This article is a US Government work and, as such, is in the public domain in the United States of America.
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.
Although the rates of invasive cervical cancer (ICC) have decreased substantially in the US since the advent of the Papanicolaou (Pap) test, Appalachian women remain at increased risk compared with the nation as a whole. The ICC incidence rates were compared in 5 Appalachian states with population-based cancer registries to investigate variability within the Appalachian region.
Alabama, Kentucky, Ohio, Pennsylvania, and West Virginia were selected for the analysis on the basis of their having high-quality cancer registry data for 1998 through 2003. Incidence rates were calculated by state and by Appalachia/non-Appalachia, urban/rural, and black/nonblack within each state, following the standard case definition and inclusion criteria used in this supplement. Data from the Behavioral Risk Factor Surveillance System (BRFSS) were used to characterize the prevalence of Pap testing and smoking.
The ICC incidence rates varied among the 5 states, being highest in West Virginia (10.9 of 100,000) and Kentucky (10.7 of 100,000), and lowest in Ohio (8.2 of 100,000). The Appalachian regions of Kentucky, West Virginia, and Ohio had considerably higher rates than those of Alabama and Pennsylvania. These variations reflected patterns in the rates of poverty, education, smoking, and Pap testing.
The variability in ICC risk across subgroups of Appalachia should be considered in the planning of preventive strategies, including reduction in risk factors and promotion of screening and vaccination. Cancer 2008;113(10 suppl):2974–80. Published 2008 by the American Cancer Society.
The Papanicolaou (Pap) test has been an extremely effective screening tool in detecting abnormal cervical cells that can lead to cervical cancer. Because of the widespread use of the Pap test, the rate of invasive cervical cancer (ICC) has decreased substantially in the last several decades in the US.1 Nevertheless, cervical cancer mortality rates have been consistently higher among subgroups of the population, including Appalachian women.2–4 The Appalachian area is generally characterized as being white, rural, poor, and poorly educated.5 However, the Appalachian area is quite extensive and diverse, spanning 13 states ranging from New York in the north to Mississippi in the south.6 Previous research has shown that rates of cervical cancer incidence and mortality vary considerably across the Appalachian states.2, 3, 7 An analysis of cervical cancer incidence data from 3 Appalachian states (Kentucky, Pennsylvania, and West Virginia) demonstrated that rural status, education, and race were significant predictors of higher rates of cervical cancer.7
The purpose of the current study was to expand and update the previous research7 to include 5 Appalachian states, using the most current 6-year incidence data available from the National Program of Cancer Registries (NPCR). This article complements another article included in this supplement featuring a regional analysis of incidence rates across another area of the US, namely, the US-Mexico border region.8 Both regions cross state lines, and state-level analyses may not reveal the high incidence rates across the affected area, in this case, Appalachia.
MATERIALS AND METHODS
The calculations of age-adjusted ICC rates for the 5 states included in this study were based on 1998 through 2003 NPCR data coded C53.0-C53.9 (cervix uteri), as described in detail in the article by Watson et al in this supplement.9 Briefly, NPCR data from 5 population-based cancer registries were used: Alabama (AL), Kentucky (KY), Ohio (OH), Pennsylvania (PA), and West Virginia (WV). Each of these tumor registries meets high‒quality standards for the study period, including ≥90% case ascertainment and ≤3% missing data regarding age, race, and sex.9 These 5 Appalachian states were selected on the basis of the availability of complete cancer registry data for 1998 through 2003. Of the remaining 8 Appalachian states, only New York and South Carolina met the criteria for inclusion. However, those 2 states were excluded from the analysis for the following reasons: New York City has such a large population that it overwhelms any analysis of non-Appalachian New York (Appalachian counties comprised 5.2% of New York State's population during the study period, whereas New York City comprised 42.2%); South Carolina, conversely, has only 6 of its 46 counties in Appalachia (13%). The 5 selected states represent different population distributions across demographic factors such as race, income, and education; range geographically from north to south; and include 57% of the Appalachian population, thus constituting a good cross-section of the region. The study area is depicted in Figure 1.
