Early‐onset colorectal cancer in young individuals

1adolescent and young adult

2CpG island methylator phenotype

3chromosomal instability

4colorectal cancer

5early‐onset colorectal cancer

6familial adenomatous polyposis

7fecal occult blood test

8Lynch syndrome

9microsatellite and chromosome stable

10MutYH‐associated polyposis

11metatastic colorectal cancer

12mismatch repair

13microsatellite instability

14NTHL1‐associated polyposis

15overall survival

16polymerase proofreading‐associated polyposis

17progression‐free survival

18Response rate

19surveillance, epidemiology and end results program

20Colorectal cancer (CRC) is the third most common cancer and cause of cancer death worldwide in both genders (Siegel et al., 2018). At diagnosis, the median age of patients with colon cancer is 68 and 72 years in men and women, respectively; the median age of patients with rectal cancer is 63 years in both genders (American Cancer Society. Colorectal Cancer Facts & Figures, 2017; Siegel et al., 2018). In recent years, the overall CRC incidence and mortality in the USA and Europe has dropped (Malvezzi et al., 2018; Siegel et al., 2018). Since the mid‐2000s, CRC incidence in the USA has decreased by 2–3% per year in men and women (American Cancer Society. Colorectal Cancer Facts & Figures, 2017; Siegel et al., 2018). This reduction has been mainly correlated with the spread of screening tests [fecal occult blood test (FOBT) and colonoscopy] allowing detection and excision of premalignant lesions, and to a higher awareness of CRC risk factors among the population (Holme et al., 2013; Siegel et al., 2018; Welch and Robertson, 2016). Since 2012 in Europe, the CRC mortality rate decreased by 6.7% in men and 7.5% in women, whereas between 2008 and 2016, CRC incidence increased by 6% annually (Malvezzi et al., 2018; Vuik et al., 2018). Overall survival (OS) at 5 years from diagnosis is around 60% considering all stages of disease (Siegel et al., 2018; Van Cutsem et al., 2014). Metastatic CRC, despite therapeutic advances, exhibits poor prognosis; with the present state of the art, only 14% of patients are alive after 5 years from diagnosis (Siegel et al., 2018; Van Cutsem et al., 2014).

21In 2010, CRC among patients < 50 years old accounted for 4.8% and 9.5% of colon and rectal cancers, respectively (Bailey et al., 2015; Silla et al., 2014). Interestingly, recent studies on several continents have reported an increase in CRC incidence in this age subset and especially in individuals < 40 years (Bailey et al., 2015; Bleyer et al., 2017; Exarchakou et al., 2018; Hessami Arani and Kerachian, 2017; Troeung et al., 2017), including also the metropolitan area of Milan, Italy (A.‐G. Russo, A. Andreano, A. Sartore‐Bianchi, G. Mauri, A. Decarli & S. Siena, personal communication). To date, the actual magnitude and the underlying etiologies of this increase are unclear. Diagnostic and therapeutic protocols dedicated to early‐onset CRC (EO‐CRC) in young individuals are a currently unmet clinical need. Also, there is no consensus on whether EO‐CRCs are undistinguishable or are a distinct molecular/immunologic entity from CRC in older patients (Bleyer et al., 2017; Luzzatto and Pandolfi, 2015; Tomasetti and Vogelstein, 2015; Tricoli et al., 2016).

22In this article, we critically review clinical and molecular characteristics of EO‐CRC based on retrospective analyses, clinical trials, books and reviews currently available in the scientific literature. A systematic review approach has been applied for prognosis and standard metastatic colorectal cancer (mCRC) molecular biomarkers (RAS and BRAF) prevalence among EO‐CRC.

23We first performed a narrative review of clinical and pathological characteristics of EO‐CRC, retrieving articles through PubMed database, American Society of Clinical Oncology, European Society for Medical Oncology, and United European Gastroenterology meeting libraries, websites www.cancer.org and https://seer.cancer.gov/statfacts, and the book Cancer in Adolescents and Young Adults – Chapter 13. Secondly, according to PRISMA Criteria Statement 2009 (Fig. 1; Moher et al., 2009) we systematically retrieved articles describing EO‐CRC prognosis compared with older counterparts, and ther prevalence of RAS and BRAF mutations through PubMed database using [young(Title) OR younger(Title) OR early‐onset(Title) OR adolescent(Title) OR association of age(Title)] AND [colon(Title) OR rectal(Title) OR colorectal(Title)] AND [cancer(Title)] as search terms. Articles published in languages other than English, published earlier than year 2000, articles with a follow up shorter than 2 years, articles including patients older than 50 years of age in the younger group, and articles not comparing young with old patients in the same cohort were not included in the systematic review. This search was implemented by further manual revision of the bibliography of the included articles.

