Introduction

Prostate cancer is a very common type of cancer in the male population. In 2018, the global number of cases and deaths from prostate cancer were 1,276,106 and 358,989, respectively [1]. Although several risk factors for the development and progression of prostate cancer has been identified, such as family history and genetic factors [2], advanced age [2], and ethnicity [2], these factors are not modifiable. Previous studies have also revealed potential modifiable risk factors, such as obesity [3], dietary factors [4], and physical inactivity [5]. Moreover, in recent years, there has been an increasing interest in the contextual (neighborhood) effect on prostate cancer [6, 7].

Neighborhood social capital is considered to be a neighborhood feature that may be modifiable and high neighborhood social capital could thus promote good health [8] and potentially reduce social disparities in prostate cancer. Neighborhood social capital has been frequently operationalized as a collective dimension of society that is external from an individual [8], and it is established through social relationships that can improve the efficiency of society by facilitating coordinated actions [9]. Neighborhood social capital has three perspectives: linking, bonding, and bridging social capital [10]. Few studies have examined associations between low neighborhood linking social capital and cancer [11, 12]; in addition, to our knowledge, no large-scale follow-up study has examined the potential effect of neighborhood linking social capital on prostate cancer.

Several studies have suggested potential mechanisms relating neighborhood social capital and individual health: the diffusion of knowledge on health promotion, the maintenance of healthy behavioral norms through informal social control, and psychological processes that provide effective support [8, 13, 14]. Therefore, we hypothesized that lower levels of neighborhood linking social capital may be associated with a greater incidence of prostate cancer among Swedish men aged 50 years and older. Besides, we investigated whether lower levels of neighborhood linking social capital are associated with higher mortality of patients with prostate cancer.

Methods

Data sources

This cohort study started on January 1, 2002 and proceeded until the first event of prostate cancer, death from prostate cancer, death from any other cause, emigration, or the end of the study period on December 31, 2015.

Data used in this study were retrieved from several national registers that contain information on the entire Swedish population. We used datasets containing nationwide individual and neighborhood information, including comprehensive demographic and socioeconomic data. This study used the Total Population Register, the Hospital Discharge Register (1964–2015), the Outpatient Register (2001–2015), the Cancer Register (1958–2015), and the National Registry of Causes of Death (1961–2015); the latter was used to identify the date and cause of death. Individuals were tracked using personal identification numbers, which are assigned to each resident of Sweden. Each personal identification number was replaced with a serial number to ensure the confidentiality of all individuals.

All individuals were geocoded to their neighborhoods of residence for the assessments of linking social capital, and small area market statistics (SAMS) were used to define neighborhoods. The SAMS boundaries, which are small administrative areas in Sweden with an average population of about 1000 residents, were drawn to include similar types of housing construction in a neighborhood. We included 7264 SAMS units in the present study.

We excluded individuals with unknown neighborhood information (n = 5,846, 0.5%) and who were diagnosed with prostate cancer between January 1, 1999 and December 31, 2001 (n = 14,729, 1.2%) in order to “wash out” individuals with prevalent prostate cancer. In total, 1,196,563 men aged ≥ 50 years were included in the analyses.

Neighborhood linking social capital

Linking social capital refers to connections between individuals/groups who interact across explicit power or authority gradients in society [10, 13, 15]. A recent review found that linking social capital can be assessed by voting and trust in legal, political, or government institutions [16]. Several studies conducted in Sweden used voting rates in local government elections as a proxy of neighborhood linking social capital [11, 12, 17,18,19]. Voting during local government elections is believed to be a relatively stable variable over time, and the participation rates in local government elections can be considered a good indicator of neighborhood linking social capital.

Neighborhood linking social capital was conceptualized as the number of people in the neighborhood (SAMS) who voted in local government elections divided by the number of people in the neighborhood who were entitled to vote at baseline (2002). Neighborhoods were divided into the following three groups based on the proportions of residents who voted: (1) low, (2) intermediate, and (3) high. Group 1 comprised 20% of the entire Swedish population living in neighborhoods with the lowest proportions of voters (≤ 74.0%). Group 2 comprised 60% of the entire Swedish population living in neighborhoods with intermediate proportions of voters (74.1–82.0%), and group 3 comprised 20% of the entire Swedish population living in neighborhoods with the highest proportions of voters (> 82.0%).

