Genetic risk information for common diseases may indeed be already useful for prevention and early detection

Several genetic risk tests have been launched recently for common diseases such as breast cancer, prostate cancer, atrial fibrillation and myocardial infarction. Most agree that the genetic markers used are well-validated in the sense that the genotyping can be performed accurately and the markers used have been widely replicated as significant risk factors in several populations including tens of thousands of patients and controls. The relative risks conferred by the genetic risk profiles in these diseases equals or exceeds that of most conventional risk factors and are independent of them, including paradoxically, immediate family history. However, some have questioned the wisdom of introducing such tests when there are more sequence variants to be discovered or they would like to see more evidence confirming clinical utility vs. potential harms. Indeed, variants discovered in the future using genome-wide sequencing methods once costs come down will complement the common variants already reported and we agree that information from ongoing and future clinical utility studies will further establish how genetic risk tests complement conventional risk factors. Nevertheless, these tests further refine our assessment of risk and represent tools for more cost-effective prioritization of scarce resources for early detection and prevention of some of the most costly and deadly common diseases. Furthermore, we believe, it is the physician who is best positioned to interpret and act on these new genetic risk tests for common diseases. In the 1980s and 1990s, there was an avalanche of disease gene discoveries for many of the rare Mendelian diseases such as cystic fibrosis, Duchenne's muscular dystrophy and Fragile X mental retardation. However, these were diseases generally caused by high-penetrance variants in a single gene. The disease genes contain determinative variants with very high penetrance – for example, patients who have the trinucleotide repeat mutation within the Huntington's disease gene, will all develop the characteristic movement disorder, chorea, as well as dementia, no matter what they do earlier in life. Conversely, patients who do not possess mutations in the Huntington's disease gene will never develop Huntington's disease, no matter what their environment is. Now, we have a new avalanche of genetic discoveries, this time in the common diseases. Common diseases generally do not follow a determinative genetic inheritance pattern like the Mendelian diseases; if anything, they may skip one or more generations or have no obvious family history. On the basis of large twin and family studies, we know that there is a significant genetic component to the risk of all common diseases; in fact, our genealogical analyses of extended families of many common diseases show that this component extends far beyond the nuclear family including up to 2nd cousins and much of it is not accounted for by family history of the disease in first degree relatives 1–6. This paradox can be explained by the confluence of multiple genetic factors and the environment. Over the last few years, we and others have carried out numerous hypothesis-independent genome-wide association and linkage studies using up to a million genetic markers including single nucleotide polymorphisms (SNP), leaving almost no common variation of the genome unturned. In many of the most common diseases, several genetic markers that confer significant risk have been found and widely replicated by a large number of major research groups. Just our group alone, using the Icelandic population as the discovery population and replicating the findings in collaboration with over 300 investigators around the world, has reported over 150 confirmed SNPs covering 30 common diseases and traits. These discoveries have expanded our knowledge of the pathogenesis of common diseases. Currently, there are several common diseases where the discovered sequence variants account for a substantial proportion of overall risk and that are independent of conventional risk factors including immediate family history. While more sequence variants are expected to be discovered, we have already reached the point where genetic markers may be used to define some higher risk individuals before they become patients and display symptoms and signs of disease. These genetic markers reflect the patient's intrinsic risk and are based on measuring the most stable molecule in the human body, DNA, and unlike many protein and RNA-based tests, do not show biological variation day-to-day or according to fasting or fed states. Furthermore, germline DNA markers do not need to be re-measured throughout the patient's life – a given set of variants can be measured once at any time in an individual's life using a blood sample or simple inner cheek swab. Therefore, genetic profiles may be useful additions to conventional risk assessments in certain clinical situations. Without a doubt, the best way to deal with cancer is to detect it early. Prostate cancer would be considered a preventable disease if we found a way to detect and treat intermediate and high-grade tumours before they spread beyond the prostatic capsule. There is no chemotherapeutic agent that works for prostate cancer – once it spreads, a high percentage of patients die unnecessarily and the high