What Do Women in the U.S. Know about Human Papillomavirus and Cervical Cancer?

Cervical cancer screening with the Pap test is well integrated into the U.S. health care system and is widely accepted by women. However, advances in our understanding of cervical cancer etiology, specifically, the identification of human papillomavirus (HPV) as a necessary cause (1), have led to the development, evaluation, and recommendation of HPV-based technologies for cervical cancer prevention and control. Current clinical guidelines recommend using the HPV DNA test in women ages 30 and older for primary screening (2, 3) and in all ages for follow-up to abnormal cytology (4). More recent recommendations include administering the HPV vaccine in adolescent and young adult females (ages 9-26) to prevent the transmission of certain carcinogenic HPV types (5).

The changing landscape of technologies to prevent and detect cervical cancer offers new options to women and health care providers. Accurate knowledge about HPV and its link to cervical cancer are therefore critical in order to make appropriate, evidence-based health care choices. Limited available data suggest that women know little about HPV (6). As HPV-based technologies diffuse into the general population, tracking which groups of women remain unfamiliar with HPV and would, therefore, benefit from educational messages will be essential. Health communication researchers have an important window of opportunity to ensure that misinformation does not occur and women understand the state-of-the-science. Although knowledge is not a direct predictor of health behavior (7), health behavior theories hypothesize that it is a distal factor mediated by attitudes, risk perceptions, social influence, and self efficacy (8). Thus, acquiring knowledge is a key first step to the success of any health education intervention.

Most research in the U.S. on knowledge of HPV and its link to cervical cancer have been conducted in special populations such as adolescent and young adults (9-17) or individuals attending sexually transmitted infection (STI) and primary care clinics (18-23). These studies suggest that HPV awareness is low. Information about recruitment methods for the sole national survey of women living in the U.S. is limited; therefore, its generalizability is unclear; ref. 32. Thus, little is known about HPV awareness among adult women and those at average risk for developing cervical cancer (6). Furthermore, few studies have examined factors associated with HPV knowledge (23, 25, 26). Assessment of the public's knowledge of HPV in nationally representative samples was identified as a high-priority research aim in 1999 by the Centers for Disease Control and Prevention, but has yet to be achieved (6).

Disparities in cervical cancer screening persist in the U.S. (27). To prevent further disparities in cervical cancer outcomes, identifying women who are most likely to benefit from educational messages about the dangers of persistent HPV infections is necessary. Clear and consistent information will help women make decisions about the usefulness of new technologies like the HPV DNA test and the HPV vaccine.

The purposes of this report are to (a) assess awareness of HPV and knowledge of its link to cervical cancer in a national sample of U.S. women, and (b) to explore the association of accurate HPV knowledge with factors relevant to health communication. Results should aid in developing effective communication about HPV and identify opportunities to facilitate appropriate adoption of state-of-the-science technologies to reduce the burden of cervical cancer.

Data were from the National Cancer Institute's 2005 Health Information National Trends Survey (HINTS) conducted between February and August 2005. HINTS collects nationally representative data on the U.S. population's knowledge of, methods of searching for, and use of cancer-related information as well as performance of behaviors that prevent cancer or detect it early. A random-digit-dial sample of all working residential telephones in the U.S. was used to recruit participants. Telephone exchanges were geographically stratified and those with higher percentages of African-Americans and Hispanics were oversampled. During the screening process, one adult aged 18 years or older from each household was randomly selected for the extended interview. Interviews were conducted in English or Spanish with 5,394 adults. The overall response rate of 20.8% and was calculated by combining the American Association for Public Opinion Research's standard definitions (28) for screening interview response rate (34%; i.e., the proportion of households in which eligibility was ascertained) and the extended interview rate (61.3%; i.e., the proportion of eligible households in which an adult completed the full survey; ref. 29). More information about the HINTS 2005 data collection procedures can be found at the National Cancer Institute web site (http://hints.cancer.gov/instrument.jsp). Only female respondents were asked questions about HPV; thus, male respondents were excluded from all analyses.

The HINTS 2005 included items that assessed sociodemographics, health status, personal and familial cancer history, general and specific-cancer knowledge, attitudes, beliefs, health communication, and cancer screening behaviors.

Outcome variables. Female respondents were asked if they had heard of human papillomavirus or HPV. If a woman responded “no,” she was not asked HPV knowledge questions and was not included in the analysis of the following dependent variables. Knowledge of the link between HPV and cervical cancer was assessed with a single item (“Do you think HPV causes cervical cancer?”). Knowledge of other aspects about HPV was assessed with five questions, specifically, whether HPV: is a STI, will often go away without treatment, can cause abnormal Pap tests, is rare, and affects a woman's ability to get pregnant. The latter two items were distracters and were reverse coded. A summary scale of the five items was created in which higher scores indicated more accurate knowledge.

