Does Vaccination Protect against Human Papillomavirus-Related Cancers? Preliminary Findings from the United States National Health and Nutrition Examination Survey (2011–2018)

Most oropharyngeal and anogenital cancers are caused by human papillomavirus (HPV). Although HPV vaccines showed high efficacy against oropharyngeal and anogenital HPV infections, and cancer precursors in randomized clinical trials, there are limited data on the effectiveness of HPV vaccination against HPV-related cancers. We aimed to evaluate the association of HPV vaccination with HPV-related cancers among a nationally representative sample of United States adults, aged 20–59 years. In a cross-sectional study combining four cycles from the National Health and Nutrition Examination Survey, from 2011 through 2018, we used a survey-weighted logistic regression model, propensity score matching and multiple imputations by chained equations to explore the association of HPV vaccination with HPV-related cancers. Among 9891 participants, we did not find an association of HPV vaccination with HPV-related cancers (adjusted OR = 0.58, 95% CI 0.19; 1.75). Despite no statistically significant association between HPV vaccination and HPV-related cancers, our study findings suggest that HPV-vaccinated adults might have lower odds of developing HPV-related cancers than those who were not vaccinated. Given the importance of determining the impact of vaccination on HPV-related cancers, there is a need to conduct future research by linking cancer registry data with vaccination records, to obtain more robust results.

There are highly efficacious HPV vaccines against the high-risk HPV infection types, recommended at the age of 9-14 for girls and boys [11]. The Advisory Committee on

Study Variables
Two survey questions were used to develop the binary outcome variable: reported diagnosis of HPV-related cancers. The first question was: "Have you eve told by a doctor or other health professional that you had cancer or a malignancy kind?" [29]. The second question, asked to those who responded "yes" to th question, was: "What kind was it?". Participants who reported being diagnose larynx, windpipe, mouth, tongue, lip or cervical cancers (anal, penile, vulvar, and v cancers were not self-reported) were categorized as HPV-related cancers, while tho were not diagnosed with cancer or were diagnosed with other cancer type categorized as non-HPV-related cancers.
The primary exposure variable was HPV vaccination history. It was directly ob from the survey question: "Have you ever received one or more doses of th vaccine?" [29]. Those participants who reported receiving at least one dose categorized as HPV vaccinated, while those who did not receive any dose

Study Variables
Two survey questions were used to develop the binary outcome variable: a selfreported diagnosis of HPV-related cancers. The first question was: "Have you ever been told by a doctor or other health professional that you had cancer or a malignancy of any kind?" [29]. The second question, asked to those who responded "yes" to the first question, was: "What kind was it?". Participants who reported being diagnosed with larynx, windpipe, mouth, tongue, lip or cervical cancers (anal, penile, vulvar, and vaginal cancers were not self-reported) were categorized as HPV-related cancers, while those who were not diagnosed with cancer or were diagnosed with other cancer types were categorized as non-HPV-related cancers.
The primary exposure variable was HPV vaccination history. It was directly obtained from the survey question: "Have you ever received one or more doses of the HPV vaccine?" [29]. Those participants who reported receiving at least one dose were categorized as HPV vaccinated, while those who did not receive any dose were categorized as not HPV vaccinated. Based on previous literature and epidemiological reasoning, the following covariates were considered as potential confounders of the association of HPV vaccination history with HNGC: age [3,30,31], ethnicity [30,31], education [3,22,31], household income [22], whether participants were born in the U.S. or not [31], and NHANES cycle. Potential risk factors for HNGC included marital status, any history of consuming 4-5 alcohol drinks daily, having smoked at least 100 cigarettes over their lifetime, self-reported diet, a history of being overweight, a history of diabetes, moderate or vigorous physical activity at work, and routine access to healthcare services "when sick or need advice about health" [31]. The assumed causal relationships are depicted in Figure 2.
Vaccines 2022, 10, x FOR PEER REVIEW  4 of 16 daily, having smoked at least 100 cigarettes over their lifetime, self-reported diet, a history of being overweight, a history of diabetes, moderate or vigorous physical activity at work, and routine access to healthcare services "when sick or need advice about health" [31]. The assumed causal relationships are depicted in Figure 2.

