The advent of immune checkpoint inhibitors (ICIs) has fundamentally transformed the therapeutic landscape of oncology, leading to durable responses across multiple tumor types and redefining long-term survival expectations. ¹ In parallel with these clinical advances, substantial efforts have been devoted to identifying biomarkers capable of predicting treatment response, resistance, and toxicity. To date, however, most biomarker strategies in immuno-oncology remain anchored in a fundamentally static paradigm—relying on measurements obtained at a single time point, such as PD-L1 expression or baseline neutrophil-to-lymphocyte ratio (NLR). ^{1– 3} While clinically convenient, this paradigm treats the immune system as a snapshot, potentially misrepresenting its core nature as a dynamic, time-evolving system. This static approach stands in stark contrast to a foundational principle of physiology: accumulating evidence demonstrates that the processes governing immune surveillance and anti-tumor immunity are intrinsically dynamic and temporally structured. Immune cell trafficking, cytokine secretion, antigen presentation, and checkpoint pathway activity are not constant but fluctuate over time under the control of endogenous circadian clocks. ^{4– 6} These rhythms are generated by molecular clock networks operating within both immune cells and peripheral tissues, coordinating systemic immune readiness with environmental and metabolic cues. ^{4, 5} Consequently, a single-point measurement captures an arbitrary phase of a continuous biological oscillation, not a stable biological truth. Despite this well-established temporal organization of immunity, the dimension of time is largely omitted from the epistemic framework of immunotherapy biomarkers. Single measurements are frequently assumed to be representative and stable, even though spatial heterogeneity and temporal variability are well documented for key markers such as PD-L1. ^{2, 3} This creates a fundamental mismatch: we are attempting to describe a dynamic system with static descriptors. This mismatch may contribute to inconsistent predictive performance, apparent discordance between clinical trials, and the frequent observation that patients with similar baseline profiles experience markedly different outcomes. In this context, negative or equivocal biomarker results may not necessarily reflect biological irrelevance but rather a categorical error in measurement philosophy—the under-sampling of a time-dependent system. This conceptual gap becomes practically evident in emerging clinical data. Retrospective analyses have suggested that the time of day at which ICIs are administered may influence clinical outcomes. ^{7– 9} Although these findings remain exploratory, they reinforce the notion that immunotherapy efficacy is modulated not only by molecular targets but also by host temporal biology. Importantly, such time-of-day effects are system-level phenomena unlikely to be captured by any single baseline biomarker, regardless of its molecular sophistication. Parallel developments in longitudinal biomarker research have highlighted the value of immune trajectories over isolated values. Dynamic changes in inflammatory markers during treatment often provide richer prognostic information than baseline measurements alone. ^{10, 11}

However, most longitudinal approaches still focus on linear trends or slopes without explicitly modeling the inherent rhythmicity or temporal architecture of the immune system. As a result, the higher-order pattern — the coherent temporal organization itself—remains unmeasured and uninterpreted. To address these limitations, a paradigm shift is warranted: from point-based quantification to system-level, time-aware characterization. Rather than asking whether a given marker is “high” or “low” at baseline, a more biologically faithful question is: “Does the immune system exhibit coherent, stable, and rhythmically organized behavior over time?” Such properties cannot be inferred from isolated measurements but require longitudinal observation and analytical tools designed to reveal temporal structure. In this perspective, we introduce the concept of Circadian Immune Coherence (CIC) as a systems-level descriptor of immune organization over time.

CIC refers to the degree to which immune parameters maintain consistent rhythmic structure, phase alignment, and temporal stability across observation periods. Importantly, this concept does not replace existing biomarkers but reframes them within a temporal architecture, emphasizing relational patterns and system behavior rather than absolute values. A system with high CIC may represent a more resilient and adaptable physiological state, whereas low or unstable coherence may signal systemic dysregulation or vulnerability. Advances in time-series analysis, including spectral and Fourier-based methods, offer the necessary methodological toolkit to interrogate such coherence within longitudinal immune data. ¹² These approaches enable the decomposition of complex immune signals into rhythmic components, allowing the detection of periodicity, phase shifts, and loss of synchronization that define system-level organization.

Integrating these tools into immunotherapy research may therefore reveal latent biological structure with direct conceptual and clinical relevance. In this perspective, we synthesize evidence from circadian immunology, ^{4– 6} emerging clinical timing data, ^{7– 9} and longitudinal biomarker research ^{10, 11} to argue that static biomarkers often fall short because they omit time as a core variable of the system. By framing circadian immune coherence as a missing but essential systemic property for immunotherapy evaluation, we aim to provide a conceptual and methodological foundation for more biologically aligned biomarker science and time-aware clinical trial design.

