Introduction

Carcinoids are neuroendocrine tumors with the majority occurring in the gastrointestinal tract (67%). Incidence of carcinoid tumors is about 2.47 (0.87 for small intestine alone) per 100,000 population per year with the incidence having increased 3-fold over the past 40 years [1, 2]. Multiple factors likely contribute to the rise in incidence of small intestinal neuroendocrine (carcinoid) tumors (SINT) including better and more accurate detection, tobacco use, and changes in environmental exposures [3–6]. Tobacco use has declined since its peak in the 1960’s, however the lag time from tobacco exposure to cancer later in life is consistent with the recent rise in incidence of SINT [7]. Although SINT are slow growing tumors, they are often not diagnosed until advanced stage when prognosis is dramatically diminished. Advances in imaging have allowed for earlier detection, however, the asymptomatic and rare nature of this tumor makes broad screening impractical. The key to reversing the trend is to better understand risk factors, mitigate these risk factors, identify individuals at risk, and implement early diagnosis tools.

Because this is a rare cancer, epidemiologic studies have been limited. Carcinoid tumors comprise approximately 40% of all small intestinal primary tumors and are similar in frequency to adenocarcinomas of the small intestine [8–10]. Studies, including our recent publication, have consistently shown a high relative risk of 6- to 11-fold in first-degree relatives, with siblings having twice the relative risk of parents [9, 11, 12]. The increase in concordance between siblings versus parents suggests that some aspect of a shared environment compounds the risk of developing SINT. Cigarette smoking and alcohol have been associated with an elevated risk of SINT, but reports are varied. In a meta-analysis including patients from the U.S.A. and Europe, an odds ratio of 1.40 (1.06–1.86) was observed for smoking and 1.04 (0.63–1.72) for alcohol [6]. In Utah, we observed an odds ratio of 1.44 (1.11–1.86) for smoking and 1.62 (1.05–2.49) for alcohol and risk of SINT [13]. Utah has one of the lowest smoking and alcohol consumption rates in the United States due to the majority of the population being members of the Church of Jesus Christ of Latter-Day Saints (a.k.a. Mormon) which has proscriptions against drinking and smoking, however the rising incidence of SINT are comparable to national reports [1, 14], suggesting other risk factors are involved. Environmental exposure to carcinogens are considered to be a driving force in increased incidence of some cancers such as SINT [15].

A recent report using data from SEER suggests that the increasing incidence of colorectal cancer in younger adults is consistent with a “birth cohort effect and points to early life exposures, accumulated throughout the life course—that may increase cancer risk” [16]. Similar to the increasing incidence of colorectal cancer in younger adults, SEER data (1975–2016) shows a 3.4 annual percent increase in SINT incidence under age 50y (SEER*Stat 8.3.6). However, diagnoses of SINT over age 50y are also increasing (2.4 annual percent increase) while colorectal cancers in this age group have been decreasing. One of the more dramatic examples of early life exposure and elevated risk of solid cancers is a study of atomic bomb survivors who were exposed in utero or as young children (0–5 years) [17]. One of the challenges of studying the effects of early life conditions on subsequent adult cancer risks is the lack of location and environmental data during the relevant exposure period. For example, all SINT cases diagnosed in Utah were born before 1982 with 57% having been born before 1940. Data resources at the University of Utah offer a unique opportunity to explore the relationship between early life environmental exposures and subsequent cancer risks. Multiple studies have used these state-wide resources to consider early life conditions including socioeconomic status, parental mortality, religion and birth order on long-term health outcomes including cancer, mortality and systemic inflammation [18–20]. Early life exposures based on place of birth has not, to our knowledge, been explored using the Utah state-wide resources.

Extensive data is available through the Utah Population Database (UPDB), a state-wide resource containing all individuals with a life event in Utah (e.g. birth, marriage, death, driver license renewal or hospitalization) with records linked to statewide cancer data [21, 22].

The objective of this study was to test for the occurrence and location(s) of SINT clusters based on birth location and assess the associations between early life environment exposures and SINT cases. We examined geographic clustering of SINT cases based on birth locale and exposure based on industries present at the time and place of birth.

Materials and methods

Study was approved by the Institutional Review Board of University of Utah and the Utah Resource for Genetic and Epidemiologic Research Board (the governing body of UPDB use).

Exposure assignment based on birth locale history

The relationship between early life exposures and the risk of SINT was examined using a case-control design as described above. Time-varying historical profiles, covering the time span of births (1883 to 1982) were constructed for each of the 405 birth locales. These profiles included information on potential exposures such as the major manufacturing and industrial operations, mining and smelting operations, and transportation of coal and ore. Agricultural activities (i.e. orchards, truck crops, dry farming and livestock) and settlement size and connectedness to major metropolitan areas were also determined for each birth locale throughout the range of birth years. City was designed to capture whether the subject was born in a significant economic or political settlement of the area for the time period of the birth. Rural was designed to capture areas that were not in proximity to the major metropolitan centers. Locales that are neither city nor rural are small towns that while not contiguous with a metropolitan area, are also not isolated from an urban area. This information was abstracted from multiple historical documents including the Utah Centennial County History Series, a set of comprehensive histories for each county, mining and railroad histories, and US Environmental Protection Agency documentation [28, 29, 35–39]. For each of these activities we used the data in the historical documents to estimate the year that the activity began and the year it ceased. In most cases, the birth locales were relatively small and documents about the history of the area clearly described the major industries and economic activities along with a general description of the years these activities began and ended. We also created a comprehensive georeferenced database of all major mines and smelters and used this to identify birth locales within these industries.

