Studies investigating the impact of direct-to-consumer testing

Since the completion of the human genome sequence, many companies offering genetic testing have been established⁽ Reference Hunter, Khoury and Drazen ^{17 )}. DTC testing can be defined as testing that is initiated by the individual, with the results provided directly to individuals without the involvement of a healthcare provider⁽ Reference Bloss, Madlensky and Schork ^{18 )}. There are some examples in the literature investigating the effect of DTC testing on behaviour change (Table 1).

Table 1.

Summary of studies investigating the impact of direct-to-consumer (DTC) testing on motivating behaviour change

Kaufman et al. carried out an online survey of DTC customers of 23andMe, deCODEme and Navigenics⁽ Reference Kaufman, Bollinger and Dvoskin ^{19 )}. A random selection of customers were invited to take part with 33 % completing the survey (n 1048). As part of the questionnaire, individuals were asked information regarding demographics, health behaviours, motivations for purchasing the test, and opinions and actions taken in response to their results⁽ Reference Kaufman, Bollinger and Dvoskin ^{19 )}. The investigators reported that nearly half of those surveyed had sought additional information about a health condition tested. Furthermore, 28 % of individuals had informed a healthcare professional of their results, and 9 % had follow-up laboratory tests. Additionally, 16 % of respondents had made changes to their medication and/or supplement intake and one-third reported being more careful with their diet⁽ Reference Kaufman, Bollinger and Dvoskin ^{19 )}. Many of these findings were significantly associated with their response to a question regarding their perceived risk of colon cancer, which suggests that those who consider themselves at high risk of cancer, were more likely to report making health behaviour changes⁽ Reference Kaufman, Bollinger and Dvoskin ^{19 )}. A limitation to this survey was the lack of comparison with a control group who did not receive genetic testing. It is also important to note that only one-third of customers completed the survey, which may have biased the results.

Egglestone et al. conducted an online survey of individuals who had purchased a DTC genetic test (consumers, n 189), and these were compared with those who were thinking about purchasing a test, or who were awaiting their results (potential consumers, n 86)⁽ Reference Egglestone, Morris and O'Brien ^{20 )}. The aim of the study was to investigate consumer perceptions of the effect of genetic risk information, obtained via DTC genetic tests, on their health behaviour and health anxiety. Individuals were contacted through social media and asked a series of open-ended and close-ended questions (forty-four questions for consumers, thirty-nine questions for non-consumers). It was found that 27·3 % of consumers reported changes in their health behaviour, which were all either positive or neutral⁽ Reference Egglestone, Morris and O'Brien ^{20 )}. Furthermore, a change in health anxiety was reported by 24·6 % of consumers, of which, 85·3 % were a reduction. A major strength of this survey and the one conducted by Kaufman et al.⁽ Reference Kaufman, Bollinger and Dvoskin ^{19 )} was the fact that they were conducted on real-life DTC consumers, rather than those who received the genetic testing free or at a reduced rate. This is in contrast to a study conducted by Bloss et al. who recruited employees from health and technology companies who agreed to purchase the Health Compass at a discounted rate⁽ Reference Bloss, Schork and Topol ^{21 )}. Those taking part in the study underwent health assessments at baseline and a 3-month follow-up (when they had received their genetic results) via SurveyMonkey⁽ Reference Bloss, Schork and Topol ^{21 )}. Of those that completed the follow-up (n 2037), no significant differences were observed between baseline and follow-up in terms of anxiety, dietary fat intake or exercise behaviour⁽ Reference Bloss, Schork and Topol ^{21 )}. The investigators also re-tested these subjects at 12 months and similarly, no differences were reported in terms of anxiety, fat or exercise⁽ Reference Bloss, Wineinger and Darst ^{22 )}. Therefore, results with regard to the impact of DTC testing on changes in diet and lifestyle behaviours are mixed. It may be postulated that paying the full price for the genetic test may have a role to play in the impact of DTC testing on behaviour.