Circadian rhythms represent a fundamental organizing principle of immune physiology, coordinating leukocyte trafficking, cytokine secretion, antigen presentation, and checkpoint pathway activity across the 24-hour cycle. ^{4– 6, 13} These rhythms are driven by cell-autonomous molecular clocks synchronized by systemic cues, resulting in predictable temporal patterns of immune readiness and responsiveness. While individual immune parameters have been shown to oscillate over time, most immunotherapy research continues to treat these dynamics as background variability rather than biologically meaningful structure.

To address this gap, we propose Circadian Immune Coherence (CIC) as a system-level descriptor of immune organization over time. CIC refers to the degree to which immune parameters exhibit stable, aligned, and rhythmically consistent temporal behavior across longitudinal observation. Importantly, CIC is not defined by the absolute value of a given biomarker, but by the relational and temporal properties of immune dynamics.

We conceptualize circadian immune coherence as a system-level property with three core aspects that we have found useful when working with longitudinal immune data. One is rhythmic integrity—essentially whether immune signals display a genuine periodic structure or are dominated by stochastic noise. Another is phase alignment, referring to whether different immune components tend to peak and trough in a coordinated manner within an individual. The third aspect, temporal stability, concerns whether these rhythmic relationships persist over clinically meaningful intervals, rather than fragmenting or drifting unpredictably over time. Together, these features define whether the immune system behaves as a coordinated temporal system or as a desynchronized collection of signals.

This framework differs fundamentally from conventional dynamic biomarkers that focus on trends, slopes, or early on-treatment changes. ^{10, 11, 14} While such approaches acknowledge that immune markers evolve over time, they typically do not interrogate whether this evolution follows an organized temporal structure. In contrast, CIC explicitly treats time as an intrinsic biological variable, rather than a nuisance parameter to be adjusted for or averaged out.

From a biological perspective, immune coherence is consistent with principles observed in other physiological systems, where synchronization and rhythmic coordination are critical for efficient function. ¹⁵ Loss of coherence, even in the presence of preserved mean values, has been associated with reduced adaptability and pathological states. Applied to immuno-oncology, this suggests that two patients with similar baseline biomarker levels may differ substantially in their immune system’s capacity to mount and sustain an effective anti-tumor response, depending on the coherence of their underlying immune rhythms.

Emerging clinical observations indirectly support this concept. Variability in immunotherapy outcomes associated with timing of drug administration ^{7– 9, 16} implies that host immune state at the moment of treatment is not interchangeable across the day. Similarly, longitudinal analyses demonstrating prognostic value of immune trajectories over baseline measurements ^{10, 14, 17} suggest that temporal behavior contains biologically relevant information. CIC provides a unifying framework capable of integrating these findings without reducing them to isolated associations.

Operationally, CIC does not require the introduction of entirely new biomarkers. Instead, it reframes existing immune parameters—such as inflammatory indices, leukocyte subsets, or cytokine levels—within a temporal architecture. What changes is not what is measured, but how measurements are interpreted. Rather than single thresholds, CIC emphasizes pattern recognition across time, enabling distinction between coherent oscillatory behavior and erratic or drifting immune dynamics.

Time-series analytical approaches, including spectral and Fourier-based methods, are particularly well suited for quantifying CIC. ^{12, 15, 18} These techniques allow decomposition of longitudinal immune data into rhythmic components, facilitating assessment of periodicity, phase consistency, and synchronization. Unlike traditional regression-based analyses, spectral methods are explicitly designed to detect hidden temporal structure, even in noisy biological data.

Within this conceptual model, high circadian immune coherence may reflect a resilient and adaptable immune system capable of responding predictably to immunomodulatory interventions. Conversely, low or unstable coherence may indicate vulnerability to immune exhaustion, dysregulated inflammation, or inconsistent therapeutic response. Importantly, CIC is envisioned as a continuous property, rather than a binary classification, allowing graded stratification and longitudinal monitoring.

By formalizing circadian immune coherence as a measurable system-level attribute, this framework provides a biologically grounded bridge between circadian immunology, longitudinal biomarker research, and immunotherapy trial design. In the following sections, we explore how time-series methodologies can operationalize this concept and discuss its implications for future biomarker development and time-aware clinical trials.

