Treatment Pathways

Treatment pathway analysis describes the sequence of therapies that patients in a cohort actually receive over time. It answers the question: "After diagnosis (or treatment initiation), what drugs do patients receive, in what order, and how often do they switch, add, or discontinue therapies?" Treatment pathway analysis is essential for understanding real-world treatment patterns and adherence to clinical guidelines.

---

How Pathway Analysis Works

The analysis proceeds through several stages:

1. Target cohort identification: Start with a target cohort (e.g., newly diagnosed T2DM patients) that defines the population and observation window.

2. Event cohort matching: For each patient in the target cohort, identify which event cohorts (one per drug or drug class) overlap with their target cohort membership period. Each event cohort defines a continuous treatment era for that therapy.

3. Sequence construction: For each patient, construct an ordered sequence of treatment steps based on the chronological order of event cohort eras. The combination window determines whether overlapping therapies are classified as "combination" (taken simultaneously) or "sequential" (one after another).

4. Path aggregation: Aggregate the individual patient sequences into pathway frequencies --- how many patients followed each unique treatment path.

5. Results presentation: Display the aggregated pathways as a ranked table showing the most common treatment sequences.

---

Creating a Treatment Pathway Analysis

1. Navigate to Analyses and select the Pathways tab.
2. Click New Analysis and select Pathway.
3. Configure the analysis on the design page.

Design Configuration

Target Cohort: Add the cohort representing the population to analyze. This is typically an incident (new-user) cohort for a condition or treatment.

Event Cohorts: Add multiple event cohorts, each representing one therapy or drug class to track. For a diabetes treatment pathway study, you might add:

Event Cohort	Description
Metformin users	Patients with active metformin exposure
Sulfonylurea users	Patients with active SU exposure
DPP-4 inhibitor users	Patients with active DPP-4i exposure
GLP-1 RA users	Patients with active GLP-1 receptor agonist exposure
SGLT2 inhibitor users	Patients with active SGLT2i exposure
Insulin users	Patients with active insulin exposure

Each event cohort should be independently defined and generated before configuring the pathway analysis.

Settings:

Parameter	Description	Default
Combination window	Days within which two overlapping drug eras are classified as a combination therapy rather than sequential	1
Max depth	Maximum number of treatment steps to track in the sequence	5
Max path length	Maximum number of treatment steps to display	5
Minimum cell count	Suppress pathways with fewer than N patients (privacy)	5

Understanding the Combination Window

The combination window is the most impactful parameter in pathway analysis. It determines how overlapping drug exposures are classified:

• Window = 1 day (strict): Only drugs with eras that start on the same day are classified as combination therapy. Most switches will appear as sequential steps.
• Window = 30 days: Drugs with eras overlapping for at least 30 days are classified as combination. This accommodates refill gaps and pharmacy overlap.
• Window = 90 days: Very lenient --- drugs active at any point within a 90-day sliding window are considered combination.

Choosing the combination window

Match the combination window to the prescription pattern in your data:

• 30-day supply databases (US commercial claims): Use 30 or fewer days
• 90-day supply databases (Medicare, mail-order): Use up to 90 days
• EHR data without supply days: Use a shorter window (7--14 days) since exposure gaps are common

---

Viewing Results

After execution, the results page shows the aggregated treatment pathways in tabular form:

Column	Description
Pathway	Ordered sequence of treatment steps (e.g., "Metformin --> Metformin + Sitagliptin --> Insulin")
Count	Number of patients who followed this exact sequence
% of Target	Percentage of the target cohort following this pathway

Pathways are ranked by patient count (most common first).