Figure 2.
Directed acyclic graph examining the association of human papillomavirus (HPV) vaccination history with HPV-related cancers. Age, ethnicity, education, household income, and whether participants were born in the U.S. or not, and National Health and Nutrition Examination Survey (NHANES) cycle are considered as confounders; adjusting for these blocks all backdoor pathways (indicated in gray boxes). Marital status, any history of consuming 4-5 alcohol drinks daily (alcohol), having smoked at least 100 cigarettes over their lifetime (smoking), self-reported diet (healthy eating), a history of being overweight, a history of diabetes, moderate or vigorous physical activity at work, and routine access to healthcare services "when sick or need advice about health" (healthcare access) are assumed to be risk factors for HPV-related cancers (shown in blue, except "HPV-related cancers", which is the outcome). The variable in green with sign " " is the primary exposure variable (HPV vaccination). The variable with " " sign is the outcome (HPV-related cancers).

Primary Analysis
To account for the complex sampling design, survey weights, strata and clusters were used in analyzing data. Survey weights were divided by 4 to account for combining four NHANES cycles. In the descriptive analysis, categorical variables were presented as sample frequencies, while percentages were presented as weighted estimates for the target population. Associations between categorical variables and the outcome variable were tested using Rao-Scott chi-square test for complex survey design studies [32]. A Figure 2. Directed acyclic graph examining the association of human papillomavirus (HPV) vaccination history with HPV-related cancers. Age, ethnicity, education, household income, and whether participants were born in the U.S. or not, and National Health and Nutrition Examination Survey (NHANES) cycle are considered as confounders; adjusting for these blocks all backdoor pathways (indicated in gray boxes). Marital status, any history of consuming 4-5 alcohol drinks daily (alcohol), having smoked at least 100 cigarettes over their lifetime (smoking), self-reported diet (healthy eating), a history of being overweight, a history of diabetes, moderate or vigorous physical activity at work, and routine access to healthcare services "when sick or need advice about health" (healthcare access) are assumed to be risk factors for HPV-related cancers (shown in blue, except "HPV-related cancers", which is the outcome). The variable in green with sign " x FOR PEER REVIEW 4 of 16 daily, having smoked at least 100 cigarettes over their lifetime, self-reported diet, a history of being overweight, a history of diabetes, moderate or vigorous physical activity at work, and routine access to healthcare services "when sick or need advice about health" [31]. The assumed causal relationships are depicted in Figure 2. vaccination history with HPV-related cancers. Age, ethnicity, education, household income, and whether participants were born in the U.S. or not, and National Health and Nutrition Examination Survey (NHANES) cycle are considered as confounders; adjusting for these blocks all backdoor pathways (indicated in gray boxes). Marital status, any history of consuming 4-5 alcohol drinks daily (alcohol), having smoked at least 100 cigarettes over their lifetime (smoking), self-reported diet (healthy eating), a history of being overweight, a history of diabetes, moderate or vigorous physical activity at work, and routine access to healthcare services "when sick or need advice about health" (healthcare access) are assumed to be risk factors for HPV-related cancers (shown in blue, except "HPV-related cancers", which is the outcome). The variable in green with sign " " is the primary exposure variable (HPV vaccination). The variable with " " sign is the outcome (HPV-related cancers).

Primary Analysis
To account for the complex sampling design, survey weights, strata and clusters were used in analyzing data. Survey weights were divided by 4 to account for combining four NHANES cycles. In the descriptive analysis, categorical variables were presented as sample frequencies, while percentages were presented as weighted estimates for the target population. Associations between categorical variables and the outcome variable daily, having smoked at least 100 cigarettes over their lifetime, self-reported diet, a history of being overweight, a history of diabetes, moderate or vigorous physical activity at work, and routine access to healthcare services "when sick or need advice about health" [31]. The assumed causal relationships are depicted in Figure 2. vaccination history with HPV-related cancers. Age, ethnicity, education, household income, and whether participants were born in the U.S. or not, and National Health and Nutrition Examination Survey (NHANES) cycle are considered as confounders; adjusting for these blocks all backdoor pathways (indicated in gray boxes). Marital status, any history of consuming 4-5 alcohol drinks daily (alcohol), having smoked at least 100 cigarettes over their lifetime (smoking), self-reported diet (healthy eating), a history of being overweight, a history of diabetes, moderate or vigorous physical activity at work, and routine access to healthcare services "when sick or need advice about health" (healthcare access) are assumed to be risk factors for HPV-related cancers (shown in blue, except "HPV-related cancers", which is the outcome). The variable in green with sign " " is the primary exposure variable (HPV vaccination). The variable with " " sign is the outcome (HPV-related cancers).

