Examining the association between socioeconomic status and potential human papillomavirus-associated cancers§

Authors

  • Vicki B. Benard PhD,

    Corresponding author
    1. Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Division of Cancer Prevention and Control, Atlanta, Georgia
    • Epidemiology and Applied Research Branch, Division of Cancer Prevention and Control, Mailstop K-55, NCCDPHP, CDC, 4770 Buford Hwy NE, Atlanta, GA 30341
    Search for more papers by this author
    • Fax: (770) 488-4639.

  • Christopher J. Johnson MPH,

    1. Cancer Data Registry of Idaho, Boise, Idaho
    Search for more papers by this author
  • Trevor D. Thompson BS,

    1. Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Division of Cancer Prevention and Control, Atlanta, Georgia
    Search for more papers by this author
  • Katherine B. Roland MPH,

    1. Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Division of Cancer Prevention and Control, Atlanta, Georgia
    Search for more papers by this author
  • Sue Min Lai PhD, MS, MBA,

    1. Kansas Cancer Registry, University of Kansas Medical Center, Kansas City, Kansas
    Search for more papers by this author
  • Vilma Cokkinides PhD,

    1. Department of Epidemiology and Research Surveillance, American Cancer Society, Atlanta, Georgia
    Search for more papers by this author
  • Florence Tangka PhD,

    1. Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Division of Cancer Prevention and Control, Atlanta, Georgia
    Search for more papers by this author
  • Nikki A. Hawkins PhD,

    1. Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Division of Cancer Prevention and Control, Atlanta, Georgia
    Search for more papers by this author
  • Herschel Lawson MD,

    1. Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Division of Cancer Prevention and Control, Atlanta, Georgia
    Search for more papers by this author
  • Hannah K. Weir PhD

    1. Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Division of Cancer Prevention and Control, Atlanta, Georgia
    Search for more papers by this author

  • The opinions or views expressed in this supplement are those of the authors and do not necessarily reflect the opinions or recommendations of the journal editors, the American Cancer Society, Wiley-Blackwell, or the Centers for Disease Control and Prevention.

  • The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.

  • §

    This article is a US government work and, as such, is in the public domain in the United States of America.

Abstract

BACKGROUND.

This study examined the association between county-level measures of socioeconomic status (SES) and the incidence rate of human papillomavirus(HPV)-associated cancers, including cervical, vulvar, vaginal, anal, penile, and oral cavity and oropharyngeal cancers.

METHODS.

The authors collected data from cancer registries for site-specific invasive cancer diagnoses between 1998 and 2003, inclusive, among adults aged >20 years at the time of diagnosis. County-level variables that included education, income, and poverty status were used as factors for socioeconomic status. Measures of rural-urban status, the percentage of the population that currently smoked, and the percentage of women who reported having ever had a Papanicolaou (Pap) test were also studied.

RESULTS.

Lower education and higher poverty were found to be associated with increased penile, cervical, and vaginal invasive cancer incidence rates. Higher education was associated with increased incidence of vulvar cancer, male and female anal cancer, and male and female oral cavity and oropharyngeal cancers. Race was an independent predictor of the development of these potentially HPV-associated cancers.

CONCLUSIONS.

These findings illustrate the association between SES variables and the development of HPV-associated cancers. The findings also highlight the importance of considering SES factors when developing policies to increase access to medical care and reduce cancer disparities in the United States. Cancer 2008;113(10 suppl):2910–8. Published 2008 by the American Cancer Society.

