Introduction

The proportion of colorectal cancer (CRC) in persons under the age of 50 years, known as early-onset CRC (EOCRC), has steadily increased from 6 to 11% of the total over the past 25 years in the USA, while the incidence of CRC has fallen in persons ≥ 50 years or age, late–onset CRC (LOCRC) [1,2,3]. The rising incidence in young adults has generated a recent recommendation from the American Cancer Society to screen average-risk adults for CRC beginning at age 45 years [4]. It is not known whether EOCRC is caused by the same factors that cause LOCRC or whether there are unique causes that alter the clinical features [5,6,7,8,9]. Although many distinct features of EOCRC are appreciated, a comprehensive approach to early diagnosis has been elusive. This review was designed to analyze the salient clinical and genetic characteristics of EOCRC and to seek clues to assist in risk management for EOCRC and explanations for the differences from LOCRC.

Clinical Distinctions Between EOCRC and LOCRC (Table 1)

Since screening is not routinely performed in people < 50 years old, 86–98% of these tumors are found because of alarm symptoms such as bleeding and abdominal pain. Additional important differences between EOCRC and LOCRC include that EOCRC is more likely to be located in the distal colon (Fig. 1) and to present more frequently at an advanced stage of disease, the latter of which is associated with a paradoxically shorter interval from the onset of symptoms to diagnosis [5,6,7,8,9,10]. Although persons with EOCRC respond with equal or better outcomes to stage-specific cancer treatment compared to that in LOCRC, the overall survival is worse, a feature attributed to the increased presentation at a later stage of disease [5, 9, 11].

Table 1 Comparison of characteristics between persons with early-onset versus late-onset, screen-detected and late-onset, symptom-detected colorectal cancer.
Full size table
Fig. 1
figure 1

Adapted from: Strum WB and Boland CR (2019, in press) [9]. Data not previously presented

Locations of cancers in EOCRC. The prevalence of colorectal cancer location by specific colon segments among persons with early onset is illustrated in the diagram and demonstrates the predominance of distal colon origins.

Full size image

Family History

A family history of CRC in persons with EOCRC is vital information, but has limitations in identifying persons at risk, a shortcoming also noted with LOCRC. Meta-analysis of family history data indicates that for all ages, a positive family history conveys an increased risk of up to fourfold depending upon the age of onset for the CRC and the number of relatives affected [12]. The most reliable results come from: (a) persons with a history of more than one first-degree relative (FDR) with CRC or (b) the onset of CRC in the FDR being < 60 years [13]. In the context of this meta-analysis and other reports, persons with EOCRC have at least one FDR with CRC between 7 and 26% of cases and at least one second-degree relative in 12–22% [6,7,8,9]. Using family history data, an estimate of the EOCRC population that would be appropriate for early screening at or before the age of 40 years is around 7% [12, 14]. Consequently, a family history of CRC in the EOCRC population, as in the LOCRC population, identifies a minority of persons with CRC [12, 14,15,16].

Genetic Distinctions Between EOCRC and LOCRC

A critical question is whether EOCRC is part of the continuum of CRC just occurring in younger people, or whether it includes groups of cancers that are biologically different from LOCRC. It has been repeatedly shown that no more than 15–20% of EOCRC tumors have microsatellite instability (MSI), reflecting Lynch syndrome and Lynch-like syndrome [17,18,19,20,21,22,23], which is not far from the 10–15% reported in studies of unselected cases of CRC [24,25,26], with minimal variation among various ethnic groups [27]. The older group of CRCs with MSI include a large proportion (~ 75%) that represent the non-familial sporadic methylation-induced silencing of the DNA mismatch repair (MMR) gene, MLH1 [26], whereas the younger population has relatively few of those, and more hereditary cancer accounting for the MSI. This difference is not great but tumors with MSI would predictably skew the outcome in EOCRC in a less aggressive direction and likely lead to tumors that are more proximal in location [28], neither of which is the case in EOCRC.