24An unequivocal definition of ‘EO‐CRC’, or ‘young adult CRC’, is currently needed, as no clear and widely accepted consensus is available in literature or guidelines. According to a non‐pediatric oncology definition, the definition generally comprises all CRCs diagnosed before the screening age, i.e. < 50 years of age. Most screening programs start from this age chosen based on cost‐effective analyses of healthcare system sustainability. In contrast, in Adolescent and Young Adult (AYA) Oncology, it comprises CRC patients diagnosed at 15–29 years of age (Bleyer et al., 2017). Nevertheless, in the context of some AYA clinical trials, the age range has been extended to 50 years by the Children's Oncology Group (Bleyer et al., 2017). The US National Cancer Institute (NCI) Progress Review Group of AYA Oncology in 2006 proposed the age range of 15–39 years (Bleyer et al., 2017; Coccia et al., 2018). The definition of ‘very EO‐ CRC’ has also no clear consensus in the literature, with extreme variability among different publications. Therefore, definition of age groups among CRC patients is currently based on non‐specific epidemiologic screening or clinical trial accrual criteria. Current age‐group subdivision is indeed a limitation for interpreting results obtained from published molecular and clinical studies among EO‐CRCs. In this review, we included all publications dealing with EO‐CRC diagnosed before 50 years of age.

25A reasonable solution to improve future retrospective analyses of previously published studies in such a population of EO‐CRCs might be to consider age a continuous variable rather than using arbitrary predefined age cut‐offs (Lieu et al., 2014). The open question remains as to how to define age cut‐off for clinical trials.

26According to the Surveillance, Epidemiology and End Results Program (SEER) database in the USA, around 5% of all CRC are diagnosed in patients < 45 years old (Colorectal Cancer ‐ Cancer Stat Facts, 2018). Rectal cancer is diagnosed in up to 18% of cases < 50 years old in men and women alike (Ahnen et al., 2014). EO‐CRC is more commonly diagnosed among minorities and uninsured populations (You et al., 2012).

27At the beginning of the 21st century, several studies showed evidence that CRC incidence was changing across different age groups. Indeed, in the USA an increase in the number of CRC diagnoses in the population younger than 50 years has been described and, in particular, this increase was observed in patients aged 20–35 years (Bailey et al., 2015; Bhandari et al., 2017; Edwards et al., 2014; Meyer et al., 2010; O'Connell et al., 2003). Epidemiological evidence indicates that since 1994, CRC incidence in those < 55 years is increasing by 2% per year (Figs 2 and 3; Fast Stats, 2018; Siegel et al., 2018). Bailey et al., based on CRC incidence trends between 1975 and 2010 in the USA, predicted a nearly doubled incidence rate of CRC by 2030 in the population aged < 35 years, and the highest increase is expected for recto‐sigmoid cancers between the ages of 18 and 35 years (Bailey et al., 2015). By 2030 in the USA, 10% of all colon and 22% of all rectal cancers are expected to be diagnosed in patients < 50 years old, a provocative prediction if compared with 4% and 9% for colon and rectal cancer, respectively, 10 years ago (Fig. 3; Bailey et al., 2015). These data indicate that EO‐CRC is a current public health issue in the USA (Ahnen et al., 2014; Bailey et al., 2015) and elsewhere (Fig. 4; Abou‐Zeid et al., 2017; Brenner et al., 2017; Exarchakou et al., 2018; Gandhi et al., 2017; Hessami Arani and Kerachian, 2017; Malekzadeh et al., 2009; A.‐G. Russo, A. Andreano, A. Sartore‐Bianchi, G. Mauri, A. Decarli & S. Siena, personal communication; Troeung et al., 2017; Vuik et al., 2018). Following this trend, in the USA it has been proposed that CRC screening should start at 45 years rather than 50 years, but further data are needed to explore this option in real life (Peterse et al., 2018).

