Systematic literature searches were performed through February 20, 2013 in Medline, PubMed, and Centre for Reviews and Dissemination. Searches were broadly scoped, using terms for CRC, family history, natural history, and screening adherence.(supplemental material) Two investigators independently reviewed identified abstracts and articles against a priori specified inclusion criteria. Disagreements were resolved through consensus or input of a third reviewer. Detail of search strategies, study selection, analysis and results are available in the supplemental materials.

Inclusion criteria for all questions included age of the person at risk ≥18 and study reported in English language. Exclusion criteria for all questions included illness-associated CRC (e.g. Crohn disease), inherited CRC syndromes, and studies with unmeasured or poorly defined family history criteria. Only studies reporting family history in terms of both number and degree of relatedness for affected family members were included.

For risk, prevalence, and adherence, we included only population-based estimates. Case control studies were excluded to minimize bias by enrichment with people with family history of CRC. For natural history we included only studies that would allow assessment by family history of age of onset of adenoma or CRC, number of adenomas or tumors, advanced adenoma, and adenoma/tumor location. Case-only and case-control studies were permitted for this question.

We reviewed all included studies for the independence of their study populations and years of data collection. For studies from the same data source we included papers that best fit the relevance to the study questions.

Due to the large number of studies and heterogeneity of outcomes and family history definitions, we further refined our inclusion criteria using a best evidence approach, a staged method for prioritizing evidence from all potentially relevant to a set best suited to answering a question. Such approaches can be appropriate for large, heterogeneous literatures to enhance applicability of the evidence.¹⁴ For risk and prevalence we defined our best evidence set as population-based studies with a minimum sample size of 30,000, a number chosen by team consensus, and excluding screening studies as these may over-represent people with family history. For natural history we limited to studies where colonoscopy was conducted in a minimum sample size of 500 people with CRC or adenoma. For adherence, we identified an existing comprehensive review that contained data highly relevant to our study question¹⁵ and included this as a source of primary evidence,¹⁶ supplemented by subsequent US population-based studies.

Histological characteristics of tumors and adenomas of interest to this review were informed by the World Health Organization histology resource¹⁷ in consultation with clinical experts (supplemental material). We abstracted estimates of prevalence or relative risk, with confidence intervals where provided. For the adherence question we limited data abstraction to adults age 50+ to reflect current screening recommendations. We did not conduct meta-analyses due the heterogeneity of outcomes measures and family history definitions. We critically appraised the best evidence set for threats to internal validity from selection; attrition, detection, and reporting bias, adapting criteria from previously published approaches.^15,18

Summary: The prevalence of having at least one first degree relative (FDR) was estimated between 3.1% and 10%. The prevalence of having at least two FDRs was much lower (0.3% to 0.34%). Having a FDR with an early age at diagnosis (before age 45 or 50 years) was uncommon, around 0.3%, compared with having a FDR with a later diagnosis (above age 60 or 65 years), which was over 3%. Few data were available on racial and ethnic-specific prevalence of family history.

Summary: The relative risk of developing CRC varied from 0.89 (for people with no family history) to nearly a 20-fold risk (for people with likely inherited syndromes), with risk levels in between, with increasing family history burden. Risk of CRC was higher when the relative was diagnosed at an earlier age. CRC risk also depended on the age of the person at risk: people with positive family history in their 30s or 40s demonstrated a higher relative risk compared to their age-matched peers than people with the same positive family history at an older age.

Nine studies of unique populations ranging in size from 30,353 to 7 million individuals were included.^{2,19–21,23–27} Five studies were conducted in the US,^{2,19,20,24,25} four others were from Britain,²³ Sweden,²⁷ Japan,²¹ and China.²⁶ Most studies considered the risk of developing colorectal cancer in those with family histories of CRC compared to those with no family history. One very large study considered standardized incidence ratios for colorectal cancer people with family history compared with the general population.¹⁹ Summarized results for risk are shown in Table 2; detailed tables are in the supplementary materials.

The Utah Population Database (UPDB) is a state-wide population-based resource of genealogies of the original Utah pioneers and their modern-day descendants. This analysis used Utah Cancer Registry data (from 1952) linked to the UPDB in a subset of 2.3 million persons who were part of 3 generations of Utah genealogy data and descendants of original Utah pioneers.

