001 to p = .21). Moreover, the Sobel test concluded that the catalytic event indeed predicted the increase in Internet search activity, which then predicted the increase in VAERS reporting for HPV4 (ß = .61; p < .0001) (Sobel statistic = 2.64; two-tailed p = .008). To ensure this was a unique phenomenon, we tested the same mediation pattern (i.e., catalytic event → Internet search activity → VAERS reporting) on MNQ, but it did not reach statistical significance (Sobel statistic = 1.31; two-tailed p = .19).
As hypothesized, there was a temporal association between the frequency of media reports and the frequency of VAERS reports for HPV4, whereas none was found for MNQ. During the initial 2.5 years after FDA licensure, the frequency of media reports related to MNQ remained stable and low. In contrast, for the HPV4 vaccine, there was an intense spike in media coverage during Month 9 after FDA licensure, followed by a period of decline in which the frequency of HPV4-related media reports remained similar to MNQ. Starting just before Month 9 and continuing forward, there was an increase in the number of HPV4-associated adverse event reports to VAERS. This increase coincided initially with the increase in print media coverage and, subsequently, Internet search activity. Moreover, our research indicates that adverse events reports remained high despite waning print media coverage as interest continued to flourish on the Internet.
The HPV vaccine received a great deal of mainstream media attention in the months after its licensure. However, it does not appear that this attention focused solely on the potential side effects of vaccination. Previous content analyses have indicated that whereas the news media may have provided some information about the efficacy of the vaccine and the link between HPV and cervical cancer [
] found that mentions of the sexual nature of transmission and the link between HPV and cervical cancer were more frequent in news reports after February 2007, when Governor Rick Perry enacted his mandate. Therefore, although the news reports could have been more informative, they did not seem to emphasize the risk of adverse events. Although we did not assess the content of the print media reports in this study, previous literature does not seem to indicate that reports were unbalanced or overwhelmingly positive or negative. This gives us further evidence to believe that the sheer volume of interest in HPV vaccination by the media in the months after FDA approval (particularly in February 2007), rather than the content of the messages themselves, was related to VAERS reporting patterns.
News media activity can set the agenda for public discussion regarding a topic, and today much of that discussion takes place on the Internet. In essence, the Internet acts as a cheap, accessible forum for users to relay and debate information from print media sources. Although most HPV vaccine–related news articles on the Internet are also neutral [
], anti-vaccination activists and parents who are worried about the potential side effects of vaccination can find outlets for their messages through Web-based applications that allow users to share comments or content with others (e.g., Facebook, Twitter, Wikipedia, YouTube, and LinkedIn) [
]. These applications, collectively known as Web 2.0, aid in the continuity of discussion among peers with varying degrees of (accurate and inaccurate) information on vaccination, and may ultimately influence vaccination decisions. In addition, there are a host of anti-vaccination Web sites that include claims of vaccines causing illness or adverse events [
], 95% of anti-vaccination Web sites also claimed that adverse events were underreported because of physicians' delay or failure to report them. This type of information may spur audiences to report to VAERS any adverse events or illness proximal to vaccination, whether or not it has a possible vaccine connection [
Our results do not necessarily suggests that media coverage spurs people to report inaccurate or false adverse events, but rather that media coverage may increase awareness of the VAERS program, the potential to experience side effects, and/or the importance of reporting possible side effects. Media consumers may be more likely to notice side effects because they are told to look for them and are advised to report concerns. However, the rise in adverse event reporting was observed only for HPV4, not MNQ, which suggests that the increase in reporting likely resulted from the spike in HPV4-related media coverage rather than increasing vaccine uptake over time. Moreover, this observed Weber effect seems reflective of increased publicity rather than patient exposure, dosage forms, or changes to the manufacturing process. Finally, we cannot rule out that the difference in frequency of VAERS reports between MNQ and HPV4 was not a real consequence of differing rates of side effects experienced by those vaccinated. However, based on their respective package inserts, HPV4 and MNQ both report low rates of serious adverse events (.8% vs. 1.0%, respectively). Differences were noted for several mild side effects such as malaise (HPV4, 1.4%; MNQ, 21.9%), injection site pain (HPV4, 83.9%; MNQ, 59.2%), swelling (HPV4, 25.4%; MNQ, 10.8%), and joint pain (HPV4, 1.2%; MNQ, 17.4%); however, they were not always elevated in the same vaccine [
]. Moreover, although we cannot prove whether the rates of actual side effects differed between vaccines, we do not believe that the higher rates reported for HPV4 are associated with the fact that HPV4 requires more doses (3 vs. 1). During the time our data were collected, HPV4 was recommended only for females and many women had not yet completed the three-dose series, whereas MNQ was recommended for both males and females and had a slightly higher uptake rate [
]. Our study demonstrated a correlation between the number of HPV-related print media reports, Internet search activity, and VAERS reporting. Although we did not adjust for the content of the news reports and/or Internet search activity, previous studies have shown a mixture of positive and negative messages regarding vaccination in both data sources. It may be that any increase in print media or Internet search activity, regardless of the content of its messages, will increase VAERS reporting, but it may also be the case that VAERS reporting is related to how the issue was framed in the media or on the Internet. Moreover, future research should focus on how the frequency of negatively framed messages affects VAERS reporting. In addition, our findings suggest the need for public health education efforts on the Internet regarding potential vaccine side effects and appropriate use of the VAERS system, where the discourse may be muddled by anti-vaccination messages and general misinformation provided by the public.
There were no sources of funding, either direct or indirect, for this study.
Long term protection against cervical infection with the human papillomavirus: Review of currently available vaccines.