Correlate variables. Continuous variables were grouped into ordinal categories to facilitate inclusion in the multivariate logistic regression analysis. Sociodemographic variables examined included age (18-29, 30-64, 65-75), race/ethnicity (non–Hispanic White, Hispanic, Black, other), education (less than high school, high school degree, some college, college degree), income (<$⁠

Access to health care was measured with three variables: insurance status (yes, no), visit health care provider in past year (yes, no), and comfortable speaking English (a little/not at all/does not speak versus somewhat/very/completely/English is native language).

Attention to and trust in health information was assessed with two composite variables. Respondents were asked whether they paid attention to information about health issues from three sources (newspapers/magazines, television news programs, and the Internet). Then, they were asked—on a four-point Likert scale—how often they paid attention. For print and television media, responses were dichotomized as once or more per week and less than once per week. For the Internet, responses were coded as 1+ or 0 per month. A summary scale was created (range, 0-3) with higher scores indicating more exposure.

Participants were also asked how much they would trust information about health issues from seven sources (doctor, family or friends, newspapers, magazines, radio, the Internet, and television). Based on a coding scheme developed to examine the association between this construct and mammography using HINTS 2003 data (30), we created a summary index score (range, 0-7) indicating the number of sources women “did not trust at all.” Then, a dichotomous variable was created with 0 representing trusting all sources “a lot, some, or a little”, and 1 reflecting that the woman said she trusted one or more sources “not at all.” Trust was coded in the negative direction because we hypothesized that women who expressed no trust would be less likely to have heard of or have knowledge about HPV.

Cancer history was assessed with three dichotomous variables—cervical cancer family history, other cancer family history, and personal history of cancer other than cervical cancer. Women reporting a personal history of cervical cancer were excluded from the analyses.

Cervical cancer screening history was assessed with two questions about the timing of the most recent Pap test and the one before the most recent. Response options were a year ago or less, more than 1 but not more than 3 years ago, more than 3 years ago but not more than 5 years ago, and over 5 years ago. To determine a respondent's screening pattern, a three-level, composite variable was created: 1, never had a Pap test or not one in the past 3 years; 2, had a Pap test in the past 3 years and not one in the year before; or 3, on a regular schedule (had two Pap tests 1 year apart with the most recent in the past 3 years). The following factors were also included: ever had a hysterectomy, ever had an HPV infection, whether the main reason for the most recent Pap test was because of previous abnormal findings, and whether a woman had heard about the change in clinical guidelines for Pap test frequency from once a year to every 3 years (2, 31).

Women reporting a history of cervical cancer (n = 76) and women older than age 75 (n = 473) were excluded from all analyses because cervical cancer screening recommendations differ for them (2, 31). A total of 3,108 women ages 18 to 75 years old with no history of cervical cancer completed the interviews. Women missing values for either of the dependent variables were also excluded from analyses (n = 32). The final sample size was 3,076 respondents. Women selecting “refused” or “don't know” for a survey item measuring a correlate variable were counted as missing for that variable and were excluded from relevant univariate and all multivariate analyses.

For univariate analyses, we used χ² statistics to assess associations. In the multivariate logistic regression analyses, a significance level of P ≤ 0.25 from the univariate analyses was used as the cutoff for selecting correlate variables (32). To determine if collinearity among the correlate variables was affecting the fit of the multivariate model, we checked for large regression coefficients and SEs (32). Because of the sampling design of HINTS 2005, SUDAAN 9.0.1 was used to calculate weighted population estimates and confidence intervals in all analyses.

Knowledge about HPV among U.S. women ages 18 to 75 years old was relatively low; 40% of women (n = 1,248) reported that they had ever heard of HPV (Table 1). Among those who had heard of it, less than half knew that HPV causes cervical cancer. Higher percentages of women knew that HPV is a STI, that it causes abnormal Pap tests, and that it is not rare (64%, 79%, and 70%, respectively). Very few knew that HPV often resolves without treatment (4%) and that it does not affect chances of getting pregnant (13%). Less than 2% correctly answered all five items of the HPV summary knowledge scale; slightly more than 50% answered three or more items items correctly.

Weighted frequencies of sociodemographic, access to health care, health information, cancer history, and cervical cancer screening history for all respondents (n = 3,076), those who had heard of HPV (n = 1,248), and those who knew that HPV caused cervical cancer (n = 605) are provided in Table 2. For all respondents, the majority were White, had some college or a college degree, were married, had a household income >$50,000, had health insurance, visited with a health care provider in the past year, and felt comfortable speaking English. Approximately 19% of the women had had a hysterectomy, 76% got regular Pap tests, and ∼21% were aware of the recent change in cervical cancer screening guidelines to less frequent Pap tests (2, 31).