28Similarly, recent data from Europe indicate an annual 1.5% increase in rectal cancer incidence between 1990 and 2008, and an annual 7.4% increase in colon cancer incidence between 2008 and 2016 (Vuik et al., 2018). In the Western world, it is reasonable to assume that access to screening procedures, i.e. FOBT and recto‐sigmoidoscopy or colonoscopy in the population > 50 years, can explain only partially the decreasing trend in CRC incidence (Edwards et al., 2014; Welch and Robertson, 2016). Indeed, the trend towards CRC incidence increase among young individuals has been reported not only in the USA and Europe but also in Iran and Egypt (Abou‐Zeid et al., 2017; Hessami Arani and Kerachian, 2017; Malekzadeh et al., 2009).

29Finally, based on current data and projections for the next years it is reasonable to predict an increase of CRC cases diagnosed worldwide in individuals < 50 years for reasons that need to be investigated appropriately.

30Excluding patients adhering to specific screening programs, CRC diagnosis in young individuals is usually carried out when symptoms appear (Hill et al., 2007; Karnak et al., 1999; Riaz et al., 2017). Painless bleeding could anticipate other CRC symptoms by 2–3 years (Riaz et al., 2017). However, CRC diagnosis in young adults is carried out on average 6 months later than symptom onset due to low level of suspiciousness by probands and clinicians, sense of invincibility in young adults, and lack of medical insurance (Bleyer et al., 2008; Hill et al., 2007; O'Connell et al., 2004b).

31In 61% of patients < 50 years (Kneuertz et al., 2015) and up to 76% of patients < 30 years (Indini et al., 2017; Kam et al., 2004; Khan et al., 2016), CRC is diagnosed as stage III or IV, strikingly different from older CRC patients (46–50% diagnosed as stage III or IV; Ferrari et al., 2008; Khan et al., 2016; Kneuertz et al., 2015). EO‐CRCs are more frequently poorly differentiated G3 tumors, left‐sided and rectal (Chang et al., 2012; Indini et al., 2017; Kam et al., 2004; Khan et al., 2016; Kneuertz et al., 2015; Sultan et al., 2010; Wang et al., 2015a). Signet ring cell CRC, accounting for < 1% of all CRC, among younger patients account for 3–13% of cases, especially in those younger than 30 years (Chang et al., 2012; Khan et al., 2016; Kneuertz et al., 2015; Wang et al., 2015a; Yantiss et al., 2009). Young patients usually have a Charlson comorbidity index of 0 and no other medical issues (Kneuertz et al., 2015).

32Non‐Mediterranean Western diet, obesity, little physical activity, high consumption of red and processed meat, and low fiber intake are the most relevant risk factors for developing CRC (Castelló et al., 2018; Edwards et al., 2014; Hessami Arani and Kerachian, 2017; Slattery et al., 1998, 2003; Thune and Lund, 1996). There is a recently increased incidence of obesity in the USA, especially among young patients, and, being a well‐known risk factor for CRC, this may have played a role in reducing the age of CRC onset (Aleksandrova et al., 2013; Bailey et al., 2015; Bassett et al., 2010; Hedley et al., 2004). The net contribution of these factors cannot be assessed (Bailey et al., 2015; Kim et al., 2018).

33Inflammatory bowel diseases increase of by two‐ to threefold the risk of CRC compared with general population, especially when diagnosed in early age (Triantafillidis et al., 2009). Known hereditary cancer‐predisposing syndromes or familial CRC is higher among EO‐CRCs (see next paragraph; Hampel et al., 2005; Liang et al., 2003; Losi et al., 2005; Sinicrope, 2018). Low adherence to specific screening programs in individuals with known cancer syndromes or familial CRC is also a major point in countries with a private health system or among populations with a low socioeconomic level (Hampel et al., 2005; Shin et al., 2017; U.S. Preventive Services Task Force, 2002; Vale Rodrigues et al., 2018). Prior abdominal irradiation (i.e. radiotherapy for curable pediatric malignancies): colonoscopy is recommended starting from 35 years of age or a decade following > 30 Gy radiation treatment to the pelvis (Hill et al., 2007).

34Inflammatory bowel diseases increase of by two‐ to threefold the risk of CRC compared with general population, especially when diagnosed in early age (Triantafillidis et al., 2009).

35Known hereditary cancer‐predisposing syndromes or familial CRC is higher among EO‐CRCs (see next paragraph; Hampel et al., 2005; Liang et al., 2003; Losi et al., 2005; Sinicrope, 2018).

36Low adherence to specific screening programs in individuals with known cancer syndromes or familial CRC is also a major point in countries with a private health system or among populations with a low socioeconomic level (Hampel et al., 2005; Shin et al., 2017; U.S. Preventive Services Task Force, 2002; Vale Rodrigues et al., 2018).