Primary Analysis
To account for the complex sampling design, survey weights, strata and clusters were used in analyzing data. Survey weights were divided by 4 to account for combining four NHANES cycles. In the descriptive analysis, categorical variables were presented as sample frequencies, while percentages were presented as weighted estimates for the target population. Associations between categorical variables and the outcome variable " sign is the outcome (HPV-related cancers).

Primary Analysis
To account for the complex sampling design, survey weights, strata and clusters were used in analyzing data. Survey weights were divided by 4 to account for combining four NHANES cycles. In the descriptive analysis, categorical variables were presented as sample frequencies, while percentages were presented as weighted estimates for the target population. Associations between categorical variables and the outcome variable were tested using Rao-Scott chi-square test for complex survey design studies [32]. A survey-weighted multiple logistic regression model was built to test the association of HPV vaccination history with HPV-related cancers using the complete case dataset. We used a directed acyclic graph with the backdoor criterion to select a minimum sufficient adjustment set for the model [33]. Previous literature suggested that age, ethnicity, education, household income, and whether individuals were born in the U.S. were associated with HPV vaccination and with HPV-related cancers [3,31]. We also assumed that timing would be associated with both the exposure and the outcome; thus, in addition to the aforementioned confounders, NHANES cycle was also added to the model. After retaining the variables blocking the backdoor paths in the model, potential risk factors for the outcome were selected using the backward elimination approach, based on the Akaike Information Criterion [34]. We used the Archer-Lemeshow test for design-based regression models [35] and the Area Under the Receiver Operating Characteristics Curve [36] to test the model's goodness-of-fit. A p-value less than 0.05 was considered a statistically significant finding.

Secondary Analysis (Propensity Score Matching)
Additionally, we performed propensity score-matched analysis to test the robustness of the results; this is considered a secondary confounding adjustment approach to the conventional logistic regression analysis when there are fewer events per confounder [37]. Propensity score (PS) matching was applied using a 1:1 nearest-neighbor method (without replacement), with a caliper width of 0.2 of the standard deviation of the logit of the propensity score [38]. We estimated PS using a logistic regression model including confounding variables, risk factors for the outcome (as shown in Figure 2), and survey features as recommended by Dugoff et al. [39]. We examined covariate balance between matched HPV-vaccinated and non-HPV-vaccinated groups using standardized mean difference (SMD). SMD < 0.1 was indicative of adequate covariate balance between the matched groups [40,41]. A crude survey-weighted logistic regression for the outcome was conducted to estimate the average treatment effect on the treated in the matched subsample.

Sensitivity Analysis
While the complete case analysis was more a conservative approach to address missing or invalid responses in covariates, the analysis actually excluded more than 4000 participants. We could not assume that the data was missing completely at random because distributions of sociodemographic characteristics of the study participants-including sex, ethnicity, whether individuals were born in the U.S. or not, income, smoking history, and physical activity-differed between the complete case dataset and the dataset with missing values ( Table 1). Assuming that the missing data can be explained by the observed data, i.e., missing at random assumption [42], and attempting to increase a statistical power of the analysis, we imputed missing data for covariate variables using the "multiple imputations then deletion" approach [43]. The missing values were imputed by chained equations using the "mice" package [44]. We used 20 imputations. Binary and polytomous logistic regression models were applied to impute missing values for binary and multi-level categorical variables, respectively. We used all variables to predict missing values, including the complex survey design features [45]. Survey-weighted outcome regression analysis was performed by forcing confounding variables in the imputation models, while the risk factors for the outcome were selected using AIC-based backward elimination by "stacking" all multiply imputed datasets [46]. Estimates from the imputation models were pooled using Rubin's rules [44]. All statistical analyses were performed using the R-software version 4.1.1 (Vienna, Austria) [47]. Table 1. Descriptive characteristics of U.S. adults aged 20-59 years, stratified by those who were diagnosed with HPV-related cancers and those who were not diagnosed with HPV-related cancers, using data from the National Health and Nutrition Examination Survey (NHANES), 2011-2018.