prevalence of bone metastasis leads to a miserable and very costly quality of life during the long years before death. However, if caught and treated early, prostate cancer is curable. Every cancer found early vs. at a later stage saves the healthcare system hundreds of thousands of dollars. Widespread screening using the imperfect biomarker, prostate specific antigen (PSA) has contributed to a 35% drop in the population age-adjusted mortality rate per 100 000 (according to the NIH supported SEER database http://seer.cancer.gov/statistics). This translated to a drop in annual prostate cancer deaths in the US from 40 000 in 1990 to 27 000 last year. The role of PSA and digital rectal examination screening in this dramatic drop in cancer death was supported by the PSA screening randomized clinical trial in Europe which showed a 20–27% drop in mortality in the screening vs. no-screening groups 7 (the NCI-run PLCO study so far has not demonstrated a difference in mortality; however, this trial has fewer observed endpoints and because of poor compliance, only 85% of the screening group had any PSA screening while 50% of the 'standard of care control group' was also screened as intensively by PSA, severely reducing the power of the study 8. Many diagnosed prostate cancer patients will die of other causes than the cancer because prostate cancer does not progress to death as fast as many other cancers and most patients are older than 60. This has led to a call from some for active surveillance for evidence of progression in patients with the intermediate grade cancer before recommending definitive treatment 9. However, a randomized clinical trial of patients diagnosed with intermediate grade prostate cancer in Scandinavia showed that the mortality rate in the watchful waiting group was twice that of the surgical treatment group when cancer was diagnosed before 65 years of age 10. Despite screening for prostate cancer in the US, it remains the second deadliest cancer for men. In contrast to many other common diseases, there are not many factors available to risk stratify men for prostate cancer. The only conventional risk factors are family history of early-onset prostate cancer in first degree relatives and ethnicity (especially African-Americans). Therefore, fewer than 5% of white males are considered higher risk (about twofold) because of family history and by default, 95% are labelled as average risk. Higher risk men are recommended to have earlier screening American Cancer Society (ACS), American Urological Association (AUA) and to be biopsied at borderline levels of PSA National Comprehensive Cancer Network (NCCN). Using large-scale genome-wide linkage and case-control association studies on prostate cancer, we and others have discovered 25 common germline variants that contribute to the risk of prostate cancer 11–21. Each of these markers has been replicated by us and others in 8–15 populations totalling 20 000 patients and 50 000 controls. The risk compared to the general population at each marker has been estimated using these large datasets and it ranges from 1·14 to 2·65 in those who are homozygous. The risks may be combined by simple multiplication. This is justified because no interaction has been found between the markers and no model fits the data better than the multiplicative one (log-additive) 16,17,20 (Fig. 1). The nine significant markers found by the initial genome-wide association studies (GWAS) studies have a greater effect than the 16 new markers that have been found later through large-scale meta-analyses 20. The genetic profile defines risk spanning 0·2 to over fourfold compared to the general population risk (Fig. 1). Two of these markers also define patients who have modestly increased likelihood of aggressive vs. less aggressive cancer at diagnosis 11,15. Prostate cancer risk across the general population based on a genetic risk profile. The 25 validated SNP markers together define relative risks from 0.2 fold to over fourfold compared to the general population. The x-axis shows the percentile in the general male population and the y-axis shows the relative risk at each percentile. Based on the current 25 replicated SNP markers, the upper 15th percentile of the population averages 2·1 fold risk for prostate cancer, top 5 percentile 2·8 fold, and the top one percentile 3·9 fold risk 20. Given that white individuals have a 17% lifetime risk (SEER database), this translates into 34, 47 and 66% lifetime risk respectively. However, this test is not determinative. High-risk patients are not destined to develop prostate cancer and low-risk patients are not immune from it. Rather, this is a risk test more analogous to LDL-cholesterol for cardiovascular risk than a genetic test for Huntington's disease. For that reason, many physicians will be able to use such tests in clinical practice without a re-education programme or a requirement for genetic counsellors (although the latter group may indeed give support to physicians). The physician is the ideal health care provider to assess and put into context all risk factors as well as to suggest early prevention or detection strategies to their patients, just as they already do today. It is easier to emphasize to physicians that this is a risk test like many other clinical risk factors than to train tens of thousands of genetic counsellors and educate them on how to manage risk factors in patients. This genetic profile