37Prior abdominal irradiation (i.e. radiotherapy for curable pediatric malignancies): colonoscopy is recommended starting from 35 years of age or a decade following > 30 Gy radiation treatment to the pelvis (Hill et al., 2007).

38In the general population of CRC there are three main pathways of carcinogenesis involved in the onset and development of CRC: chromosomal instability (CIN), microsatellite instability (MSI), and CpG island methylator phenotype (CIMP).

39Among EO‐CRCs, as well as in the general population, CIN is the most common cause of CRC (Goel et al., 2010; Markowitz and Bertagnolli, 2009). In a variable percentage of cases depending on disease stage, CRC is caused by MSI (Lanza et al., 2006; Tran et al., 2011; Zaanan et al., 2018). There are two possible mechanisms leading to MSI‐CRC (Sinicrope, 2018):

40hereditary germline mutations occurring in mismatch repair (MMR) genes; tumor somatic hypermethylation of MLH1.

41hereditary germline mutations occurring in mismatch repair (MMR) genes;

42tumor somatic hypermethylation of MLH1.

43In all MSI‐CRC patients, referral for genetic counseling is recommended to screen Lynch syndrome (LS; Palomaki et al., 2009). LS predisposes to EO‐CRC and the percentage of MSI tumors has been observed to rise to up to 27%, especially in patients < 30 years (Khan et al., 2016). In contrast to the overall CRC population, among patients younger than 30 years the location of the primary tumor and histology of MSI‐CRCs do not differ from MSS‐CRC (Khan et al., 2016).

44These mechanisms of carcinogenesis are not mutually exclusive and may coexist (Armaghany et al., 2012; Snover, 2011). A small proportion of MSI‐CRCs can also present CIN, whereas around 50% of MSS‐CRC are CIN‐negative (Banerjea et al., 2009; Sinicrope et al., 2006; Tang et al., 2004). The latter subgroup has been defined as Microsatellite and Chromosome Stable (MACS). These tumors are more frequently rectal or left‐sided, characterized by a poor prognosis and poor recognition by the immune system (Banerjea et al., 2009; Sinicrope et al., 2006; Tang et al., 2004). MACS‐CRCs have been described more frequently among young patients with a family history of CRC, but data in the literature are conflicting (Chan et al., 2001; Rex et al., 2012). MACS‐CRC has been associated with LINE‐1‐hypomethylation and CIMP‐low. Among MACS‐CRC, BRAF mutation or absent MLH1 expression are rare (Antelo et al., 2012; Cai et al., 2008; Silver et al., 2012). In addition, recent results from an extended somatic molecular investigation in a cohort of left‐sided and rectal EO‐CRC showed higher mutation rates of NF1, POLE, SMAD4 and BRCA2 (Puccini et al., 2018). Furthermore, mutations in left‐sided EO‐CRC genes involved in histone modification, such as KDM5C, KMT2A, KMT2C, KMT2D and SET2D, were reported more frequently (Puccini et al., 2018).

45A prominent cause of EO‐CRC is the presence of a germline oncogene mutation, giving rise to a hereditary cancer syndrome. Prevalence of hereditary CRC syndromes among EO‐CRCs is influenced by the different age groups analyzed in different studies (Mork et al., 2015; Pearlman et al., 2017; Stoffel et al., 2018), with a higher prevalence among patients < 35 years old (Mork et al., 2015).

46Four potential genetic scenarios should be considered among EO‐CRCs: known hereditary cancer syndromes, de novo germline hereditary cancer mutations, familial colorectal cancer and non‐hereditary and non‐familial CRC.