Studies investigating the impact of disease genetic risk disclosure with phenotypic/familial risk presence

Some studies have investigated the impact of genetic risk disclosure on those individuals who were already at a phenotypic and/or familial risk of a particular disease, many of which are shown in Table 2. One of the first of these studies investigated the hypothesis that confirmation of an individual's diagnosis of familial hypercholesterolaemia by the presence of genetic mutation, reduces patients’ perceptions of control over the disease and adherence to risk-reducing behaviours⁽ Reference Marteau, Senior and Humphries ^{23 )}. Individuals already diagnosed with familial hypercholesterolaemia and their first- or second-degree relatives, were randomised to receive routine clinical diagnosis or routine clinical diagnosis and genetic testing. The investigators found that those who had the mutations believed less strongly in the efficacy of diet in reducing cholesterol levels (P = 0·02) and demonstrated a trend in believing more strongly in the efficacy of cholesterol-lowering medication (P = 0·06). Therefore, the authors concluded that genetic testing does not affect the extent to which individuals believe that they have control over their health, but does affect their perceptions of how control is most effectively obtained⁽ Reference Marteau, Senior and Humphries ^{23 )}. Hollands et al. investigated the effect of communicating risk assessment for Crohn's disease based on genotype on motivating smoking cessation for 24 h or longer, assessed at 6 months⁽ Reference Hollands, Whitwell and Parker ^{24 )}. Smokers with first-degree relatives with Crohn's disease were randomised to receive either a postal booklet detailing their risk assessment of developing Crohn's disease based on family history and smoking status, or with additional NOD2 genotyping⁽ Reference Hollands, Whitwell and Parker ^{24 )}. Following receipt of the booklet, a research counsellor contacted all participants to reiterate the information and give advice regarding smoking cessation. At 6 months, there were no differences between the groups in terms of the numbers of participants who stopped smoking for 24 h or the number of those who made an attempt to quit⁽ Reference Hollands, Whitwell and Parker ^{24 )}.

Table 2.

Studies investigating the impact of disease genetic risk disclosure with phenotypic/familial risk presence

DM, diabetes mellitus.

The effect of genetic risk information on those at an increased phenotypic risk of diabetes has also been investigated⁽ Reference Grant, O'Brien and Waxler ^{25 )}. Overweight individuals at phenotypic risk of diabetes were randomised to either receive genetic testing or not. For those individuals who received genetic testing, thirty-six risk alleles previously associated with type 2 diabetes were analysed and individualised genetic risk assessments were then calculated to assign participants to a higher or lower diabetes genetic risk. These individuals received a 15 min structured one-to-one genetic counselling session, explaining their results and encouragement to make certain behavioural changes to reduce overall diabetes risk⁽ Reference Grant, O'Brien and Waxler ^{25 )}. Following this, counselled intervention participants and untested controls took part in a diabetes prevention programme, conducted by an experienced dietitian for 12 weeks⁽ Reference Grant, O'Brien and Waxler ^{25 )}. There were no differences found with regard to self-reported motivation, programme attendance or weight loss when higher-risk recipients and lower-risk recipients were compared with those in the control group who received no genetic testing⁽ Reference Grant, O'Brien and Waxler ^{25 )}. When the higher- and lower-risk groups were compared, higher-risk result recipients more often reported that the genetic counselling made them more motivated to participate in the 12-week programme (P = 0·003) and make lifestyle changes (P = 0·008); however, programme attendance and weight loss were not statistically different between these two groups⁽ Reference Grant, O'Brien and Waxler ^{25 )}. In contrast to this, Vassy et al. conducted a nationally representative survey of young adults, who were given hypothetical scenarios of receiving genetic susceptibility results for heart disease, type 2 diabetes and stroke⁽ Reference Vassy, Donelan and Hivert ^{26 )}. These individuals were asked about their interest in such testing, anticipated likelihood of improving diet and physical activity with high- or low-risk test results and readiness to change. Responses were analysed based on the presence of established disease-risk factors⁽ Reference Vassy, Donelan and Hivert ^{26 )}. The investigators reported that those individuals with high phenotypic risk reported increased likelihood of improving their diet and physical activity in response to high-risk results, compared with those with low diabetes risk for diet and physical activity (Table 2). Furthermore, Vernarelli et al. ⁽ Reference Vernarelli, Roberts and Hiraki ^{27 )} reported that individuals who were informed that they had at least one copy of the risk increasing (E4+) gene, had 4·75 times the odds of reporting a change in dietary supplement use than those with less risky genotype (E4−)⁽ Reference Vernarelli ^{28 )}. However, there were no differences found in terms of changes in overall diet, exercise or medications.