Birth locale profiles included 16 dichotomous exposure indicator variables denoting the presence of each potential exposure or characteristic with starting and ending year noted. The exposures and definitions are described in Table 1 and detailed information regarding the sources of historical information are presented in S1 Table. A birth locale could have more than one of these exposure indicators at any point in time, and the exposures attributed to a locale varied over time. The exposure indicators were assigned to cases and controls based on their birth locale and year.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Industries, agriculture and settlement type assigned to birth locale.

https://doi.org/10.1371/journal.pone.0231991.t001

Conditional logistic regression was used to estimate both crude odd ratios (OR) and adjusted odds ratios (aOR) with adjustment for indication of smoking and alcohol consumption and family history of SINT. The analysis was stratified based on three historic time periods, 1883 to 1929 (Period 1), 1930 to 1944 (Period 2) and 1945 to 1982 (Period 3) to account for changes over time in potential exposures associated with a specific industry. For example, worker or community exposures associated with coal mining may have changed due to improvements in occupational safety practices or extraction and processing improvements. The time periods were defined based on major economic transitions that affected many industrial processes, namely the Great Depression and the end of World War II. Two cases and seventeen controls were excluded from the adjusted analysis due to lack of family history, defined as having no second-degree relatives in UPDB. Given the small sample size and exploratory nature of this analysis we did not adjust for multiple comparisons.

Results

A birth locale was determined for 390 cases and 3710 matched controls (Table 2). Race, indication of tobacco use, and indication of alcohol use have similar proportions between cases and controls. As expected, a greater proportion of cases had a family history of SINT. In general, the distribution of tumor location and stage are similar to our Utah data of all cases (n = 745), including those born outside of Utah as well as national SEER data. The mean and median age of diagnosis in this study was two years younger than reported for national SEER data, but identical to our overall Utah data [1]. Tumor grade was reported on only 29% of the cases, with the majority being well differentiated.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Demographics of cases and controls.

https://doi.org/10.1371/journal.pone.0231991.t002

For our spatial analysis, 405 birth locales were used representing all cases and controls born in Utah from 1883 until 1982. Most of the locales (60.5%) were the birth place of a single SINT case. Three locales, Salt Lake City (neighborhood undetermined, n = 54), Ogden (n = 21), and Logan (n = 12) were the birth place for 22% (87/390) of the SINT cases.

One significant isotonic (spatial) cluster occurred in Utah. The isotonic cluster has three steps as well as an overall relative risk (RR) for the whole cluster as shown in Fig 1. For each step, the RR is the risk within the cluster step as compared to the area outside of the cluster, i.e., the rest of Utah. The first step consists of one birth locale containing four cases, and is at the center of the cluster with a RR of 8.58. The relative risk was attenuated in the second step of the isotonic cluster, which had an RR of 6.31 and included four locales and three additional cases. In Step 3, the RR decreased to 1.53 and contained eight locales and five additional cases. The overall RR of the isotonic cluster was 2.86 and had the same areal extent as Step 3 (Fig 1). Twelve SINT cases were born in this geographic area, whereas 4.28 cases would be expected (p = 0.016). Interestingly, all of the cases were born prior to 1940 (1897–1935). As shown in Fig 1, this region was historically a coal mining and shipping region with little agriculture or manufacturing [35].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Geographic analysis of SINT cases and controls.

The isotonic cluster signal and steps. The overall cluster has a 2.86 relative risk of SINT as shown in dark red. Overlaid on the Utah map are study defined birth locales for Utah (green circles) and the percent of currently abandoned coal mines by county.

https://doi.org/10.1371/journal.pone.0231991.g001

The association of environmental characteristics with SINT was investigated using a time-stratified case-control analysis of 16 exposure categories of the birth locale (Table 3). Those born in a city with a large construction project during Period 1 (1883–1929) had a significantly greater risk of SINT (OR = 1.98, 95% confidence interval (CI) 1.09 to 3.60, p = 0.024). During Period 1 there was also a suggestive but not significant increases in odds of SINT cases being born in a locale with mining activities, including coal mining (1.95 with 95% CI of 0.97 to 3.92, p = 0.061), hard rock mining (1.38 with 95% CI of 0.97 to 1.97, p = 0.075), and uranium mining (2.70 with 95% CI of 0.86 to 8.48, p = 0.088). An elevated, but not significant, odds ratio was found for birth locales near orchards in Period 2, 1930 to 1944, a time when chemical pesticides were introduced (3.94 with 95% CI of 0.76 to 20.31, p = 0.101). A significantly lower odds ratio was found for individuals born in a rural setting between 1945 and 1982 (OR = 0.58 with 95% CI = 0.36 to 0.96, p = 0.033). Adjusting for family history of SINT, alcohol consumption and smoking did not significantly alter these results (S2 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Odds ratios for SINT from exposure to industries and settlements.

