Precision medicine increasingly uses molecular information to guide clinical treatment decisions. Genetic testing can identify inherited susceptibility, predict certain drug responses, and reveal molecular targets for treatment. In cancer, genomic profiling has enabled therapies directed against specific tumor mutations. Yet much of precision medicine still relies on molecular information obtained at a single point in time.

A patient undergoes molecular testing, and the results guide treatment selection. The disease may subsequently progress, treatment resistance may develop, or the patient may experience adverse effects. If diseases evolve, why does precision medicine remain largely centered on static molecular profiling?

Beyond Static Molecular Profiling

Germline genetic variants generally remain unchanged throughout life and can inform drug selection or dosing whenever a corresponding medication is considered. Many other molecular characteristics are dynamic. Gene expression, protein activity, metabolic processes, and immune responses change with disease progression and treatment. Cancer provides a particularly clear example because tumor cells can acquire additional mutations or undergo changes in clonal composition.

A targeted therapy may initially suppress tumor cells carrying a particular mutation. Over time, resistant populations can emerge through additional genetic alterations or alternative signaling pathways. The original mutation may remain detectable, but targeting it alone may no longer control the disease. Repeated molecular profiling can identify emerging resistance mechanisms and provide information for subsequent treatment selection.

How Treatment Changes the Disease

Disease progression is influenced not only by the underlying biological mechanisms but also by the treatments already administered. Targeted cancer therapies can impose selective pressure on tumor populations. Immune-modulating treatments can change inflammatory activity, and repeated exposure to a medication may alter its subsequent effectiveness or tolerability.

Two patients with similar molecular profiles at diagnosis may follow very different clinical courses after receiving the same initial therapy. One may maintain disease control, while the other develops resistance or treatment-limiting toxicity. Even if their original genomic findings remain similar, the next treatment may differ. A dynamic Precision Medicine model can integrate longitudinal molecular information with treatment history and clinical response to guide subsequent therapeutic decisions.

From Molecular Measurements to Treatment Decisions

Longitudinal molecular profiling can track changes in disease mechanisms during treatment and provide information for subsequent therapeutic decisions. The clinical utility of these measurements depends on identifying molecular changes that inform treatment response, resistance mechanisms, and subsequent therapeutic options.

Longitudinal monitoring can be tailored to the disease, available biomarkers, and therapeutic decisions being considered. In oncology, circulating tumor DNA or repeat biopsy may identify acquired resistance. For diseases in which affected tissues are inaccessible, longitudinal monitoring may depend more heavily on clinical and physiological measurements combined with existing genomic information.

Artificial Intelligence and Sequential Treatment Decisions

Adaptive treatment becomes increasingly complex when decisions depend on accumulated molecular and clinical information. A patient's treatment history may include multiple medications, variable responses, adverse effects, and changes in disease activity. Artificial intelligence can integrate longitudinal molecular profiles, clinical trajectories, and treatment histories to support individualized sequential treatment decisions.

The scientific difficulty lies in moving from prediction to treatment selection. Sequential treatment optimization requires models that estimate patient outcomes under alternative therapeutic strategies. Observational clinical data are particularly challenging because treatment selection is influenced by disease severity and previous responses. Patients receiving more intensive therapy may have poorer outcomes because they were already more severely ill. Reliable treatment recommendations require estimating the effects of alternative therapeutic decisions on patient outcomes.

Toward Adaptive Precision Medicine

We advocate extending precision medicine from individualized treatment selection to individualized treatment trajectories. The initial treatment would be selected using the patient's molecular and clinical characteristics. Subsequent decisions would incorporate updated information about disease activity, treatment response, emerging resistance, and tolerability. The appropriate decision may be to continue the existing treatment, adjust its intensity, introduce another intervention, or change therapeutic strategy. Treatment adjustments would be guided by the clinical significance of molecular changes and the expected benefit of available therapeutic options.

Conventional clinical trials frequently assess a particular intervention or predefined regimen. An adaptive treatment strategy may generate different therapeutic sequences for patients as their molecular and clinical conditions evolve. This creates a new regulatory challenge for agencies such as the U.S. Food and Drug Administration (FDA): how can the safety and effectiveness of an entire therapeutic trajectory be evaluated when treatment decisions vary across patients and over time? One approach is to evaluate the adaptive decision system as a whole, with predefined treatment options, decision rules, and clinical endpoints. Prospective trials could compare patient outcomes under adaptive treatment strategies with those under standard clinical care, assessing both the benefits and risks accumulated throughout the treatment course.

 Regulatory approval would therefore need to address the validated boundaries of the adaptive system, including which treatment decisions are permitted, how molecular and clinical information guides those decisions, and how the system's performance is monitored over time. Physician oversight would remain an essential component of clinical implementation.

Methods for Evaluating Adaptive Treatment Strategies

Sequential Multiple Assignment Randomized Trials (SMART) provide a practical method for studying adaptive treatment strategies. In these trials, patients can be randomized at successive treatment decision points based on their responses to earlier interventions. This design allows investigators to compare multiple treatment sequences within a single study and estimate which strategies produce better outcomes for different patient groups.

Statistical approaches such as Q-learning and A-learning can analyze these data to identify effective sequential treatment rules. Longitudinal observational studies can provide additional evidence through causal inference methods that account for time-varying confounding and treatment history. These methods can also support the evaluation of treatment strategies across larger and more diverse patient populations.

Evidence synthesis could combine patient-level longitudinal data from multiple studies, harmonize treatment decision points and clinical outcomes, and use causal inference methods to estimate the effectiveness of alternative treatment sequences. Such an approach could extend evidence-based medicine to adaptive therapeutic strategies and provide a methodological foundation for their regulatory evaluation. The remaining research challenge is scaling these approaches to high-dimensional molecular data, many possible treatment sequences, and heterogeneous patient populations.

We discuss this direction in our recently Perspective, Precision Medicine 3.0: Integrating Genomic Stratification with Adaptive Therapeutic Trajectories.  (https://doi.org/10.3390/jgbg1020016)