Cancer site–specific polygenic risk scores (PRS) effectively identify individuals at high risk of individual cancers, but the effectiveness of PRS on overall cancer risk assessment and the extent to which a high genetic risk of overall cancer can be offset by a healthy lifestyle remain unclear. Here, we constructed an incidence-weighted overall cancer polygenic risk score (CPRS) based on 20 cancer site-specific PRSs. Lifestyle was determined according to smoking, alcohol consumption, physical activity, body mass index, and diet. Cox regression by sex was used to analyze associations of genetic and lifestyle factors with cancer incidence using UK Biobank data (N = 442,501). Compared with participants at low genetic risk (bottom quintile of CPRS), those at intermediate (quintiles 2 to 4) or high (top quintile) genetic risk had HRs of 1.27 (95% confidence interval, 1.21–1.34) or 1.91 (1.81–2.02) for overall cancer, respectively, for men, and 1.21 (1.16–1.27) or 1.62 (1.54–1.71), respectively, for women. A joint effect of genetic and lifestyle factors on overall cancer risk was observed, with HRs reaching 2.99 (2.45–3.64) for men and 2.38 (2.05–2.76) for women with high genetic risk and unfavorable lifestyle compared with those with low genetic risk and favorable lifestyle. Among participants at high genetic risk, the standardized 5-year cancer incidence was significantly reduced from 7.23% to 5.51% for men and from 5.77% to 3.69% for women having a favorable lifestyle. In summary, individuals at high genetic risk of overall cancer can be identified by CPRS, and risk can be attenuated by adopting a healthy lifestyle.

Significance:

A new indicator of cancer polygenic risk score measures genetic risk for overall cancer, which could identify individuals with high cancer risk to facilitate decision-making about lifestyle modifications for personalized prevention.

Cancer, a leading cause of death worldwide, is a result of complex interplay between genetic and environmental factors. The heritability of cancer has been estimated to be 33% according to a Nordic twin's study (1). To date, a large number of genetic loci have been identified as susceptibility markers for certain cancers by genome-wide association studies (GWAS). Although these loci individually have relatively modest effects on cancer risk, a polygenic risk score (PRS) combining multiple loci together as an indicator of genetic risk has been proved to effectively predict incidence of site-specific cancer (2–4). Recent advances in precision public health have found that, compared with rare monogenic mutations, polygenic risk scores can identify larger fraction of population at comparable or greater disease risk for common diseases, which poses opportunities for clinical utility (5–7). However, it is largely unknown whether a PRS can predict overall cancer risk and identify individuals at high genetic risk for potential personalized prevention.

Unhealthy lifestyles, including tobacco smoking, alcohol drinking, obesity, physical inactivity, and unhealthy dietary pattern, are known to be associated with an increased risk of cancer (8). More than 40% of all cancer cases and deaths are attributable to potentially modifiable risk factors, predominantly unhealthy lifestyles (9). On the contrary, a healthy lifestyle is also known to be associated with an increase in the total life expectancy and life expectancy free of cancer (10).

Much evidence has indicated that cancer risk of individuals at high genetic risk can be attenuated by adherence to a healthy lifestyle, such as cancers of the breast (11), colorectum (12), and stomach (13). However, these studies generally focused on single-cancer risk rather than overall cancer risk. Thus, the extent to which overall cancer risk of individuals with a high genetic risk can be offset by healthy lifestyle remained unclear.

In this study, we constructed an incidence-weighted cancer polygenic risk score (CPRS) for overall cancer risk derived from 16 and 18 site-specific cancer PRSs for men and women, respectively, and then evaluated the effectiveness of CPRS in predicting risk of overall cancer in the UK Biobank (UKB). We also assessed the utility of the CPRS over the life course and estimated the extent to which a healthy lifestyle was associated with a reduced overall cancer risk across groups with a different genetic risk defined by the CPRS.

Study design and participants

As shown in Fig. 1, we systemically retrieved and summarized genetic loci from published GWASs of site-specific cancers in populations of European ancestry by Jan 1, 2020. Then, we constructed PRS for each cancer and subsequently aggregated into CPRSs of overall cancer for men and women, respectively. At last, we evaluated the effectiveness of the CPRSs and the impact of healthy lifestyle on overall cancer risk in the UKB.

Figure 1.

Study design and workflow.

Figure 1.

Study design and workflow.

Close modal

The UKB is a large population-based cohort study with a detail protocol publicly available (14). In brief, approximately 500,000 participants aged 40–70 were recruited from 22 assessment centers across England, Scotland, and Wales between 2006 and 2010 at baseline. Each eligible participant completed a written informed consent form and provided information on lifestyle and other potentially health-related information through extensive baseline questionnaires, interviews, and physical measurements. Meanwhile, a blood sample was also collected and used for genotyping. The UK Biobank study has approval from the multicenter Research Ethics Committee, the National Information Governance Board for Health and Social Care in England and Wales, and the Community Health Index Advisory Group in Scotland (http://www.ukbiobank.ac.uk/ethics/).

Genotyping and imputation

Detailed information on genotyping process and DNA arrays used in the UKB study has been described elsewhere (14, 15). Briefly, participants were genotyped using the Applied Biosystems UK BiLEVE Axiom Array (807,411 markers tested for 49,950 participants) or Applied Biosystems UK Biobank Axiom Array (825,927 markers tested for 438,427 participants) by Affymetrix. These arrays share more than 95% of single-nucleotide polymorphisms (SNP) tested. Imputation was performed with SHAPEIT3 and IMPUTE3 based on merged UK10K and 1000 Genomes phase3 panels.

PRS calculation and CPRS construction

For each cancer site, risk-associated SNPs and corresponding effect sizes were derived from the largest published GWASs in terms of sample size (see Supplementary Methods). A PRS was created for each cancer site following an additive model as described previously (2). In short, the dosage of each risk allele of each individual was summed after multiplication with its respective effect size of site-specific cancer. Except for nonmelanoma skin cancer and those without relevant GWAS or significant genetic loci published to date, we derived PRSs for 20 cancer types for further analyses (Supplementary Table S1).

To generate an indicator of genetic risk for overall cancer, we constructed the CPRS as follows:

formula

Where is the cancer polygenic risk score of ith individual, hk is the age-standardized incidence of site-specific cancer k in the UK population, and PRSi,k is the aforementioned PRS of site-specific cancer k (Supplementary Methods). Given the different spectrum of cancer incidence between men and women, we constructed sex-specific CPRS accordingly.

