L4 Genetics Direct to consumer genetic tests - metabolism and athletics

Overview

The lecture applies a clinical filter to direct-to-consumer (DTC) genetic testing. It starts by revising the genetic diagnostic toolkit used in clinical practice (single gene, panel, exome, genome), then contrasts that with what DTC companies actually do: SNP genotyping of common variants, algorithmic comparison against GWAS risk markers, and a consumer dashboard. From there it separates what DTC can do reliably (specific monogenic variants) from what it cannot (immature polygenic risk scores), works through how a relative risk becomes the absolute percentage a consumer sees, and applies the same filter to sports and injury genetics and to pharmacogenomics. The final third covers consequences: clinician workload, data as the business model, forensic use, insurance discrimination in Aotearoa New Zealand, and the regulatory and BMJ guidance that tells a clinician what to do when a patient arrives with a DTC report.

Lecture objectives

1. Revise the main types of genetic tests: single gene, panels, exome, whole genome.
2. Discuss what medically relevant DTC tests can be done: monogenic (for example cystic fibrosis) and polygenic disease risk (for example obesity, diabetes).
3. Discuss the positives and negatives of DTC genetic testing, including who is tested and when, personal autonomy, and potential risks to data privacy. The answers given were consumer understanding, clinician time, and insurance implications.

The genetic diagnostic toolkit

Four clinical test modalities, with scope, indicative New Zealand cost, and the case each suits.

Modality	Scope	Cost (NZ)	Ideal clinical use case
Single gene testing	Specific pathogenic variation	~$100	Known specific conditions, for example haemochromatosis
Targeted gene panel	Panel of variations linked to a disease phenotype	~$300to$400	Phenotype with known genetic heterogeneity, for example retinitis pigmentosa
Whole Exome Sequencing (WES)	Resequencing the ~2% of the genome coding for proteins	~$400to$500	Unidentified variations in an affected child
Whole Genome Sequencing (WGS)	High-resolution comprehensive analysis	~$1,000	Complex, unsolved diagnostic challenges

Adjunct tools: chromosome microarray analysis (aCGH) for CNVs, genotyping for risk association, and RNA analysis for oncology.

What DTC companies sell

• 23andMe marketed health tests from 2013. Reports split into trait reports and health/medical reports.
• Trait reports are largely non-medical curiosities: ability to match musical pitch, asparagus odour detection, back hair, bald spot, bitter taste, bunions, cheek dimples, cilantro taste aversion, cleft chin, dandruff, earlobe type, early hair loss, earwax type, eye colour, fear of heights, fear of public speaking, finger length ratio, flat feet, freckles, hair photobleaching, hair texture, hair thickness, ice cream flavour preference, light or dark hair, misophonia, mosquito bite frequency, motion sickness, newborn hair, photic sneeze reflex, red hair, skin pigmentation, stretch marks, sweet versus salty, toe length ratio, unibrow, wake-up time, widow’s peak. Several are men only (back hair, bald spot, early hair loss).
• Health/medical reports include: age-related macular degeneration, alpha-1 antitrypsin deficiency, BRCA1/BRCA2 (selected variants), coeliac disease, chronic kidney disease (APOL1-related), familial hypercholesterolaemia, G6PD deficiency, hereditary amyloidosis (TTR-related), hereditary haemochromatosis (HFE-related), hereditary prostate cancer (HOXB13-related), hereditary thrombophilia, late-onset Alzheimer’s disease, MUTYH-associated polyposis, and Parkinson’s disease.

Scale and price, the “allure of the $149 genome”:

• Over 26 million ancestry tests completed across all companies by 2019.
• Over 12 million tests processed by 23andMe alone.
• $149entrypointforaHealth+Ancestryservice,downfromtheoriginal2006priceof$999.
• Store pricing shown: Ancestry Service $119reducedto$79; Health + Ancestry $229reducedto$148; 23andMe+ Premium $298reducedto$178 (annual membership with ongoing personalised reports); 23andMe+ Total Health $99/month (includes next-generation sequencing). The Health + Ancestry advert was marked FSA and HSA eligible.
• Popular framing of predictive medicine was illustrated by the book Outsmart Your Genes (Brandon Colby, MD).