Variables
No n (%) a n (%) a n (%) a n (% To the question "How healthy is your overall diet?", responses "excellent", "very good" or "good" were categorized to "healthy", while responses "fair" or "poor" were categorized as "poor". † Overweight was defined as those who have ever told by a doctor that they were overweight or those whose body mass index was 25 or higher in the past 10 years. ‡ To the question "Is there a place that you usually go when you sick or you need advice about your health?", responses "Yes" and "There is more than one place" were categorized as "Yes", and "There is no place" as "No".

Sample Characteristics and Univariate Analysis
Overall, 0.7% (72/9891) participants reported being diagnosed with HPV-related cancers in the complete case sample (Table 1). Sixy-eight out of 72 HPV-related cancers were cervical cancers. The prevalence rate of HPV-related cancers decreased in the last two NHANES cycles (Figure 3). On the other hand, HPV vaccination prevalence rates were increasing from 5.5% to 13.3% (Rao-Scott chi-square test, p < 0.05) between 2011 and 2018. 76.8% of all HPV vaccinated participants were in the age group 20-29 years (Table S1). HPV vaccination rates were comparable between all participants and those participants who were not diagnosed with HPV-related cancers (9.4% and 9.4%, respectively), while participants who were diagnosed with HPV-related cancers had the lowest HPV vaccination rate (2.6%, p < 0.01) ( Table 1). There were statistically significantly differences between participants who were diagnosed with HPV-related cancers and those who were not diagnosed with HPV-related cancers in age (p = 0.001), sex (p < 0.001), education (p < 0.01), ethnicity (p = 0.01), household income (p = 0.001), having smoked at least 100 cigarettes over their lifetime p < 0.001), and routine access to healthcare services (p < 0.01).

Primary Analysis
After AIC-based stepwise backward elimination, the survey-weighted final multiple logistic regression model, assessing the association of HPV vaccination history with HPVrelated cancers, was adjusted for age, education, ethnicity, marital status, whether individuals were born in the U.S. or not, income, having smoked at least 100 cigarettes, moderate or vigorous physical activity at work, a history of consuming 4/5 alcohol drinks daily, a history of being overweight, routine access to healthcare services and NHANES

Primary Analysis
After AIC-based stepwise backward elimination, the survey-weighted final multiple logistic regression model, assessing the association of HPV vaccination history with HPV-related cancers, was adjusted for age, education, ethnicity, marital status, whether individuals were born in the U.S. or not, income, having smoked at least 100 cigarettes, moderate or vigorous physical activity at work, a history of consuming 4/5 alcohol drinks daily, a history of being overweight, routine access to healthcare services and NHANES cycle ( Table 2). In this model, there was no statistically significant association between HPV vaccination history and HPV-related cancers (adjusted OR = 0.58, 95% CI 0.19; 1.75). ) and other covariates with less 1% missing data (routine access to healthcare services, self-reported diet, a history of being overweight, marital status, having smoked at least 100 cigarettes over their lifetime, education, moderate or vigorous physical activity at work, location of birth, history of diabetes) were imputed using "multiple imputation then deletion" approach. 20 imputations were used. Each imputed dataset contained 13,993 observations. Estimates were pooled using the Rubin's rules.

Secondary Analysis
891 HPV-vaccinated participants were matched with 891 non-HPV-vaccinated participants using PS matching approach. Covariates between the matched groups were adequately balanced-SMDs were <0.1 for all covariates (Figure 4). In the surveyweighted simple logistic regression in the matched dataset (Table 2), no statistically significant association of HPV vaccination history with HPV-related cancers was observed (crude OR = 0.40, 95% CI 0.10; 1.69).