To characterize the burden of ICC across the study region, we classified subregions within and across the 5 states as Appalachian/non-Appalachian and rural/urban, based on their county-level classification. Each county was classified as Appalachian/non-Appalachian using the Appalachian Regional Commission (ARC) classification,6 and as either rural or urban, according to the Department of Agriculture's 2003 Rural-Urban Continuum Codes10; codes 0-3 were grouped as urban, and codes 4-9 as rural, as done previously.7 Only 16 cases meeting the inclusion criteria were missing county of residence and therefore were excluded from these analyses. Rates were also calculated by race, dichotomized as black and nonblack because of the small number of nonblack minorities in much of the study region. Cases missing a designation for race (n = 179) were excluded from the calculations of rates by race.
Census 2000 data were used to characterize the study region by the proportion of residents living in poverty and their educational level across each of the states, and for each of the subgroups of analysis (Appalachian status, urban/rural, race groups), based on the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) program-modified population estimates. SEER*Stat software11 was used to calculate the age-adjusted rates and 95% confidence intervals (using the method of Tiwari et al12) across the same categories.
To further explore some of the observed contrasts across the different states using available secondary data sources, we used 2004 data from the Behavioral Risk Factor Surveillance System (BRFSS) on smoking and Pap test usage, 2 behavioral variables strongly associated with the risk of ICC (smoking as a biologic factor, and Pap test usage as a screening-related factor). Smokers have been consistently found to have approximately twice the risk of ICC as nonsmokers.13–15 Although it has now been established that infection with an oncogenic human papillomavirus (HPV) type is a necessary condition for the development of ICC, it is not a sufficient cause, and smoking is clearly a contributing factor. In fact, women who test positive for HPV and smoke have approximately 5 times the risk of ICC as nonsmokers who test positive.16 We categorized female survey respondents as smokers if they reported being current smokers, and we classified them as having had a recent Pap test if they had received 1 within the last 3 years. Women were excluded from the analysis if they reported having undergone a hysterectomy and if data were missing for both the smoking and screening questions. The BRFSS data used in this descriptive analysis are publicly available,17 but the county of residence is excluded if there are <50 respondents in any given county. For the 5 states of our analysis, the county variable was missing for 14% of respondents in PA and OH, and for 25% to 30% in the remaining 3 states. Given this sizeable proportion of counties, we did not stratify the analysis by Appalachian/non-Appalachian or by rural/urban status, because both classifications are at the county level. However, we were able to calculate the prevalence of smoking and of recent Pap tests by race, income, education, and age groups. The BRFSS data were analyzed using Stata 10.0 statistical software (StataCorp, College Station, Tex).18
Given the well-established relation between lower education, lower income, and black race and the increased risk of ICC, it is relevant to present in some detail the demographic composition of the study population, as shown in Table 1. Overall, the general characteristics of the population in the study region are quite similar to those of the US as a whole, particularly in terms of education and poverty but also with respect to the percentage of blacks in the population. However, there is a wide range of variability within the study area, both among states and among the Appalachian regions within each state. For example, the 2 states with the largest populations, OH and PA, have higher rates of education, as measured by the percentage of the adult population who did not complete high school, than the other 3 states. A similar pattern was observed for poverty levels, with lower rates of poverty found in OH (10.6%) and PA (11.0%) than in the other 3 states (15.8%-17.9%). With respect to race, the percentage of blacks ranged from 3.1% in WV to 25.9% in AL, with the other 3 states distributed in between (7.3% in KY, 9.9% in PA, and 11.4% in OH).
Table 1. General Characteristics of the 5-State Study Area
The data in Table 1 also demonstrate variability within the states. Overall, there is clear disparity between Appalachian and non-Appalachian KY with regard to education (37.5% vs 21.3% who did not complete high school) and poverty levels (24.4% vs 12.4%), and to a lesser extent in OH, whereas in PA the rates do not differ much by Appalachian status. In AL, the reverse was observed for poverty (14.4% for Appalachia and 19.2% in non-Appalachia), with no differences noted with regard to education. This comparison could not be assessed in WV, because the entire state is Appalachian, but its indices of education and poverty fall between those of Appalachian KY and Appalachian OH. With respect to race, all the states have a higher proportion of blacks in their non-Appalachian regions, and except for Appalachian AL, the other states have <5% total blacks across their respective, combined Appalachian counties (with KY having the lowest at 1.6%). KY and WV have the largest percentage of rural counties (71% and 62%, respectively) compared with the other 3 states (52%-58%).