An individual's socioeconomic status (SES) is 1 of the most powerful predictors of health in industrialized and developing countries.1 SES can be viewed as a complex product of social influences and individual factors that reflect access to care and health resources, environmental exposures and stressors, and health behaviors.2 Common factors of SES include income,3–7 education,3–7 occupation,7, 8 place of residence,9 poverty level,5, 7, 8 and race.2–5 For the purpose of defining SES more comprehensively, these measures are frequently combined, because each indicator alone is considered inadequate to reflect SES.10, 11 The same social factors that are known to influence general health outcomes can also impact cancer incidence and mortality.3, 4, 12, 13 In the United States, an ecological analysis of SES patterns on cancer surveillance data from 1975-1999 demonstrated the presence of a distinct inverse relation between the percentage of population below the poverty level and overall cancer mortality.14

The body of research that links SES to potentially human papillomavirus (HPV)-associated cancers is sparse and has primarily focused on the HPV to cervical cancer relation.9, 15–25 In a sample of women living in Brazil, poverty and access to care inequities were found to be predictive of oncogenic-type HPV and, ultimately, of cervical cancer.21 A meta-analysis based upon 57 studies found that there was a 2-fold increased risk of invasive cervical cancer among individuals of low social status compared with individuals of high social status.22 This finding appeared to be global, as it was replicated in data from the Americas, Africa, and Asia. The predictive power of social class is likely due, in large part, to variability in access to screening, a factor known to mitigate the association between high-risk HPV and subsequent development of cervical cancer.22

Evidence has established that HPV is a necessary cause for the development of many cancers, with other cofactors involved.26–35 The present study extended previous research by examining the relation between SES and the incidence of potential HPV-associated cancers, namely vulvar, vaginal, anal, penile, and oral cavity and oropharyngeal cancers, and multiple census-based factors of SES. Measures of rural-urban status, the percentage of the population that currently smoked, and the percentage of women who reported having ever had a Papanicolaou (Pap) test were also examined as confounders in the association between SES and potentially HPV-associated cancers.

MATERIALS AND METHODS

Data Sources

Cancer incidence data

For this analysis, the authors used data from the Centers for Disease Control and Prevention, National Program of Cancer Registries, and the National Cancer Institute's Surveillance, Epidemiology, and End Results Program, as described in more detail in the methods chapter of this Supplement36 combined with data for site-specific invasive cancer diagnoses between 1998 and 2003, inclusive, among adults aged >20 years at the time of diagnosis.

County SES data

For the purposes of this analysis, education, income, and poverty status variables were used as factors for SES. High school education, median household income, and poverty status were derived from the US. Census 2000.37 High school education and poverty status were available at the county level by race, and for Hispanics and non-Hispanic whites. These variables were defined according to the race-specific data for blacks, Asian/Pacific Islanders, and American Indian/Alaska Natives. White Hispanics were assigned the value for Hispanics, and non-Hispanic whites were assigned the value for non-Hispanic whites. Those with missing data for the race-specific, county-level SES variables and those with missing or other race were assigned the county-specific SES measures for all races combined. High school education was defined as the percentage of the county population reporting having acquired a high school education, and counties were categorized as <75%, 75% to 85%, and >85%. The median household income was reported at the county level for categories <$35,000, $35,000 to $49,999, and ≥$50,000. Poverty was defined by the percentage of the county population below the federal poverty level. We categorized counties as <10%, 10% to <20%, or ≥20% of the total county population below poverty level.

County-based measures of rural-urban status

The incidence rates were examined by metropolitan, suburban, and rural counties of the United States, using the US. Department of Agriculture urban-rural continuum codes, which are based on information from the US. Census 2000.38 Three states, Illinois, Minnesota, and Hawaii, were not included in this analysis. Illinois and Minnesota do not report substate cancer incidence, and county-specific population estimates were not available for Hawaii.

County-specific BRFSS data

County-level estimates from aggregated 1998-2003 Behavioral Risk Factor Surveillance System (BRFSS) data were included for the purpose of deriving the percentage of the population that currently smoked and the percentage of women who reported having ever had a Pap test. The response rate for BRFSS over this time period is 52%. These variables were considered potential confounders of the SES-HPV relation for some cancer sites and thus were included in those multivariate models (Pap test—cervical and vaginal; smoking—cervical and oral cavity and oropharyngeal). BRFSS estimates were limited to persons aged ≥20 years and were weighted by probability of selection and demographic characteristics (age, sex, and county and/or race in some states). County estimates based on <30 respondents were replaced with the state estimate (17.1% of counties for the current smoking variable and 28.7% of counties for the Pap test variable).

