Randomised trial of population‐based BRCA testing in Ashkenazi Jews: long‐term secondary lifestyle behavioural outcomes

2.1. Design

North London‐based self‐referred consenting AJ individuals >18 years were randomised (1:1) using a computer‐generated random‐number algorithm to either population‐screening (PS arm) or FH‐based testing (FH arm). All underwent pre‐test genetic counselling. ³ Counsellors were blinded to group allocation. All participants in the PS arm and only those fulfilling standard FH‐based criteria in the FH arm underwent genetic testing for the three Jewish BRCA pathogenic variants (PVs), called founder mutations: 185delAG(c.68_69delAG), 5382insC(c.5266dupC) and 6174delT(c.5946delT). Individuals with previous BRCA testing, and first‐degree relatives (FDR) of BRCA‐carriers were excluded. Individuals with PVs received post‐test counselling. These details have been described in earlier reports on psychological (primary) outcomes ³ , ⁷ (90% power for detecting a HADS‐score difference = 1.2 between study arms; SD = 5.9; α = 0.05). We now report on long‐term (up to 3‐years) secondary outcomes of lifestyle behaviour: (a) diet (b) alcohol intake, (c) smoking, (d) exercise, as well as (e) cancer risk perception and (f) routine breast screening uptake. We report between‐group (FH and PS arms) differences adjusted for any baseline difference collectively across 1, 2 and 3 years. We also assessed time effects for each group (within‐FH group and within‐PS group) from baseline over 1, 2 and 3 years. We explored the association of gender, income, BRCA status and FH of cancer and age on these outcomes and adjusted for them in the analysis. Customised questionnaires were used to collect socio‐demographic, FH data and outcome data. Data were collected at baseline and then annually for 3 years.

Standardised items obtained from the UK National Diet & Nutrition Survey (NDNS); Health Survey England (HSE), Harvard Center for Cancer‐Prevention's Harvard Cancer Risk Index (HCRI), and CDC Behavioral Risk‐Factor Surveillance System (BRFSS) were used to assess these outcomes. The NDNS is a rolling programme collecting information on food consumption, nutrient intake and nutritional status of the UK general population funded by Public Health England (PHE) and the UK Food Standards Agency (FSA). HSE is an annual survey monitoring changes in the health and lifestyles of people across the UK. Cancer‐related food consumption was measured with three items adapted from the NDNS: ‘Please indicate how often, on average, you eat each of the following foods: red meat, fruit, vegetables'; with seven response options (ranging from ‘Never’ to ‘More than once a day’), which were categorised for analysis to: <once/week, ≥ once/week and ≥ once/day. Vitamin consumption was assessed with an item adapted from NDNS: ‘Are you taking any vitamin supplements?’ (no/yes). Alcohol consumption was assessed with one item adapted from the NDNS: ‘How often have you had an alcoholic drink during the last 12 months?’ (response options: ≤ twice/year; every 2 months; ≤ twice/month; 1–2 days/week; ≥3 days/week), plus one item adapted from the Harvard Center for Cancer Prevention ‘Your Disease Risk’ website, which is based on the HCRI (Colditz,2000; https://siteman.wustl.edu/prevention/ydr/): How many servings of alcohol do you have on a typical day? (0/1/2/3 or more).

Physical activity was assessed with one item from the CDC‐BRFSS (http://www.cdc.gov/brfss/questionnaires/pdf‐ques/2006brfss.pdf): ‘During the past month, other than your regular job, did you participate in any physical activities or exercises such as running/golf/gardening/walking for exercise?’ (yes/no); one item from the HCRI: ‘Do you walk (or do other moderate activity) for at least 30 minutes most days, or at least 3 hours/week?’ (yes/no), and one item adapted from the HSE: ‘Thinking about your job in general, would you say that you are very/fairly/not very/not at all physically active’.

Smoking was assessed with items adapted from the CDC‐BRFSS: ‘Do you now smoke cigarettes every day, some days, or not at all?’ and ‘During the past 12 months, have you stopped smoking for 1 day or longer because you quit or were trying to quit smoking?’ (yes/no).