The ICC age-adjusted incidence rates for the study period (1998-2003), stratified by race, urban/rural, and Appalachian status, are presented in Table 2. All rates are shown per 100,000 women. The rate for the entire study area was 8.8; however, state-specific rates varied, with WV and KY having the highest rates (10.9 and 10.7, respectively), followed by AL (9.0), PA (8.5), and OH (8.2).
Table 2. Incidence Rates (Per 100,000 Women) of ICC in Selected Appalachian States by Race, Rural/Urban Area, and Appalachian Status, 1998-2003
Rates are per 100,000 population and are age-adjusted to the 2000 US standard population (19 age groups [Census P25-1130]). Data were obtained from 5 population-based cancer registries that participate in the National Program of Cancer Registries (NPCR) and/or the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) Program and meet high-quality data criteria. These registries cover approximately 57% of the US population in Appalachia.
All 5 states
The comparisons by Appalachian status demonstrated variability across the region. In the study area as a whole, the rates were somewhat higher in Appalachia than in non-Appalachia (9.5 vs 8.4), and this difference reached statistical significance, as determined by the lack of overlap of the 95% confidence intervals. However, there was no consistent trend noted across all the states. The highest incidence rate of ICC was found in Appalachian KY (12.5), a rate that was 25% higher than in the non-Appalachian part of the state. The greatest contrast was observed in OH, in which the Appalachian rates were 36% higher than in non-Appalachia. Conversely, there was no difference noted in AL or PA with respect to Appalachian status. As previously stated, WV is entirely Appalachian, and therefore this factor could not be assessed.
The findings across urban-rural classification were similar. Overall, the rates were significantly higher in rural counties than in urban counties (10.2 vs 8.4). This contrasting trend was observed across all the study states except PA, and it reached statistical significance in OH (10.1 vs 7.7), as evidenced by the lack of overlap of the 95% confidence intervals.
With respect to race, we compared only blacks and nonblacks because the number of cases of other ethnic and/or minority groups was too small in some of the regions to yield stable rates. The rates of ICC were approximately 35% higher among blacks across the study region (11.3 vs 8.4). This contrast persisted across each of the states in this study, except for WV, in which the estimates were unstable because of the low number of cases among blacks (N = 18 over the 6-year study period). The racial difference was significant in AL (56% higher among blacks), followed by PA (49% higher) and KY (48% higher), but lower in OH (13%) with overlap of the 95% confidence intervals.
The analysis of the 2004 BRFSS data with respect to 2 salient ICC risk factors—smoking and lack of frequent Pap testing—demonstrated variability across states and also across demographic groups within each state (Table 3). The self-reported prevalence of current female smoking was highest in WV (28.1%), followed by KY (27.0%) and OH (26.0%), and lowest in PA (23.4%) and AL (22.3%). In addition, AL was the only state in which smoking rates across all levels of income were lower than the combined rates of the 5 states. Black women were less likely than nonblacks to be smokers in OH and AL than in the other 3 states. The negative effects of lower educational attainment and lower income level on the frequency of Pap testing were found to be similar across all the 5 states, as was the decreasing gradient observed with age. It is interesting to note that OH and AL had the greatest proportion of women aged ≥65 years who had recently been screened.
Table 3. Distribution of Current Smoking Rates and Recent Papanicolaou Testing* by State and by Race, Income, and Education Based on BRFSS 2004 Data, Selected Appalachian States16
Women were excluded from the analysis if they reported having undergone a hysterectomy or if data were missing from either the smoking or screening question.