Example Results

Pathway	Count	% of Target
Metformin only	12,340	31.2%
Metformin --> Metformin + DPP-4i	4,560	11.5%
Metformin --> Sulfonylurea	3,890	9.8%
Metformin --> Metformin + SGLT2i	2,780	7.0%
Metformin --> Insulin	2,100	5.3%
Sulfonylurea only	1,890	4.8%
No treatment	1,560	3.9%
Metformin --> Metformin + GLP-1 RA	1,230	3.1%
Other pathways (< 3% each)	9,150	23.4%

Visualization roadmap

Sankey diagram and Sunburst visualization for treatment pathways are planned for a future release. These interactive visualizations will show patient flow through treatment sequences as connected streams (Sankey) or nested arcs (Sunburst). The current release provides full pathway data in tabular form.

---

Common Pathway Patterns

Understanding the vocabulary of treatment patterns helps interpret results:

Pattern	Definition	Example
Monotherapy	Patient stays on a single agent throughout	Metformin only
Add-on (augmentation)	Second agent added without stopping the first	Metformin --> Metformin + DPP-4i
Switch	First agent discontinued, second started	Metformin --> Sulfonylurea
Intensification	Escalation from one class to a more intensive class	Oral agents --> Insulin
Step-down	De-escalation from intensive to less intensive therapy	Insulin --> Metformin only
Combination start	Initial therapy is a combination	Metformin + SGLT2i (first line)
Treatment gap	Period with no active therapy between steps	Metformin --> [gap] --> Sulfonylurea

These patterns map directly to treatment guideline recommendations and can reveal how closely real-world practice aligns with evidence-based algorithms.

---

Multi-Source Comparison

Run the same pathway analysis across multiple data sources to understand geographic and population-level variation in treatment patterns:

Source	Top Pathway	2nd Pathway	3rd Pathway
US Commercial Claims	Metformin (35%)	Met --> Met+DPP4i (12%)	Met --> SU (10%)
UK CPRD	Metformin (42%)	Met --> SU (15%)	Met --> Met+SGLT2i (8%)
Korea NHIS	Metformin (28%)	Met+DPP4i (18%)	DPP4i only (11%)

Cross-source comparison reveals how prescribing culture, guideline adoption, formulary coverage, and healthcare system structure influence treatment decisions.

---

Interpreting Results

What Pathways Reveal

• Guideline adherence: Do most patients start with recommended first-line therapy?
• Treatment inertia: How long do patients stay on initial therapy before intensification?
• Real-world combination use: Which combinations are actually used in practice?
• Equity patterns: Do pathway distributions differ by age, gender, or other demographics?

Common Pitfalls

• Short follow-up: If many patients have short observation periods, they may appear as "monotherapy only" simply because intensification was not yet observed. Filter for patients with minimum follow-up.
• Event cohort gaps: If event cohorts are not comprehensive (e.g., missing insulin), patients using those therapies appear as "no treatment" or have truncated pathways.
• Data lag: Prescription claims may lag behind actual prescribing by weeks or months, causing apparent treatment gaps.

Missing therapies bias pathways

If an important therapy class is not represented by an event cohort, patients using that therapy will appear to have a treatment gap or shorter pathway. Ensure all relevant therapies in the therapeutic area are represented as event cohorts.

---

Best Practices

1. Define comprehensive event cohorts: Include all major therapy classes for the condition of interest. Missing classes create misleading pathway gaps.

2. Use incident target cohorts: Start with newly diagnosed or new-user populations to capture the full treatment trajectory from the beginning.

3. Calibrate the combination window: Run the analysis with multiple window values (e.g., 1, 30, 90 days) to understand how the classification changes. Report the chosen window and justification.

4. Require minimum follow-up: Consider adding an inclusion rule requiring minimum observation after index (e.g., 365 days) to ensure patients have sufficient time for treatment changes to be observed.

5. Report pathway coverage: State what percentage of the target cohort is represented by the reported pathways. If the "Other" category exceeds 30--40%, consider simplifying by grouping related therapies.

6. Validate against clinical knowledge: Share results with clinical domain experts to confirm that observed patterns make medical sense. Unexpected pathways may indicate data quality issues or cohort definition problems.