Comparative cancer risk was assessed with (a) ‘Compared with other people of your age and sex, do you think your chances of getting cancer at some point in your life are: much lower, lower, about the same, higher, much higher?’ (Sutton, 1994); and (b) ‘On a scale from 0 to 100, where 0 is no chance at all and 100 is absolutely certain, what do you think are the chances that you will get cancer sometime during your lifetime?’ (Smith, 2004). Use of a mammogram was assessed with items adapted from the CDC‐BRFSS: Have you ever had a mammogram?’ [yes/no/not sure]; ‘If yes, how long has it been since you had your last mammogram?’ [<1 year, >1 year, not sure].

2.2. Statistical analysis

The responses to the lifestyle questions at ‘baseline’, 1 year, 2 years and 3 years were analysed using generalised linear mixed models, where a random intercept term represented the unexplained heterogeneity corresponding to each subject. Specifically, a logistic mixed‐model was fitted for the binary items and an ordinal logistic regression model fitted for ordered multiple response items. Questions on breast screening (mammogram) were re‐formulated to be a binary response variable that asked whether the subject had the screening/mammogram within the last 12 months and responses were analysed using logistic mixed‐models. The analysis of mammogram outcomes was restricted to women >50 years. Each time‐point was included as a fixed‐effect and interacted with the group term (‘family history’ or ‘population screening’), resulting in all eight group and time mean values being freely estimable. In addition, the model was adjusted for gender (men versus women), BRCA founder‐mutation status (positive or negative or unknown), marital status (married/cohabiting versus widowed/divorced/single), income (<£10,000, £10,000 to <£20,000, £20,000 to <£30,000, £30,000 to <£40,000, £40,000 to <£50,000, and >£50,000), education (degree level/above versus no formal qualification/GCSE/O‐level/CSE/NVQ1/NVQ2/A‐level education), family history (low‐risk versus high‐risk) and age.

A linear combination of the relevant time and group parameters for each model was used to estimate the accumulated difference between groups over time (years 1, 2 and 3) in the logit scale, accounting for any baseline difference. A similar linear combination of group‐specific parameters for each model was calculated to estimate the total change from baseline (over years 1, 2 and 3) for each group in the logit = log(odds) scale. This is the cumulative difference between groups over years 1, 2 and 3 (accounting for the baseline difference). As 13 variables were analysed, results were adjusted for multiple testing. We were conservative and used a Bonferroni correction, with P‐values <0.0039 (0.05/13) considered to be significant. Potential group differences over the four time‐points were also explored visually to help interpret the model parameters for group when interacted with time. To visualise any trends, the STATA margin command was used to make predictions over the sample for each of the eight group‐by‐time interactions. Specifically, for the binary items, the marginal prediction was the population averaged (that is, after integrating over the random effect for subject) probability of a positive response. For the ordered items, the same predictions were made for each of the possible outcome categories. These marginal predictions with their confidence intervals were then plotted. Statistical analyses used STATA‐17.0 (StataCorp LP).

2.2.1 | Core outcome sets (COS)

There are no COS for population BRCA testing at present.

2.2.2 | Patient & public involvement (PPI)

The GCaPPS trial was preceded by a year‐long extensive community engagement process with all denominations in the Jewish community, involving a broad range of stakeholders. This included representatives from the Liberal, Reform, Masorti/Conservative, Orthodox and Unaffiliated community denominations. We engaged with a range of stakeholders including religious leaders, Rabbis, Boards of Deputies, Jewish charities, Jewish Medical Association, cancer charities and patient support groups. This exercise enabled exchanges of ideas and understanding of underlying concerns related to BRCA testing and the research. It provided community inputs into study protocol development, communication strategy, development of participant/patient‐facing materials, study conduct/delivery. There was PPI/public representation on trial steering and management committees. The extensive PPI exercise was instrumental in generating support, awareness and successful completion of the study.

4.1. Main findings

To the best of our knowledge this is the only RCT study on lifestyle impact comparing population‐based BRCA testing with FH‐based testing. Our findings are reassuring, as they do not show a significant difference or any detrimental impact on overall lifestyle factors, such as dietary fruit, vegetable or meat intake, alcohol, smoking or physical activity and exercise, with population‐based BRCA testing compared with the standard FH‐based testing. We found an increase in vitamin intake within the population‐based BRCA testing arm. Men eat fewer fruits, vegetables, vitamins and consume more alcohol and meat than women. Fruit, vegetable, alcohol and vitamin consumption increased, whereas meat intake decreased with age. Non‐carriers perceived themselves at lower risk than BRCA‐carriers and overall cancer risk perception decreased with time, but there was no difference between the PS and FH approaches. Routine mammogram screening uptake declined with time for average‐risk women over 50 years, but there was no difference between the FH and PS arms.