High school graduate
The results of the analysis presented herein are consistent with those reported previously, showing that the rates of ICC are higher among populations that are rural,7, 19–22 are poor,22, 23 and have a low educational attainment.7, 23 In addition, each year the ARC designates certain counties in Appalachia as distressed, a term used to denote the most economically depressed counties based on poverty, unemployment, and income. The proportion of counties that are distressed in each of the 5 study states, as projected by ARC for 2008, are: 1 of 52 or 1.9% in PA; 2 of 37 or 5.4% in AL; 4 of 29 or 13.8% in OH; 13 of 55 or 23.6% in WV; and 37 of 51 or 72.6% in KY.6 It is noteworthy that the proportion of distressed Appalachian counties by state closely follows the relative rates of ICC across the Appalachian regions of the 5 states. This pattern is also consistent with the notion that although Appalachia as a whole is often regarded as a disadvantaged region with respect to economic and social factors as well as health risks,5, 24, 25 it encompasses a large and heterogeneous area of the country. This has been documented particularly for cancer, as reflected by the variations in cancer incidence rates noted across Appalachia,26 including marked differences in the rates of ICC.7 This was also evident in the current analysis, with distinct patterns of risk observed across the northern, central, and southern regions of Appalachia, as described also in the study by Wingo et al.26
The higher rates of poverty and lower levels of education, in turn, reflect other underlying factors that contribute to the increased risk of ICC. In addition to the well-established necessary role of HPV infection in the development of ICC, smoking has been clearly identified as a contributory factor, conferring approximately double the risk of ICC among women who smoke. In general, there is a pattern of increased smoking prevalence with decreased income and educational levels,27, 28 also suggested in the results of the current analysis. Pap test usage for the early detection of suspicious or precancerous lesions in the cervix, or the detection of early-stage ICC, follows similar patterns across income and educational groups. A lower prevalence of Pap testing also tends to occur among rural populations.19, 29–32
However, the interplay between socioeconomic factors, smoking, and Pap testing is complex because women who smoke, as a whole, may be more likely to engage in risky behavior, including a higher risk of exposure to HPV infection through a larger number of sexual partners.33–36 The census data presented in Table 1 shows that KY, WV, and AL have similar rates of education and poverty, yet the rates of ICC are considerably lower in AL. This may be explained at least in part by the finding that AL has lower smoking rates than KY and WV, as well as higher rates of Pap testing, and this contrast is greatest among the low-income, low-education groups. This may reflect public health intervention practices that may have been more common or more effective in AL, as well as possible cultural differences in the Deep South, but further work is needed to elucidate these state-regional differences.
With respect to race, black women in KY had the highest ICC rate (15.1); however, the racial contrast between black and nonblack women peaked in AL, in which the ICC rate was 56% higher in black women. This is somewhat puzzling, given that blacks in AL had much lower rates of smoking than nonblacks and blacks in the other states. They also had higher rates of Pap testing. The contrast could be at least partially explained by the limitations of the BRFSS, which may not accurately represent all subgroups of the population, either by the relative proportion of women who have landline telephones or by differences between respondents and nonrespondents, given the national 2004 median BRFSS response rate of only 52.7% (range, 45.1%‒61% among the 5 states in our analysis).37 The accuracy of responding to the questions regarding Pap testing also may introduce bias, either by social desirability pressure or poor knowledge of Pap testing.38 Other factors not accounted for in this analysis may also contribute to some of the differences observed in the current study, such as rates of follow-up for abnormal Pap results, exposure to other ICC risk factors, or interaction effects not revealed by this analysis.39–41 Finally, the distribution of the black population across the study area differed, suggesting there may be other socioeconomic, lifestyle, or health access factors underlying the observed differences in ICC rates. AL as a whole has a higher poverty level than KY, PA, and OH, and it has a much greater percentage of blacks than the other 4 states, particularly when compared with KY and WV. This contrast in race distribution is even greater across the state-specific Appalachian regions.
In conclusion, we found variability in the risk of ICC across the 5 Appalachian states included in the current analysis, which stretched from PA in the north to AL in the south, with rates found to be substantially higher in KY and WV and lower in OH, PA, and AL. We also observed variations across the Appalachian-only counties, with the highest rate found in KY and the lowest in PA. The variations across the subregions appeared to reflect the socioeconomic differences of their population, which in turn may be at least partly explained by differing prevalence of ICC risk factors, such as smoking and screening practices.