Therefore, evidence is mixed with respect to impact of genetic testing on behaviour change in the presence of familial or phenotypic disease risk. It can be postulated that the particular disease risk is likely to have an effect on the impact of genetic knowledge, i.e. the perceived seriousness of the disease may affect motivation to change.

Studies investigating the impact of genotype-based personalised nutrition

Studies investigating the hypothesis of delivering genotype-based PN advice on motivating key behaviour changes such as diet, lifestyle and physical activity are presented in Table 3. One of the first studies to test this concept was by Arkadianos et al. in a weight loss clinic in Greece⁽ Reference Arkadianos, Valdes and Marinos ^{29 )}. Patients with a history of weight loss failures were offered a nutrigenetic test examining twenty-four variants in nineteen genes involved in metabolism (ACE, APOC3, CBS, CETP, COLIAI, GSTMI, GSTPI, GSTTI, IL6, LPL, MTRR, MTHFR, MTR, NOS3, PPARG, SOD2, SOD3, TNFα and VDR), using the Sciona MyCellf kit. These individuals were then matched for age, sex, starting BMI and number of times they had visited the clinic with patients who had not taken the test. Both groups followed a low glycaemic index Mediterranean diet, recommended exercise routines and regular follow-up visits to the clinic. For those in the nutrigenetic group, the diet and exercise programme was modified based on the individual genetic results. For example, if an individual had a variation in MTHFR, MTRR, MTR or CBS, they were advised to take a supplement containing 800 μg folic acid, 15 mg vitamin B₆ and 20 μg vitamin B₁₂ ⁽ Reference Arkadianos, Valdes and Marinos ^{29 )}. The investigators reported that after 6 months the two groups performed the same in terms of weight loss⁽ Reference Arkadianos, Valdes and Marinos ^{29 )}. However, at 12 months, the nutrigenetic group were more likely to maintain some weight loss compared with the control group, with an average BMI loss of 1·93 kg/m² in the nutrigenetic group compared with an average BMI gain of 0·86 kg/m² in the control group (P < 0·023). It is important to note that the individuals were put on altered diets based on their individual genetic results and therefore, this study design cannot distinguish between whether knowledge of genotype-motivated individuals to make appropriate dietary and lifestyle changes or whether the particular diet chosen for each individual promoted a better biological response, which resulted in sustained weight loss compared with the control group.

Table 3.

Studies investigating the impact of genotype-based personalised nutrition

In contrast to these positive findings, a US research group investigated the effect of genetic-guided weight loss advice in comparison with a standard weight loss diet⁽ Reference Frankwich, Egnatios and Kenyon ^{30 )}. The investigators randomised fifty-one obese or overweight US veterans to either a nutrigenetic-guided diet (balanced, low-carbohydrate, low-fat or Mediterranean) based on seven genes (APOA2, ADIPOQ, FTO, KCTD10, LIPC, MMAB and PPARG) or standard balanced diet for 6 months. No differences were observed in the percentage of participants who lost 5 % of their body weight at 2 or 6 months between the groups⁽ Reference Frankwich, Egnatios and Kenyon ^{30 )}. As secondary analysis, the researchers also investigated the role of adherence in the participants’ weight loss and found that, within the nutrigenetic-guided group, there was a significant relationship between adherence and weight loss (P = 0·002) and those in the top quartile had lost a significant amount of weight compared with the lowest quartile (P = 0·03)⁽ Reference Frankwich, Egnatios and Kenyon ^{30 )}. This may suggest that genotype-based PN advice may be of benefit to those who are most adherent to their recommended diet plan⁽ Reference Frankwich, Egnatios and Kenyon ^{30 )}. Similar to the study conducted by Arkadianos et al. ⁽ Reference Arkadianos, Valdes and Marinos ^{29 )}, this study does not test the hypothesis of the impact of genotype-based personalised advice on motivating behaviour change but rather the potential biological response to a nutrigenetic-guided weight loss diet.