DTC versus clinical genetics

Dimension	Direct-to-consumer	Clinical genetics
Primary method	Genotyping on SNP chips mapping ~1 to 2 million specific common variants	Exome (~2% of genome) or whole genome sequencing
Access and support	Self-directed, convenience-driven, zero pre-test counselling	Requires medical referral, embedded professional genetic counsellors
Primary focus	Ancestry, lifestyle traits, polygenic health predispositions	Diagnosing specific, highly penetrant, actionable disease mutations
Data destination	Massive proprietary corporate databases (30% consent to research)	Highly protected confidential medical records

The mechanism: spit to dashboard

1. Collection. The consumer mails a saliva sample to a CLIA-certified laboratory.
2. Genotyping. DNA is extracted and run on a microarray chip. This does not read the whole genome; it checks only ~1 to 2 million specific common variants (SNPs).
3. Algorithmic filtering. Raw variants are computationally compared against known risk markers from published genome-wide association studies (GWAS).
4. Consumer dashboard. Results are stylised into consumer-friendly percentages and risk categories, masking the complex mathematics underneath.

The consumer-facing version of the same pathway is: order a kit, spit into the tube and post it, the CLIA-certified lab analyses the DNA in approximately 10 weeks, then log in and explore. The mock dashboard displayed risk categories as 50% elevated, 30% typical, 20% reduced.

The spectrum of genetic risk

Variants can be plotted by effect size against allele frequency.

• Rare alleles with high effect, for example Mendelian diseases: hard to find, but clinical impact is definitive.
• Common variants with low effect, for example GWAS findings: easy to find via DTC chips, but negligible predictive power individually.
• DTC tests hunt primarily in the common/low-effect corner. They identify thousands of common variants with modest effects to build a probabilistic model, rather than finding single definitive disease markers.

Monogenic versus polygenic testing

Monogenic (definitive):

• Concept: one specific gene mutation directly causes the condition.
• Examples: cystic fibrosis (CFTR carrier), haemochromatosis.
• DTC capability: high reliability for specific variants, though it may miss rare mutations not on the exact SNP chip. Highly actionable clinically.

Polygenic (probabilistic):

• Concept: thousands of low-effect variants combined mathematically into a Polygenic Risk Score.
• Examples: type 2 diabetes, severe obesity.
• DTC capability: immature science, relying on millions of markers with varying predictive power.

Polygenic scores do separate the extremes of a population. In the obesity scorecard data shown (Khera), over 25+ years of follow-up the top 10% by score reached roughly 15 to 16% severely obese, the middle 80% roughly 6%, and the bottom 10% only about 1%.

The engine: relative risk to absolute risk

The dashboard number is a simple multiplication of a population baseline by a variant relative risk:

$\text{Absolute risk}=\text{Base population risk}\times \text{Variant relative risk}$

Worked example from the slides: base population risk 3% (frequency of gout for men of European descent) multiplied by a relative risk of 2.0x (assigned to the SLC2A9 variant from GWAS literature) gives a 6% absolute disease risk, which is the figure shown to the consumer.

The consumer sees the marketing warning "double the risk" and often misses the context that the absolute probability moved only from 3% to 6%.

The generic DTC flowchart runs: genotypes from saliva, matched to known risk markers, giving a relative risk, multiplied by average population disease risk, giving the absolute disease risk displayed on screen. A separate branch off the genotypes is labelled unknown genetic factors, which the calculation does not capture.

Transcript flag (slide 10): the inset flowchart is low resolution. Its example values read as relative risk 3.5, population risk 10% and displayed figure 25%, but those small numerals are not fully crisp.

Sports genetics: elite performance and injury prevention

• The promise: identifying talent via DNA, supported by the remarkably high heritability of elite athletic status (~0.7).
• The reality: common variants marketed for power (ACTN3) or endurance (ACE) have very little positive predictive value for overall sports success. Predicting gold medals is currently impossible.
• The actual utility, applying the clinical filter, lies in identifying rare variants to reduce injury risk: screening for sickle cell trait, identifying APOE E4 variants, and spotting rare cardiomyopathies to prevent sudden cardiac events on the field.
• Genes marketed in this space, from the gene doping figure (with gene doping itself crossed out): endurance performance ACTN, ACE, CYP2D6, AMPD1, CK-MM, HFE, HOS3, UCP2, VEGPA; power performance ACTN3, ACE, IFG1, CK-MM, ILG, PPARγ.