Quantifying circadian immune coherence demands a methodological departure from the conventional analytical toolkit of immuno-oncology. Standard approaches—such as comparing baseline values, calculating on-treatment deltas, or fitting linear slopes—are rooted in a reductionist paradigm that seeks to distill complexity into a single number or trend. ^{10, 11, 14} By treating time-related variation primarily as noise around a mean signal, these methods are conceptually blind to temporal structure. They can describe change, but they cannot characterize temporal organization—the very essence of coherence.

Time-series analysis offers a fundamentally different epistemological stance: it treats sequential immune measurements not as independent data points, but as a continuous temporal signal emitted by a dynamical system. Within this framework, the analytical goal shifts from assessing magnitude (“high” vs. “low”) to deciphering pattern (“organized” vs. “chaotic”). Among time-series methods, spectral and Fourier-based approaches are uniquely suited to this task, as they are explicitly designed to decompose complexity into its constituent rhythmic components. ^{12, 15, 18} This makes them not just another statistical tool, but the necessary operational language for a theory of immune coherence.

Fourier decomposition provides the core mechanism for this shift in perspective. By transforming data from the time domain to the frequency domain, it allows us to ask: “What rhythmic frequencies dominate this immune signal?” This is a qualitatively different question from “What is its average value or slope?” In practical terms, this transformation can reveal underlying circadian or ultradian periodicity even in noisy, irregularly sampled clinical data, because it is agnostic to waveform shape and robust to missing values—a critical advantage for real-world biomedical applications.

When applied through the lens of CIC, Fourier-based analysis interrogates three pillars of temporal organization:

Periodicity: Does the signal oscillate with a biologically plausible rhythm (e.g., ~24-hour), or is it dominated by stochastic fluctuation?

Spectral Concentration: Is the signal’s power focused within a narrow frequency band (indicating a stable, dominant rhythm), or is it dispersed (indicating a lack of organized oscillation)?

Phase Stability: Does the timing of peaks and troughs remain consistent over successive cycles, or does it drift, indicating a loss of temporal anchor?

This approach democratizes circadian analysis. It can be applied to routine clinical lab parameters—complete blood counts, inflammatory ratios, cytokine panels—without necessitating impractical, high-frequency circadian sampling protocols. The power of the method lies in its ability to extract latent temporal information from existing longitudinal data that is currently discarded as noise. ¹²

To move from a static snapshot to a dynamic movie of immune coherence, complementary techniques build upon the Fourier foundation: •

Windowed (Short-Time) Fourier Transforms can track how rhythmic properties evolve during therapy, capturing the phase drift or fragmentation of signals that may precede clinical progression or toxicity.

The integration of machine learning further amplifies this framework. ¹⁹ When models are trained not on static values but on spectral features (dominant frequency, phase stability, spectral entropy), they can identify novel temporal phenotypes that are invisible to conventional biomarkers. Crucially, in this “Chrono-Fourier” paradigm, AI serves not as a black-box predictor, but as a pattern-recognition engine operating within a biologically constrained feature space—the space of rhythms and temporal relationships.

Conceptually, this entire methodological shift redefines the meaning of “variability.” What appears as uninformative noise in a time-series plot may, in the frequency domain, be clearly identified as the absence of a coherent rhythm—a profound biological signature of dysregulation. Conversely, a stable, low-amplitude circadian oscillation reflects a highly organized system, even if its absolute values never cross conventional clinical thresholds.

Importantly, circadian immune coherence should not be conflated with mere rhythmic complexity or signal regularity. A highly complex or oscillatory immune signal is not necessarily coherent. Coherence implies structured temporal coordination—stability of dominant frequencies, constrained phase relationships, and reproducibility over time—rather than maximal variability or oscillatory richness.

From a systems perspective, CIC represents an emergent property of immune organization, arising from the coordinated interaction of cellular trafficking, cytokine signaling, endocrine modulation, and host–environment synchronization. Fourier-based metrics do not impose coherence onto the data; they quantify the degree to which such organization is already present or progressively lost.

This distinction is critical for biological interpretation. Loss of coherence may occur without dramatic changes in absolute biomarker levels, reflecting a breakdown of temporal coordination rather than depletion or activation of specific immune components. Conversely, preserved coherence in the setting of modest biomarker fluctuations may indicate resilient immune regulation capable of sustaining effective antitumor responses.