Healthy lifestyle factors

We adopted five healthy lifestyle factors according to the World Cancer Research Fund/American Institute of Cancer Research [WCRF/AICR (Arlington, VA)] recommendations (https://www.aicr.org/cancer-prevention/; ref. 16), that is, no current smoking, no alcohol consumption, regular physical activity, moderate BMI (body mass index, 18.5∼30), and a healthy diet pattern (Supplementary Table S2). “No current smoking” was referred to never smokers or former smokers who had quit smoking at least 30 years. “No alcohol consumption” was referred to those who never used alcohol. “Regular physical activity” was referred to those who exercised for at least 75 minutes of vigorous activity per week or 150 minutes of moderate activity per week (or an equivalent combination) or engaged in vigorous activity once and moderate physical activity at least 5 days a week (17). “A healthy diet pattern” was referred to those who consumed an increased amount of fruits, vegetables, whole grains, fish, and a reduced amount of red meats and processed meats (17). The lifestyle index ranged from 0 to 5, with a higher index indicating a healthier lifestyle.

Outcomes

Incident cancer cases in the UKB were ascertained through record electronic linkage with the National Health Service central registers and death registries in England, Wales, and Scotland. Complete follow-up was updated to October 31, 2015 for Scotland, and to March 31, 2016 for England & Wales. Cancer cases were coded using the 10th Revision of the International Classification of Diseases and obtained from data field 40006 and 40005 of the UKB.

Statistical analysis

Cancer risk in participants of UKB was assessed from baseline up to the date of diagnosis, death, loss to follow-up, or date of complete follow-up, whichever occurred first. Multivariable Cox proportional hazards regression analyses were performed to assess associations of genetic and lifestyle factors with cancer incidence, including a dose–response relationship (Ptrend), by HRs and 95% confidence intervals (CI). Schoenfeld residuals and log-log inspection were used to test the assumption of proportional hazards. We compared HRs for participants at low (the bottom quintile of CPRS), intermediate (quintiles 2–4), and high (the top quintile) genetic risk, as described previously (11, 13, 17, 18). Similarly, we determined cancer risk for participants at favorable (4–5 healthy lifestyle factors), intermediate (2–3 healthy lifestyle factors), and unfavorable (0–1 healthy lifestyle factor) lifestyle. We also used multivariable logistic regression models to assess associations between the CPRSs and individual lifestyle factors with adjustment for age, family history of cancer, Townsend deprivation index, height, and the first 10 principal components of ancestry, where applicable (19). Absolute risk was calculated as the percentage of incident cancer cases occurring in a given group. We calculated absolute risk reduction as the difference in cancer incidences among given groups, extrapolated the difference in 5-year event rates among given groups. The 95% CIs for the absolute risk reduction were derived by drawing 1,000 bootstrap samples from the estimation dataset. We fitted an adjusted survival curves including both prevalent cancer cases and incident cancer cases with age as the underlying timescale, which was used to estimate the average lifetime risk by age 75 with CIs obtained from normal approximation (4). We performed additive interaction analysis between genetic and lifestyle categories by using two indexes: the relative excess risk due to interaction (RERI) and the attributable proportion due to interaction (AP; ref. 20). The 95% CIs of the RERI and AP were estimated by bootstrap (N = 5,000), which would contain 0 if there was no additive interaction. All the abovementioned analyses were performed for men and women separately.

Participants with missing data on any of the covariates were multiple imputed, and independent analyses were also performed based on complete data for sensitivity analyses. Besides, to examine the robustness of our results, we conducted several further sensitivity analyses: (i) reconstructing CPRSs by combining PRSs of individual cancers directly without weighting by cancer incidence; (ii) reconstructing CPRSs by standardizing the mean of each site-specific cancer PRS to 1; (iii) reconstructing CPRSs by excluding SNPs that appear or in strong LD (r2 > 0.6) in more than one site-specific PRS; (iv) reclassifying genetic risk levels based on quartiles (bottom, 2–3, and top quartiles defined as low, intermediate, and high genetic risk, respectively) or tertiles (corresponding to low, intermediate, and high genetic risk) of CPRSs; (v) reevaluating the effect of genetic risk based on participants of unrelated British ancestry; and (vi) excluding incident cases of any cancer occurring during the first year of follow-up. All P values were two-sided and P < 0.05 was considered statistically significant. All statistical analyses were performed with R software, version 3.5.1 (R Project for Statistical Computing).

Participants

After removing individuals who had withdrawn their consent to participate, had been diagnosed with cancer at baseline, failed to be genotyped, had a nonmelanoma skin cancer, or reported a mismatch sex with genetic data, we included 442,501 participants in the final analysis, including 202,842 men and 239,659 women (Table 1; Supplementary Fig. S1). There were 26,320 incident cancer cases, with 13,480 men and 12,840 women (Supplementary Table S3), during a median follow-up of 7.07 years (IQR: 6.36–7.71).

Table 1.

Incident cancer cases and cancer-free participants by baseline characteristics in the UK Biobank cohort.