https://doi.org/10.1371/journal.pone.0231991.t003

Discussion

There is substantial interest and little information on the impact of early life or parental environmental exposures on adult health [15]. Early life exposures have been linked to the risk of subsequent development of prostate cancer and there is evidence that this may be related to epigenetic effects of environmental exposures [40]. Assessing early life impacts on rare cancers, such as SINT, is difficult as retrospective exposure data are needed for sufficient numbers of cases. Often, there is limited data on potential exposures during early life due to a lack of information on birth location and/or environmental conditions at those locations many decades into the past. Using a unique data resource, the UPDB, and historical information to reconstruct potential exposures we explored the link between early life exposures and the risk of developing SINT later in life. We found statistical evidence of clustering based on birth location, in locales coincident with areas where coal mining was the primary economic activity at the time of birth. Coal mining began in Utah in the 1850s and this region was historically dominated by coal mining and shipping with little agriculture or manufacturing [35]. Many mining towns were located at the mine site. These coal-company towns were slowly abandoned post-1942 as miners were no longer required to live in company housing and automobiles became more commonplace for commuting from nearby towns. From the 1950s to 1970s, many of the small coal mines in this area closed, as shown in Fig 1, and the population moved to more urban areas. As noted in the results, all of the cases in this spatial cluster were born before 1940. Additionally, the case-control analysis did find suggestive evidence of an elevated risk among individuals born in locales with coal, hard rock and uranium mining during Period 1 (1883–1929) when extraction practices would have led to high community exposures. Prior geospatial analyses have found residential proximity to coal mines in Illinois and Kentucky a risk factor for multiple cancers most notably colorectal and lung [41, 42]. These studies, however are not based on birth location, but residence at time of diagnosis. Although we speculate that coal mining was driving the observed cluster signal, other coincident factors could have been associated with elevated risk. A study from Canada found a higher incidence of neuroendocrine tumors in rural areas versus urban areas of residence [43]. The study’s average age at diagnosis was 61 years which would translate to approximate birth years of 1936–1950. Our data on rural settlements suggests that the time period 1930–1944 had slightly elevated odds ratio (1.13) but this was not significant. An interesting parallel is our observation of a 3.9 increased odds of SINT in the orchard industry for this same second period which might point to the advent of chemical pesticides.

Limitations of this study were our use of a single residential location to characterize early life exposures, determination of historical community exposures, multiple hypothesis testing issues, and small sample size. Place of birth has benefits in that it is often the residence throughout childhood, so it is a reasonable snapshot of early life exposures. While ascertaining potential exposures in the early part of the twentieth century is challenging, several factors suggest that exposures may have been more homogenous and consistent over time. For example, consistent residential mobility was generally less than the present day, settlements had only one or two primary economic activities (e.g., mining or manufacturing), and these activities did not change rapidly over short periods of time. Future studies may be able to use complete residential histories made possible by the UPDB to characterize exposures over the life course as only a small proportion of all available addresses (11%) are attributable to birth certificates. We were also concerned that relatedness of cases, a known risk factor, could overlap with geographic clustering. When analysis of exposure was adjusted for family history of SINT there were no significant changes, suggesting that the risk exposures we identified were independent of familial risk. Although exposure analysis was adjusted for indication of tobacco use, data on tobacco use prior to 1996 was limited to contributing cause of death information recorded on a Utah death certificate. Thus, it is likely that for some subjects, indication of tobacco use was misclassified, particularly for case and control subjects who died before 1996 in which exposure data were limited.

Our study applied a novel set of methods to use in-depth historical research to characterize the early life environment of cases and controls. Because birth certificate records are a comprehensive resource across time, very few of these rare cancers were excluded due to a lack of birth locale. Future studies that capture the cumulative lifetime exposures for an individual, while complex, could provide valuable insights but would be particularly challenging for studies of rarer cancers that may lack sufficient numbers. Unlike birth certificates, other sources of residential history are comparatively incomplete and are limited in scope in terms of years covered, thus many cases would likely be excluded. The methods used here, along with cumulative exposures, could practically be applied to the more common cancers or in large populations of rarer cancers, where numbers would be sufficient for study.

Conclusion

Using cases and birth cohort controls, we report geographic clustering of a birth locale associated with greater than 2-fold elevated risk of SINT, located central to historic coal mining communities. A retrospective analysis of industry and mining exposures further suggest birth/early life residence near construction industry, uranium mining and coal mining confers an elevated risk, specifically during the earliest time period when there were no regulations to reduce exposures to workers or the public.

Supporting information

Acknowledgments

We thank the University of Utah Center for Clinical and Translational Science (funded by NIH Clinical and Translational Science Awards), the Pedigree and Population Resource, University of Utah Information Technology Services and Biomedical Informatics Core for establishing the Master Subject Index between the Utah Population Database, the University of Utah Health Sciences Center and Intermountain Health Care.