Statistical Analysis

Age-adjusted incidence rates

We computed age-adjusted incidence rates for men (including penile, anal, and oral cavity and oropharyngeal) and women (including cervical, vaginal, vulvar, anal, and oral cavity and oropharyngeal) by SES factors, including education, income, poverty status, and measures of rural-urban status. Patient age at diagnosis was categorized in 5-year intervals for ages 20-84 years, with a final category of >85 years. This analysis was limited to adults aged ≥20 years because the SES and risk-factor data did not pertain to those <20 years of age. Other reports in this Supplement present overall rates by race and Hispanic ethnicity; therefore, race and Hispanic ethnicity were presented in the multivariate modeling only.

Sex-specific and site-specific incidence rates and 95% confidence intervals (CI) were calculated for the period 1998-2003 by county-level SES variables. All rates were expressed as per 100,000 and were standardized to the US. Census 2000 standard population by 5-year age groups by using the direct method. Modified gamma intervals were calculated as 95% CI.39 Statistical testing for differences in incidence rates between SES groups was based on the modified F-test intervals around the rate ratio.40 No adjustments were made for multiple comparisons.

Multilevel modeling

Poisson multilevel mixed models using SAS PROC GLIMMIX (SAS version 9.1, SAS Institute, Inc., Cary, NC) were fitted to model the incidence density ratio (IDR) of potentially HPV-associated cancers. The data had a hierarchical multilevel structure, with the lowest level consisting of race by Hispanic ethnicity by county cells. Cells were nested within counties, which were nested within states. The BRFSS measures were included as continuous variables and rescaled in such a way that the modeled incidence density ratios were interpretable as change in incidence rate per 5% change in BRFSS risk factor/screening behavior.

Interactions between race and each SES variable were examined to determine if the relation between SES and cancer incidence rate was similar across race groups. State of residence at the time of diagnosis was included as a random effect in the models. Conditional on the state random effect rates and the fixed effects at the county and cell levels, the observed cancer case counts were assumed to be independent Poisson variables. A covariance parameter was included in the models to account for spatial autocorrelation. Tobler's first law of geography states, “Everything is related to everything else, but near things are more related than distant things.”41 Because nearby counties may have similar cancer rates, the model residuals may not be independent, and residual error may be underestimated, making confidence intervals spuriously narrow. A low-rank radial smoother (13 knots based on a k-d tree structure using nearest neighbors) was used with the county centroid latitude and longitude coordinates measured in decimal degrees to account for spatial autocorrelation.42, 43 In our study, the spatial autocorrelation covariance parameter was nonsignificant in each model, but was retained in the models to avoid potential bias.

The natural log of the expected cell count was used as an offset in the models. Expected counts were based on age-specific rates for all races combined and all states combined, thus accounting for potential confounding by age. The models were optimized by use of the Newton-Raphson technique with ridging, starting from generalized linear model estimates.43, 44 All variables included in each model are shown in the results tables.

RESULTS

Age-adjusted Incidence Rates

Table 1 includes the age-adjusted incidence rates by census-based SES measures and cancer sites among men. Lower education, lower income, higher poverty, and rural residence were all associated with increased penile cancer incidence rate. Higher education, mid-level income, and metropolitan residence were associated with increased anal cancer incidence rate. Higher education, low-level to mid-level income, and residence in metropolitan or suburban areas were associated with an increased oral cavity and oropharyngeal cancer incidence rate. Ten to twenty percent poverty status was associated with a decreased anal and oral cavity and oropharyngeal cancer incidence rate.

Table 1. Age-adjusted Incidence Rates* by County-Level Socioeconomic Status Measures and Cancer Site Among Men Aged 20 Years and Older, United States, 1998-2003
CharacteristicPenisAnusOral Cavity and Oropharynx
CasesRate95% CICasesRate95% CICasesRate95% CI
  • *

    Rates are per 100,000 population and age-adjusted to the 2000 US standard population.