Germline Mutations in EOCRC

Several large retrospective studies of germline testing using multigene panels have demonstrated the genetic factors involved in CRC in persons under the age of 50 years [21,22,23]. A review of 193 persons under the age of 36 years, selected by having undergone genetic counseling at MD Anderson Cancer Center in Houston and tested for the known high-penetrance CRC genes, reported that 35% had a hereditary cancer syndrome, including 23 with Lynch syndrome, 22 with germline mutation-negative CRCs with defective DNA MMR activity (dMMR), 16 with familial adenomatous polyposis, two with constitutional MMR deficiency, two with biallelic MutYH mutations, and one with Li-Fraumeni syndrome. As expected, those with a hereditary syndrome were more likely to present at an earlier stage and to have a family history of CRC, although 19% of 67 patients with a documented germline mutation in a hereditary cancer gene did not have a family history of cancer [20].

A study from Ohio was published of 450 persons with EOCRC from 2013 to 2016 first screened for dDNA MMR activity, followed by germline testing using a 25-gene cancer panel. Forty-eight (10.7%) had dMMR activity, of whom 37 (8.2% of all persons evaluated and 77% of those with dMMR activity) had mutations in Lynch syndrome-associated genes. Analyses of the remaining 402 patients with MMR-proficient tumors uncovered 32 (8%) with one or more mutations in 9 high-penetrance CRC-associated genes, 13 high-moderate penetrance genes not traditionally associated with CRC, and 10 low-penetrance CRC-associated genes [21].

Four hundred and thirty persons with EOCRC from the clinical genetics service of a tertiary cancer center at the University of Michigan were studied using a variety of commercially available multigene cancer panels (containing up to 124 genes) or a custom-designed next-generation sequencing panel of 67 genes. The participants were selected after having undergone genetic counseling between 1998 and 2015; 111 (26%) had a FDR with CRC. Fewer than half of the tumors were screened for dMMR activity, and 41 of the tumors (10% of the entire cohort) had MSI [22]. Germline sequencing was performed in 315 persons and 79 (25%) had a germline mutation associated with hereditary cancer. An additional 21 persons had DNA variants of unknown significance. Fifty-six persons (17.8%) were identified as Lynch syndrome, 10 persons (3.2%) as FAP, and 13 persons (4.1%) had other cancer-associated genetic mutations. Of these 79 persons, about half had a FDR with CRC. Among 117 patients with uninformative clinical evaluations, next-generation sequence analysis using a multigene panel detected actionable germline variants in six patients (5%), increasing the total to 85 persons (19.8%) with a germline mutation associated with a hereditary cancer [22].

These studies of EOCRC indicate that about 10–20% of the tumors have dMMR activity. About half these are linked to germline mutations in DNA MMR genes (Lynch syndrome), and the other half have double somatic mutations but no known germline mutation in MMR genes [29, 30], often referred to as Lynch-like syndrome [31]. Another 2–3% are linked to FAP and its variants, and about 2–3% with other hereditary cancer genes not historically associated with CRC [16,17,18,19]. As mentioned, the presence of Lynch syndrome would neither tend to make the clinical outcomes worse or the tumors more distal, suggesting the problem is more complex than just the involvement of Lynch Syndrome.

Somatic Genetic Alterations in EOCRC

Genetic causes for the young age of onset in the remaining 80% of cases of EOCRC are not known. Explanations include the possibility that additional genetic and epigenetic alterations are involved in EOCRC but have not been identified. There is no evidence of global hypermethylation in EOCRC (such as would be seen in the CpG island methylator phenotype, or CIMP), but a subset of these tumors have LINE-1 hypomethylation that occurs in the promotors of LINE-1 sequences, which make up about 17% of the human genome. Expression of these sequences is normally suppressed by promoter methylation. The presence of LINE-1 hypomethylation is highly associated with EOCRC and associated with an appreciably worse clinical outcome [32]. In addition, there are intronic oncogenes (including MET, RAB3IP, CHRM3), located in some of the LINE-1 sequences, and these undergo simultaneous hypomethylation along with the LINE-1 promoters, resulting in re-expression of the oncogene mRNA and protein [33]. One report indicates that NOMO1 is deleted (somatically) in EOCRC [34]. A variety of changes in gene expression are reported in EOCRC, including a subset that are both diploid and microsatellite stable, but none explains the unique clinical features or provides a clue regarding a specific etiology of this variant of CRC [35,36,37,38]. The development of the CpG island methylator phenotype in sessile serrated adenomas is considered a sign of dysplastic progression, predominates in older populations, and might possibly be relevant for persons in the 40–49 year range since its appearance begins to accelerate at age 40 years [39].