Men (n = 202,842)Women (n = 239,659)All (n = 442,501)
Incident cancerNo cancerIncident cancerNo cancerIncident cancerNo cancer
Variablen = 13,480 (%)n = 189,362 (%)n = 12,840 (%)n = 226,819 (%)n = 26,320 (%)n = 416,181 (%)
Age at baseline, mean (SD), years 61.27 ± 6.14 55.94 ± 8.21 58.70 ± 7.35 55.87 ± 8.03 60.01 ± 6.88 55.91 ± 8.11 
Height, mean(SD), centimeters 175.18 ± 6.76 175.67 ± 6.86 162.51 ± 6.32 162.45 ± 6.31 169.00 ± 9.11 168.46 ± 9.30 
Socioeconomic status quintilea 
 1 (least deprived) 2,776 (20.59) 37,976 (20.05) 2,569 (20.01) 45,303 (19.97) 5,345 (20.31) 83,279 (20.01) 
 2–4 7,933 (58.85) 112,163 (59.23) 7,806 (60.79) 137,325 (60.54) 15,739 (59.80) 249,488 (59.95) 
 5 (most deprived) 2,771 (20.56) 39,223 (20.71) 2,465 (19.20) 44,191 (19.48) 5,236 (19.89) 83,414 (20.04) 
Family history of cancer 
 Yes 5,271 (39.10) 63,513 (33.54) 5,168 (40.25) 78,578 (34.64) 10,439 (39.66) 142,091 (34.14) 
 No 8,209 (60.90) 125,849 (66.46) 7,672 (59.75) 148,241 (65.36) 15,881 (60.34) 274,090 (65.86) 
Healthy lifestyle factors 
 No current smoking 5,739 (42.57) 99,244 (52.41) 7,236 (56.36) 140,813 (62.08) 12,975 (49.30) 240,057 (57.68) 
 No alcohol consumption 280 (2.08) 5,440 (2.87) 744 (5.79) 13,292 (5.86) 1,024 (3.89) 18,732 (4.50) 
 Normal BMI 9,831 (72.93) 140,807 (74.36) 9,351 (72.83) 171,982 (75.82) 19,182 (72.88) 312,789 (75.16) 
 Regular physical activity 8,607 (63.85) 122,437 (64.66) 7,938 (61.82) 143,604 (63.31) 16,545 (62.86) 266,041 (63.92) 
 Healthy diet 2,505 (18.58) 37,373 (19.74) 3,688 (28.72) 67,193 (29.62) 6,193 (23.53) 104,566 (25.13) 
Healthy lifestyleb 
 Favorable 822 (6.10) 14,477 (7.65) 1,357 (10.57) 28,514 (12.57) 2,179 (8.28) 42,991 (10.33) 
 Intermediate 8,492 (63.00) 126,193 (66.64) 8,509 (66.27) 154,252 (68.01) 17,001 (64.59) 280,445 (67.39) 
 Unfavorable 4,166 (30.91) 48,692 (25.71) 2,974 (23.16) 44,053 (19.42) 7,140 (27.13) 92,745 (22.28) 
Cancer polygenic risk scorec, mean (SD) 37.24 ± 1.38 36.91 ± 1.33 30.89 ± 1.15 30.69 ± 1.13 34.14 ± 3.42 33.52 ± 3.33 
 Low 1,984 (14.72) 38,585 (20.38) 2,086 (16.25) 45,846 (20.21) 4,070 (15.46) 84,431 (20.29) 
 Intermediate 7,729 (57.34) 113,975 (60.19) 7,486 (58.30) 136,309 (60.10) 15,215 (57.81) 250,284 (60.14) 
 High 3,767 (27.95) 36,802 (19.43) 3,268 (25.45) 44,664 (19.69) 7,035 (26.73) 81,466 (19.57) 
Men (n = 202,842)Women (n = 239,659)All (n = 442,501)
Incident cancerNo cancerIncident cancerNo cancerIncident cancerNo cancer
Variablen = 13,480 (%)n = 189,362 (%)n = 12,840 (%)n = 226,819 (%)n = 26,320 (%)n = 416,181 (%)
Age at baseline, mean (SD), years 61.27 ± 6.14 55.94 ± 8.21 58.70 ± 7.35 55.87 ± 8.03 60.01 ± 6.88 55.91 ± 8.11 
Height, mean(SD), centimeters 175.18 ± 6.76 175.67 ± 6.86 162.51 ± 6.32 162.45 ± 6.31 169.00 ± 9.11 168.46 ± 9.30 
Socioeconomic status quintilea 
 1 (least deprived) 2,776 (20.59) 37,976 (20.05) 2,569 (20.01) 45,303 (19.97) 5,345 (20.31) 83,279 (20.01) 
 2–4 7,933 (58.85) 112,163 (59.23) 7,806 (60.79) 137,325 (60.54) 15,739 (59.80) 249,488 (59.95) 
 5 (most deprived) 2,771 (20.56) 39,223 (20.71) 2,465 (19.20) 44,191 (19.48) 5,236 (19.89) 83,414 (20.04) 
Family history of cancer 
 Yes 5,271 (39.10) 63,513 (33.54) 5,168 (40.25) 78,578 (34.64) 10,439 (39.66) 142,091 (34.14) 
 No 8,209 (60.90) 125,849 (66.46) 7,672 (59.75) 148,241 (65.36) 15,881 (60.34) 274,090 (65.86) 
Healthy lifestyle factors 
 No current smoking 5,739 (42.57) 99,244 (52.41) 7,236 (56.36) 140,813 (62.08) 12,975 (49.30) 240,057 (57.68) 
 No alcohol consumption 280 (2.08) 5,440 (2.87) 744 (5.79) 13,292 (5.86) 1,024 (3.89) 18,732 (4.50) 
 Normal BMI 9,831 (72.93) 140,807 (74.36) 9,351 (72.83) 171,982 (75.82) 19,182 (72.88) 312,789 (75.16) 
 Regular physical activity 8,607 (63.85) 122,437 (64.66) 7,938 (61.82) 143,604 (63.31) 16,545 (62.86) 266,041 (63.92) 
 Healthy diet 2,505 (18.58) 37,373 (19.74) 3,688 (28.72) 67,193 (29.62) 6,193 (23.53) 104,566 (25.13) 
Healthy lifestyleb 
 Favorable 822 (6.10) 14,477 (7.65) 1,357 (10.57) 28,514 (12.57) 2,179 (8.28) 42,991 (10.33) 
 Intermediate 8,492 (63.00) 126,193 (66.64) 8,509 (66.27) 154,252 (68.01) 17,001 (64.59) 280,445 (67.39) 
 Unfavorable 4,166 (30.91) 48,692 (25.71) 2,974 (23.16) 44,053 (19.42) 7,140 (27.13) 92,745 (22.28) 
Cancer polygenic risk scorec, mean (SD) 37.24 ± 1.38 36.91 ± 1.33 30.89 ± 1.15 30.69 ± 1.13 34.14 ± 3.42 33.52 ± 3.33 
 Low 1,984 (14.72) 38,585 (20.38) 2,086 (16.25) 45,846 (20.21) 4,070 (15.46) 84,431 (20.29) 
 Intermediate 7,729 (57.34) 113,975 (60.19) 7,486 (58.30) 136,309 (60.10) 15,215 (57.81) 250,284 (60.14) 
 High 3,767 (27.95) 36,802 (19.43) 3,268 (25.45) 44,664 (19.69) 7,035 (26.73) 81,466 (19.57) 

aSocioeconomic status were assessed with the Townsend deprivation index, which combines information on social class, employment, car availability, and housing.

bHealthy lifestyle was defined as favorable (4 to 5 healthy lifestyle factors), intermediate (2 to 3 healthy lifestyle factors), and unfavorable (0 to 1 healthy lifestyle factor).

cCancer polygenic risk score was defined as low (the bottom quintile), intermediate (quintiles 2 to 4), and high (the top quintile) by sex.