is independent of immediate family history. It is likely that family history will be accounted for by rarer variants of larger effect that have not been found yet. However, this profile already accounts for three times more prostate cancer than family history and has a comparable level of risk. Therefore, this test complements family history by adding another 15% of the population who have twice the risk to the 5% of the population at twofold risk already based on family history alone, making this genetic risk profile the largest risk factor for prostate cancer to date. Together, higher risk patients comprising 20% of the population would account for about 40% of the prostate cancer burden and therefore more aggressive screening of this group could lead to a decrease in prostate cancer morbidity, mortality and the high costs of managing later stage prostate cancer (Fig. 2). Defining men with higher risk of prostate cancer based on either (i) family history of early prostate cancer alone or (ii) family history or upper 15th percentile of genetic risk based on 25 SNP markers, or both. The genetic risk profile is largely independent of family history – there is 1% overlap on this scale, but not shown. Note that the genetic risk profile defines another 15% of men without family history who have an average relative risk equivalent to that conferred by family history of early prostate cancer. Serum PSA and rectal exam screening is recommended by ACS to begin at age 50; however, they recommend earlier PSA screening in higher risk patients. Even more importantly, given the modest sensitivity of PSA for prostate cancer detection when using the conventional cut-off of 4 ng mL–1, high-risk patients with high normal PSA may benefit from a more aggressive evaluation with ultrasound and biopsy rather than watchful waiting or losing the patient to follow-up; this approach is already recommended by NCCN for higher risk patients based on family history and PSA between 2·6 and 4·0 ng mL–1. Adding this important risk factor to PSA evaluation may also improve the specificity of PSA as the prior probability of prostate cancer is much higher in some patients and the genetic test shows no correlation to benign prostatic hyperplasia (BPH), the major source of false-positive PSA elevation. Large clinical utility studies demonstrating the expected increase in PSA specificity and PPV are underway (i.e. higher positive biopsy or repeat biopsy rates in patients with higher genetic risk). Conversely, the patients in the lowest 35 percentile risk have half the risk (average RR of 0·5) and are expected to have a lower positive biopsy rate; therefore, these patients may be managed by watching for PSA rise over time rather than biopsy immediately, saving the health care system 1000–2000 dollars per deferred biopsy. About 80% of prostate biopsies are negative for cancer; more careful risk assessment might decrease the number and cost of invasive procedures in lower risk patients, while leading to higher yields among the higher risk patients. Several common variants have recently been discovered conferring risk of the common forms of breast cancer 22–28. Rare mutations in the BRCA1 and BRCA2 genes are often measured to define risk in women with strong family history of the early onset form of breast cancer. However, 95% of breast cancers do not fit this pattern – the vast majority of cancer patients do not have family history, much less a family history of early breast cancer. As in prostate cancer, the common variants for breast cancer fit a multiplicative model 29; the 12 strongest variants define risk of the common forms ranging from 0·4 fold to fourfold risk compared with general population (Fig. 3). Each of these sequence variants have been replicated in 5–25 populations totalling almost 100 000 patients and controls. The upper 5 percentile of the white female population has a twofold risk of the general population and about 0·5% has a threefold risk. Assuming a 12% lifetime risk (SEER database), this translates into a 24–36% lifetime risk, certainly not high enough to trigger prophylactic mastectomies, but high enough to suggest earlier and more intensive screening for the common forms of breast cancer. It is important to note that 5% of women are diagnosed before 40, the age at which mammography is recommended to begin in the US for women who do not have higher risk (20% occur before age 50, the start age of mammography in the UK and recently recommended by the US Preventive Services Task Force amid much controversy!). Most of these women do not have strong family history of breast cancer and are generally negative for BRCA1 and BRCA2 mutations, yet they are not often screened for breast cancer! Nevertheless, for every cancer diagnosed before spread, hundreds of thousands of dollars may be saved in terms of salvage therapy and management of the complications of metastasis. Common forms of breast cancer risk across the general population based on a genetic risk profile. The 12 validated SNP markers together define relative risks from 0.4 fold to fourfold compared to the general population. The x-axis shows the percentile in the general female population and the y-axis shows the relative risk at each percentile. Some women who are considered of average risk based on conventional risk factors will be reclassified as higher risk based on this genetic risk test. Higher risk patients may benefit from