This has a direct implication for the core failure mode of static biomarkers: two patients with identical baseline profiles can have diametrically opposed temporal architectures. One may show a sharp spectral peak at 24 hours with stable phase (high CIC), while the other’s spectrum may be flat or chaotic (low CIC). This divergence in system organization—not in component value—may be the key to explaining heterogeneous treatment outcomes and the notorious irreproducibility of many biomarker associations.

Therefore, Chrono-Fourier approaches are not proposed as a replacement for existing tests, but as an integrative analytical layer that contextualizes them within the dimension of time. They operationalize CIC as a continuous, quantifiable system property, transforming it from a metaphor into a measurable variable. This, in turn, opens the door to truly time-aware precision medicine: longitudinal monitoring of immune coherence, patient stratification by temporal phenotype, and adaptive treatment scheduling synchronized with the patient’s intrinsic biological rhythms.

Adopting circadian immune coherence as a core framework necessitates a fundamental recalibration of two pillars of translational immuno-oncology: the science of biomarker development and the architecture of clinical trials. If immune function is constitutively temporal, then the prevailing practice of deriving biomarkers from single time-points commits a systematic error of omission. The resulting misclassification is not a failure of biological relevance, but of methodological completeness—we are measuring a state, not a system. Therefore, incorporating time is not an added complexity; it is a necessary correction to an incomplete model. 1.

From Static Dichotomies to Dynamic Phenotypes: A New Epistemology for Biomarkers

The dominant paradigm in biomarker development seeks binary classifiers—fixed thresholds that sort patients into “high” or “low” risk categories. This approach, while pragmatically seductive, is philosophically reductive. It imposes a static taxonomy upon a dynamic process, implicitly assuming temporal invariance where none exists. Circadian immune coherence proposes an alternative: classification by temporal phenotype. Here, the defining characteristic is not the magnitude of a marker at an arbitrary time-zero, but the pattern of its behavior over time—its rhythmic stability, phase alignment, and systemic coherence. A “temporal phenotype” captures the functional architecture of the immune system: Is it a well-orchestrated symphony with predictable rhythms, or a desynchronized ensemble? This phenotype directly interrogates systemic properties like resilience, adaptability, and capacity for coordinated response—the very properties that determine success or failure in immunotherapy. ^{10, 14, 17} Crucially, this is not a call to discard existing biomarkers, but to re-contextualize them within their native temporal dimension. A parameter like PD-L1 expression or NLR gains new meaning when viewed not as a static value, but as a moving point within a rhythmic trajectory. Two patients with identical “moderate” baseline NLR may belong to different ontological categories: one with a stable, coherent circadian rhythm in leukocyte subsets (high CIC), and another with a chaotic, aperiodic pattern (low CIC). Their similar snapshot obscures a fundamental difference in systemic organization, with profound implications for prognosis and therapeutic strategy.

The framework of circadian immune coherence, while compelling in its biological rationale, occupies a necessary frontier of inquiry between conceptual innovation and empirical validation. Although originally formulated as a generative hypothesis, its recent application to longitudinal real-world clinical datasets (manuscript under review) demonstrates operational feasibility and supports its transition toward a candidate framework for systematic investigation. A clear-eyed assessment of its limitations remains essential to define a credible and responsible path forward.

The foundational evidence for time-dependent effects in immunotherapy remains largely correlative and retrospective. Associations between treatment timing, immune rhythms, and clinical outcomes emerge from heterogeneous datasets in which unmeasured confounders—such as treatment selection, comorbidities, lifestyle factors, and supportive medications—are intrinsically interwoven with temporal signals. ^{7– 9} While the reproducibility of these observations across independent cohorts strengthens their biological plausibility, it does not in itself establish causality. Accordingly, circadian immune coherence should not yet be regarded as a ready-for-clinic diagnostic tool, but rather as a structured framework guiding hypothesis-driven validation and prospective experimentation.

Ideal circadian phenotyping relies on high-frequency, phase-anchored sampling—an experimental standard rarely attainable outside specialized chronobiology laboratories. In contrast, real-world clinical data are inherently sparse, irregular, and pragmatically collected. Although time-series approaches such as Fourier-based analysis exhibit robustness to such imperfections, a critical threshold of sparsity exists below which biologically meaningful rhythmic structure becomes indistinguishable from stochastic noise. Bridging this gap requires a dual strategy: (1) prospective studies incorporating denser temporal sampling where feasible, and (2) advanced analytical methods—including imputation frameworks and state-space models—capable of inferring latent rhythmic organization from real-world longitudinal data streams.