Sensitivity Analysis
25.7% of data a history of consuming 4/5 alcohol drinks every day, 6.6% of data for household income and less than 1% of data for other covariates were missing (see the Table 2 footnotes for details). The proportion of HPV-related cancers was lower (0.9%, not shown in Table 1) in the missing data than in the complete data. There were also differences in sex, education, ethnicity, location of birth, household income, having ever smoked at least 100 cigarettes over a lifetime, a history of being overweight, and moderate or vigorous physical activity at work between the complete dataset and the missing dataset. After applying the "multiple imputations then deletion" method, each imputed dataset contained 13,993 individuals. Pooled estimates from imputed models adjusting for age, education, ethnicity, marital status, whether they born in the U.S. or not, income, having ever smoked at least 100 cigarettes, moderate or vigorous physical activity at work, ever drank 4/5 alcohol drinks every day, a history of being overweight, routine access to healthcare services and NHANES cycle revealed no statistically significant association of

Sensitivity Analysis
25.7% of data a history of consuming 4/5 alcohol drinks every day, 6.6% of data for household income and less than 1% of data for other covariates were missing (see the Table 2 footnotes for details). The proportion of HPV-related cancers was lower (0.9%, not shown in Table 1) in the missing data than in the complete data. There were also differences in sex, education, ethnicity, location of birth, household income, having ever smoked at least 100 cigarettes over a lifetime, a history of being overweight, and moderate or vigorous physical activity at work between the complete dataset and the missing dataset. After applying the "multiple imputations then deletion" method, each imputed dataset contained 13,993 individuals. Pooled estimates from imputed models adjusting for age, education, ethnicity, marital status, whether they born in the U.S. or not, income, having ever smoked at least 100 cigarettes, moderate or vigorous physical activity at work, ever drank 4/5 alcohol drinks every day, a history of being overweight, routine access to healthcare services and NHANES cycle revealed no statistically significant association of HPV vaccination history with HPV-related cancers (adjusted OR = 0.47, 95% CI 0.15; 1.45; see in Table 2).

