Swimlane Plots in Oncology Reporting

Introduction

Oncology clinical trials often generate a large amount of longitudinal patient-level information. A single patient may receive multiple treatment cycles, experience dose interruptions or reductions, undergo repeated tumor assessments, achieve a response, progress, discontinue treatment, and remain in follow-up. Tables can summarize these events, but they can be difficult to interpret when the goal is to understand the sequence and duration of events for each individual patient.

A swimlane plot solves this problem by assigning one horizontal lane to each patient and using the horizontal axis to represent time. Treatment exposure becomes a horizontal bar, while clinically important events are placed at their corresponding time points.

What Does a Swimlane Plot Show?

The basic structure is straightforward. Each patient occupies one horizontal row, and the x-axis represents a clinically meaningful time scale such as days, weeks, months, or time since first dose.

Figure 1: Basic Oncology Swimlane Plot

Figure 1 demonstrates the central principle: the visual encoding should allow the reader to reconstruct the patient's clinical trajectory without consulting multiple tables.

Choosing the Time Origin

The time origin should be clinically meaningful and should be stated explicitly in the figure title, subtitle, or footnote. Common choices include:

What Belongs in the Lane?

The answer depends on the purpose of the figure. A CSR figure should generally emphasize information that helps interpret treatment exposure and clinical outcome rather than attempting to reproduce every event in the database.

Response Encoding

Response symbols should be simple enough to interpret at a glance. A common conceptual scheme is to distinguish response milestones from progression and treatment-disposition events.

The exact symbols are a graphic design choice; what matters is consistency and an unambiguous legend. If response categories such as CR, PR, SD, and PD are shown, define them in the figure legend or accompanying text.

Figure 2: Treatment Modifications and Assessments

Representing Dose Reductions and Holds

A treatment bar should not imply uninterrupted exposure if a patient had clinically meaningful treatment interruptions. One useful approach is to divide the exposure bar into segments.

For example, if a patient receives the planned dose through week 16, has a dose reduction through week 24, and then returns to the original dose, the lane can contain three contiguous segments rather than one undifferentiated bar.

Figure 3: First Response, Best Response, Progression, and Follow-Up

Ordering the Patients

Patient ordering has a major effect on how quickly a reader can interpret the figure. Random ordering may be statistically neutral, but it often produces a visually noisy display.

Common ordering strategies include:

• Best response: responders first, followed by stable disease and progressive disease.
• Time to progression: patients ordered by progression time.
• Treatment duration: longest treatment exposure first.
• Patient identifier: useful when traceability is the primary objective.
• Stratification factor: useful when comparing treatment cohorts or biologically relevant subgroups.

A Useful Mathematical Representation

Although a swimlane is primarily a graphical display, it helps to think of each patient's record as a collection of time intervals and event times.

For patient \(i\), let \(T_{i0}\) denote treatment start and \(T_{i1}\) denote treatment end. Treatment duration can then be represented conceptually as:

For a response event occurring at \(R_i\), time to first response can be expressed as:

Likewise, if progression occurs at \(P_i\), time to progression from treatment start is:

The swimlane does not replace formal time-to-event analyses. Instead, it provides a patient-level visual context for the same longitudinal information.

Data Structure for Programming

A robust swimlane program is much easier to maintain when treatment intervals and clinical events are stored separately. A conceptual patient-level event dataset might look like this:

patient_id   event_type       start_day   end_day   response   reason
001          treatment             0        180      —         —
001          dose_reduction       84        126      —         AE
001          response              56         56      PR        —
001          progression         180        180      PD        —
002          treatment             0        120      —         —
002          response              42         42      PR        —
002          progression         120        120      PD        —
003          treatment             0        240      —         —
003          response              70         70      SD        —
003          response             140        140      PR        —
003          follow_up            240        240      —         ongoing

This structure separates interval data from point events. That separation is particularly helpful when a patient has multiple dose changes, treatment holds, or response assessments.

Programming Logic

Example R Implementation

The following is a simplified example using ggplot2. The exact implementation should be adapted to the analysis dataset, display conventions, and production standards for the study.

library(ggplot2)

ggplot() +
  geom_segment(
    data = treatment,
    aes(x = start_day, xend = end_day,
        y = patient, yend = patient),
    linewidth = 7,
    lineend = "round"
  ) +
  geom_point(
    data = response,
    aes(x = event_day, y = patient, shape = response),
    size = 3
  ) +
  geom_point(
    data = progression,
    aes(x = progression_day, y = patient),
    shape = 18,
    size = 3
  ) +
  geom_point(
    data = discontinuation,
    aes(x = discontinuation_day, y = patient),
    shape = 124,
    size = 5
  ) +
  scale_x_continuous(name = "Days from first dose") +
  scale_y_discrete(name = NULL, limits = patient_order) +
  theme_minimal()

Handling Patients With No Progression

Patients without progression require special attention because the end of their displayed lane may represent different things: treatment discontinuation without progression, death without documented progression, database cutoff, or ongoing follow-up.

Common Mistakes

Validation Checklist

When Should You Use a Swimlane Plot?

Swimlanes are particularly effective when the clinical story is inherently longitudinal and patient-level. They are often useful for early-phase oncology studies, studies with small or moderate sample sizes, studies with substantial treatment modification, and settings where individual response trajectories are clinically informative.

They are less effective when the number of patients is so large that individual lanes cannot be read. In those situations, aggregated displays, Kaplan–Meier plots, response summaries, or tabular presentations may communicate the primary result more effectively.

Swimlane Plots vs. Other Oncology Figures

Final Takeaways

1. A swimlane plot is a patient-level longitudinal visualization, not merely a treatment-duration chart.
2. The horizontal axis should have a clearly defined and clinically meaningful time origin.
3. Treatment exposure should form the visual backbone, with response, progression, and disposition events layered on top.
4. Dose reductions and treatment holds should be visually distinguishable when they matter to interpretation.
5. Patient ordering should be prespecified and reproducible.
6. Every displayed event should trace back to a controlled analysis derivation.
7. The figure should complement, rather than replace, formal efficacy and safety analyses.

References

Eisenhauer, E.A., et al. (2009). New response evaluation criteria in solid tumours: Revised RECIST guideline (version 1.1). European Journal of Cancer, 45(2), 228–247.

Food and Drug Administration. Clinical Trial Endpoints for the Approval of Cancer Drugs and Biologics. Guidance for Industry.

National Cancer Institute. Common Terminology Criteria for Adverse Events (CTCAE).

This tutorial uses illustrative patient-level data and graphics. The example figures are not derived from an actual clinical trial.