  • Data are from 36 population-based cancer registries that participate in the National Program of Cancer Registries and the Surveillance Epidemiology, and End Results Program and meet high-quality data criteria. These registries cover approximately 76.5% of the US population.

  • P < .05 for comparison of the group of interest to the reference (ref.) level.

Overall45831.141.11-1.1860831.421.39-1.4631,4167.267.18-7.35
% ≥High school education
 <7511431.601.51-1.7110701.311.23-1.4051726.696.50-6.88
 75 to <8515491.171.11-1.2318021.321.25-1.3810,1117.237.09-7.38
 ≥85 (ref.)18880.950.90-0.9932031.521.47-1.5816,1127.517.40-7.63
Median household income
 <$35,00012151.411.34-1.5011991.381.30-1.4665207.307.12-7.48
 $35,000-$49,99926031.101.06-1.1537381.491.44-1.5419,0797.547.44-7.65
 ≥$50,000 (ref.)7620.980.91-1.0511381.291.21-1.3657966.476.30-6.64
% Below poverty level
 <10 (ref.)24510.990.95-1.0337561.451.41-1.5019,6377.457.34-7.55
 10 to <2013041.251.19-1.3314401.331.26-1.4075846.856.69-7.01
 ≥208251.611.49-1.738791.381.29-1.4841747.307.07-7.53
Urban-rural status
 Rural6431.451.34-1.574811.080.99-1.1929906.486.25-6.72
 Suburban3641.231.10-1.363791.251.13-1.3922737.367.06-7.67
 Metropolitan (ref.)35731.091.06-1.1352151.481.44-1.5226,1327.367.27-7.45

Table 2 includes the age-adjusted incidence rates by SES measures and cancer site among women. Lower education, lower income, and higher poverty were associated with increased cervical and vaginal cancer incidence rates. Rural and suburban residence was also associated with higher cervical cancer rates. Higher education and lower poverty were associated with increased vulvar, anal, and oral cavity and oropharyngeal cancer incidence rates in women. Lower income was associated with increased vaginal and vulvar cancer incidence rates. Metropolitan residence was associated with a decreased vulvar cancer incidence rate.

Table 2. Age-adjusted Incidence Rates* by County-Level Socioeconomic Status Measures and Cancer Site Among Women Aged 20 Years and Older, United States, 1998-2003
CharacteristicCervixVaginaVulvaAnusOral Cavity and Oropharynx
CasesRate95% CICasesRate95% CICasesRate95% CICasesRate95% CICasesRate95% CI
  • *

    Rates are per 100,000 population and age-adjusted to the 2000 US standard population.

  • Data are from 36 population-based cancer registries that participate in the National Program of Cancer Registries and the Surveillance Epidemiology, and End Results Program and meet high-quality data criteria. These registries cover approximately 76.5% of the US population.

  • P < .05 for comparison of the group of interest to the reference (ref.) level.

Overall59,80812.4112.31-12.5133400.640.62-0.6612,5152.402.36-2.4410,8342.142.10-2.199,4641.881.84-1.92
% ≥High School Education
 <7518,40218.3718.10-18.647870.880.82-0.9418051.981.88-2.0717871.951.86-2.041,4771.611.53-1.69
 75 to <8518,79112.4812.30-12.6611500.670.63-0.7144282.602.53-2.6835182.152.08-2.222,9761.811.75-1.88
 ≥85 (ref.)22,5409.939.80-10.0613970.530.51-0.5662712.422.36-2.4855252.232.17-2.295,0052.021.96-2.08
Median household income
 <$35,00014,01614.8114.56-15.068090.730.68-0.7827012.512.41-2.6123002.202.11-2.291,9231.831.75-1.91
 $35,000-$49,99935,54812.5212.39-12.6519710.640.61-0.6774962.442.39-2.5065782.222.17-2.275,7291.941.89-1.99
 ≥$50,000 (ref.)10,1699.929.73-10.115540.530.48-0.5723072.202.11-2.2919521.891.81-1.981,8061.781.70-1.86
% Below poverty level
 <10 (ref.)28,89610.4410.32-10.5717960.550.53-0.5881312.532.47-2.5967482.202.15-2.266,0251.961.91-2.01
 10<2016,12313.2013.00-13.419050.680.64-0.7330962.362.28-2.4528722.242.16-2.322,2941.791.72-1.87
 ≥2014,71419.0018.69-19.326330.950.88-1.0312771.841.74-1.9512101.741.64-1.841,1391.641.54-1.74
Urban rural status
 Rural604313.6213.28-13.983770.690.62-0.7613672.602.46-2.7510902.142.01-2.288711.691.58-1.81
 Suburban405212.8912.49-13.302560.690.60-0.789412.602.43-2.777872.242.08-2.406591.851.71-2.00
 Metropolitan (ref.)49,63812.2512.14-12.3527010.630.60-0.6510,1962.362.32-2.4189532.142.10-2.187,9281.901.86-1.95