Discussion
In this nationally representative study, the primary analysis showed no association of HPV vaccination with HPV-related cancers. Although the relationship was not statistically significant, participants who were immunized with at least one dose of a HPV vaccine had 42% odds of HPV-related cancers (adjusted OR = 0.58, 95% CI: 0.19; 1.75) than those who were not HPV vaccinated. When conducting the propensity score-matched analysis and the analysis on multiple imputed datasets to assess the association between vaccination and HPV-related cancers, we also did not find a statistically significant association. However, a similar trend toward reduction of odds among those who were vaccinated (60% and 53%, respectively) was observed in the alternative analysis and the sensitivity analysis. The alternative and sensitivity analyses also provided narrower confidence intervals for OR, as they were able to better deal with the small number of HPV-related cancer cases per covariate in the model and potentially increase statistical power by imputing missing data.
A similar result was reported in a population-based study from Sweden [21]. Utilizing the Swedish Total Population Register, researchers compared incidence rates of HPV-associated cervical cancer between vaccinated and unvaccinated populations from 2006 to 2017. They found that the incidence rate ratio (IRR) was 0.51 (95% CI 0.32; 0.82) for vaccinated relative to unvaccinated populations; after adjusting for more covariates, the association became even stronger: IRR = 0.37 (95% CI, 0.21; 0.57) [21]. Our findings are also consistent with results from a Finnish follow-up study [20]. Linking women who were previously enrolled in HPV vaccine clinical trials to the Finish Cancer Registry, researchers were able to compare HPV-associated cancer rates between vaccinated and unvaccinated cohorts [20]. They observed ten cancer cases caused by HPV (eight cervical cancers, one oropharyngeal cancer and one vulvar cancer) among previously unvaccinated cohorts, while no cancer cases were observed in vaccinated cohorts. Thus, the study reported an HPV vaccine efficacy estimate of 100% (95% CI 16; 100) [20]. Similarly, a study from England reported relative incidence rate reductions in cervical cancer among women who were offered the vaccine as follows: between age 16-18 years: 34% (95% CI 25%-41%); between age 14-16 years: 62% (95% CI 52%-71%); and between age 12-13 years: 87% (95% CI 72%-94%), in comparison to unvaccinated women [48]. Another study from Denmark found a statistically significant reduction in incidence rate of cervical cancer in women who were vaccinated at age 16 years or younger-86% (95% CI 47%; 96%)-while failing to find any reduction at ages 17 years or older [49]. Lastly, the comprehensive HPV vaccination program in Australia is expected to halve cervical cancer rates by 2035 [50].
We noticed a reduction in HPV-related cancer cases in the last two NHANES cycles, relative to the previous 2011-2012 and 2013-2014 cycles. Similarly, a trend analysis of HPV-associated cancers between 1999 and 2015 using data from cancer registries in the U.S. showed a decline in incidence rates of cervical cancer in all age groups (1.6% decrease per year), vaginal cancer among <40 years old (2.8% decrease per year) and anal cancer among <40-year-old men (2.9% decrease per year) [18]. In Australia, since the introduction of the National HPV Vaccination Programme-first among girls in 2007, then among boys in 2013-HPV vaccine type prevalence and incidence of genital cancer precursors have declined among vaccinated women [51].
We also found a significant increase in HPV vaccination rates from 2011 through 2018. Similarly, other studies have shown an increase in HPV coverage among adolescents [52,53] and adults aged 19-26 years in the U.S. [54]. As the efficacy and effectiveness of the HPV vaccine against HPV prevalence were previously reported [55,56], nationally representative studies reported a decline of HPV prevalence in the U.S. Results from the NHANES 2003-2010 cycles showed a substantial reduction in HPV vaccine type prevalence by 56% decrease among young females [56]. Another NHANES study compared HPV vaccine type prevalence between the pre-vaccination era (2003)(2004)(2005)(2006) to post-vaccination era (2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014) [55]. The study results found that there was a 71% reduction in HPV prevalence among 14-to 19-year-old females and a 61% reduction among 20-to 24-year-old females between pre-and post-vaccination eras [55].
Although the efficacy and effectiveness of the HPV vaccine against HPV prevalence, HPV-associated anogenital warts and cervical cancer precursors is well supported by the evidence [16,17,19,[56][57][58], our results are important for understanding and extending knowledge about the HPV vaccine effectiveness against HPV-related cancers, evidence of which is currently lacking. While several high-income countries (e.g., Norway, Italy and Denmark) [59][60][61] and a few middle-income countries (e.g., Uzbekistan) [62] have reached and maintained high HPV vaccination coverage, the U.S. is experiencing challenges to maintain an adequate vaccination rate [30,63]. On the other hand, only less than a quarter of low-income and less than third of lower-middle-income countries had introduced the HPV vaccination [64]. Given HPV vaccination rates are relatively low nationally [30,63] and internationally [65] and overall hesitation to introduce HPV vaccine in low-and middle-income countries [64], the study findings can be used by policymakers and health professionals to further encourage HPV vaccination uptake among adolescents aged 11-15 years and to support decision-making about HPV vaccination among adults eligible to be vaccinated and parents considering immunizing their children.