Analyses from the UPDB using detailed family pedigrees found distributions of standardized incidence ratios for CRC from 0.89 (95% CI 0.87–0.91) for those with no first degree relatives with CRC to 19.86 (95% CI 9.29–43.24) for those with five or more first degree relatives, who likely have inherited syndromes.¹⁹ Significant differences in CRC incidence were present between those with confirmed negative family history (0.89), one first degree relative date of diagnosis unknown or after age 60 (1.91–1.99), one first degree relative diagnosed before age 60 (2.69), and 3 or 4 first degree relatives (4.41). The Nurses’ Health Study (NHS) found increased risk for CRC from 1+FDR to 2+ FDR.² In the UPDB having a FDR diagnosed before age 60 years increased personal CRC risk. Other US data were consistent with an earlier diagnosis of CRC in a first degree relative conferring higher personal CRC risk, but were not consistently defined (i.e., used different ages to stratify earlier diagnosis) to contribute to precise risk estimates by age of relative’s CRC diagnosis.^2,24

In the Swedish Family Cancer database, data from 7 million people representing over 2 million families was linked with the Swedish Cancer registry.²⁷ The individuals represented offspring born after 1934, who had at least two siblings, and their parents. Having a parent with CRC was associated with a doubling of risk, and a tripling of risk if the parent was diagnosed younger than age 60.²⁷ One Chinese study looked at risk associated with affected siblings versus parents and did not suggest greater impact of sibling CRC over parental CRC.²⁶

Three studies provided relative risks stratified by the age of the person at risk. In the Swedish database, the presence of a parental history of CRC more than quadrupled CRC risk in adults aged 30–39 years, compared to others the same age with no family history.²⁷ Within each age decile of the person at risk, parental cancer history was associated with increased relative risks compared to people without an affected parent of the same age. Relative risks remained elevated until at least age 70. Data from the Health Professionals Followup Study and Nurses’ Health Study populations was generally consistent.² British data on prevalent, rather than incident cancer, suggest decreasing risk with increasing age of the person at risk but confidence intervals were wide; also this study population did not include people under 50 where the risk may be most exaggerated..²³ Overall, the preponderance of data suggests that history of one or more FDR with CRC is associated with a smaller and diminishing incremental increase in relative risk in those aged 50 years and older, as the prevalence of family history in the population increases. Thus, the relative impact of family history on preventable cancers in those under ages 50–55 years will be much greater than in older adults (see supplemental tables).

Adenoma summary: There are very few data with which to make a strong conclusion about adenoma and family history. Data from two studies suggest a higher prevalence of adenomas in people with a positive family history of CRC compared with people with no family history. There was a higher prevalence of two or more adenomas in people with a positive family history of CRC compared with people with no family history. No evidence suggested differential adenoma location by family history status. There is no evidence to suggest a differential prevalence of advanced adenoma in people with a family history of CRC.

Cancer summary: Very limited data were available on family history, natural history and CRC. There was no evidence for difference in age at CRC diagnosis or stage at diagnosis by family history status. There was mixed evidence on tumor location depending on family history status: Some evidence suggested that those with a positive family history are more likely to have distal tumors, whereas others reported no difference.

We identified 9 unique studies: 3 with relevant adenoma outcomes^28–30, and six with relevant CRC outcomes.^31–36

Summary: Individuals with a positive family history are 1.4 to 3.3 times more likely to be adherent to colorectal cancer screening recommendations than individuals with no family history. One study, which objectively measured both family history and screening behavior suggested a 7%–8% absolute increase in screening adherence in people with positive family history.

Our findings suggest that the prevalence of family history may be lower than the commonly discussed estimate of 10%, which is likely based on self-reported data. This may reflect our emphasis on studies that focused on population samples rather than screening studies, to minimize bias due to selective volunteering of patients and families with positive family histories. Though the prevalence of second degree relatives was higher than that of first degree relatives,¹⁹ family history considering various family patterns involving second or third degree relatives was never associated with more than a doubling of CRC risk. This suggests that for clinical purposes, determining the history of cancer in all first degree relatives may be both sufficient and the most feasible. Earlier age of FDR diagnosis with CRC was relatively consistently associated with further increased risk, but current data made the degree of additional risk difficult to quantify. Only one study reported prevalence of family history by race but suggested prevalence at odds with actual increased CRC prevalence in African Americans.²⁰ If these data are correct, family history may be underreported, less-well known, or not a primary influence in the excess CRC burden in African Americans. One study suggested higher prevalence of family history in females, suggesting gender differences in reporting consistent with other studies.^18,49