47Around 22% (10–33% according to different studies) of patients diagnosed with EO‐CRC are affected by hereditary cancer syndromes (Fig. 5). This proportion is significantly higher when compared with the 2–5% of hereditary cancer syndromes among the general CRC population (Chang et al., 2012; Jasperson et al., 2010; Mork et al., 2015). The most frequent hereditary cancer syndrome is hereditary non‐polyposis colorectal cancer, also known as LS (Mork et al., 2015; Pearlman et al., 2017; Sinicrope, 2018; Stoffel et al., 2018). LS patients have a lifetime risk of developing CRC of 70%, and in 40% of cases the onset of CRC is before age 40 (Lynch et al., 2008). LS is the most frequent CRC hereditary syndrome among patients younger than 50 years old and accounts for around one‐third of EO‐CRC in patients younger than 35 years old (Mork et al., 2015; Pearlman et al., 2017; Stoffel et al., 2018). Germline mutations in MLH1 and MSH2 genes are the most frequent, but MSH6 and PMS2 gene can present pathogenic mutations as well (Evans et al., 2007; Lynch et al., 2015). In addition, mutations occurring in EPCAM gene can also silence the MSH2 promoter region, leading to LS (Kempers et al., 2011; Ligtenberg et al., 2009, p. 2). In some cases of LS, homozygous or compound heterozygous (bi‐allelic) constitutive MMR gene mutations have been identified (Durno et al., 2005, 2015; Wimmer and Etzler, 2008). These lead to a very low age of CRC onset, usually earlier than 16 years old (Durno et al., 2005, Wimmer and Etzler, 2008). In contrast, a relevant proportion of young patients with MSI‐H tumors, usually MLH1 hypermethylated, do not harbor germline alterations in MMR genes (Pearlman et al., 2017; Zbuk et al., 2009).

48CRC may develop earlier in patients affected by adenomatous polyposis syndromes: familial adenomatous polyposis (FAP), polymerase proofreading‐associated polyposis (PAPP), MutYH‐associated polyposis (MAP) and NTHL1‐associated polyposis (NAP). FAP is the second most frequent hereditary cancer syndrome (Mork et al., 2015). Lifetime risk of developing CRC in FAP patients is 100% with a median age of CRC onset of 39 years old (Zbuk et al., 2009). Different APC mutations have been associated with different phenotypes, with mutations occurring in codon 1309 being associated with severe polyposis in the first decade of life (Caspari et al., 1994). FAP screening is recommended from the age of 10 years old and the treatment of choice is total proctocolectomy (Warrier and Kalady, 2012; Winawer et al., 2003). PAPP is an autosomal dominant genetic disease leading to increased risk of cancer and by germline mutations in POLE and POLD1 genes (Palles et al., 2013). PAPP confers a highly penetrant predisposition to develop polyposis (usually < 100) and EO‐CRCs (Palles et al., 2013; Valle et al., 2014). In addition, mutations occurring in POLE and POLD1 can occur de novo and in patients without polyposis (Valle et al., 2014). Thus, alterations in these genes could be found in all familial CRCs and are not restricted to polyposis cases (Valle et al., 2014). Differently, MAP and NAP are two autosomal recessive genetic diseases caused by germline mutation in MutYH and NTHL1 DNA glycosylase gene of the base excision repair (BER) pathway (Weren et al., 2015, 2018). These syndromes are characterized by the onset of 10–50 polyps around the age of 40 that cause an increased risk of CRC (Weren et al., 2015, 2018). MAP confers a 28‐fold higher lifetime risk of developing CRC at usually < 60 years old with a reported penetrance of 40–100% (Lubbe et al., 2009; Weren et al., 2018).

49Other cancer syndromes, such as Li‐Fraumeni syndrome, characterized by germline TP53 mutations, have been related to EO‐CRC; they account for less than 1% of cases (Yoshida et al., 2017; Yurgelun et al., 2015; Fig. 5).

50In patients with known hereditary syndromes, adherence to screening programs should be strongly encouraged, as in LS it has been demonstrated to reduce the risk of death by 65% (Järvinen et al., 2000). However, screening program adherence is low in youths, including in Western countries, and has to be enhanced to reduce late diagnosis and cancer deaths (Kim et al., 2017; Shaikh et al., 2018). Furthermore, a preventive aspirin regimen has been reported to significantly reduce CRC incidence among LS patients (Burn et al., 2011).

51Many recent studies have tried to broaden the spectrum of germline alterations predisposing to EO‐CRC to expand the population admitted to screening programs. Several genes have been investigated, e.g. SMAD4, BRCA1‐2, ATM, BUB1‐2, BRF1 and CHEK2 (Bellido et al., 2018, p. 1; Hahn et al., 2016; Mur et al., 2018; Pearlman et al., 2017; Stoffel et al., 2018; Yurgelun et al., 2015); however, at around 2–10% they account for only a small proportion of EO‐CRC overall (Pearlman et al., 2017; Stoffel et al., 2018). Although rare, these genetic alterations should not be neglected, especially for those that can potentially be treated with drugs, such as BRCA2 alterations (Fong et al., 2009; Ledermann et al., 2012; Mateo et al., 2015; Robson et al., 2017). As pointed out by different studies, a large percentage of EO‐CRC patients with no CRC familiarity may be affected by hereditary cancer syndromes or at least harbor gene alterations potentially predisposing to CRC (Mork et al., 2015; Stoffel et al., 2018). Only half (43/85) of EO‐CRC patients presenting germline cancer‐predisposing alterations reported CRC in a first‐degree relative, therefore being the first to be diagnosed with a cancer‐predisposing syndrome (Stoffel et al., 2018). This evidence has led some authors to recommend genetic counseling for all patients diagnosed with extreme EO‐CRC (< 35 years old; Mork et al., 2015).

