1. Start with the decision that could change
Diagnostic work-up in a person with symptoms, familial-risk assessment, carrier testing, pharmacogenomics, tumour profiling, prenatal or newborn screening, and ancestry testing answer different questions. No package is ‘best’ without an intended use. Before testing, write down the possible outcomes, what they could change and who will interpret them.
2. Screening is not diagnosis
Screening looks for increased probability in people without an established condition and usually requires confirmation. A diagnostic test examines a specific clinical hypothesis. A predictive result informs risk but often cannot determine whether disease will occur, when it will begin or how severe it will be. Carrier status for a recessive condition does not mean that the person is affected.
3. The method must match the alteration
A SNP array checks preselected positions; PCR and Sanger suit focused questions; an NGS panel deeply examines selected genes; exome sequencing mainly reads coding regions; genome sequencing is broader. Large rearrangements, repeat expansions, methylation, low-level mosaicism or RNA fusions may need separate methods. ‘Complete DNA test’ is not a technical specification.
4. Check three levels of evidence
Analytical validity asks whether the laboratory reliably detects the claimed variant. Clinical validity asks how well the variant relates to a condition or response. Clinical utility asks whether using the result improves a decision or outcome versus usual care. A link to an association study does not establish all three levels at once.
5. Read the boundary of a negative result
Documentation should list genes, regions and variant classes, coverage quality, detection limits and blind spots. A negative result means ‘not detected by what was examined at the stated sensitivity’, not ‘there is no genetic cause’. A focused consumer test may query only a few variants among many possibilities.
6. Ask for a test passport before purchase
Ask for the exact intended use, specimen, laboratory and relevant accreditation, method, panel version, coverage, detectable variant classes, classification criteria, confirmation policy, report format, secondary-findings policy, sample and data retention, deletion options and third-party transfers. Regulatory status must be checked for the specific product, intended use and country—not inferred from a company logo.
7. Marketing red flags
Treat with caution claims of a universal ‘health passport’, precise sport, career, personality, diet or lifespan prediction from a few SNPs; undisclosed methods; no limitations; proprietary risk scores without external validation; supplements sold from the same report; or treatment changes without clinical confirmation. Gene count and a polished interface do not replace evidence.
8. A quick selection route
Define the question → identify the test category → check whether the method detects the relevant variants → assess validity and utility → review possible outcomes before sampling → check privacy → plan confirmation and the next clinical step. If the result cannot change an evidence-based action, the test may still be interesting but is not medically necessary.