Performance of site-specific cancer PRS

A total of 721 SNPs were used for PRSs construction (Supplementary Fig. S2), 8.04% (58/721) of which appear or are in strong LD (r2 > 0.6) with each other in more than one cancer site (Supplementary Table S4). As a result, we created 20 site-specific cancer PRSs by using 2∼139 SNPs separately, including 16 PRSs for men and 18 PRSs for women (Supplementary Table S1; Fig. 1). The site-specific PRSs were independent from each other (correlation coefficient r2 < 0.2; Supplementary Fig. S3), and most participants (97.17%, 429,984/442,501) were at high genetic risk (the top quintile of the PRS distribution) for at least one cancer (Supplementary Table S5). Almost all of the 20 PRSs had approximately normal distributions and were significantly associated with site-specific cancer risk (P < 0.05; Supplementary Fig. S4). Relatively strong association strengths (relative risks of at least 1.5 per SD) were observed for cancer sites of colorectum, skin malignant melanoma, prostate, testis, brain and central nerves, thyroid, and lymphoid leukemia in men; and colorectum, pancreas, skin malignant melanoma, brain and central nerves, thyroid, multiple myeloma, and lymphoid leukemia in women (Supplementary Table S6). About half of the cancer-specific PRSs (7/16 for men, and 9/18 for women) were also significantly associated with the overall cancer risk but with relatively modest effects, and the strongest associations were observed for prostate cancer PRS in men and breast cancer PRS for women (Supplementary Table S7).

Associations of sex-specific CPRSs with overall cancer risk

Participants with incident cancer tended to have a higher CPRS than those without incident cancer (Fig. 2A and B), and those with a high CPRS tended to be susceptible to multiple cancers in both men and women (Supplementary Table S8). Sex-specific CPRSs were significantly associated with incident cancer risk in men [HR, 1.27 (95% CI, 1.25–1.29) per SD increase, P < 0.0001] and women [HR, 1.20 (95% CI, 1.18–1.22) per SD increase, P < 0.0001]. We also observed a significantly gradient increase in incident cancer risk from quintile 1 to quintile 5 of CPRSs (Supplementary Fig. S5). Compared with individuals at low genetic risk (the bottom quintile of CPRSs), those in the intermediate (quintiles 2 to 4) and high genetic risk (the top quintile) had a significantly higher risk of overall cancer, with HRs of 1.27 (95% CI, 1.21–1.34, P < 0.0001) and 1.91 (95% CI, 1.81–2.02, P < 0.0001) in men (Fig. 2C), 1.21 (95% CI, 1.16–1.27, P < 0.0001) and 1.62 (95% CI, 1.54–1.71, P < 0.0001) in women (Fig. 2D), respectively. These results did not change after adjustment for lifestyle factors (Supplementary Table S9). When transformed to lifetime risk, the absolute cumulative risk by age 75 was estimated to be 19.27%, 25.10%, and 38.32% in men, and 15.83%, 20.86% and 29.17% in women, for low, intermediate, and high genetic risk individuals, respectively, on average (Supplementary Fig. S6). These results did not change significantly when genetic risk was reclassified by quartiles or tertiles of the CPRS (Supplementary Table S10), or the analysis were restricted to participants of unrelated British ancestry (Supplementary Table S11). Furthermore, a similar, but somewhat weaker association was observed, when we reconstructed the CPRSs without weighting by individual cancer incidence (Supplementary Table S12), by standardizing the mean of each site-specific cancer PRS to 1 (Supplementary Table S13), or by excluding SNPs that appear or in strong LD in more than one cancer (Supplementary Table S14).

Figure 2.

Effect of genetic and lifestyle factors on the risk of incident cancer in the UKB cohort. The distrubution of cancer polygenic risk scores between participants with incident cancer and those without incident cancer in the UKB cohort for men (A) and women (B). Standardized rates of cancer events in low (bottom quintile), intermediate (quintiles 2 to 4), and high (top quintile) genetic risk groups in the UKB cohort for men (C) and women (D). Standardized rates of cancer events in favorable (4 or 5 healthy lifestyle factors), intermediate (2 or 3 healthy lifestyle factors), and unfavorable (0 or 1 healthy lifestyle factor) lifestyle groups in the UKB cohort for men (E) and women (F). HRs and 95% CIs were estimated using Cox proportional hazard models with adjustment for age, height, family history of cancer, Townsend deprivation index, and the first 10 principal components of ancestry. Shaded areas are 95% CIs.

Figure 2.

Effect of genetic and lifestyle factors on the risk of incident cancer in the UKB cohort. The distrubution of cancer polygenic risk scores between participants with incident cancer and those without incident cancer in the UKB cohort for men (A) and women (B). Standardized rates of cancer events in low (bottom quintile), intermediate (quintiles 2 to 4), and high (top quintile) genetic risk groups in the UKB cohort for men (C) and women (D). Standardized rates of cancer events in favorable (4 or 5 healthy lifestyle factors), intermediate (2 or 3 healthy lifestyle factors), and unfavorable (0 or 1 healthy lifestyle factor) lifestyle groups in the UKB cohort for men (E) and women (F). HRs and 95% CIs were estimated using Cox proportional hazard models with adjustment for age, height, family history of cancer, Townsend deprivation index, and the first 10 principal components of ancestry. Shaded areas are 95% CIs.

Close modal

Associations of lifestyle factors with cancer risk

In the UKB cohort, most participants had two or three (67.22%, 297,446/442,501) of the five healthy lifestyle factors at baseline (Table 1). All of the five healthy lifestyle factors were significantly associated with the risk of overall cancer (Supplementary Fig. S7), while none was associated with the aforementioned CPRSs (Supplementary Table S15). Besides, stronger association effects were observed for alcohol drinking in men and for BMI and physical activity in women (Supplementary Table S16). Among both men and women, similar gradient increase was observed for unhealthy lifestyle factors with overall cancer risk (Ptrend < 0 0.0001, Supplementary Fig. S8). Individuals were then categorized into favorable (4 to 5 healthy lifestyle factors), intermediate (2 to 3 healthy lifestyle factors), and unfavorable (0 to 1 healthy lifestyle factor) lifestyle categories. Participants with an unfavorable lifestyle were at higher incident cancer risk than those with a favorable lifestyle, with adjusted HRs of 1.38 (95% CI, 1.28–1.48, P < 0.0001) in men (Fig. 2E) and 1.46 (95% CI, 1.37–1.56, P < 0.0001) in women (Fig. 2F), respectively, and the associations did not change after adjustment for genetic risk (Supplementary Table S17). Similar patterns were noted in a series of sensitivity analyses with exclusion of incident cancer cases occurred during the first year of follow-up (Supplementary Table S18), or in the unimputed data (Supplementary Table S19).