Outcome variables

The outcome variables in this study were new cases of prostate cancer (yes/no) and deaths in patients with prostate cancer (yes/no), respectively. We used the Swedish Cancer Registry to identify the primary diagnoses of prostate cancer in the study population. This information was then linked to the records in the Cause of Death Register to identify deaths among patients with prostate cancer during the study period. All cancer cases in Sweden must be registered in the Swedish Cancer Registry. The completeness of cancer registration is currently close to 100%. Only primary neoplasms of the prostate classified according to the 7th revision of the International Classification of Diseases (ICD-7) were studied. The Swedish Cancer Registry has transferred all the cancer ICD codes into ICD-7 codes; in this study, code 177 was used. The outcome variable, i.e., mortality due to prostate cancer in the cause of death register was defined according to ICD-10 (codes C61).

Independent variables

The independent variables were age at baseline, marital status, family income, educational attainment, immigration status, geographical region of residence, mobility, diagnosis of chronic obstructive pulmonary disease (COPD) as a proxy for smoking, alcoholism or alcohol-related liver disease, coronary heart disease, hypertension, diabetes, obesity, and tobacco smoking.

Age

The participants were 50 years and older at baseline.

Marital status

Participants were classified as married/cohabiting or single (including divorced and widows/widowers).

Family income

Information on family income in 2002 came from the Total Population Register, which was provided to us by Statistics Sweden. We used this information to determine the distribution of family income in Sweden and then used the distribution to calculate empirical quartiles.

Educational attainment

Participants were classified based on completion of compulsory school or less (≤ 9 years), practical high school or some theoretical high school (10–12 years), or completed theoretical high school and/or college (> 12 years).

Immigration status

Individuals were born in or outside Sweden.

Geographical region of residence

Individuals were classified as living in a large city, Southern Sweden, or Northern Sweden.

Mobility

This variable was defined as the length of time the individuals had lived in the neighborhood, i.e., < 5 or ≥ 5 years).

Comorbidities

Individuals with a history of COPD, which was used as a proxy for smoking, were identified in the Hospital Registry and Outpatient Register during the follow-up period according to the corresponding ICD codes (ICD-10, J40–J49). Individuals with a history of alcoholism or alcohol-related liver disease were identified according to ICD-10, codes F10 and K70. The rest of the comorbidities were identified as follows: coronary heart disease (ICD-10, I20-I25), diabetes (ICD-10, E10-E14), hypertension (ICD-10 I10-I19), obesity (ICD-10, E65-E68), and tobacco smoking (ICD-10, F17, T65.2, Z71.6, Z72.0).

Statistical analyses

Multilevel logistic regression models were used to estimate odds ratios (ORs) with 95% confidence intervals (95% CIs). In this study, multi-level Cox proportional hazards models were not used, because the extensive data set was too large to run on available software. However, multilevel logistic regression models are a good approximation of Cox proportional hazards models under circumstances such as ours, i.e., a large sample size, low incidence, risk ratios of moderate size, and a relatively short follow-up [20]. The analyses were performed using MLwiN version 3.02 (University of Bristol, Bristol, UK). Random intercept multilevel logistic regression models were used to allow for the clustering of individuals within neighborhoods and to estimate the variance in the risk for prostate cancer that is attributable to neighborhood characteristics. This approach was used to estimate the intraclass correlation coefficient (ICC), thereby determining and comparing the proportion of variance in the outcome attributable to the differences between the individuals in different and same neighborhoods [21]. The ICC was estimated using the latent variable method as exemplified by the formula

$$\text{ICC}\, = \,\text{V}_{\text{n}} /\left( {\text{V}_{\text{n}} \, + \,\uppi ^{2} /3} \right),$$

where Vn is the variance between neighborhoods and π2/3 is the estimated variance between individuals. The proportion of the second-level variance explained by different variables was calculated as

$$\text{V}_{\text{EXPLAINED}} \, = \,\left( {\text{V}_{0} \, - \,\text{V}_{1} } \right)/\text{V}_{0} \, \times \,100,$$

where V0 is the second-level variance in the initial model and V1 is the second-level variance in the different models.

Results

The characteristics of the study participants are presented in Table 1. Among the 1,196,563 Swedish men aged 50 years and older, the cumulative incidence (per 100) and cumulative mortality (per 100) from prostate cancer were 8.22 (95% CI 8.17–8.27) and 1.80 (95% CI 1.78–1-83), respectively, during the follow-up period.

Table 1 Distribution of prostate cancer events, cumulative incidence (per 100), prostate cancer deaths and cumulative mortality (per 100) of prostate cancer, N = 1,196,563
Full size table

The association between neighborhood linking social capital and prostate cancer incidence is presented in Table 2. A significant association was observed between neighborhood linking social capital and prostate cancer incidence, as individuals living in neighborhoods with low linking social capital were less likely to be diagnosed with prostate cancer than those living in neighborhoods with high linking social capital, after adjusting for potential confounders (Model 4: OR 0.90; 95% CI 0.88–0.93). Similarly, individuals living in neighborhoods with intermediate linking social capital were also less likely to be diagnosed with prostate cancer (Model 4: OR 0.94; 95% CI 0.92–0.96).