Hietaranta-Luoma et al. investigated the effect of APOE genotype information on behavioural changes related to diet, including fat quality, vegetable and fruit consumption, alcohol consumption and exercise levels in a 12-month intervention study⁽ Reference Hietaranta-Luoma, Tahvonen and Iso-Touru ^{31 )}. Participants were randomised to either the control group who received general information or intervention group who were tested for the APOE genotype and given a tailored health message related to their APOE genotype, its effect on CVD risk and importance of dietary and lifestyle changes based on the extended parallel process model⁽ Reference Hietaranta-Luoma, Tahvonen and Iso-Touru ^{31 )}. Dietary changes were assessed at baseline, 10 weeks, 6 months and 12 months. Those with the APOE E4+ genotype reported better dietary fat quality compared with those in the E4− group and control group (P < 0·05) at 10 weeks and 6 months; however, this was not observed at 12 months⁽ Reference Hietaranta-Luoma, Tahvonen and Iso-Touru ^{31 )}. Nielsen et al. also reported changes in sodium intake following disclosure of the ACE genotype⁽ Reference Nielsen and El-Sohemy ^{32 )}. Here, the investigators recruited healthy individuals to take part in an online randomised controlled trial, where individuals were randomised to receive either genotype-based PN advice or general healthy eating advice. Those in the intervention group were genotyped for five dietary-related genes including CYP1A2, GSTT1, GSTM1, TAS1R2 and ACE. The investigators reported no differences between the groups at 3 months, but at 12 months, those with the risk version of the ACE gene reported significantly lower dietary sodium intakes compared with those with the non-risk ACE genotype or control group (P = 0·008). However, no differences were observed in terms of any of the other dietary markers associated with the genes investigated⁽ Reference Nielsen and El-Sohemy ^{32 )}.

It is interesting to note the differences in the duration of follow-up within these studies, as the effect of genotype-based personalised advice was observed at 12 months in studies conducted by Arkadianos et al.⁽ Reference Arkadianos, Valdes and Marinos ^{29 )} and Nielsen & El-Sohemy⁽ Reference Nielsen and El-Sohemy ^{32 )}, but not reported in those studies of shorter duration⁽ Reference Frankwich, Egnatios and Kenyon ^{30 )}. However, Hietaranta-Luoma et al. reported positive changes early on in the intervention, but these was no longer evident at 12 months⁽ Reference Hietaranta-Luoma, Tahvonen and Iso-Touru ^{31 )}. At present, the evidence for the benefit of genotype-based PN is not strong, and more studies are warranted before any clear conclusions can be deduced on personalised advice based on genotype and motivating behaviour change.

Recent findings in personalised nutrition: the Food4Me study

In the previous section, studies investigating the effect of genotype-based personalised advice on motivating behaviour change were examined. However, PN advice is not exclusive to genotype information, and in fact, PN advice can be delivered using dietary intake information as well as phenotypic markers and genotype⁽ Reference Gibney and Walsh ^{4 )}. Therefore, the Food4Me proof-of-principle study set out to investigate the effect of varying levels of PN advice on health outcomes in comparison with general healthy eating advice⁽ Reference Celis-Morales, Livingstone and Marsaux ^{33 )}. Individuals were recruited across seven centres in Europe to take part in a 6-month online PN intervention study where participants were randomised into one of four groups: control group who received general healthy eating guidelines; level 1 group who received PN advice based on dietary intake; level 2 group who received PN advice based on diet and phenotypic; level 3 group who received PN advice based on diet, phenotype and genotype information.

This study was designed to emulate existing PN services, and therefore, all data were self-collected and self-reported by the participants⁽ Reference Celis-Morales, Livingstone and Marsaux ^{33 )}. To facilitate this, participants were sent instructions and equipment via post. Online videos were also available demonstrating how to perform the measurements. Dietary information was collected using an online FFQ, specifically designed for the purposes of the study⁽ Reference Forster, Fallaize and Gallagher ^{34 ,} Reference Fallaize, Forster and Macready ^{35 )}. Anthropometric measurements, including BMI, waist, hip and thigh circumferences were self-collected by the volunteers. Participants collected dry blood spots and posted them back to their local research centre. These dry blood spot cards were analysed for cholesterol, glucose, a range of fatty acids and carotenoids⁽ Reference Celis-Morales, Livingstone and Marsaux ^{33 )}. Each participant also collected a sample of their DNA using a buccal swab and a panel of thirty-three SNP were analysed⁽ Reference Celis-Morales, Livingstone and Marsaux ^{33 )}. Participants received tailored feedback reports appropriate to their intervention level (as above) at months 0, 3 and 6, using decision trees for a systematic approach across seven research centres in line with behaviour change techniques⁽ Reference Celis-Morales, Livingstone and Marsaux ^{33 )}. Level 3 participants received feedback on five dietary-related SNP, including FTO, TCFL2, APOE, FADS1 and MTHFR.