Joint effect and interaction of genetic and lifestyle factors on overall cancer risk

The overall incident cancer risk associated with both genetic risk and unhealthy lifestyle increased in a dose-response manner (Fig. 3A and B). Of participants with a high genetic risk and an unfavorable lifestyle, the incidence rates of overall cancer per 100,000 person-years were estimated to be 1545.90 (95% CI, 1456.93–1638.81) in men and 1199.99 (95% CI, 1116.92–1287.57) in women versus 478.88 (95% CI, 393.36–577.37) in men and 536.01 (95% CI, 468.41–610.58) in women with low genetic risk and a favorable lifestyle. More than double risks [HR, 2.99 (95% CI, 2.45–3.64) in men, P < 0.0001; 2.38 (95% CI, 2.05–2.76) in women, P < 0.0001] were observed in participants with a high genetic risk and an unfavorable lifestyle, compared with those with a low genetic risk and a favorable lifestyle. In addition, an additive interaction between genetic and lifestyle factors on overall cancer risk was observed in women but not in men (Supplementary Table S20). Specifically, the RERI was 0.51 (95% CI, 0.23–0.78), accounting for 22% of the risk in women with both a high genetic risk and an unfavorable lifestyle.

Figure 3.

Risk of incident cancer according to genetic and lifestyle categories in the UKB cohort for men (A) and women (B). The HRs were estimated using Cox proportional hazard models with adjustment for age, height, family history of cancer, Townsend deprivation index, and the first 10 principal components of ancestry.

Figure 3.

Risk of incident cancer according to genetic and lifestyle categories in the UKB cohort for men (A) and women (B). The HRs were estimated using Cox proportional hazard models with adjustment for age, height, family history of cancer, Townsend deprivation index, and the first 10 principal components of ancestry.

Close modal

Benefits of adherence to a healthy lifestyle with overall incident cancer

In further stratification analyses by genetic risk category with unfavorable lifestyles as the reference group, we confirmed that favorable lifestyle was significantly associated with a lower incident cancer risk across genetic risk groups (Table 2). Among participants at a high genetic risk, the standardized 5-year incident cancer rates were 7.23% and 5.77% for men and women with an unfavorable lifestyle versus 5.51% and 3.69% for those with a favorable lifestyle, respectively. Similarly, among participants at a low genetic risk, the standardized 5-year incident cancer rates decreased from 4.14% and 3.42% for men and women, respectively, with an unfavorable lifestyle to 2.24% and 2.53% for those, respectively, with a favorable lifestyle. Similar patterns were noted by reclassifying genetic risk levels according to quartiles or tertiles of the CPRS (Supplementary Tables S21 and S22), or restricted the analysis to participants of unrelated British ancestry (Supplementary Table S23).

Table 2.

Incident cancer risk associated with lifestyle categories by genetic risk level.a

Low genetic riskIntermediate genetic riskHigh genetic risk
SexLifestyle categoryUnfavorableIntermediateFavorableUnfavorableIntermediateFavorableUnfavorableIntermediateFavorable
Men No. of cases/Person-years 636/71,950 1,239/185,033 109/22,761 2,420/215,687 4,833/557,959 476/62,667 1,110/71,803 2,420/183,140 237/20,098 
 HR Ref. 0.83 (0.76–0.92) 0.59 (0.48–0.73) Ref. 0.82 (0.79–0.87) 0.73 (0.66–0.80) Ref. 0.91 (0.85–0.98) 0.80 (0.69–0.92) 
 (95% CI)          
 P value  2.05 × 10−4 6.35 × 10−7  1.18 × 10−14 3.54 × 10−10  8.66 × 10−3 1.55 × 10−3 
 Ptrend 5.57 × 10−8 1.80 × 10−17 3.92 × 10−4 
 Absolute risk (%)-5 years (95% CI) 4.14 (3.79–4.48) 3.13 (2.93–3.33) 2.24 (1.81–2.67) 5.23 (5.00–5.45) 4.03 (3.90–4.16) 3.54 (3.21–3.86) 7.23 (6.78–7.69) 6.18 (5.90–6.46) 5.51 (4.80–6.23) 
 Absolute risk reduction (%)-5 years (95% CI) Ref. 1.00 (0.60–1.37) 1.90 (1.40–2.47) Ref. 1.19 (0.96–1.43) 1.69 (1.31–2.06) Ref. 1.06 (0.59–1.56) 1.72 (0.95–2.58) 
Women No. of cases/Person-years 475/65,638 1,386/226,242 225/41,977 1,734/195,610 4,938/677,990 814/123,685 765/63,750 2,185/223,688 318/41,458 
 HR Ref. 0.83 (0.74–0.92) 0.69 (0.59–0.81) Ref. 0.82 (0.78–0.87) 0.71 (0.65–0.77) Ref. 0.81 (0.75–0.88) 0.62 (0.54–0.70) 
 (95% CI)          
 P value  3.76 × 10−4 5.11 × 10−6  1.57 × 10−14 2.24 × 10−10  1.02 × 10−6 5.74 × 10−13 
 Ptrend 1.63 × 10−6 8.92 × 10−18 1.05 × 10−13 
 Absolute risk (%)-5 years (95% CI) 3.42 (3.09–3.74) 2.89 (2.72–3.07) 2.53 (2.19–2.87) 4.22 (4.00–4.43) 3.47 (3.35–3.58) 3.13 (2.91–3.35) 5.77 (5.34–6.20) 4.70 (4.47–4.92) 3.69 (3.27–4.10) 
 Absolute risk reduction (%)-5 years (95% CI) Ref. 0.52 (0.16–0.85) 0.89 (0.44–1.34) Ref. 0.75 (0.53–0.97) 1.08 (0.80–1.39) Ref. 1.07 (0.62–1.54) 2.08 (1.50–2.66) 
Low genetic riskIntermediate genetic riskHigh genetic risk
SexLifestyle categoryUnfavorableIntermediateFavorableUnfavorableIntermediateFavorableUnfavorableIntermediateFavorable
Men No. of cases/Person-years 636/71,950 1,239/185,033 109/22,761 2,420/215,687 4,833/557,959 476/62,667 1,110/71,803 2,420/183,140 237/20,098 
 HR Ref. 0.83 (0.76–0.92) 0.59 (0.48–0.73) Ref. 0.82 (0.79–0.87) 0.73 (0.66–0.80) Ref. 0.91 (0.85–0.98) 0.80 (0.69–0.92) 
 (95% CI)          
 P value  2.05 × 10−4 6.35 × 10−7  1.18 × 10−14 3.54 × 10−10  8.66 × 10−3 1.55 × 10−3 
 Ptrend 5.57 × 10−8 1.80 × 10−17 3.92 × 10−4 
 Absolute risk (%)-5 years (95% CI) 4.14 (3.79–4.48) 3.13 (2.93–3.33) 2.24 (1.81–2.67) 5.23 (5.00–5.45) 4.03 (3.90–4.16) 3.54 (3.21–3.86) 7.23 (6.78–7.69) 6.18 (5.90–6.46) 5.51 (4.80–6.23) 
 Absolute risk reduction (%)-5 years (95% CI) Ref. 1.00 (0.60–1.37) 1.90 (1.40–2.47) Ref. 1.19 (0.96–1.43) 1.69 (1.31–2.06) Ref. 1.06 (0.59–1.56) 1.72 (0.95–2.58) 
Women No. of cases/Person-years 475/65,638 1,386/226,242 225/41,977 1,734/195,610 4,938/677,990 814/123,685 765/63,750 2,185/223,688 318/41,458 
 HR Ref. 0.83 (0.74–0.92) 0.69 (0.59–0.81) Ref. 0.82 (0.78–0.87) 0.71 (0.65–0.77) Ref. 0.81 (0.75–0.88) 0.62 (0.54–0.70) 
 (95% CI)          
 P value  3.76 × 10−4 5.11 × 10−6  1.57 × 10−14 2.24 × 10−10  1.02 × 10−6 5.74 × 10−13 
 Ptrend 1.63 × 10−6 8.92 × 10−18 1.05 × 10−13 
 Absolute risk (%)-5 years (95% CI) 3.42 (3.09–3.74) 2.89 (2.72–3.07) 2.53 (2.19–2.87) 4.22 (4.00–4.43) 3.47 (3.35–3.58) 3.13 (2.91–3.35) 5.77 (5.34–6.20) 4.70 (4.47–4.92) 3.69 (3.27–4.10) 
 Absolute risk reduction (%)-5 years (95% CI) Ref. 0.52 (0.16–0.85) 0.89 (0.44–1.34) Ref. 0.75 (0.53–0.97) 1.08 (0.80–1.39) Ref. 1.07 (0.62–1.54) 2.08 (1.50–2.66) 