Table 2 Odds ratios and 95% confidence intervals for cumulative incidence (per 100) of prostate cancer
Full size table

We also performed additional analyses to examine the association between neighborhood linking social capital and screening for malignant prostate neoplasms (Table 3). After adjusting for potential confounders, individuals living in neighborhoods with low linking social capital were less likely to be screened compared with those living in neighborhoods with high linking social capital (Model 4: OR 0.71; 95% CI 0.59–0.86). Similarly, a significant association was observed between intermediate neighborhood linking social capital and screening for malignant prostate neoplasms (Model 4: OR 0.82; 95% CI 0.70–0.96).

Table 3 Odds ratios and 95% confidence intervals for encounter for malignant prostate neoplasm screening
Full size table

The association between neighborhood linking social capital and mortality in patients with prostate cancer is presented in Table 4. After adjusting for potential confounding factors, patients with prostate cancer living in neighborhoods with low (Model 4: OR 1.15; 95% CI 1.08–1.23) and intermediate (Model 4: OR 1.09; 95% CI 1.03–1.14) linking social capital were more likely to die from prostate cancer compared with those living in neighborhoods with high linking social capital.

Table 4 Odds ratios and 95% confidence intervals for cumulative mortality (per 100) of patients with prostate cancer
Full size table

Discussion

We found that low and intermediate levels of neighborhood linking social capital were associated with lower odds of prostate cancer incidence than the reference category (high linking social capital). However, the lower categories of neighborhood linking social capital were associated with higher mortality in patients with prostate cancer. These inconsistent findings could be explained by differences in health checkup attendance. Our study suggests that those who lived in neighborhoods with lower social capital could have been less likely to undergo cancer screening for malignant prostate neoplasms. Thus, the incidence of prostate cancer in neighborhoods with lower social capital might have been an underestimation of the “true” incidence. Therefore, more efforts may be needed to increase prostate cancer screening in men living in neighborhoods with lower linking social capital.

Although the causal mechanisms cannot be understood, several possible explanations behind our findings could be suggested. One possible explanation is that individuals living in higher social capital neighborhoods could encourage others to take part in health checkups more easily due to close social networks within such neighborhoods [8]. For example, a previous study conducted in the United States found that higher social capital was associated with greater adherence to clinical breast examination and mammography screening [22]. In addition, a study conducted in Denmark found that a higher level of neighborhood social capital was associated with a higher probability of participating in the health checkup phase of a population-based lifestyle intervention [23].

Another possible explanation is that individuals living in higher social capital neighborhoods are more likely to maintain social order when they witness deviant behavior [8]. This hypothesis is based on a theory on the occurrence of vandalism [24] but is equally applicable and relevant to the prevention of underage smoking and alcoholism [8]. Studies examining the association between lifestyle-related factors, such as smoking and alcoholism, and prostate cancer have so far found inconsistent results [2]; therefore, further analyses which include interaction terms between social capital and lifestyle-related factors may be worthwhile.

Our results also showed that lower socioeconomic status (SES), i.e., income and educational attainment, were associated with a lower incidence of prostate cancer. A diagnosis of prostate cancer is highly influenced by the availability of prostate cancer screening via, e.g., Prostate-Specific Antigen testing [25]. Considering these results, both lower neighborhood linking social capital and individual SES have independent effects on prostate cancer risk; hence, better social networks among medically underserved populations may be helpful in preventing prostate cancer in Sweden.

Our study findings have important implications for further research on social capital and prostate cancer; nonetheless, it had some limitations. First, our results could be influenced by other unmeasured risk factors, such as body mass index (BMI). However, we attempted to adjust for BMI by analyzing hospitalization for obesity. Second, the modifiable area unit problem has been suggested as a potential limitation when using aggregated data [26]. However, the SAMS neighborhoods in our study included similar types of buildings, which imply that SAMS neighborhoods are comparatively homogeneous in Sweden. Third, this study did not consider changes of residence during the follow-up. However, we included mobility before baseline to reduce possible selection bias.

Our study also has many strengths. This large cohort included 1,196,563 men aged over 50 years in Sweden, which increased the generalizability of our results. Moreover, we could eliminate spurious associations due to same-source bias because our neighborhood- and individual-level variables were obtained from different sources. Finally, the prospective design of this study may, in part, reflect some causality.

Conclusions

Our findings show a lower incidence of and higher mortality from prostate cancer in neighborhoods with lower levels of linking social capital in addition to a lower screening. These findings suggest that improved screening efforts, as well as better social networks, may be needed in neighborhoods with lower levels of linking social capital to decrease health disparities in prostate cancer among men.