52The opposite situation is represented by patients characterized by a relevant familial CRC not affected by hereditary CRC syndromes. In particular, this is the case for patients fulfilling Amsterdam Criteria I or II but not affected by LS (Garre et al., 2015; Vasen et al., 1991, 1999). Today, the Amsterdam criteria have largely been replaced by the Bethesda criteria, which are more sensitive but less specific in identifying MSH2 and MLH1 germline mutations carriers, thus broadening the spectrum of familial CRC patients (Piñol et al., 2005; Umar et al., 2004). Amsterdam or Bethesda criteria‐positive patients without a diagnosis of LS are diagnosed with Familiar Colorectal Cancer Syndrome Type X (Nejadtaghi et al., 2017; Umar et al., 2004; Vasen et al., 1991, 1999). This peculiar population may be an ideal group in whom to extend molecular screening in order to find out potentially actionable gene alterations hopefully to confer survival benefit, if receiving appropriately targeted agents (Garre et al., 2015; Nejadtaghi et al., 2017). In this regard, Garre and coworkers documented that, in a population of EO‐CRC patients fulfilling Amsterdam criteria II, the prevalence of BRCA2 germline mutations, including point and frame mutations, was 60% (29/48; Garre et al., 2015), leading to the speculation that these patients might derive benefit from platinum‐ or PARP‐inhibitor‐based regimens in the metastatic setting (Garre et al., 2015; Mylavarapu et al., 2018).

53In addition, CRC incidence in this subgroup may increase in the next years due to the activity of known carcinogenic agents in patients harboring a hypothetical genetic predisposition to develop CRC. At this regard, in a Japanese cross‐sectional study it has been recently reported that alcohol consumption might be associated with earlier CRC onset in patients affected by LS (Miguchi et al., 2018).

54It is very surprising to note that although hereditary and familial cancer syndromes are more frequent among young individuals, half of EO‐CRC patients have neither (Mork et al., 2015; Stoffel et al., 2018). This implies that, since these patients are not included in screening programs, they are often diagnosed in later stages of the disease. Among this subset of patients, without known hereditary cancer syndromes, different alterations in TNFR1, EIF4E, LTBP4, CYR61, UCHL1, FOS and FOS B genes have been demonstrated between early‐ and late‐onset CRC. However, definite conclusions are still far from being reached, since study populations were limited to cohorts of 10–39 patients and the gene panels tested were heterogeneous (Berg et al., 2010; Hong et al., 2007; Kirzin et al., 2014).

55In addition to somatic tumor mutations, tumor immune infiltration is currently under investigation in EO‐CRC. Recently, in EO‐CRC a different expression of immune activation genes (i.e. CLC and IFNAR1) has been reported when compared with the older counterparts (Ågesen et al., 2011). Overall, it is conceivable that among EO‐CRCs, more patients could potentially benefit from immunotherapy in the metastatic setting because of a higher prevalence of MMR, POLE and POLD1 aberrations, leading to a higher tumor mutational burden (Pai et al., 2017). However, data on peculiar features of interactions between the immune system and tumors in EO‐CRC are still to be addressed.

56Environmental reasons for EO‐CRC are still unknown. In the last decades, dramatic environmental and behavioral changes may have contributed to lowering the age of CRC onset. Changes in microbiome induced by cesarean delivery or appendicectomy might play a role. Microbiome may also be altered by modern dietary regimens including colorants and preservatives. These substances may also play a role of direct carcinogenesis on intestine cells. The extensive use of antibiotics in agriculture and medicine, although immediately beneficial, could reasonably alter the gut microbiome. Reduced breast‐feeding may alter the development of the immune system and its capacity for cancer surveillance. However, we still do not know how these agents may alter cells pathways causing CRC. Large epidemiological and preclinical studies are warranted.