aCox proportional hazards regression is adjusted for age, height, family history of cancer, Townsend deprivation index, and the first 10 principal components of ancestry.

Genetic risk and healthy lifestyle act in concert to influence incident cancer risk. In this study, we systematically created site-specific PRSs for 20 cancer types, and constructed CPRSs to assess the effect of genetic risk on overall incident cancer risk in the UKB cohort. The results of this study indicated that a high genetic risk defined by CPRSs was associated with an increased incident cancer risk independent of lifestyle in the large community-based population. Participants with a high genetic risk and an unfavorable lifestyle had the greatest incident cancer risk, compared with those with a low genetic risk and a favorable lifestyle. Within and across genetic risk groups, adherence to a healthy behavioral lifestyle was consistently associated with a reduced absolute incident cancer risk.

The increasing availability of genomics and biobanks offers opportunities to assess genetic risk of cancer. PRSs have been developed to measure genetic risk for a series of site-specific cancers (2–4, 11–13, 21). Recently, a pan-cancer integration analysis showed an improved efficiency after integrating PRS into personalized cancer risk assessment (22). Allison and colleagues constructed a composite PRS for 13 diseases and 12 mortality risk factors by combining PRS of each trait directly in a recent analysis (23). However, genetic risk score for overall incident cancer risk have not been reported to date. To our knowledge, this study is the first to provide convincing evidence that CPRSs can be used to assess the effect of genetic risk on overall incident cancer risk. The PRSs of site-specific cancer, incorporating with clinical risk predictors, could be used to improve risk classifications and optimize established screening programs, such as cancers of the breast (24), colorectum (25), lung (26), and prostate (27), where clinical trials are already underway (28). Moreover, our results indicate that the CPRSs could improve a person's awareness of their inherited susceptibility to cancer as a whole and encourage them to participate in precision prevention strategies. Genetic-based risk assessment can help define cancer risks for individuals and families and facilitate decision-making about risk management options (29).

Previous study has demonstrated that around 4 in 10 cases of cancer could be prevented in the UK each year through lifestyle changes (30). In consistent with the finding, as well as those of previous studies of site-specific cancers [including cancers of the breast (11), colorectum (12), and stomach (13)], we demonstrated that adherence to a healthy lifestyle could reduce overall cancer risk within and across genetic risk groups and that participants at a high genetic risk would also benefit from adopting healthy lifestyles. Of interest, we observed an additive interaction between genetic and lifestyle factors on overall cancer risk in women, suggesting that women at high genetic risk may benefit more from healthy lifestyles. Particularly, adherence to favorable physical exercises and a moderate BMI would be more helpful to women, whereas stopping drinking alcohol would be more important for men. Taken together, our findings as well as previous evidence not only suggest the importance of interpreting genetic risk in individual-based risk assessment but also provide collective support for public efforts in promoting a healthy lifestyle for everyone, which will eventually lead to a reduction of overall cancer burden (18, 31).

Our work also demonstrated that few participants could be exempted from high genetic risk of any cancer, and participants at a high risk of CPRSs were susceptible to more cancer types. Fortunately, the detrimental effect of genetic risk can be largely offset by a healthy lifestyle. Genetic risk exists at birth, whereas lifestyle usually formulates in midlife. Therefore, effective behavioral interventions for individuals at a high genetic risk, especially before the formation of unhealthy lifestyle in early life, were probably to be effective precautions for cancer prevention. Nevertheless, additional evidence is needed to evaluate the impact of disclosing genetic information to and risk communication with individuals prior to the deployment in different healthcare settings. Furthermore, challenge remains in refining the CPRS and accounting for only a small fraction of susceptibility loci that have been identified for most cancers by now (32).