4.2. Strengths and weaknesses

Our study has a number of advantages, including population‐based ascertainment, inclusion of men and women, randomised design, adequate questionnaire response rate and long‐term follow‐up. Its limitations include lack of qualitative data and inferences from BRCA testing in the AJ‐population not being directly generalisable to the non‐Jewish general population.

4.3. Interpretation

Although genetic testing can be used to facilitate healthy lifestyle behaviours, a clear cause‐and‐effect association has yet to be demonstrated. ²⁰ Lifestyle factors such as diet (reduced vegetable/fruit), reduced physical activity/exercise, smoking and alcohol can also increase risk of chronic diseases such as heart disease, stroke, diabetes, lung disease and cancer. ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ That these were not negatively affected by population‐based BRCA testing is reassuring. Our finding of increased vitamin intake in the population testing arm is contrary to two earlier small non‐randomised cohort studies in high‐risk AJ and general population women that showed no short‐term impact on vitamin intake with BRCA testing. ⁹ , ¹⁰ However, those studies had small sample sizes. Additionally, given the multiple analyses undertaken, our result of increased vitamin intake could be purely a chance finding, as it was not statistically different from the vitamin intake in the FH arm after adjusting for multiple testing. Although we did not specifically assess this, both high prevalence and increased use ²¹ of complementary and alternative medicine (CAM) has been reported up to the 1‐year follow‐up following BRCA testing in high‐risk individuals. However, the efficacy or benefit of CAM on cancer risk remains unestablished. According to the UK FSA, half of adults (48%) take food supplements (multivitamins being the commonest form) regularly, with another third having taken them in the past. ²² People take vitamins to maintain/improve their overall health, as a perceived ‘boost’ or to replace nutrients lost due to diet, age or a health condition. Although sales of vitamin supplements have increased over the years, the long‐term health consequences of vitamin consumption are unknown. Consumption in moderation is acceptable, but taking high‐dose supplements is not always effective for disease prevention, and may even be harmful to health. ²³

The decrease in long‐term job‐related physical activity we found is consistent with the reduction in physical activity linked to increasing age in the population. That this was not accompanied by a significant reduction in physical exercise associated with sport or moderate activities such as walking is reassuring. Our findings of lack of health behavioural impact on diet, physical activity, alcohol and smoking is consistent with other reports from high‐risk women. ⁹ , ¹⁰ Although a recent report from a population‐based sample suggested that UK women largely anticipated positive engagement with a healthy lifestyle following ovarian/breast cancer risk disclosure, this was not completely borne out in our study. ²⁴ This could be due to population differences in these studies or the intention–behaviour gap with intent did not completely translate into actions leading to actual lifestyle change. ²⁵ Our findings are consistent with a recent systematic review which suggests that overall personalised risk information does not lead to a strong effect on health‐related behaviours or sustainable changes to lifestyle. ⁸ However, although genetic testing itself may not lead to a significant lifestyle impact, there are some data suggesting that subsequent interventions such as attending a well‐structured patient retreat offering education/information, management updates and support can lead to short‐term positive lifestyle benefits in higher risk women. ²⁶

Some reports indicate a positive impact on lifestyle following ‘direct‐to‐consumer’ (DTC) genetic testing, in up to 23% participants, on factors such as diet, exercise, vitamin intake, alcohol intake and smoking. ²⁷ , ²⁸ However, these were mainly short‐term effects and often related to genetic information with uncertain clinical utility or validity (e.g. SNP profiling). Long‐term data and information on the impact of DTC testing associated with high‐penetrance cancer genes are needed. Some DTC studies show that providing personalised nutrition‐related genetic information can have a positive impact on some dietary components compared with general population‐based dietary advice, ²⁹ but others show no beneficial impact on physical activity or other lifestyle factors. ³⁰ , ³¹