Multilevel Modeling

The modeling results are shown by cancer site for men (Table 3) and women (Table 4), with no significant interactions. White race, lower education, and higher poverty were all associated with increased penile cancer incidence rates after adjustment for all other variables (Table 3). Asian/Pacific Islander race, Hispanic ethnicity, and residence in a rural or suburban setting were all associated with decreased anal cancer incidence among males. Those with a county-level median household income of <$50,000 had an increased anal cancer incidence rate relative to those with an income of ≥$50,000.

Table 3. Multilevel Poisson Modeling* Results by Cancer Site for Men Aged 20 Years and Older in the United States, 1998-2003
CharacteristicPenisAnus
IDR95% CIPIDR95% CIP
  • *

    Model results show all variables included in each model.

  • Data are from 36 population-based cancer registries that participate in the National Program of Cancer Registries and the Surveillance, Epidemiology, and End Results Program and meet high-quality data criteria. These registries cover approximately 76.5% of the US population.

Race  <.0001  <.0001
 Asian/Pacific Islander vs white0.400.30-0.52 0.140.10-0.19 
 Black vs white0.640.54-0.76 1.140.98-1.33 
Ethnicity  .0629  <.0001
 Hispanic vs non-Hispanic1.170.99-1.38 0.520.44-0.62 
% ≥High school education  .0002  .0144
 <75 vs ≥851.321.15-1.53 1.000.87-1.15 
 75 to <85 vs ≥851.161.07-1.26 0.910.84-0.98 
Median household income  .8618  <.0001
 <$35,000 vs ≥$50,0001.020.89-1.17 1.341.19-1.51 
 $35,000 to $49,999 vs ≥$50,0001.020.93-1.13 1.301.20-1.41 
% Below poverty level  <.0001  .7699
 10 to <20 vs <101.171.05-1.29 1.030.94-1.13 
 ≥20 vs <101.591.33-1.90 1.040.87-1.23 
Urban-rural status  .3634  .0002
 Rural vs metropolitan1.080.97-1.20 0.800.71-0.89 
 Suburban vs metropolitan1.040.92-1.17 0.880.79-0.99 
Table 4. Multilevel Poisson Modeling* Results by Cancer Site for Women Aged 20 Years and Older in the United States, 1998-2003
CharacteristicVaginaVulvaAnusOral Cavity and Oropharynx
IDR95% CIPIDR95% CIPIDR95% CIPIDR95% CIP
  • *

    Model results show all variables included in each model.

  • Data are from 36 population-based cancer registries that participate in the National Program of Cancer Registries and the Surveillance, Epidemiology, and End Results Program and meet high-quality data criteria. These registries cover approximately 76.5% of the US population.