Women who had heard of HPV were more likely to: be younger than 65 years old, be non–Hispanic White, have some college or a college degree, have a higher income, visit a health care provider in the past year, be exposed to two or more sources of health information, engage in a regular Pap test schedule, and be aware of the change in screening guidelines than those who had not heard of HPV (Table 2). Women who had heard of HPV were also less likely to lack trust in one or more sources for health information and to have had a hysterectomy. All of the women who reported testing positive for HPV also reported that they had heard of it. Because of lack of variability, HPV-positive status was not included in the multivariate model for “heard of HPV.” Three factors (being married, familial cervical cancer history, and no personal cancer history) were marginally associated (P < 0.25) with having heard about HPV.

Among women who had heard of HPV, those who knew about the causal link between HPV and cervical cancer were more likely to: be Hispanic, have some college or a college degree, report being HPV-positive, and have had abnormal Pap results. Five other factors (younger age, very low or high income, having health insurance, familial non–cervical cancer history, and comfortable speaking English) were marginally associated.

Two socioeconomic status indicators were measured in HINTS—education and household income. We decided to include only education in the multivariate analyses because (a) educational attainment is a robust measure of socioeconomic status that is relatively stable over the adult life span regardless of health status, and (b) income was not reported by many respondents, thus, the use of income as a covariate would have resulted in a substantial loss of participants.

The odds ratios and the 95% confidence intervals (CI) from the multivariate models are shown in Table 3. In the multivariate model for HPV awareness (i.e., heard of HPV), the following factors remained positively associated: younger age, being non–Hispanic White, higher educational attainment, exposure to more than one source of health information, getting regular Pap tests, and being aware of the change in screening guidelines. Lack of trust in health information from all sources remained negatively associated. In contrast, the only factors positively associated with accurate knowledge of the HPV-cervical cancer link were abnormal Pap and positive HPV test results.

Awareness of HPV has increased over the past decade, but knowledge of its link to cervical cancer remains low. The earliest study of HPV knowledge (1992) that we identified found that 13% of women attending a southeastern university had ever heard of HPV and only 8% of them knew that it was associated with cervical cancer (15). In 2000, a population-based survey of women ages 18 to 65 living in the U.S. found that only 28% had ever heard of HPV and only 41% of those knew that it was sometimes associated with cervical cancer (24). Our study, in a national sample of women ages 18 to 75 years old, found that ∼40% of women had heard of HPV, but less than half of those also knew that it caused cervical cancer. Additionally, only 50% of women who had heard of HPV could correctly answer at least three of the five items on a summary scale measuring knowledge of other aspects of HPV. These low levels of knowledge are consistent with findings from other studies of U.S. adult women (10-14, 23, 24), and are not surprising given that the HPV DNA test was only recommended for primary cervical cancer screening in 2002 to 2003 (2, 3). Although HPV is the most common STI in the U.S. (33), most infections are transient (median duration, 8 months; ref. 34); thus, most women probably do not receive treatment and therefore miss an opportunity to learn about the infection from a health care provider.

Caution should be exercised when comparing our results to past HPV knowledge studies. Several of these earlier studies did not assess whether the study population had ever heard of HPV before assessing knowledge (10, 14, 16, 19, 20, 35, 36). Others did not make clear whether they used a skip pattern after asking respondents about whether they had heard of HPV (18, 37), whereas some studies continued to ask questions about HPV knowledge even after respondents stated they had never heard of HPV (11, 12, 15, 25, 38). Question wording, sequence, and skip patterns have been shown to affect prevalence estimates (39, 40). Additionally, several studies did not describe whether they used a prompted (recognition) or an open-ended (recall) question format. This methodologic difference could explain why a 2004 study of a representative population sample of British women, which used an open-ended format, found extremely low levels of HPV knowledge (e.g., only 1% knew that HPV caused cervical cancer; ref. 35). Respondents were more likely to guess with the prompted format, which tends to yield higher scores of cancer knowledge (8, 41). This tendency suggests that our findings, which were based on a prompted question format, may overestimate U.S. women's true levels of knowledge about HPV and cervical cancer.

Although some studies have looked at whether sociodemographic and cervical cancer screening variables were associated with HPV knowledge (23, 25, 26), ours was the first to use a national sample of U.S. women and the first to include health communication and cancer history variables. Our main finding was that factors associated with HPV awareness differed substantially from those associated with HPV-cervical cancer knowledge. Only one variable was consistently associated with both outcomes—HPV-positive status. All of the women who reported testing positive for HPV had heard of it and they were also 3.5 times more likely to know that HPV causes cervical cancer. This finding is consistent with past research (25, 26). Although HINTS 2005 data are cross-sectional and prevent us from establishing a temporal sequence, our findings may suggest that education is more likely to occur after a woman has experienced an adverse consequence from an HPV infection. Also supportive of this hypothesis are the higher levels of general HPV knowledge found among women attending colposcopy clinics (19, 22). However, not all women who tested positive knew about the HPV-cervical cancer link, which suggests that some women do not remember the purpose of an HPV test. Another explanation is that they are confusing HPV with another STI diagnosis like herpes simplex virus. A longitudinal analysis would clarify whether HPV status predicts accurate knowledge.

