Introduction: The Cyclical Puzzle of Seasonal Affective Disorder

Seasonal Affective Disorder (SAD) is a subtype of recurrent major depression that follows a predictable seasonal pattern. Its hallmark is the onset of depressive episodes during the autumn or winter months, with full remission occurring in the spring or summer. The prevalence of SAD ranges from 1% in mild climates to nearly 10% in high-latitude regions such as Scandinavia or Canada, where winter daylight is severely limited. The condition is closely tied to the shortening photoperiod—the length of daylight each day—which disrupts the body’s internal circadian rhythm and alters neurotransmitter systems, particularly serotonin and dopamine. These biological disruptions produce classic symptoms: hypersomnia, carbohydrate cravings, weight gain, lethargy, and a profound lack of motivation.

Because these symptoms recur with clockwork regularity each year, researchers have long sought mathematical descriptions that can capture and predict the seasonal swing. Among the most versatile tools is the cosine function, a periodic wave that naturally mirrors the sinusoidal rise and fall of day length. By fitting a cosine curve to symptom data, clinicians can quantify the amplitude of the seasonal effect, pinpoint when symptoms will peak, and evaluate how treatments like light therapy flatten or shift the curve. This article explores how cosine modeling works, how it applies to light therapy, and how it is being extended with modern data streams to personalize care.

The Biological Basis: Why SAD Follows a Periodic Pattern

The primary driver of SAD is the annual variation in photoperiod. As days shorten in autumn, the suprachiasmatic nucleus (SCN)—the brain’s master clock—detects reduced light input via the retinohypothalamic tract. This triggers a cascade of changes: melatonin secretion begins earlier in the evening and persists longer into the morning, phase‑delaying the sleep‑wake cycle. Simultaneously, serotonin transporter binding increases in the winter, reducing synaptic serotonin availability and contributing to depressive symptoms. The resulting mood pattern is strikingly consistent: symptoms typically begin in September or October, worsen through December and January, and resolve by March or April as daylight lengthens.

This regularity makes SAD an ideal candidate for periodic modeling. Unlike many psychiatric conditions—which can fluctuate unpredictably due to random life events—SAD exhibits a strong annual signal. The cosine function, with its single peak and trough per cycle, is the simplest mathematical representation of such a pattern. More complex phenomena, such as the slight lag between the winter solstice and the nadir of mood (which often occurs in late January), can be captured by adjusting the phase shift. This tractability has made cosine models a staple of chronobiological research and clinical forecasting.

Building the Core Cosine Model

A standard cosine model for SAD symptom severity over a 12‑month period is expressed as:

S(t) = A · cos(2π (t − φ) / 12) + C

Here, t is time in months (with January as month 1), and the remaining parameters define the shape and position of the seasonal curve.

Parameter Deep Dive

  • A (Amplitude) – The half‑difference between the peak and trough of symptom severity. A higher amplitude indicates a stronger seasonal effect. Genetic predisposition, latitude of residence, and baseline mental health all influence an individual’s amplitude. For instance, a person living near the Arctic Circle may have an amplitude twice that of someone in Florida.
  • φ (Phase shift in months) – Shifts the cosine wave left or right along the time axis. In SAD, φ typically lies between 1 and 2 months, placing the peak of symptoms in January or February—after the solstice—because the biological response to reduced light takes several weeks to build. Some individuals may have a phase shift closer to zero, corresponding to earlier peaks.
  • C (Baseline constant) – The average symptom level across the entire year, representing the component of depression that is unrelated to season. For patients with comorbid non‑seasonal depression, C may be elevated year‑round.
  • Period – Fixed at 12 months for annual cycles. In theory, other periods could be modeled (e.g., a 6‑month cycle for some bipolar patients), but the standard SAD model assumes a single annual peak.

Fitting the Model to Data

Clinicians collect symptom scores—often using the Structured Interview Guide for the Hamilton Depression Rating Scale, Seasonal Affective Disorders version (SIGH‑SAD)—at weekly or monthly intervals for at least one full year. These data are then fitted to the cosine equation using least‑squares regression or Fourier analysis. The resulting parameters provide a quantitative snapshot of the patient’s seasonal pattern. A high R² value (e.g., >0.7) indicates that much of the symptom variance is attributable to the seasonal component. This fit can then be used to forecast future episodes and to assess treatment response by comparing parameters before and after an intervention.

Modeling Light Therapy: Two Mechanistic Approaches

Light therapy is the most evidence‑based treatment for SAD. Typically, patients are exposed to a bright white light source (10,000 lux) for 30–60 minutes each morning. This exposure suppresses melatonin, phase‑advances the circadian clock, and increases serotonin synthesis. The clinical effect can be incorporated into the cosine model in two principal ways.

Amplitude Reduction Model

The simplest modification subtracts a treatment effect term T from the amplitude:

S(t) = (A − T) · cos(2π (t − φ) / 12) + C

  • T represents the reduction in seasonal amplitude attributable to consistent light therapy. If T is sufficiently large (T ≥ A), the seasonal oscillation is completely erased, and symptoms become constant (a flat line). In clinical trials, T typically corresponds to a 40–60 % amplitude reduction compared to placebo.
  • This model assumes that light therapy does not change the timing of the peak nor the baseline non‑seasonal depression level. That assumption holds reasonably well for many patients: the winter dip becomes shallower, but the same phase and baseline still apply.