Race  <.0001  <.0001  <.0001  <.0001
 Asian/Pacific Islander vs white0.600.46-0.78 0.180.14-0.23 0.240.20-0.29 0.320.26-0.39 
 Black vs white1.241.03-1.49 0.650.58-0.73 0.840.75-0.95 1.090.96-1.24 
Ethnicity  .8552  <.0001  .0015  <.0001
 Hispanic vs non-Hispanic0.980.81-1.19 0.620.54-0.70 0.820.72-0.93 0.450.39-0.52 
% ≥High school education  .0080  .0035  .3795  .0023
 <75 vs ≥851.261.07-1.48 1.101.00-1.21 1.070.97-1.19 1.050.94-1.16 
 75 to <85 vs ≥851.131.03-1.25 1.091.04-1.14 1.030.97-1.08 0.930.88-0.98 
Median household income  .2435  .0006  <.0001  .0691
 <$35,000 vs >$50,0001.040.89-1.22 1.080.99-1.18 1.121.02-1.22 1.050.95-1.16 
 $35,000 to $49,999 vs ≥$50,0001.090.98-1.21 1.111.05-1.17 1.151.09-1.22 1.071.01-1.14 
% Below poverty level  .0365  .5849  <.0001  .0992
 10 to <20 vs <101.151.02-1.29 1.010.95-1.08 1.050.99-1.13 1.060.98-1.13 
 ≥20 vs <101.281.04-1.58 1.070.94-1.22 0.860.75-0.97 0.970.84-1.12 
Urban-rural status  .8321  .1015  .2011  .0016
 Rural vs metropolitan0.970.85-1.10 0.930.87-1.00 0.930.86-1.01 0.860.79-0.93 
 Suburban vs metropolitan1.020.89-1.17 0.950.88-1.02 0.980.90-1.06 0.940.87-1.03 
% Ever had Pap test  .3801         
 By 5 percentage point increase1.030.96-1.11    
% Current smokers           .0070
 By 5 percentage point increase   1.031.01-1.06 

Lower education and higher poverty were associated with an increased vaginal cancer incidence rate (Table 4). Blacks had a higher adjusted rate of vaginal cancer than whites, while Asian/Pacific Islanders had a significantly lower incidence rate. Hispanics had a decreased vulvar cancer incidence rate. Blacks and Asian/Pacific Islanders had lower vulvar incidence rates than whites. Those with a county-level median household income of <$50,000 and those residing in counties with high school education rates <85% had higher rates of vulvar cancer incidence than did those in the highest income and education categories. Blacks and Asian/Pacific Islanders had lower anal incidence rates than whites. Those in counties with lower median household income had higher rates of anal cancer. There was also a statistically significant relation between poverty and anal cancer incidence rate; women in counties with ≥20% poverty had lower incidence rates than women in counties with <10% poverty. Hispanic ethnicity, Asian/Pacific Islander race, and rural residence were associated with lower rates of female oral cavity and oropharyngeal cancer. Those residing in counties with high school education rates of 75% to <85% had a lower incidence rate than those with ≥85% high school education. Current smoking was associated with an increased oral cavity and oropharyngeal cancer incidence rate.

Table 5 reports the sex and cancer site among men and women with significant interactions between race and measures of rural-urban status. Hispanic ethnicity, lower education, lower income, and a higher percentage of poverty were associated with higher rates of cervical cancer incidence after adjusting for other demographic and socioeconomic differences (Table 5; female cervix). There was a statistically significant interaction between race and measures of rural-urban status on the cervical cancer incidence rate (P < .0001). Asian/ Pacific Islanders had a lower cervical cancer incidence rate compared with whites in metropolitan areas. However, in rural areas, Asian/Pacific Islanders had significantly higher adjusted rates of cervical cancer compared with whites. Blacks had an increased cervical cancer incidence rate compared with whites in both rural and suburban areas. A 5% increase in current smoking prevalence at the county level was associated with a 2% increase in cervical cancer incidence.

Table 5. Multilevel Poisson Modeling* Results by Sex and Cancer Site Among Men and Women Aged 20 Years and Older in the United States, 1998-2003
CharacteristicFemale CervixMale Oral Cavity and Oropharynx
IDR95% CIPIDR95% CIP
  • *

    Model results show all variables included in each model and include significant race by area interaction in each model (P < .0001 for cervix; P = .0008 for male oral cavity and oropharyngeal).