Having had an abnormal Pap result was associated with knowing that HPV causes cervical cancer, but was not associated with having heard about it. This distinction further supports the idea that women gain knowledge when they experience cervical abnormalities or HPV positivity. In the HINTS data, we do not know where women acquired information about HPV. Women who tested positive could have received information from their health care providers or actively sought out information about HPV from other sources. Future research should investigate cognitive processing and information-seeking behaviors among women receiving abnormal Pap or positive HPV test results.

Our findings also suggest that familiarity with HPV does not guarantee accurate knowledge about its link to cervical cancer. Thus, health communication researchers have two tasks when crafting messages: (a) to increase recognition of the name, human papillomavirus, and its acronym, and (b) to increase the depth of knowledge of the potential consequences of HPV infection. Several sociodemographic subgroups (older women, less educated) seem to be largely unaware of HPV and would benefit from educational messages. Increasing the depth of knowledge among women who have heard about HPV is a more difficult task, but is crucial to prevent misinformation among women who have not received positive test results.

Interestingly, women who did not trust one or more sources of health information were less likely to report having heard about HPV. This finding may suggest that individuals who lack trust in the typical sources (e.g., doctor, media) do not attend to or recall new health information. Future research should investigate whether exposure to health information and trust in sources interact to affect awareness of HPV and receptivity to health education interventions.

Our study had some limitations. The overall response rate of the HINTS 2005 was significantly lower than HINTS 2003 and is mainly attributable to the screener response rate (34% and 55%, respectively); however, the extended interview response rate held steady at ∼60%. The strong decline in response rates to random-digit-dial surveys seems to be continuing (42) and is probably exacerbated by technologies that allow individuals to screen and block incoming calls (e.g., Privacy Manager; ref. 43) and by adults who use only cell phones (6.7% of households in early 2005; ref. 44). Recent methodologic investigations suggest that lower response rates do not necessarily indicate larger biases in survey estimates (42, 45). HINTS 2005 tried to reduce the potential for screener nonresponse bias through weighting adjustments of the following characteristics: time of the telephone numbers' release into the field, whether an invitation letter could be mailed to the household, percentage of college graduates estimated for the telephone exchange, and percentage of minorities estimated for the telephone exchange. Regardless of the weighting adjustment, our data may not be generalizable to all women living in the U.S. In addition, researchers should exercise caution when comparing our results to studies in other countries. Another possible limitation mentioned earlier is that the prompted question format probably encouraged guessing; however, nothing in the literature suggests that guessing is differential across sociodemographic groups. Therefore, we expect that bias was towards the null and probably underestimated the magnitude of the patterns of association. Finally, items for the HINTS survey were not developed with a particular theoretical framework in mind. The survey measured a limited set of psychosocial correlates, thus limiting our ability to examine whether theoretical models like the Cognitive-Social Health Information Processing Model (46) or Leventhal's Common-Sense Model (47) explain how HPV knowledge influences cervical cancer prevention behavior.

Individuals are constantly presented with new health care research that updates previous knowledge, conflicts with prior knowledge, or provides entirely new options for diagnosis and treatment (48). Media coverage about and direct-to-consumer marketing for the HPV vaccine should raise women's awareness of HPV's link to cervical cancer. However, Anhang et al.'s 2004 study showed that news media coverage of HPV from 1995 to 2002 has been incomplete and sometimes misleading (49). Another study found that patients searching for information about vaccines on the Internet are likely to encounter sophisticated antivaccination web sites (50). Thus, key communication challenges are to prevent misinformation, minimize confusion, and track the diffusion of knowledge. Because HPV is both a carcinogenic virus and a STI, HPV-based technologies may suffer from stigma which could adversely affect women's use of them and the public's acceptance of HPV vaccination programs (26).

A first step in addressing these communication challenges is to identify women least likely to have accurate HPV knowledge and to develop clear and appropriate messages for them. Our data indicate that awareness of HPV and cervical cancer among U.S. women is low, especially among those who are older, less educated, and less exposed to health information. If HPV testing and vaccination are the future for cervical cancer control, then accurate understanding of persistent HPV infection and its relationship to cancer is crucial. If targeted to and accessible by the right audiences, these efforts may reduce current health disparities in mortality and make possible the eradication of cervical cancer among U.S. women.

The authors appreciate the assistance of Bradford Hesse, Richard Moser, and Lila Finney Rutten.