Skip to content

Hero image: A woman and multidisciplinary clinical team reviewing a personalized treatment plan together

All insights

Precision health

Sex-Aware Precision Medicine: Better Evidence Before Personalized Care

11 min readEvidence synthesis
Read the evidence

The question in focus

How sex, gender, reproductive stage and diverse representation can improve research, prescribing and precision medicine without biological oversimplification.

Evidence at a glance

86

medicines were evaluated in a review linking sex differences in drug exposure and adverse reactions

Women had higher pharmacokinetic values for 76 of the 86 medicines; among 59 with identifiable adverse reactions, exposure differences predicted the direction of sex-biased reactions in 88% of cases. This was a literature synthesis, not a dosing rule for every drug. [1]

Sex differences in pharmacokinetics predict adverse drug reactions in women

Precision medicine must distinguish sex and gender

Sex-related biology can involve chromosomes, hormones, anatomy, body composition, immune function and metabolism. Gender can shape exposures, work, caregiving, violence, access, communication and treatment. Neither variable explains every difference, and both interact with age, ancestry, environment and socioeconomic conditions. The Endocrine Society advises that sex and gender should not be used interchangeably. [5]

A precision model should therefore begin with a causal question rather than a pink version of standard care. Is a difference driven by pharmacokinetics, reproductive stage, disease prevalence, diagnostic thresholds, clinician behaviour or unequal access? The answer determines whether the appropriate response is a dose study, new reference range, inclusive trial, service redesign or social intervention.

  • Define each variable and how it is measured.
  • Avoid treating sex as a universal binary proxy for all biology.
  • Avoid treating gender categories as fixed biological mechanisms.

Drug response is a concrete use case

A 2020 analysis examined 86 medicines with published sex-specific pharmacokinetic data. Women had higher exposure-related values for 76. Among 59 medicines with identifiable adverse reactions, the direction of sex-biased pharmacokinetics predicted the direction of sex-biased reactions in 88% of cases. The authors noted that women experience adverse drug reactions nearly twice as often as men. [1]

These findings do not support lowering every dose for every woman. They support drug-specific research that considers absorption, distribution, metabolism, elimination, body size, hormonal state, kidney and liver function, co-medication and the relationship between exposure and response. An individual's prescribed dose should follow product information, clinical guidance, monitoring and qualified judgment.

  • Look for sex-disaggregated pharmacokinetic and safety data.
  • Monitor benefit and adverse effects after initiation or dose change.
  • Study pregnancy, lactation and menopause when relevant to use.

Participation has improved, analysis still matters

NIH reports that female participants represented 57.8% of enrollment in NIH-supported clinical research in fiscal year 2025. [3] This aggregate is encouraging but does not show whether representation matches burden in each condition, whether pregnant people are included appropriately, or whether studies can estimate treatment differences.

An analysis of 20,020 interventional trials registered from 2000 to 2020 found variation by clinical area and underrepresentation relative to disease burden in several fields. [4] Representation should be evaluated against the intended-use population and relevant disease burden. Precision claims also require prespecified analyses, adequate sample size and honest reporting when subgroup estimates are uncertain.

  • Report enrollment and outcomes by relevant sex and gender variables.
  • Explain exclusions based on pregnancy potential or reproductive status.
  • Do not claim a subgroup effect from an underpowered post hoc comparison.

Preclinical evidence sets the foundation

NIH's Sex as a Biological Variable policy expects vertebrate animal and human studies to factor sex into design, analysis and reporting or justify a single-sex approach. It responded partly to over-reliance on male animals and cells, which can obscure mechanisms relevant to all people. [2]

Good design does not require every experiment to test every interaction. It requires a scientifically reasoned plan, balanced or factorial design where appropriate, transparent reporting and replication. Cell sex, hormonal conditions, estrous or menstrual variables and environmental factors should be recorded when they could affect the result, while avoiding unnecessary complexity that makes a study uninterpretable.

  • Report the sex of cells and animals when known.
  • Analyze rather than merely include both sexes.
  • Replicate unexpected interactions before clinical translation.
  • Connect preclinical findings to a plausible human pathway.

Precision systems require transparent validation

Biomarkers and algorithms can combine genomic, laboratory, imaging, sensor and clinical data, but high dimensionality increases overfitting risk. Models need external validation, calibration, missing-data analysis and evaluation across relevant life stages and groups. A model that includes sex may improve prediction, worsen stereotypes or simply encode existing care disparities depending on design.

FDA guidance encourages analysis and communication of sex-specific data in medical-device clinical studies where appropriate. [6] Clinicians and patients need to know whether sex-specific performance is established, uncertain or not assessed. Precision medicine is credible when it reduces uncertainty in a decision and improves outcomes, not when it merely adds more personal data.

  • Compare the model with a simpler baseline.
  • Assess whether sex-specific thresholds improve net clinical benefit.
  • Monitor performance as population and practice change.
  • Retain human review when output uncertainty is high.

What the evidence cannot yet answer

  • Observed sex differences can arise from biology, gendered exposure, access, measurement or combinations of these.
  • Literature reviews of pharmacokinetics do not provide universal individual dosing instructions.
  • Aggregate trial enrollment can hide condition-level underrepresentation.
  • Subgroup models require adequate events and external validation to avoid unstable estimates.

Questions worth taking into care

  1. What mechanism makes sex or gender relevant to this decision?
  2. Was the exact medicine, device or model evaluated by relevant subgroup?
  3. Are reproductive stage and co-medications considered when appropriate?
  4. Does personalization improve outcomes compared with a simpler approach?
  5. Are uncertainty and potential structural bias communicated clearly?

Source record

Evidence used in this review

Sources were selected for clinical authority, methodological relevance and traceability. Links open the original guidance, public-health record or research publication.

  1. [1]
  2. [2]
  3. [3]
  4. [4]
  5. [5]
  6. [6]
Editorial standard

This evidence synthesis is for general information. It does not diagnose a condition or replace care from a qualified health professional. Treatment choices depend on individual history, examination, local guidance and informed preference. Emergency or rapidly worsening symptoms need urgent local medical assessment.