SEP 1 municipalities were concentrated in the central region of Puerto Rico, and the SEP 5 level municipalities were concentrated in and around the San Juan metropolitan area (Figure). The municipalities with the lowest SEP were Maricao, Guánica, and Las Marías, and the municipalities with the highest SEP were Guaynabo, Trujillo Alto, and Carolina.

Men

During 2000-2004, the incidence of cancer for all sites among men was significantly lower in SEP 1 areas than in the SEP 5 areas (Table 1). Incidence of esophagus and stomach cancers were higher in SEP 1 than in SEP 5. During 2000-2004, incidence of the following cancers was lower in SEP 1 than in SEP 5: colorectal, kidney and renal pelvis, liver and intrahepatic bile duct, lung and bronchus, non-Hodgkin lymphoma, and prostate; a similar pattern was observed in the earlier time period (Table 1).

Women

During 2000-2004, the incidence of cancer for all sites among women was significantly lower in SEP 1 areas than in SEP 5 areas (Table 2). The incidence of esophagus and stomach cancers were higher in SEP 1 than in SEP 5 for the earlier period only. During 2000-2004, the incidence of the following cancers was lower in SEP 1 than in SEP 5: breast, colorectal, non-Hodgkin lymphoma, and thyroid; a similar pattern was observed in the earlier time period.

Men

During 2000-2004, there was no difference in all-site cancer mortality among men in SEP 1 areas and those in SEP 5 areas (Table 3). Mortality due to esophagus and stomach cancers was higher in SEP 1 than in SEP 5. During 2000-2004, mortality was lower for the following cancers in SEP 1 than in SEP 5: colorectal, liver and intrahepatic bile duct, and pancreas. A similar pattern was observed in the earlier time period for colorectal and liver and intrahepatic bile duct cancers.

Women

During 2000-2004, the cancer mortality for all sites among women was significantly lower in SEP 1 areas than in SEP 5 areas (Table 4). During 2000-2004, mortality for breast cancer was lower in the SEP 1 than the SEP 5 areas; similar patterns were observed in the earlier time period.

Our study shows that more deprived municipalities are concentrated in the central region of the Island, and less deprived municipalities are concentrated in the San Juan metropolitan area. People living in the mountainous center of Puerto Rico may lack access to health services because of economic, cultural, environmental, social, and physical barriers (7,15-17). In Puerto Rico, as in the continental United States, one of the most important determinants of access to medical care is health insurance coverage (18). Puerto Rico has a population of 3.9 million people, and approximately 91.6% are insured by public or private health insurance. Approximately 8.4% of the population is uninsured, similar to estimates of the proportion of uninsured people in United States (19). Two health care systems coexist in Puerto Rico: the private system consisting of private health insurers and Medicare (parts A and B), and the public system (Reforma) that serves more than half of the population. Although Reforma is administered by a single government agency, different insurance companies serve the 8 health regions of Puerto Rico, creating a heterogeneous model of health services. The private sector is the largest provider of services in Puerto Rico, although the government is responsible for most high-risk cases because most tertiary care services are provided by the state through the facilities of the Río Piedras Medical Center. According to the Registry of Hospitals and Health Facilities for 2002-2004, Puerto Rico has 67 hospitals, 16.7% of which are public. The distribution of hospitals varies by health region, and most are concentrated in the San Juan metropolitan region (20), an area that has high SEP. Thus, differences in access to care throughout Puerto Rico may also explain some of the differences observed in the incidence and mortality of cancer by SEP area.

We observed a positive socioeconomic gradient for overall cancer incidence, with higher cancer rates in SEP 5, and a difference of approximately 13% between high and low SEP levels for 2000-2004. This difference could be explained by the lack of access to and use of medical care in the SEP 1 areas. Most of the SEP 1 areas are in the central region of Puerto Rico, where fewer clinical facilities are available (16). Thus, the lower incidence of cancer in these regions could be partially explained by underdiagnosis.

Consistent with patterns in the incidence of all sites, incidence of the most common cancer types in Puerto Rico (breast, colorectal, and prostate) was higher in the SEP 5 areas, where most mammography facilities, urologists, and gastroenterologists are clustered (16,21). Given that these cancers are susceptible to screening and thus to potential overdiagnosis (21,22) in regions with better access to health care facilities, the proportion of cases underdiagnosed may be lower than in regions with poorer access to care (21,22). Future studies should explore whether the clustering of health care facilities in Puerto Rico influences disease detection in this population.

The other cancer types that showed higher incidence in SEP 5 areas were kidney, liver, lung in men, thyroid in women, and lymphoma. Differences in the occurrence of risk factors for these cancer types between the SEP levels could partially explain this pattern (23,24). Contrary to our data, according to the 2002 Behavioral Risk Factor Surveillance System (BRFSS) in Puerto Rico, the prevalence of smoking was higher among residents with lower incomes and lower levels of education, as in the continental United States. Nonetheless, income is only one of the variables included in our SEP index, and thus, these patterns may not be comparable. More research is necessary to evaluate whether in Puerto Rico SEP influences smoking behavior and other behavioral risk factors.

Higher rates of liver cancer among men in the SEP 5 areas may reflect the higher prevalence of alcohol consumption in this group. In a recent study in Puerto Rico, the prevalence of hepatitis C virus infection did not vary according to education and annual family income (25,26). In addition, the incidence of 1 cancer associated with human papillomavirus, penile cancer, was higher in the SEP 1, although this difference was significant only during 1 period. This finding is consistent with those of some studies that show that people who live in counties with higher poverty have higher rates of penile cancer (27).

Overall cancer mortality did not differ greatly by SEP, but mortality for specific cancer sites did. In general, we found higher mortality for stomach cancer for both sexes, and for esophagus cancer in men, in SEP 1 than in SEP 5. This result is consistent with those of other studies that show that mortality for stomach and esophagus cancers is high in areas with low socioeconomic status (7). The principal infectious agent associated with stomach cancer morbidity and mortality worldwide is Helicobacter pylori (25,28). This infection is associated with socioeconomics factors such as low income and education level (25,28). For esophagus cancer, the higher prevalence of behavioral risk factors in the SEP 1, including tobacco use, consumption of hard liquor, and a deficient diet (29), could partially explain the higher mortality. For example, BRFSS data indicate that during 2004 in Puerto Rico, the prevalence of current smoking was higher and the consumption of fruits and vegetables was lower in people with lower income and lower educational attainment (19).

This study has several limitations. Puerto Rico cancer data do not permit analysis for smaller geographic areas, such as subcounty level, census tract, or block groups, because the address information is not available at this level for most of the cases. However, municipality-level data provide an appropriate socioeconomic and political context for the formulation and implementation of public health and social policies. Area-based socioeconomic variations in cancer incidence and mortality cannot be considered as proxies for socioeconomic differentials at the individual level. Such consideration may lead to the ecological fallacy (7). Whereas area socioeconomic patterns for several of the cancer outcomes are generally consistent with those at the individual level, the area-level effects we found may differ in magnitude from individual socioeconomic effects. These differences may be partly due to the compositional heterogeneity of the areas examined, particularly municipalities, which, unlike subcounty levels or census tracts, may contain substantial socioeconomic variability (2,7).

In conclusion, this study identifies disparities in cancer incidence and mortality that could be due to health care access and utilization rather than actual disease incidence; our findings warrant further investigation.