earlier screening by mammography and perhaps more intensive screening using breast MRI, which is 2–3 times more sensitive at picking up early breast cancer than mammography alone. ACS currently recommends breast MRI for women with 20% or greater lifetime risk (relative risk of 1·65) on the basis of conventional risk factors and suggests that physicians discuss risks and benefits of screening for women with 15% or more lifetime risk 30. ASCO recommends chemoprevention with tamoxifen and raloxifene for women with 5-year risk greater than 1·65% based on a Gail score. The current set of markers is independent of the factors used in the Gail score including family history, age at menarche and pregnancy history. Therefore, lifetime and 5-year risk may be better estimated by multiplying the risk from the genetic profile with the independent risk derived from conventional scales. As in prostate cancer, the negative biopsy rate of women with for cancer is over utility studies expected in specificity of breast negative are Conversely, patients with lower risk may have an greater negative biopsy rate; perhaps a better use of our resources would be of breast biopsy and the it costs much less than the 1000–2000 biopsy. of these 12 variants also the likelihood of positive breast cancer vs. negative cancer in the the patient cancer These are the first germline markers for the common forms of cancer that future as to the more such as that the who is at higher risk for breast cancer is also more likely to have positive cancer may further chemoprevention with a like tamoxifen and raloxifene that to positive cancer. The initial genome-wide association studies of myocardial and disease found only a single with significant – that within the the genes About of the general population of the risk variants on with fold risk compared to the of the population which forms This risk is in and in to the risk of LDL-cholesterol vs. lowest and is independent of conventional risk factors. we and others have in large meta-analyses that discovered several other SNPs of effect with have in our genetic risk test the and other widely replicated markers that are also independent of conventional risk factors such as family history, history and such as protein These variants have been widely replicated in 20 independent populations totalling over 20 000 patients and 40 000 controls. The overall genetic risk by all markers ranges from 0·4 to compared to the general population and has and greater than that of LDL-cholesterol and (Fig. risk for myocardial across the general population. for this marker profile only SNPs that have been to confer risk for independent of conventional risk factors. These validated markers a marker in the and together define relative risks from 0.4 fold to fold compared to the general population. The x-axis shows the percentile in the general population and the y-axis shows the relative risk at each percentile. patients are often into three or risk for risk of according to the using the or other risk using risk factors. The risk LDL-cholesterol for example, less than for risk and less than 100 for high risk than Two clinical utility studies reported that a showed a significant association with future independent of other risk factors The marker also reclassified of patients into risk that would lead to a in the recommended For example, of variants to the large by group in the 000 patients reclassified about 12% of intermediate and patients into higher or lower risk with LDL-cholesterol The study in the in the UK reclassified 20% of the patients into more risk that these rates will be higher with the new marker profile. Some suggest that it is early to use a genetic risk test for a common disease if we do not yet know all of the sequence variants the risk of that disease It is that variants will be discovered – the DNA used in this of studies only contain common variants and that we are rare and intermediate variants as well as variants number variants and however, the relative of the many common variants we have discovered vs. less common variants that we have not discovered is a of as we have no data to our assessment The that a much greater of genetic risk for the well common diseases remains to be found from the that only a of risk is explained by the common However, as we have reported through our using our and genealogical in common disease extends far beyond the nuclear family – much of only in a of the may be largely by rare variants with a strong while much of the genetic component of risk for most individuals who do not have family history of the disease may be to common genetic The confluence of many common risk of modest effect in some individuals high risk. The current genetic profiles can as higher risk a substantial of patients to be at average risk based on conventional factors. we of the genetic risk before using what we do know for risk would be analogous to not measuring risk factors like and for cardiovascular disease because of the that future may a higher risk patient as average risk In we generally risk factors as we and them, if they are useful for the and analyses that about the of genetic compared one with another and found between the reported for diseases in common However, they to point out that these use very of markers to define risk and that one has up with the discoveries and that we and others in the is a and our genetic profiles in our disease risk tests and