57We systematically retrieved 37 articles describing the prognosis of EO‐CRC compared with older patients (Abdelsattar et al., 2016; Blanke et al., 2011; Boyce et al., 2016; Chandrasinghe et al., 2017; Chou et al., 2011, 2017; Damodaran and Seshadri, 2016; Fu et al., 2014; Fu et al., 2013; Haleshappa et al., 2017; Hawk et al., 2015; Hubbard et al., 2012; Josifovski et al., 2004; Khan et al., 2016; Kim et al., 2016; Kneuertz et al., 2015; Kolarich et al., 2018; Li et al., 2014; Lieu et al., 2014; Manjelievskaia et al., 2017; McMillan and McArdle, 2009; Murata et al., 2016; O'Connell et al., 2004a; Orsini et al., 2015; Pokharkar et al., 2017; Quah et al., 2007; Rho et al., 2017; Rodriguez et al., 2018; Shen et al., 2018; Shida et al., 2018; Sultan et al., 2010; Vatandoust et al., 2016; Wang et al., 2015a,2015b; Yang et al., 2012; You et al., 2012; Zhao et al., 2017). EO‐CRC survival data are conflicting. Some studies indicate a poorer prognosis (Bleyer et al., 2008; Khan et al., 2016; Lieu et al., 2014; McMillan and McArdle, 2009; O'Connell et al., 2004b; Shida et al., 2018; Sultan et al., 2010), whereas others support a comparable or even better prognosis in comparison with older patients (Blanke et al., 2011; Hubbard et al., 2012; Kneuertz et al., 2015; Kolarich et al., 2018; McMillan and McArdle, 2009; O'Connell et al., 2004a; Quah et al., 2007; Rodriguez et al., 2018; Vatandoust et al., 2016; Wang et al., 2015a; You et al., 2011; Table 1). From 1973 to 2005, adult CRC survival outcome improved, whereas child and adolescent CRC survival have not (Sultan et al., 2010). A worse survival was observed in studies comparing survival between patients younger than 30 years old with those > 50 years (Khan et al., 2016; Lieu et al., 2014; Sultan et al., 2010). In particular, it was reported that the younger the patient, the worse the prognosis (Lieu et al., 2014). By contrast, publications comparing patients younger than 50 years of age with those > 50 years showed a better prognosis among the younger patients (Kneuertz et al., 2015; Kolarich et al., 2018; Vatandoust et al., 2016; Wang et al., 2015a). This suggests that in patients with very early‐onset CRC (< 35 years old) a different biological background likely underlies an earlier and faster CRC progression (Ferrari et al., 2008; Indini et al., 2017; Khan et al., 2016; Zhao et al., 2017).

58Clinical guidelines do not consider early age of onset of CRC as a criterion to drive treatment, either in the adjuvant or in the metastatic setting. To date, therapeutic options for early‐ and late‐onset CRC patients are the same according to the major oncology societies worldwide (Benson et al., 2017; Van Cutsem et al., 2014; Yoshino et al., 2018). However, many publications describe a more aggressive attitude of both clinicians and surgeons treating early‐onset stage III and IV CRC (Khan et al., 2016; Kneuertz et al., 2015). This aggressiveness has been demonstrated to not confer a significant survival benefit (Kneuertz et al., 2015). In particular, an adjuvant, more intensive schedule of chemotherapy or targeted agents failed to confer a survival benefit and, overall, a more aggressive surgical and adjuvant approach in young CRC patients should not be recommended, since it may lead to overtreatment (Alberts et al., 2012; Allegra et al., 2011; De Gramont et al., 2012; Kneuertz et al., 2015; Taieb et al., 2017). Combining neo‐adjuvant chemo‐radiotherapy with surgery failed to confer a survival advantage in rectal patients younger than 50 years of age when compared with surgery alone (Kolarich et al., 2018). Conversely, oxaliplatin added to standard chemoradiotherapy in patients affected by locally advanced rectal cancer seems to improve disease‐free survival and OS in those < 60 years of age (Hofheinz et al., 2018).

59In the metastatic setting, two recent retrospective analyses have analyzed the outcome of EO‐CRC. A retrospective study including nine phase III trials stated that patients < 40 years old at diagnosis had a poorer progression‐free survival (PFS) but not OS or response rate (RR) in comparison with patients > 50 years (Blanke et al., 2011). More recently, Lieu and coworkers considered age as a continuous variable rather than a prespecified cut‐off and pooled results of 24 first line metastatic CRC trials including young patients treated with anti‐EGFR treatment according to CAIRO2, COIN, FIRE II and PRIME studies (Lieu et al., 2014). They reported that EO‐CRC patients diagnosed in the 1920s experienced a worse PFS and OS when compared with patients diagnosed in their middle ages. They stated that prognostic effect of age did not differ according to class of targeted therapy for either OS or PFS.