  • Data are from 36 population-based cancer registries that participate in the National Program of Cancer Registries and the Surveillance, Epidemiology, and End Results Program and meet high-quality data criteria. These registries cover approximately 76.5% of the US population.

Ethnicity  <.0001  <.0001
 Hispanic vs non-Hispanic1.101.06-1.15 0.570.53-0.62 
% ≥High school education  <.0001  .0149
 <75 vs ≥851.411.35-1.47 0.950.89-1.01 
 75 to <85 vs ≥851.161.14-1.19 0.950.92-0.98 
Median household income  <.0001  <.0001
 <$35,000 vs ≥$50,0001.151.11-1.20 1.141.08-1.20 
 $35,000 to $49,999 vs ≥$50,0001.111.08-1.14 1.131.09-1.17 
% Below poverty level  <.0001  .2507
 10 to <20 vs <101.051.02-1.08 1.030.99-1.08 
 ≥20 vs <101.201.15-1.26 1.040.97-1.13 
Urban-rural status*
 Asian/Pacific Islander
  Rural vs metropolitan1.731.24-2.39 2.561.14-5.76 
  Suburban vs metropolitan1.190.85-1.67 1.980.88-4.44 
 Black
  Rural vs metropolitan1.151.05-1.26 1.060.92-1.23 
  Suburban vs metropolitan1.100.98-1.24 0.960.80-1.15 
 White
  Rural vs metropolitan0.960.93-0.99 0.840.80-0.88 
  Suburban vs metropolitan1.000.96-1.03 0.970.92-1.01 
Race*
 Rural
  Asian/Pacific Islander vs white1.501.08-2.07 0.790.36-1.77 
  Black vs white1.251.13-1.38 1.511.30-1.75 
 Suburban
  Asian/Pacific Islander vs white1.000.71-1.39 0.530.24-1.18 
  Black vs white1.161.02-1.30 1.180.97-1.43 
 Metropolitan
  Asian/Pacific Islander vs white0.830.79-0.87 0.260.23-0.29 
  Black vs white1.041.00-1.09 1.191.11-1.27 
% Ever had Pap test  .5782   
 By 5 percentage point increase1.010.99-1.03  
% Current smokers  .0004  <.0001
 By 5 percentage point increase1.021.01-1.03 1.041.02-1.05 

Hispanics and those residing in a county with <85% high school education had decreased oral cavity and oropharyngeal cancer incidence rates among males (Table 5; male oral and oropharyngeal cancer). Higher county-level current smoking prevalence and residence in a county with lower median household income were associated with increased oral cavity and oropharyngeal cancer. There was a significant interaction between race and measures of rural-urban status on oropharyngeal cancer (P = .0008). Among Asian/Pacific Islanders, there was a significantly higher incidence rate in rural versus metropolitan areas. However, among whites, the adjusted incidence rate was lower in rural counties compared with metropolitan areas. Blacks had significantly higher incidence rates compared with whites in metropolitan and especially rural areas. Asian/Pacific Islanders had a decreased incidence rate compared with whites in metropolitan areas.

DISCUSSION

Our findings show that the association between county-level SES and cancer incidence varies by sex, cancer type, and the particular measure of SES. For instance, there was a statistically significant association between having less education and a higher incidence of penile, vaginal, female oral cavity and oropharyngeal cancers, but not with other HPV-associated cancers. Lower household income was associated with higher incidence of cervical and male anal and oral cavity and oropharyngeal cancers. Greater poverty was associated with higher incidence of cervical, vaginal, and female anal cancer, and of male oral cavity and oropharyngeal cancers. In contrast, higher education was associated with increased incidence of male and female anal cancer, along with male and female oral cavity and oropharyngeal cancers.

Other studies have examined cervical cancer incidence by using county-level SES measures. Consistent with our findings, an inverse relation had been reported between county-level SES and cervical cancer, based on earlier data from a smaller number of registries.25 More recently, McDougall et al reported differences in cervical cancer by county-level SES measures, a finding that may explain persistent disparities in cervical cancer incidence by race/ethnicity.45 Our study confirms the county-level SES associations with cervical cancer incidence, and it includes a larger coverage of the United States. Studies using SES measures to examine potential HPV-associated cancers other than cervical cancer are limited.