The investigators reported that following a 6-month intervention participants who were randomised to personalised groups (levels 1, 2 and 3), were more motivated to make positive changes to their diet, compared with those who were given healthy eating guidelines⁽ Reference Celis-Morales, Livingstone and Marsaux ^{36 )}. More specifically, those in the personalised groups, consumed significantly less red meat (−5·48 g (95 % CI −10·8, −0·09), P = 0·046), salt (−0·65 g (95% CI −1·1, −0·25), P = 0·002) and saturated fat (−1·14 % of energy (95% CI −1·6, −0·67), P < 0·0001)⁽ Reference Celis-Morales, Livingstone and Marsaux ^{36 )}. Furthermore, levels 1, 2 and 3 personalised groups significantly increased their folate intake and had higher Healthy Eating Index scores, than those in the control group. However, the researchers concluded that there was no additional benefit of phenotypic or phenotypic and genotype-based personalised advice, i.e. no difference across the levels of personalisation⁽ Reference Celis-Morales, Livingstone and Marsaux ^{36 )}.

Within Food4Me, the impact of genotype-based PN advice has been investigated further in relation to APOE and MTHFR genotypes to date⁽ Reference Fallaize, Celis-Morales and Macready ^{37 ,} Reference O'Donovan, Walsh and Forster ^{38 )}. Fallaize et al. investigated the impact of knowledge of APOE genotype on dietary fat changes⁽ Reference Fallaize, Celis-Morales and Macready ^{37 )}. The researchers reported significantly higher total cholesterol concentrations in those with the risk E4+ genotype, compared with those with the non-risk E4− genotype (P < 0·05)⁽ Reference Fallaize, Celis-Morales and Macready ^{37 )}. Both APOE risk and non-risk groups significantly reduced their saturated fat intakes compared with baseline values; however, no differences were observed between the risk and non-risk groups⁽ Reference Fallaize, Celis-Morales and Macready ^{37 )}. Moreover, those participants who were told that they had the non-risk APOE genotype (E4−), had a smaller reduction in their saturated fat intake compared with participants who did not receive genotype-based personalised advice (level 2 participants), suggesting that knowledge of ‘less risky’ genotypes may actually demotivate individuals from making key dietary and lifestyle changes as previously postulated⁽ Reference Fallaize, Celis-Morales and Macready ^{37 )}.

The impact of MTHFR risk knowledge within the Food4Me cohort was also investigated⁽ Reference O'Donovan, Walsh and Forster ^{38 )}. In Food4Me, participants randomised to level 3 group were informed whether they had the risk version (CT or TT genotypes) or non-risk version (CC genotype) of the MTHFR 677CT gene. To examine the impact of knowledge of MTHFR risk, level 3 participants were split into those with the risk version and those with the non-risk version and dietary intakes of folate were compared with those who received general healthy eating advice (control group)⁽ Reference O'Donovan, Walsh and Forster ^{38 )}. Overall, no differences were observed between the MTHFR risk, non-risk and control groups for changes in total dietary folate intake from baseline to month 6. Similarly, no differences were observed between the groups in terms of changes in gram intakes of food groups that contain folate such as liver, green leafy vegetables and fortified cereals⁽ Reference O'Donovan, Walsh and Forster ^{38 )}. Further analysis was conducted investigating the differences between the MTHFR risk, MTHFR non-risk and participants in levels 1 and 2 who received non-genotype-based personalised advice, but no differences were observed. This suggests that MTHFR genotype-based PN advice did not have a more significant impact on increasing dietary folate intakes, compared with general healthy eating advice or non-genotype-based PN advice. It could be postulated that perhaps the effect of genotype-based PN advice could not be observed in this study as MTHFR genotype is not as well-known from a public health point of view compared with APOE genotype and its impact on saturated fat⁽ Reference O'Donovan, Walsh and Forster ^{38 )}.