Our review of population-based risk estimates suggested a range of risk from 0.89 (for individuals with no family history) to nearly a 20-fold risk for individuals with likely inherited syndromes, with risk levels in-between depending on the family history configuration. This is consistent with those in other reviews^50,51 suggesting increasing risk of CRC with increasing burden of family history. Our review has the added benefit of using only population-based study populations and adds results from a high quality study using a comprehensive US-based population.¹⁹ This study used the entire population as the comparator in risk estimates, allowing subpopulation estimates that ranged from a modestly reduced risk to an exaggerated risk for the very few with five or more first degree relatives. Importantly, relative risks for CRC varied between 1.9 and 4.4 in those with at least one FDR, depending on the number of relatives and age of diagnosis. It also included relative risk estimates stratified by age of the person at risk. This study represents a population with potentially lower behavioral risk factors, such as smoking and alcohol use, than the general population. Concern is frequently raised about how genetically representative this population is, but the UPDB is genetically representative of US Caucasian and northern European populations with a low level of inbreeding.^52–54 This study is of high quality and its methods should be replicated wherever possible using other data sources.

Evidence for adenoma and/or CRC risk in people with a family history of adenoma (rather than CRC) is limited; no studies of this type met our best evidence criteria. One review found only two studies, both with methodologic flaws.⁵⁵ A paper published after our search ended suggests increased risk of both CRC and adenoma in people with family history of adenoma.⁵⁶

Our best evidence suggests a higher prevalence of adenomas in people with family history of CRC in men, consistent with previous work⁵⁷ but does not suggest that adenoma progression (measured by advanced adenoma) is accelerated in people with family history. Small studies have suggested that adenoma growth may be accelerated in people with a family history of CRC.^58,59 Future, larger studies could provide additional insight on the impact of family history on adenoma progression. Evolving temporal trends in adenoma classification, such as the sessile serrated pathway, made it challenging to pool data from multiple studies.

Two studies suggested a trend toward increased multiple adenomas in people with a family history of CRC. Our best evidence did not suggest an association between adenoma location and family history. For colorectal cancer, the data were even more limited, but suggested no difference in the age or stage at diagnosis by family history status. There was very limited, conflicting evidence on tumor location and family history. Very limited data suggest that CRC progression in people with family history of CRC is similar to that in the general population.

Our review confirms the findings of a previous review suggesting a clear association between family history of CRC and adherence with both colonoscopy and other CRC screening modalities.¹⁵ Increased adherence likely reflects both organizational (physician recommendation, access) and patient-level factors such as risk perception, but it is worth noting that adherence remains low in all groups. Only one study provided data using objectively measured family history and screening; these data may provide the best estimates for adherence for those with a positive family history but are limited to a single geographic location with a somewhat different social environment.

Our best evidence approach allowed us to summarize an extensive, heterogeneous body of research, but may have resulted in our excluding some relevant data. To our knowledge this is the first review to limit estimates of prevalence of family history and CRC risk to population-based studies, which may be the least biased for informing modeling of population-based screening.

We were also limited by the heterogeneous nature of family history reporting in the literature. There is often analytical rationale for collapsing family history strata to increase sample size, but this makes it difficult to pool data across studies. We excluded 41 studies based on the lack of a basic definition that included number and degree of relatives affected and a further 43 for not providing data stratified by family history. In the remaining studies, more than half used our minimum family history definition of “at least one FDR.” Few studies reported age at diagnosis of the affected relative, verified the accuracy of self-reported data or used objective measures, or assessed or adjusted for family size.

We recommend that future research define family history as, at minimum, the number and degree of affected relatives, and that raw data be reported to maximize potential for aggregating study data. Family history studies should also report how history was assessed, whether and how it was verified, and family size. Measures of how often family history changes in ways that materially impact an individual’s CRC risk as they age would help reduce the uncertainty associated with family history at younger ages and the frequency with which family history for CRC should be updated. High-quality studies of family history and CRC in populations of non-European ancestry are also needed.

Family history is an imperfect and dynamic measure. As family sizes decrease over time, fewer relatives are available to define risk. Also, as endoscopic screening with curative intervention increases, family history can be hidden in families if affected persons are more likely to communicate a diagnosis of advanced cancer to their relatives than a polyp removed during endoscopy. However, family history of CRC remains a clinically meaningful way to identify individuals at increased risk of CRC, and may be the most feasible approach at present, given that multifactorial risk assessment tools are not yet validated for clinical practice^60–65 and the utility of genetic risk stratification is still being investigated.^66–68 In future, risk based on polygenic and/or multifactorial risk assessment may augment family history-based approaches.⁶⁹ Family history collection and reporting should continue in ways that are conducive to knowledge development.