Pharmacogenomics: CYP2C19 and common drugs

CYP2C19 poor metaboliser frequencies differ markedly by population:

Population	Poor metabolisers
Oceanians	57%
East Asians	14%
African Americans	4%
Caucasians	2%

The same enzyme acts in opposite directions for two common drugs:

• Clopidogrel, used to reduce the risk of stroke and heart attack, is a prodrug that CYP2C19 converts to its active antiplatelet form.
• Citalopram requires CYP2C19 for its clearance; failure of clearance leads to prolonged QT.

The consequence for poor metabolisers, and the question of whether at-risk populations should be tested, were posed to the class rather than answered [slide does not elaborate].

The clinical friction

The “worried well”:

• Driven by curiosity, health autonomy, and glossy DTC marketing.
• Prone to false security, with reassuring dashboard results misinterpreted as absolute immunity.
• Severe lack of the genetic literacy needed to interpret probabilistic polygenic variations.

The clinical reality:

• GPs face massive workloads with limited time or specialised knowledge to interpret 125+ new DTC condition reports.
• Medical geneticists and counsellors are in dangerously short supply for the incoming consumer volume.
• Clinicians are burdened by having to untangle and verify false positives (artefacts) generated by consumer SNP chips.

Survey figures shown: 62% of doctors say genetic testing could help them provide more personalised care, and 66% feel it could lead to better outcomes for their patients.

Transcript flag (slide 12): the two survey doughnut charts carry no stated source, population or date.

The lecture also included a deliberately blank two-column exercise asking the class to fill in the positives and negatives of DTC, and used the 1997 film GATTACA to introduce genetic selection and discrimination.

Data as the product

The saliva sample feeds three downstream destinations:

1. Commercial research, the true underlying business model. 30% of 23andMe customers consent to participate in large-scale, monetised genetic research databases.
2. Third-party sales, with lingering consumer fears over the commodification of highly personal biological data and the persistent question of whether data could be sold to another corporate buyer.
3. Forensic precedents, epitomised by the 2018 Golden State Killer case, where distant relative DNA was used to legally circumvent standard police databases.

Genetic discrimination and insurance

• Status quo: DTC data feels “off the books” compared with official medical records, but in Aotearoa New Zealand both life and health insurers have legally been able to use genetic results to discriminate against applicants.
• Clinical backlash: patients frequently delay or refuse vital clinical testing because of insurance fears. Over 80% of surveyed health professionals (n = 17/21) believe insurers’ use of genetic results should be legally regulated.
• Resolution: new legislative efforts from November 2024 aim to regulate against this practice and close the loophole.
• The funnel shown contrasts two outcomes for consumer genetic data entering insurance underwriting: discrimination under the status quo leading to a denied policy, versus regulation leading to legal protection.

Source cited: “Genetic discrimination by insurance companies in Aotearoa New Zealand: experiences and views of health professionals”.

Regulatory positions and guidance

• FDA (USA): 23andMe was prevented from issuing reports on medical conditions from 2013 to 2018, on the basis that they were medical devices. Reporting is now permitted for many conditions (~125) in the USA, but is not currently available in New Zealand.
• New Zealand guidance for GPs: “General practitioner attitudes to direct-to-consumer genetic testing in New Zealand”, New Zealand Medical Journal.
• Position statements exist from the Royal College of Pathologists of Australasia (expressing concern over DTC testing) and from the NHMRC.
• Companies also offer services through medical professionals, with clinician-facing report portals, and there is cardiovascular guidance published in Circulation (AHA journals), which the slide emphasised.

BMJ practice pointer, patient-oriented questions:

• Finding a “health risk” via DTC testing often does not mean the patient will go on to develop that health problem.
• DTC tests might report false positives (artefacts).
• Reassuring DTC results might be false negatives.
• Be confident in the provenance and interpretation of a genetic result before basing any clinical decision on it.
• If the patient meets criteria for referral to clinical genetics, refer regardless of the DTC result.

The clinical baseline

1. Do not base decisions on DTC alone. Verify the provenance and interpretation of a consumer genetic result before any clinical action.
2. Beware the artefacts. DTC tests frequently report outright false positives for rare variants because of microarray limitations.
3. False security is dangerous. Reassuring dashboard results can be false negatives, because incomplete SNP arrays do not check the entire exome.
4. Follow the criteria, not the app. If a patient meets established clinical criteria for referral, refer them regardless of the dashboard.