Recommendations Regarding Genetic Testing in EOCRC

Although genetic screening of all persons with CRC is not yet advised on a routine basis, it is becoming more applicable for affected persons under the age of 50 years, and is certainly appropriate for persons with multiple FDRs with CRC or an FDR with CRC under the age of 60 years [20,21,22, 40]. A current debate involves whether gene testing should be done with a large comprehensive multiple cancer gene panel or a smaller CRC-specific panel, since the former may uncover 1–3% of individuals with mutations in a gene not linked to familial CRC, thus provoking interpretational challenges in clinical application [41]. The proportion of cases with difficult interpretations would not appear to be sufficient to offset the benefits of genetic screening for these populations. With time, genetic screening will likely become routine for most persons with CRC, as the costs of such testing have fallen substantially in recent years, making it cost-effective.

Possible Contributors to EOCRC

Aside from the obvious risk of CRC in persons with familial syndromes associated with CRC, additional factors and combination of factors contribute to risk including race, obesity, physical inactivity, diabetes, tobacco exposure, excess alcohol ingestion, and socioeconomic status [42] (also, Colorectal cancer facts & figures 2017–2019. Atlanta: American Cancer Society, 2018 [http://www.cancer.org/acs/ groups/content/documents/document/acspc-042280]). One study from a single center suggested a significant impact of EOCRC on non-Caucasians over Caucasians [9]. However, a recent large study drawn from the Surveillance, Epidemiology and End Results (SEER) program reported that, between the 1992–1996 and 2010–2014 time periods, CRC in people 20–49 years old rose with a more complex pattern. In that timeframe, the CRC incidence rose from 7.5 to 11.0/100,000 in whites, while it rose from 11.7 to 12.7/100,000 in blacks. The increase in incidence occurred disproportionately in the rectum in the white population [10]. Genetic disparities in CRC have a disproportionate impact on African-Americans where the younger age of onset is coupled with adverse tumor characteristics including a higher proportion of KRAS mutations, a lower prevalence of MSI, a higher prevalence of elevated microsatellite alterations at selected tetranucleotide repeats (EMAST), and decreased anti-tumor cytotoxic immunity, resulting in a poorer prognosis. The negative impact carries over further with insurance coverage being less or nonexistent for African-Americans [43, 44].

Obesity has a strong link to CRC. The relationship is stronger in men than women and for colon cancer over rectal cancer [45]. Data relating obesity and physical inactivity to CRC support a consistent demonstration that physical activity is associated with a preventive effect on colon, but not rectal, cancer [46]. The molecular mechanisms relating obesity to cancer are complex and thought to include chronic inflammation in visceral fat and insulin resistance which activate oxidative stress, adipocytokines (adiponectin and leptin), sex steroids, the microbiome, and other genetic mutations and environmental factors [45, 47].

Meta-analyses of diabetes, independent of obesity, physical activity and smoking, show a significant increase in risk of colon cancer for men (RR 1.43) and women (RR 1.35) and for rectal cancer in men (RR 1.22) [48]. Diabetes in the USA is estimated to be present in 1.2% of adults between 18 and 44 years as of 2015. San Diego County is considered to be affected at lower than average rates of obesity (National Diabetes Statistics Report 2017. Centers for Disease Control and Prevention [https:/www.cdc.gov/diabetes/statistics/ Accessed March 12, 2018]). A study of 109 persons with EOCRC in the San Diego area found a 6.3% prevalence of diabetes which supports a relationship between diabetes and CRC [9], but when the persons with EOCRC were compared to 66 persons with asymptomatic LOCRC discovered by screening, no significant differences in prevalence of obesity (30% vs 36%) [9] or diabetes (6.3% vs 11%) were observed (unpublished data) (Fig. 2).