This study has several strengths, including a large sample size, a prospective design of the UK Biobank study, and standardized protocols to assess heritable and lifestyle factors simultaneously as well as a novel approach to assess genetic risk of overall cancer burden. Nevertheless, we also acknowledge several limitations. First, we only included the lead GWAS variants in the construction of site-specific cancer PRSs, which may be enhanced by adopting less stringent P value thresholding to include additional risk variants, or implementing more sophisticated PRS models that incorporate linkage disequilibrium structure with genome-wide variants, as reported recently (5, 33). Second, the dramatically different number of susceptibility loci for site-specific cancer types may bias the representativeness of CPRSs. Third, some of the lifestyle factors were self-reported, and therefore, might result in misclassification of lifestyle risk levels. Forth, lifestyle factors were assessed on a single measure at baseline, and behavioral changes during the follow-up or competing risks of other illnesses may have an effect on risk estimates. Fifth, evidences have shown that the population sample of UK Biobank differs from the general UK population because of low participation and healthy volunteer bias (34); therefore, further investigation is warranted to evaluate to what degree these findings may be generalized to the general UK population. Finally, even though the findings were achieved among UK Biobank participants with diverse ethnic backgrounds and reevaluated among those of unrelated British ancestry in sensitivity analysis, the generalizability of our findings should be further assessed in more diverse populations when available.

In summary, we constructed the genetic indicator CPRS for overall cancer risk and demonstrated that detrimental effect of genetic risk on incident cancer risk could be attenuated by adherence to a healthy lifestyle. These results indicate that public policies for improving food and physical environment, as well as relevant policies on tobacco control and alcohol restriction, are critical to reduce the burden of cancer in the general population.

No disclosures were reported.

M. Zhu: Conceptualization, formal analysis, supervision, funding acquisition, validation, writing–original draft, writing–review and editing. T. Wang: Data curation, formal analysis, validation, visualization, writing–original draft, writing–review and editing. Y. Huang: Data curation, formal analysis, writing–review and editing. X. Zhao: Data curation, writing–review and editing. Y. Ding: Data curation, writing–review and editing. M. Zhu: Data curation, writing–review and editing. M. Ji: Data curation, writing–review and editing. C. Wang: Methodology, writing–review and editing. J. Dai: Methodology, writing–review and editing. R. Yin: Supervision, writing–review and editing. L. Xu: Supervision, writing–review and editing. H. Ma: Supervision, project administration, writing–review and editing. Q. Wei: Conceptualization, supervision, writing–review and editing. G. Jin: Conceptualization, formal analysis, funding acquisition, validation, writing–original draft, writing–review and editing. Z. Hu: Conceptualization, supervision, project administration, writing–review and editing. H. Shen: Conceptualization, resources, supervision, funding acquisition, project administration, writing–review and editing.

The authors thank the investigators and participants in UK Biobank for their contributions to this study. This research was conducted using the UK Biobank Resource (Application Number: 60169). UK Biobank has received ethics approval from the Research Ethics Committee (ref. 11/NW/0382). This work was supported by National Natural Science Foundation of China (81820108028, 81521004 to H. Shen); Natural Science Foundation of Jiangsu Province (BK20180675 to M. Zhu); National Key Research and Development Program of China (2016YFC1302703 to G. Jin; 2018YFC1315002 to M. Zhu); 333 High-Level Talents Cultivation Project of Jiangsu Province (BRA2018057 to G. Jin); CAMS Innovation Fund for Medical Sciences (2019RU038 to G. Jin); National Science Foundation for Post-doctoral Scientists of China (2018M640466 to M. Zhu).