The reduction in cancer risk perception observed with time following BRCA testing is consistent with other reports in the literature from high‐risk studies and cancer risk in non‐carriers in population‐based studies. ¹⁹ , ³² The lack of difference we found between FH and PS approaches has not been reported before. The lack of impact of population‐based BRCA testing on baseline breast screening behaviours is reassuring. Average risk women in the UK undergo 3‐yearly mammograms through the national breast screening programme. Our findings suggest that most women continue to undergo the routine mammograms and although mammogram uptake decreases with age in women over 50 years, this was not different between women undergoing FH‐ or PS‐based BRCA testing (Table 1; Table S1). This is congruent with breast screening recommendations for women. ³³

The gender‐based differences in lifestyle related to diet, alcohol and physical activity seen in our study are consistent with differences in health behaviour observed between men and women from the general population. ³⁴ , ³⁵ , ³⁶ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ , ⁴² Behavioural general population studies have shown that gender differences about beliefs of the importance of healthy eating explain the differences in food choices between men and women. Our results showing that men eat fewer fruits, vegetables and vitamins but consume more meat and alcohol than women, are consistent with the view that part of the reason men report making less healthy food choices is that health is a less important motivation to them in the food domain. ³⁸ , ³⁹ Physical activity statistics published by the HSE ⁴² and British Heart Foundation are in keeping with our results and show that the mean number of hours per day of moderate or vigorous workplace activity is over twice as high for men as for women (men = 3.8 hours/day, women = 1.7 hours/day). ⁴³ Similarly, age‐based associations found with diet, alcohol and exercise are akin to normative data from the general population. ³⁹ , ⁴² , ⁴⁴ , ⁴⁵ A large general population study of drinking behaviour in 35 countries across six continents (Europe, Asia, Africa, Australia, North America, South America) between 1997 and 2007 using a standardised questionnaire revealed a higher prevalence of high‐frequency drinking in older age groups of drinkers and is in keeping with our results. ³⁹ Our findings support existing literature showing intake of fruit/vegetables/vitamins to be higher but meat consumption lower among older individuals. ⁴⁵ , ⁴⁶ Our outcomes of higher levels of education being associated with a healthier lifestyle in terms of diet and physical activity are consistent with the reported association of lower education levels with decreased knowledge of genes/cancer risk factors ¹ and the perceived lack of benefit of lifestyle changes on cancer prevention. ²⁴

Lower levels of education/literacy are linked to lower socio‐economic status, which in turn is associated with a greater sense of powerlessness, belief in chance, poorer social support, limited awareness and lower engagement with screening/prevention. These findings are consistent with the theory of planned behaviour (TPB), where a person's attitudes, subjective norms and perceived behavioural control are key to determining an individual's decision to follow a particular behaviour. ⁴⁷ Successful behaviour is further contributed to by actual behavioural control, which refers to the extent to which a person has the skills, resources and other prerequisites (e.g. income, education, social factors) needed to perform the behaviour in question. ⁴⁸ Further research is needed to identify reasons for and barriers to a potential intention–behaviour gap ²⁴ limiting adoption of healthier lifestyles. Research is also needed to quantify levels and types of increased vitamin intake and understand its potential implications.

Most lifestyle outcomes of population testing are comparable to those of FH/criteria‐based testing. These data coupled with other publications from the UK, Israel and Canada ¹ , ² , ³ , ⁴ , ⁵ , ⁷ , ⁴⁹ that show acceptability, feasibility, effectiveness, cost‐effectiveness and lack of detrimental impact on psychological health or quality of life, support the concept of Jewish population‐based BRCA testing. The uptake of clinical BRCA testing remains limited and only a small proportion of eligible women undergo BRCA testing. Correspondingly, only a small proportion of at‐risk BRCA‐carriers have been identified ⁵⁰ , ⁵¹ The vast majority of BRCA‐carriers who can benefit from screening/prevention remain to be identified, missing huge opportunities for precision prevention. Population testing offers the opportunity to maximise carrier identification for primary OC and BC prevention. We ⁵² , ⁵³ and others have argued for changing guidelines in favour of population‐based BRCA testing. Israel has recently changed its policy and implemented population BRCA testing. It is important that other health systems, including the NHS, follow suit.