Synthesis: the DTC ledger

Level	Positive	Negative
Individual	Deepens engagement with personal health, satisfies curiosity, empowers the “worried well”	Severe anxiety from false positives, dangerous false security from false negatives, lack of interpretive knowledge
Clinical	Occasionally uncovers genuinely actionable hidden carrier risks (for example CFTR) that would otherwise go untested	Strains under-resourced GPs, polygenic risk scores remain scientifically immature and clinically unhelpful
Systemic	Crowdsources massive unprecedented datasets driving population-level genetic research	Opens ethical minefields in data privacy, forensic overreach, and insurance discrimination

Transcript flag (slide 1): the title slide's background matrix is partly hidden behind the title panel and its row and column labels repeat, so it does not read as a coherent data table. It is recorded as visible only, with no meaning inferred.

Self-test

1. List the four main clinical genetic test modalities, with the scope and an ideal use case for each.
2. State the approximate New Zealand cost of each of the four clinical test modalities.
3. Name the three adjunct tools listed alongside the main modalities and what each is used for.
4. Describe the four steps by which a DTC company turns a saliva sample into a consumer dashboard result.
5. Explain why DTC genotyping is not equivalent to sequencing, and how many variants a typical SNP chip interrogates.
6. Distinguish DTC testing from clinical genetics on method, access and support, focus, and data destination.
7. Explain where on the effect size versus allele frequency plot DTC tests operate, and why that limits their individual predictive power.
8. Distinguish monogenic from polygenic DTC testing in terms of concept, examples, and reliability.
9. Explain what the obesity polygenic scorecard data show about the top 10%, middle 80% and bottom 10% over 25+ years of follow-up.
10. Calculate the absolute risk shown to a consumer given a 3% base population risk of gout in men of European descent and a 2.0x relative risk for the SLC2A9 variant, and explain why the “double the risk” framing is misleading.
11. Explain what the “unknown genetic factors” branch of the DTC risk flowchart implies about the completeness of the calculated risk.
12. Explain why elite athletic heritability of ~0.7 does not translate into a usable DTC test for sporting talent.
13. List the three genuinely useful rare-variant screens in sport described in the lecture and the harm each aims to prevent.
14. Distinguish the roles of CYP2C19 for clopidogrel and for citalopram, and state the consequence of failing to clear citalopram.
15. State the CYP2C19 poor metaboliser frequencies for Oceanians, East Asians, African Americans and Caucasians, and explain why this matters for population-level testing decisions.
16. Describe three features of the “worried well” and three pressures DTC testing places on clinical services.
17. Describe the three downstream destinations of DTC genetic data, including the proportion of 23andMe customers who consent to research.
18. Explain the significance of the 2018 Golden State Killer case for genomic databases.
19. Describe the insurance situation for genetic results in Aotearoa New Zealand, the clinical consequence, and the proposed resolution.
20. Summarise the FDA’s regulatory history with 23andMe medical reports and the current availability of those reports in New Zealand.
21. List the four points of the clinical baseline for handling a patient’s DTC result.
22. A patient brings a DTC report showing a reassuring result for hereditary breast cancer risk, but has a strong family history that meets referral criteria. Explain what you should do and why.
23. A patient is alarmed by a DTC report stating a rare pathogenic variant. Explain the two main reasons this may not be a true finding, and your next step.
24. Integrative: explain how the technical limitations of SNP genotyping propagate into the individual, clinical and systemic harms in the DTC ledger.