60Finally, age‐grouped comparisons assessing the efficacy of triplet vs doublet drug combinations in young patients are still lacking. Even if Lieu et al. were the first to include patients treated with anti‐EGFR drugs, more molecularly focused analyses are warranted to clarify this complex picture. Whether age of onset in addition to sidedness is a factor involved in determining anti‐EGFR drug resistance is still unknown (Boeckx et al., 2017; Holch et al., 2017).

61Similar to all‐age populations, anti‐EGFR drugs are a valid treatment option for RAS wild‐type EO‐CRC patients (Benson et al., 2017; Douillard et al., 2013; Van Cutsem et al., 2009, 2016; Yoshino et al., 2018). RAS mutations are identified in 40% of all CRC (Cancer Genome Atlas Network, 2012).

62We systematically retrieved 11 articles describing RAS and BRAF prevalence among EO‐CRC; data are conflicting (Alsop et al., 2006; Chang et al., 2012; Goel et al., 2010; Khan et al., 2016; Kirzin et al., 2014; Magnani et al., 2015; Perea et al., 2017; Rho et al., 2017; Tsai et al., 2016; Watson et al., 2016; Yantiss et al., 2009; Table 2). Most studies revealed a lower prevalence of KRAS mutations among EO‐CRC (4–31%), but others reported a similar (35–39%) or higher (54%) prevalence to that in older patients (Alsop et al., 2006; Chang et al., 2012; Goel et al., 2010; Khan et al., 2016; Kirzin et al., 2014; Magnani et al., 2015; Perea et al., 2017; Rho et al., 2017; Watson et al., 2016; Yantiss et al., 2009; Table 2).

63NRAS prevalence was reported at around 1% in a small population (n = 69) of EO‐CRC (Perea et al., 2017). The prevalence of BRAF mutations is reported to be similar among EO‐CRC and the older onset patients, ranging from 0 to 19% (Chang et al., 2012; Goel et al., 2010; Khan et al., 2016; Kirzin et al., 2014; Magnani et al., 2015; Tsai et al., 2016; Watson et al., 2016; Yantiss et al., 2009).

64In conclusion, as summarized in Table 2, there are no unequivocal data regarding the prevalence of RAS/RAF mutations in EO‐CRC.

65Most of the studies on EO‐CRC are retrospective and include a small number of patients of different ranges of ages. Despite the wide heterogeneity, the reviewed studies can be considered to generate hypotheses, even though no definitive conclusions or recommendations can be derived. Several studies demonstrate an increase incidence of EO‐CRC in different areas of the world, but the reasons behind this epidemiologic phenomenon are still to be unveiled. To counteract the greatest loss of life‐years in this younger than average CRC population, we need to achieve the best possible clinical results with the current treatments and to develop new therapies. Non‐hereditary CRC among the youngest (< 35 years old) represents the toughest challenge and this subset of EO‐CRC is expected to become even more prevalent in the next years. Available data suggest that survival is worst in patients younger than 30 years old, whereas it is comparable or even better among patients between 40 and 50 when compared with those older than 50 years. Given the clinical and molecular peculiarities of EO‐CRC, especially in patients younger than 30 years of age, a different molecular carcinogenesis might be speculated in those sporadic cases.

66A mandatory step to answering questions of recognizing EO‐CRC as a common subdivision of age groups is necessary for comparison of EO‐CRC studies. In interpreting retrospective studies, considering age as a continuous variable may be an interesting solution. In addition, clinical criteria could be useful to identify EO‐CRC patients more likely to benefit from targeted therapy, such as MSS BRCA2 mutated patients fulfilling Amsterdam criteria, or immunological treatments, such as patients harboring POLE or POLD hypermutated tumors potentially benefitting from checkpoint inhibitors. The role of environmental risk factors and the microbiome remains to be clarified. Finally, but more compelling, there is still a scarcity of clinical trials focusing on EO‐CRC and these are warranted to improve care of these patients.

67AS‐B has acted as a consultant/advisory member for Amgen, Bayer, Lilly and Merck‐Serono. AB has acted as a consultant/advisory member for Horizon Discovery, Biocartis and Trovagene. SS is an advisory board member for Amgen, Bayer, BMS, CheckmAb, Celgene, Incyte, Merck, Novartis, Roche and Seattle Genetics.