Circadian immune coherence does not arise in isolation. It represents an emergent property of host temporal biology, shaped by molecular clock function and dynamically modulated by disease burden, prior and concurrent therapies (particularly lymphodepleting regimens), concomitant medications such as corticosteroids, sleep–wake behavior, and psychosocial stress. A key challenge for future research is to move beyond observing coherence toward elucidating its determinants. This will require integrative, multilevel modeling approaches capable of partitioning variance in CIC into circadian, pathological, iatrogenic, and behavioral components. In this light, CIC may be most informative not as a static scalar, but as a dynamic readout reflecting the evolving host–temporal context during immunotherapy.

The transition from conceptual framework to validated scientific tool demands coordinated effort across several fronts:

Priority 1: Prospective, Time-Stratified Interventional Trials. The most direct test is a prospective trial where timing is not a hidden variable but a controlled element of design. Randomizing patients to receive immunotherapy within specific circadian windows (e.g., morning vs. evening infusion) while collecting serial immune samples will establish whether CIC is merely associated with outcomes or prognostically predictive of them. Such trials are the definitive step from correlation to causation.

Priority 2: The “Temporal Re-Analysis” Initiative. In parallel, a systematic, secondary analysis of existing high-value clinical trial biobanks and real-world datasets through the CIC lens can yield rapid, cost-effective insights. This effort would standardize temporal feature extraction, test the prognostic added value of CIC over static biomarkers, and identify patient subgroups where temporal signals are strongest. It turns past data into a testing ground for future hypotheses.

Priority 3: Methodological Standardization and Benchmarking. The field requires consensus on operational definitions and analytical pipelines for CIC. What are the minimal data requirements? Which spectral metrics are most robust and clinically interpretable? Establishing these standards, followed by head-to-head benchmarking against traditional biomarkers, is crucial for building a reproducible literature.

Priority 4: Integration with Multimodal Data and AI. The ultimate analytical model will fuse CIC metrics with genomic, transcriptomic, and digital phenotypic data (e.g., actigraphy). Machine learning is indispensable here, not as a black box, but as a tool for discovering latent temporal-response classes within this high-dimensional space. The goal is to move from “high CIC” to mechanistically defined “temporal endotypes.”

Immunotherapy represents a triumph of biological engineering, leveraging the inherent dynamism of the immune system as a therapeutic force. Paradoxically, the scientific lens through which we evaluate this dynamic interaction remains stubbornly static and time-agnostic. Relying on single-point biomarker measurements is akin to judging a symphony by a single, isolated note—it captures data, but misses the music. As evidence from circadian biology, longitudinal monitoring, and clinical chronotherapy converges, the conclusion is inescapable: our current biomarker paradigm is structurally insufficient to describe the system it seeks to measure.

In this perspective, we argue that this insufficiency is not merely technical, but conceptual. Circadian Immune Coherence (CIC) is proposed not as a silver-bullet biomarker, but as a necessary framework to correct this conceptual gap. CIC reframes the central question from “What is the state of the system?” to “How is the system organized in time?”- shifting focus from isolated biomarker magnitudes to relational, rhythmic properties such as integrity, alignment, and temporal stability. This represents a fundamental reorientation: from a chemistry of concentrations to a physics of oscillations.

Crucially, this reorientation is operationally feasible. Chrono-Fourier and time-series methodologies provide the analytical language to extract latent rhythmic structure from longitudinal clinical data already collected in routine practice. This enables a pragmatic evolution toward time-aware immunology without imposing new burdens on patients or healthcare systems.

We acknowledge that CIC is, at present, a powerfully generative hypothesis rather than a validated clinical tool. Its ultimate test lies in prospective, time-stratified trials and rigorous methodological benchmarking. Yet its immediate value is substantial: it provides the conceptual scaffolding to reinterpret past inconsistencies, design smarter experiments, and ask biologically faithful questions. By formally incorporating time as a core biological variable, immuno-oncology can progress from treating a dynamic system with static tools to embracing dynamism as a first principle.

Recognizing immune function as a process that unfolds in time is not a peripheral refinement—it is the essential next step in the evolution of precision cancer therapy. This perspective is a call to initiate that evolution by building a biomarker science as dynamic, adaptive, and temporally coherent as the immune system it seeks to understand.