Answers

Reveal answers

1. Single gene testing (specific pathogenic variation; known conditions such as haemochromatosis); targeted gene panel (variations linked to a disease phenotype; phenotype with known genetic heterogeneity such as retinitis pigmentosa); whole exome sequencing (the ~2% of the genome coding for proteins; unidentified variations in an affected child); whole genome sequencing (high-resolution comprehensive analysis; complex unsolved diagnostic challenges).
2. Single gene ~$100NZ;panel$300 to $400NZ;WES$400 to $500NZ;WGS$1,000 NZ.
3. Chromosome microarray analysis (aCGH) for CNVs; genotyping for risk association; RNA analysis for oncology.
4. Collection of a saliva sample posted to a CLIA-certified lab; genotyping of extracted DNA on a microarray chip; algorithmic filtering of raw variants against known risk markers from published GWAS; presentation on a consumer dashboard as stylised percentages and risk categories.
5. Genotyping reads only pre-selected positions rather than the whole sequence. The chip checks approximately 1 to 2 million specific common variants, so anything not on the array is not examined.
6. Method: SNP chip genotyping versus exome or whole genome sequencing. Access: self-directed with zero pre-test counselling versus medical referral with embedded genetic counsellors. Focus: ancestry, traits and polygenic predispositions versus specific highly penetrant actionable mutations. Data: proprietary corporate databases (30% consent to research) versus protected confidential medical records.
7. In the common allele frequency, low effect size corner, where GWAS findings sit. Such variants are easy to detect on chips but have negligible predictive power individually, so DTC builds a probabilistic model from thousands of them rather than finding definitive markers.
8. Monogenic: one gene mutation directly causes the condition (cystic fibrosis CFTR carrier, haemochromatosis); DTC is highly reliable for the specific variants on the chip and the result is clinically actionable, though rare mutations off the chip are missed. Polygenic: thousands of low-effect variants combined into a polygenic risk score (type 2 diabetes, severe obesity); the science is immature and predictive power varies.
9. By 25+ years of follow-up roughly 15 to 16% of the top 10% by score were severely obese, about 6% of the middle 80%, and about 1% of the bottom 10%. The score separates the extreme deciles over time.
10. 3% multiplied by 2.0 gives 6% absolute risk. The relative statement sounds alarming, but the absolute probability moved only from 3% to 6%, so 94% of such men still do not get gout.
11. It shows that the calculation captures only the known risk markers on the chip. Unmeasured genetic contributions are excluded, so the displayed absolute risk is incomplete rather than a total genetic risk.
12. Heritability describes how much of the variation in a population is genetic overall; it does not mean the responsible variants are known or testable. The common variants actually marketed, ACTN3 for power and ACE for endurance, have very little positive predictive value for sports success, so predicting elite outcomes is currently impossible.
13. Sickle cell trait screening; identification of APOE E4 variants; detection of rare cardiomyopathies to prevent sudden cardiac events on the field.
14. For clopidogrel, CYP2C19 activates the prodrug into its active antiplatelet form, so poor metabolism affects drug activation. For citalopram, CYP2C19 is required for clearance, and failure to clear leads to prolonged QT.
15. Oceanians 57%, East Asians 14%, African Americans 4%, Caucasians 2%. The very high frequency in some populations raises the question of whether at-risk populations should be tested before prescribing, which the lecture left open.
16. Worried well: driven by curiosity, health autonomy and glossy marketing; prone to false security, reading reassuring results as immunity; lacking the genetic literacy to interpret probabilistic results. Clinical pressures: GPs have massive workloads and limited time or specialist knowledge for 125+ DTC condition reports; geneticists and counsellors are in short supply; clinicians must untangle and verify false positives from consumer SNP chips.
17. Commercial research, the true business model, with 30% of 23andMe customers consenting to monetised research databases; third-party sales, raising fears about commodification of personal biological data; and forensic use.
18. Distant relative DNA in a consumer genomic database was used to identify a suspect, legally circumventing standard police databases, showing that consumer data has societal reach beyond the individual who submitted it.
19. Both life and health insurers have legally been able to use genetic results to discriminate against applicants. As a result, patients delay or refuse clinically indicated testing, and over 80% of surveyed health professionals (n = 17/21) want insurers’ use of genetic results legally regulated. New legislative efforts from November 2024 aim to close this loophole.
20. The FDA blocked medical condition reports from 2013 to 2018 by classifying them as medical devices. Reporting is now permitted for approximately 125 conditions in the USA, but is not currently available in New Zealand.
21. Do not base decisions on DTC alone, verify provenance and interpretation; beware false positive artefacts from microarray limitations; treat reassuring results with caution because incomplete arrays produce false negatives; and refer on established clinical criteria regardless of the dashboard result.
22. Refer to clinical genetics. Reassuring DTC results can be false negatives because the array does not check the whole exome, and referral criteria should be followed regardless of the DTC result.
23. DTC tests frequently report false positives for rare variants because of microarray limitations, and the provenance and interpretation of the result are unverified. Confirm through an accredited clinical pathway before any clinical action or reassurance.
24. A chip reads only ~1 to 2 million common variants, so it misses rare high-effect mutations (false negatives) and misreads rare variants it does attempt (false positives), while the common variants it does read carry only modest effects. Individually this produces anxiety or false security and poor interpretation; clinically it generates verification work for under-resourced GPs and polygenic scores that are not yet useful; systemically the same datasets, valuable for research, create privacy, forensic and insurance discrimination risks.