our more are with of new validated and interpret another study at of genetic markers as of from high risk to risk and high risk as with risk greater than and risk as less than so a from to relative risk would be as a in risk. However, our profiles are with of markers and as it is to have for example, from higher risk greater than to lower risk less than Indeed, in the only of the higher risk group that was for the risk in the strongest genetic risk factor for to risk when other genetic markers were of do in as markers are validated and used vs. and and genetic profiles will do so as well further the of overall risk assessment including conventional risk factors. The of the suggest that many of these markers have been replicated as common risk factors in tens of thousands of patients and tens of thousands of in and studies, we randomized clinical are and to the years or and mortality of tens of thousands of patients managed with information from the genetic factors and risk factors vs. tens of thousands of patients managed using only risk factors. that more clinical utility studies can be to or how the genetic markers improve specificity of conventional screening tests as breast and PSA as to improve our of healthcare agree that and clinical utility studies to further define the evidence and clinical are important to further support and define use of genetic risk tests as is for any other test It be out with rare no risk test or test has been in randomized clinical out for mortality before or before only and clinical not In almost all clinical utility studies, in the form of studies and as were only the substantial them into breast MRI, MRI and The later studies to define further the of the benefits over potential harms. and that there may be several potential that can come from genetic risk Furthermore, they that this be less than the more obvious benefit that can in higher risk patients who can be put the of earlier detection and earlier However, for there has indeed been that genetic can lead to and and have the potential of including and the recently reported study showed that patients and given their for the strongest risk factor for disease, the not have rates of or from those who were not given their and to any studies that show that it is for to their of genetic risk assessment for common diseases. agree with and that high of that the benefits of new potential – a is to and need to be in thousands of before a is launched for However, we that genetic risk tests for common diseases are not determinative or of a disease – when used together with conventional risk they a more of the overall risk. has several where patients who certain risk are they earlier or more intensive screening with more definitive or a or As they point genetic tests are just like any other clinical tests, physicians to define overall risk more In their of prostate cancer screening and more on the of treatment than the significant of death found in the Prostate the that the of this study that benefit of screening is at if it at Note that prostate cancer screening is a to who the more definitive biopsy. a diagnosis of high grade or intermediate grade cancer is patient and physician can the relative of and of which can be vs. early death and largely be in men than at In our there is greater potential when more intensive or prevention strategies are from patients who are at higher risk and they and their physician do not know it the patient in an stage of cancer or substantial or is Conversely, there are potential to patients who tests at a lower risk, but their physician does not know it. Therefore, they are that genetic would lead to an of healthcare we it is likely that it will lead to a more use of resources – fewer resources for lower risk patients greater yields for higher risk patients. come from of the lower screening costs at lower risk and earlier detection and prevention on the higher risk While our health care system with major of the cost of and for later stage cancer patients and of cardiovascular and it is to that less than 5% of the health care is used for with most of the to just a of the costs to the to and manage higher risk patients, may the from to earlier detection and prevention with dramatic in costs and risk tests one important that is the of more cost-effective rather than the of fits While we will to markers to improve further the of these genetic tests, many of the current risk profiles represent relative risk that or exceeds the risk of conventional risk factors. utility is the of the risk and what as a physician will do with some of patients who risk Therefore, many of these genetic risk tests are for time Over the or so there has been a substantial in the of the of there is a patient between the of and who to see a physician for a new without from the large of information on their It could be that of healthcare are more control of their to information about the genetic risk of diseases would this to to the The who the of disease is more likely to to treat it and by the a who the risk has of a disease is more likely to to the risk. It is also important to that by about the genetic risk have of diseases, are certain about who of the of our is that is to in more who It is important that can this information without to through a healthcare because such an could as a Furthermore, there is no evidence in support of the that genetic risk of common diseases is and and in and