It is not clear why certain area measures of SES are associated with incidence of some potentially HPV-associated cancers and not others, but several causal pathways can be hypothesized. Cervical infection with HPV, which is linked to both female and male sexual behavior, and access to adequate cervical cancer screening programs are likely to be important in explaining the increased cervical cancer incidence rates observed in different SES groups.46 In addition, it has been postulated that individuals with lower SES, including lower levels of education and low income, may be more likely to engage in higher risk sexual activity or have delayed access to medical screening services, leading to higher exposure and infection rates.46 This could account for not only cervical but also vaginal, vulva, penile, and anal cancers, where sexual behaviors are relevant risk factors. Although there are no recommended screening protocols for these HPV-associated cancers, precursor lesions can often be identified during general physical examinations by health care providers.47–49 Another explanation of why SES variables were not uniformly associated with all HPV-associated cancers is that the estimate of the attributable fraction of HPV infection in the development of cancer varies widely for different cancers.46 For example, almost 100% of cervical cancer and 90% of anal cancer can be attributed to HPV infection, but only 40% of cancers of the vulva, vagina, and penis, and 35% of cancers of the oral cavity and oropharyngeal area, are estimated to be attributable to HPV.46

The relations between potentially HPV-associated cancers and SES measures are complex and are further compounded by the diversity in the age, race, sex, and cultural composition of the US population. In our study, Asian/Pacific Islanders had significantly lower incidence rates of penile, vaginal, vulvar, female oral cavity and oropharyngeal cancers, as well as male and female anal cancers, than did whites. Black females had significantly higher incidence rates of vaginal cancer and significantly lower rates of vulvar and anal cancers than did white females. The magnitude of female cervical and male oral cavity and oropharyngeal cancers varied by rural-urban status across the racial groups. It is also important to note that the observed association between low SES and incidence rates of potentially HPV-associated cancers was independent of race.

Because the cancer registries do not collect individual-level measures of SES, we were limited to county-level measures in our examination of the association between SES and potentially HPV-associated cancer. We used county-level measures of poverty, median income, and percentage of high school education to characterize the socioeconomic status of the neighborhood in which individuals resided at the time of cancer diagnosis. These selections were based on prior studies examining the association between SES and cervical cancer.7, 9, 13, 45, 50 Also, HPV-associated cancer may have relatively long latency periods (10-20 years). For an evaluation of the relation between SES and incidence, the SES poverty variable might need to be from an earlier census collection to provide a better proxy measure of SES status when the cancer was actually developing. In addition, because populations are fairly mobile, the characteristics of the neighborhood where the person resided when cancer first developed may be different from the current neighborhood. Another limitation is that we were not able to examine cancer stage for this analysis.

This study is the first to report that SES is associated with the incidence of potentially HPV-associated cancers beyond cervical cancer, using nationally generated data that includes a majority of the US population. Our study sets the stage for monitoring patterns in SES measures related to HPV-associated cancers to reduce disparities by focusing on subgroups for future prevention, especially with the adoption of HPV vaccines and other emerging technologies. Although this research covers a large proportion of the US population (>76%), improved data collection through cancer registries will be essential in continued monitoring of the burden of potentially HPV-associated cancers in all areas of the United States. Because this is the first report to examine other cancers beyond cervical, ongoing studies are needed to confirm these unique findings.

In conclusion, the results of our study show an obvious need for more research. One priority need is better characterization of the factors that are related to SES status. A greater ability to stratify outcomes by a consistent group of population characteristics could help in targeting population-based interventions to increase early detection of these lesions and to educate the public and providers about the burden of disease. The clinical and public health benefits of understanding the role that SES status plays in disparities in healthy behaviors, health screening, and participation in preventive health care can be great. These findings highlight the importance of considering SES factors when developing policies to reduce the incidence of HPV-associated cancer in the United States.

Ancillary