Fig. 2
figure 2

Adapted from: Strum WB and Boland CR (2019, in press) [9]. Data not previously presented

Locations of EOCRC and screen-detected LOCRC. A comparison of proximal and distal locations of early-onset versus late-onset, screen-detected and late-onset, symptom-detected colorectal cancer demonstrates the anatomic differences which are highly significant.

Full size image

Tobacco-related carcinogens can reach the colorectal mucosa through either the alimentary tract or the circulatory system and could exert a carcinogenic effect by DNA damage or altered expression of important cancer-related genes. Studies of associations between cigarette smoking and incident CRC have demonstrated MSI-high, CIMP-positive, and BRAF-mutated CRC subtypes, suggesting that epigenetic modifications might be involved in smoking-related CRCs [49]. Although no definite links of these factors have been specific to EOCRC, a better understanding of the connection of these factors to CRC, individually and collectively, could lead to targets permitting prevention as well as early detection of CRC in the young, as well as older persons.

There are large-scale targeted deep sequencing studies of EOCRC tumors looking for mutational signatures (other than MSI, CIMP, and LINE-1 hypomethylation) that might provide clues to defects (constitutional or acquired) in DNA repair systems [34, 35, 37, 50]. One possible avenue to be explored is whether there is an interaction between the microbiome and the response of colorectal epithelium to the effects of oxidative stress or other DNA-damaging processes. Such interactions could simultaneously account for changing risks for CRC over time, if linked significantly to changes in the microbiome, possibly modified by temporal changes in diet, obesity, physical activity, or other environmental influences.

New Screening Protocol

On May 30, 2018, the American Cancer Society initiated a qualified recommendation for average-risk adults to begin screening at age 45 years [4], 5 years earlier than most prior recommendations [51,52,53,54] which was endorsed by the American College of Gastroenterology, the American Gastroenterological Association, and the American Society for Gastrointestinal Endoscopy, while recognizing that the supporting evidence is limited. If fully implemented, this expansion of the screening age might be expected to capture up to fifty percent of the EOCRC population at an earlier stage, since about half of this population presents with CRC in their mid- to late 1940s and will have CRC by the time of examination. The impact of colonoscopy on removing colorectal adenomas may be limited as colorectal adenomas tend not to accelerate in growth until the sixth decade [53, 55, 56] and the number of patients needed to screen to prevent one CRC death may be unacceptably high. Additional concerns for a favorable outcome are raised by studies of screening for CRC in asymptomatic persons over the age of 50 years that have not routinely showed improved results over the past decade [57,58,59,60,61,62].

Rational Approach to Screening

While we wait for the results of further clinical, microbial, genetic, epidemiologic, and other studies that may provide more precise etiologic explanations for the rise in CRC in young adults, incremental reductions in the rise in incidence of EOCRC could occur with certain tools immediately available. These include obtaining a detailed family history, adhering to the newly recommended age of 45 years to begin CRC screening for all average-risk adults (which could begin with noninvasive screening and advance if the initial tests are positive), continuing to screen at age 40 years or younger for persons with added risk from a positive family history, and consideration of earlier screening for persons with a combinations of risk factors. Additional recommendations include: (1) endoscopy of at least the distal colon in young persons with unexplained rectal bleeding; (2) a complete colonoscopy if a firm explanation for bleeding is not established, a polyp or tumor is found, or unexplained iron-deficiency anemia is present; (3) appropriate imaging tests for persons with unexplained abdominal pain; and (4) continued promotion of a healthy diet, maintenance of ideal body weight, cessation of tobacco exposure and excess alcohol, and a physically active life style.