The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Mucci
LA
,
Hjelmborg
JB
,
Harris
JR
,
Czene
K
,
Havelick
DJ
,
Scheike
T
, et al
Familial risk and heritability of cancer among twins in Nordic countries
.
JAMA
2016
;
315
:
68
76
.
2.
Dai
J
,
Lv
J
,
Zhu
M
,
Wang
Y
,
Qin
N
,
Ma
H
, et al
Identification of risk loci and a polygenic risk score for lung cancer: a large-scale prospective cohort study in Chinese populations
.
Lancet Respir Med
2019
;
7
:
881
91
.
3.
Lecarpentier
J
,
Silvestri
V
,
Kuchenbaecker
KB
,
Barrowdale
D
,
Dennis
J
,
McGuffog
L
, et al
Prediction of breast and prostate cancer risks in male BRCA1 and BRCA2 mutation carriers using polygenic risk scores
.
J Clin Oncol
2017
;
35
:
2240
50
.
4.
Mars
N
,
Koskela
JT
,
Ripatti
P
,
Kiiskinen
TTJ
,
Havulinna
AS
,
Lindbohm
JV
, et al
Polygenic and clinical risk scores and their impact on age at onset and prediction of cardiometabolic diseases and common cancers
.
Nat Med
2020
;
26
:
549
57
.
5.
Khera
AV
,
Chaffin
M
,
Aragam
KG
,
Haas
ME
,
Roselli
C
,
Choi
SH
, et al
Genome-wide polygenic scores for common diseases identify individuals with risk equivalent to monogenic mutations
.
Nat Genet
2018
;
50
:
1219
24
.
6.
Khoury
MJ
,
Armstrong
GL
,
Bunnell
RE
,
Cyril
J
,
Iademarco
MF
. 
The intersection of genomics and big data with public health: Opportunities for precision public health
.
PLoS Med
2020
;
17
:
e1003373
.
7.
Lewis
CM
,
Vassos
E
. 
Polygenic risk scores: from research tools to clinical instruments
.
Genome Med
2020
;
12
:
44
.
8.
Wu
S
,
Zhu
W
,
Thompson
P
,
Hannun
YA
. 
Evaluating intrinsic and non-intrinsic cancer risk factors
.
Nat Commun
2018
;
9
:
3490
.
9.
Islami
F
,
Goding Sauer
A
,
Miller
KD
,
Siegel
RL
,
Fedewa
SA
,
Jacobs
EJ
, et al
Proportion and number of cancer cases and deaths attributable to potentially modifiable risk factors in the United States
.
CA Cancer J Clin
2018
;
68
:
31
54
.
10.
Li
Y
,
Schoufour
J
,
Wang
DD
,
Dhana
K
,
Pan
A
,
Liu
X
, et al
Healthy lifestyle and life expectancy free of cancer, cardiovascular disease, and type 2 diabetes: prospective cohort study
.
BMJ
2020
;
368
:
l6669
.
11.
Arthur
RS
,
Wang
T
,
Xue
X
,
Kamensky
V
,
Rohan
TE
. 
Genetic factors, adherence to healthy lifestyle behavior, and risk of invasive breast cancer among women in the UK Biobank
.
J Natl Cancer Inst
2020
Sep 1
;
112
:
893
901
.
12.
Carr
PR
,
Weigl
K
,
Jansen
L
,
Walter
V
,
Erben
V
,
Chang-Claude
J
, et al
Healthy lifestyle factors associated with lower risk of colorectal cancer irrespective of genetic risk
.
Gastroenterology
2018
;
155
:
1805
15
.
13.
Jin
G
,
Lv
J
,
Yang
M
,
Wang
M
,
Zhu
M
,
Wang
T
, et al
Genetic risk, incident gastric cancer, and healthy lifestyle: a meta-analysis of genome-wide association studies and prospective cohort study
.
Lancet Oncol
2020
;
21
:
1378
86
.
14.
Bycroft
C
,
Freeman
C
,
Petkova
D
,
Band
G
,
Elliott
LT
,
Sharp
K
, et al
The UK Biobank resource with deep phenotyping and genomic data
.
Nature
2018
;
562
:
203
9
.
15.
Choi
SH
,
Weng
LC
,
Roselli
C
,
Lin
H
,
Haggerty
CM
,
Shoemaker
MB
, et al
Association between Titin Loss-of-Function variants and early-onset atrial fibrillation
.
JAMA
2018
;
320
:
2354
64
.
16.
Shams-White
MM
,
Brockton
NT
,
Mitrou
P
,
Romaguera
D
,
Brown
S
,
Bender
A
, et al
Operationalizing the 2018 World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) cancer prevention recommendations: a standardized scoring system
.
Nutrients
2019
;
11
:
1572
.
17.
Lourida
I
,
Hannon
E
,
Littlejohns
TJ
,
Langa
KM
,
Hyppönen
E
,
Kuzma
E
, et al
Association of lifestyle and genetic risk with incidence of dementia
.
JAMA
2019
;
322
:
430
7
.
18.
Said
MA
,
Verweij
N
,
van der Harst
P
. 
Associations of combined genetic and lifestyle risks with incident cardiovascular disease and diabetes in the UK Biobank study
.
JAMA Cardiol
2018
;
3
:
693
702
.
19.
Price
AL
,
Patterson
NJ
,
Plenge
RM
,
Weinblatt
ME
,
Shadick
NA
,
Reich
D
. 
Principal components analysis corrects for stratification in genome-wide association studies
.
Nat Genet
2006
;
38
:
904
9
.
20.
Li
R
,
Chambless
L
. 
Test for additive interaction in proportional hazards models
.
Ann Epidemiol
2007
;
17
:
227
36
.
21.
Jia
G
,
Lu
Y
,
Wen
W
,
Long
J
,
Liu
Y
,
Tao
R
, et al
Evaluating the utility of polygenic risk scores in identifying high-risk individuals for eight common cancers
.
JNCI Cancer Spectr
2020
;
4
:
pkaa021
.
22.
Linda
K
,
Rebecca
EG
,
Karl
SB
,
Meyers
TJ
,
Rashkin
SR
,
Ziv
E
, et al
Pan-cancer analysis demonstrates that integrating polygenic risk scores with modifiable risk factors improves risk prediction
.
Nat Commun
2020
;
11
:
6084
.
23.
Meisner
A
,
Kundu
P
,
Zhang
YD
,
Lan
LV
,
Kim
S
,
Ghandwani
D
, et al
Combined utility of 25 disease and risk factor polygenic risk scores for stratifying risk of all-cause mortality
.
Am J Hum Genet
2020
;
107
:
418
31
.
24.
Siu
AL
,
Force
USPSTC
. 
Screening for breast cancer: U.S. preventive services task force recommendation statement
.
Ann Intern Med
2016
;
164
:
279
96
.
25.
U.S. Preventive Services Task Force
,
Bibbins-Domingo
K
,
Grossman
DC
,
Curry
SJ
,
Davidson
KW
,
Epling
JW
 Jr
,
García
FAR
, et al
Screening for colorectal cancer: U.S. Preventive Services Task Force recommendation statement
.
Ann Intern Med
2008
;
149
:
627
37
.
26.
National Lung Screening Trial Research Team
,
Aberle
DR
,
Adams
AM
,
Berg
CD
,
Black
WC
,
Clapp
JD
, et al
Reduced lung-cancer mortality with low-dose computed tomographic screening
.
N Engl J Med
2011
;
365
:
395
409
.
27.
Ilic
D
,
Djulbegovic
M
,
Jung
JH
,
Hwang
EC
,
Zhou
Q
,
Cleves
A
, et al
Prostate cancer screening with prostate-specific antigen (PSA) test: a systematic review and meta-analysis
.
BMJ
2018
;
362
:
k3519
.
28.
Shieh
Y
,
Eklund
M
,
Madlensky
L
,
Sawyer
SD
,
Thompson
CK
,
Stover Fiscalini
A
, et al
Breast cancer screening in the precision medicine era: risk-based screening in a population-based trial
.
J Natl Cancer Inst
2017
;
109
:
djw290
.
29.
Kaphingst
KA
,
Peterson
E
,
Zhao
J
,
Gaysynsky
A
,
Elrick
A
,
Hong
SJ
, et al
Cancer communication research in the era of genomics and precision medicine: a scoping review
.
Genet Med
2019
;
21
:
1691
8
.
30.
Brown
KF
,
Rumgay
H
,
Dunlop
C
,
Ryan
M
,
Quartly
F
,
Cox
A
, et al
The fraction of cancer attributable to modifiable risk factors in England, Wales, Scotland, Northern Ireland, and the United Kingdom in 2015
.
Br J Cancer
2018
;
118
:
1130
41
.
31.
Khera
AV
,
Emdin
CA
,
Drake
I
,
Natarajan
P
,
Bick
AG
,
Cook
NR
, et al
Genetic risk, adherence to a healthy lifestyle, and coronary disease
.
N Engl J Med
2016
;
375
:
2349
58
.
32.
McGuire
AL
,
Gabriel
S
,
Tishkoff
SA
,
Wonkam
A
,
Chakravarti
A
,
Furlong
EEM
, et al
The road ahead in genetics and genomics
.
Nat Rev Genet
2020
;
21
:
581
96
.
33.
Vilhjalmsson
BJ
,
Yang
J
,
Finucane
HK
,
Gusev
A
,
Lindström
S
,
Ripke
S
, et al
Modeling linkage disequilibrium increases accuracy of polygenic risk scores
.
Am J Hum Genet
2015
;
97
:
576
92
.
34.
Fry
A
,
Littlejohns
TJ
,
Sudlow
C
,
Doherty
N
,
Adamska
L
,
Sprosen
T
, et al
Comparison of sociodemographic and health-related characteristics of UK biobank participants with those of the general population
.
Am J Epidemiol
2017
;
186
:
1026
34
.