Strengths and Limitations
The study has several strengths. Since we used a nationally representative sample, the study findings are generalizable to the adult non-institutionalized U.S. population. We also conducted secondary and the sensitivity analyses to check the robustness of our study findings. We found similar results across all analyses, suggesting the robustness of the results.
There are study limitations that need to be acknowledged. First, in our study, we do not know whether HPV-related cancers were caused by HPV infection, as the outcome data were collected based on self-reporting. Since most cases were cervical cancer (68 out of 72), it is unlikely that other factors might have contributed to their development. Second, differential misclassification bias might have been present. Other anogenital cancers, including vulvar, penile, vaginal, and anal cancers, were not separately reported in the survey, and they might have been classified together under the "Other cancers" group and not included in the outcome. Not including them in the study outcome might have weakened the association by pulling the effect estimate towards the null.
Next, since the vaccination status was also based on self-reporting, we cannot exclude the existence of the recall bias, as participants might mistakenly report their HPV vaccination history. However, the recall bias might be limited, since participants were specifically asked whether they received a vaccine against HPV under brand names Cervarix, Gardasil or Gardasil 9 [29].
Since one-dose vaccination might not provide full protection against HPV infections, it might then be inappropriate to compare at least one-dose vaccinated versus unvaccinated. Given the small number of HPV-related cancer cases in HPV vaccinated group, it was numerically challenging for us to estimate the effects of full and partial HPV vaccination in relation to non-HPV vaccination in this study.
Due to the small number of events, we were also not able to assess potential effect modifications of covariates (e.g., sex, ethnicity) on the relationship. As HPV vaccination before the first sexual onset, at age 11-12 years, is generally recommended, the vaccine might have a lower effectiveness in the older individuals, as they might have already contracted the infection. We were not able to determine whether participants were vaccinated before the first sexual onset or not, since only a few participants responded to questions about sexual behavior.
Another limitation is that non-response bias might be present in the study. More than half of all participants did not know their vaccination status or refused to respond. When we compared characteristics of participants with missing or invalid responses for HPV vaccination status to the complete case data, participants with missing HPV vaccination data had higher rates of HPV-related cancers and were likely from the NHANES 2017-2018 cycle. Additionally, the study timing and the observed rarity of events were the major limitations to finding a statistically significant association. In addition to a small number of HPV-related cancers (n = 72), most vaccinated participants were between the age of 20-39 years (92.5%). Since HPV-related cancer development may take 10-20 years after first contracting HPV, the majority of participants were not yet at risk of developing cancers, reducing the probability of detecting an association [66][67][68]. Thus, having a sufficient number of events and comparable age distributions between HPV-vaccinated and unvaccinated groups are desirable to be able to detect statistically significant findings. A statistically significant association was observed when we performed an additional analysis with an expanded outcome definition by including other potentially HPV-associated genital cancers (Table S2).
Lastly, we also acknowledge that the present study had insufficient statistical power to detect a statistically significant difference between the vaccinated and unvaccinated groups (Table S3). Future studies with large sample sizes are warranted to assess a statistically significant reduction.

Conclusions
While the analyses found no association of HPV vaccination with HPV-related cancers, the study findings suggest that HPV vaccinated people tend to have lower odds of developing HPV-related cancers than unvaccinated people. Given the importance of determining the impact of the vaccination on HPV-related cancers, there is a need to conduct future research by linking cancer registry data with vaccination records to obtain more robust results. It is understood that the effect of HPV vaccination will be observed in the next coming decades; however, preliminary findings, such as this study, could be used as additional evidence to encourage vaccination among 11-to 15-year-old adolescents and eligible adults, to reduce HPV-related cancer rates and prevent HPV-related cancer cases in the future.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/vaccines10122113/s1, Table S1: Descriptive characteristics of U.S. adults aged 20-59 years, stratified by HPV vaccination status, using data from the National Health and Nutrition Examination Survey (NHANES), 2011-2018.; Table S2: Survey-weighted logistic regression models using complete case, propensity score matched and multiply imputed datasets investigating the association of human papillomavirus vaccination history with HPV-related cancers (n = 72) and other genital cancers (including prostate, testicular, cervical and ovarian cancers, n = 68) among U.S. adults aged 20-59 years in the National Health and Nutrition Examination Survey (NHANES), 2011-2018.; Table S3

Informed Consent Statement:
We performed a secondary analysis using a publicly available database from the National Health and Nutrition Examination Survey (NHANES) (accessed on 15 November 2022: https://wwwn.cdc.gov/nchs/nhanes/Default.aspx). All participants in each NHANES cycle provided their informed consent.

Data Availability Statement:
Publicly available datasets were analyzed in this study. The National Health and Nutrition Examination Survey (NHANES) datasets can be found here: https://wwwn. cdc.gov/nchs/nhanes/ (accessed on 15 November 2022).