Phase‑Advance Model

An alternative approach models light therapy as a phase shift. Morning bright light advances the circadian rhythm, potentially moving the symptom peak earlier in the winter. The equation becomes:

S(t) = A · cos(2π (t − (φ − δ)) / 12) + C

  • δ (Phase advance in months) is a positive value that shifts the cosine curve leftward. For instance, a δ of 0.5 months would move the peak from mid‑January to early January. Empirical studies show that phase advances of 0.3–0.8 months are common, but the effect is variable across individuals and may depend on the timing of light exposure.

In reality, both amplitude reduction and phase advance often occur concurrently. A combined model—with both a T term and a δ term—can be fitted, although it requires more data points to stabilize the estimates. Some researchers also model a slope in baseline C (e.g., a gradual decrease over the winter) to capture accommodation or spontaneous remission.

Practical Applications in the Clinic

Individualizing Treatment Timing

Once a patient’s φ is known, the clinician can determine the optimal window to start light therapy. For example, if the peak occurs in early February, therapy might begin in early October to preempt the upward slope. The model can also indicate when therapy can be safely discontinued—usually after the spring equinox, when natural daylight becomes sufficient to sustain mood improvement. This is particularly useful for patients who are reluctant to use the light box year‑round.

Personalizing Dose and Duration

The amplitude reduction model provides a metric for therapy effectiveness. If after two weeks of standard morning light (30 minutes at 10,000 lux) the amplitude reduction is only 20 %, the clinician may increase exposure to 45 minutes, add a second session in the afternoon, or switch to a higher‑intensity device. Repeated cosine fits across consecutive winters can guide long‑term management and detect when tolerance develops or when seasonal patterns shift.

Large‑Scale Public Health Forecasting

Aggregated cosine models from epidemiological data can help public health authorities predict regional demand for mental health services. By fitting cosine curves to population‑level depression scores (e.g., from survey or pharmacy data) and incorporating latitude and average winter cloud cover, planners can forecast the severity of the SAD season in different cities. This enables targeted awareness campaigns, subsidized light box distribution, and resource allocation for crisis hotlines.

Limitations of the Cosine Approach

While the cosine model is intuitive and mathematically tractable, it has significant limitations that clinicians must recognize:

  • Symmetry and smoothness assumptions – Cosine waves imply a symmetrical rise and fall. In reality, many SAD patients experience a rapid onset of symptoms in autumn (steep ascent) and a slow recovery in spring (gradual descent), producing an asymmetric curve. More flexible models—such as sine waves with added harmonic terms or a skewed cosine—can capture this asymmetry.
  • Non‑sinusoidal patterns in some patients – Not all patients exhibit a single annual peak. Some have a biphasic pattern with both winter and summer depressive episodes (seasonal bipolar disorder). Others show no clear seasonal component at all. For such individuals, a cosine model either fits poorly or misleads the clinician.
  • Confounding factors – The simple model ignores environmental variables such as temperature, barometric pressure, social stressors (e.g., holiday season), and vitamin D levels. These can be added as covariates in a mixed‑effects extension, but the basic cosine curve does not account for them.
  • Data intensity – Accurate estimation of A, φ, and C requires frequent sampling—ideally weekly—over at least one full cycle. Retrospective recall is notoriously unreliable. The rise of digital mood tracking via smartphone apps has made prospective data collection more feasible, but many clinical settings still lack such infrastructure.

Despite these drawbacks, the cosine model remains a valuable first approximation. More advanced techniques, such as Fourier series with multiple harmonics, can handle non‑sinusoidal shapes while preserving the periodic core. For example, adding a second harmonic (cos(4πt/12)) allows the model to produce sharper peaks and flatter troughs. Machine learning approaches can then incorporate additional predictors (daily sunlight minutes, temperature, activity levels) to improve predictive accuracy beyond the basic cosine curve.

Future Directions: Real‑World Data and Dynamic Modeling

The proliferation of wearable devices and smartphone sensors has opened the door to continuous, objective measurement of photoperiod, sleep timing, activity, and even light exposure. These streams can be synchronized with daily mood ratings to build individualized cosine models that update in real time. A mobile app could detect that a user’s estimated amplitude is increasing earlier than expected and prompt them to start light therapy sooner. Such dynamic modeling is an active area of research; for an overview, see this review on digital phenotyping for seasonal affective disorder.

Another promising direction is incorporating meteorological data. Regions with high winter cloud cover—such as the Pacific Northwest or the United Kingdom—experience more severe SAD even at the same latitude. Adding a cloud‑cover factor as a modulating coefficient on amplitude can improve model accuracy. Social factors, such as the stress of the holiday season (which may elevate baseline C in December), can be included via dummy variables.

From a therapeutic standpoint, future models may compare different light therapy regimens—such as dawn simulation versus morning bright light—by applying different damping functions to each. Researchers can also model interactions between light therapy and antidepressant medications, which sometimes show seasonal synergy. A notable study examined the combination of fluoxetine and light therapy in SAD; find it here.

Finally, as longitudinal datasets grow, population‑level cosine models could be used to test hypotheses about the effects of climate change on SAD prevalence. Warmer winters and altered cloud patterns might shift the seasonal curve in ways that can be tracked and modeled.

Conclusion

Cosine functions provide an elegant, mathematically sound foundation for understanding and treating Seasonal Affective Disorder. By distilling the cyclic symptom pattern into just three interpretive parameters—amplitude, phase, and baseline—the model allows clinicians to quantify seasonal severity, predict symptom peaks, and personalize light therapy timing and intensity. The model is simple enough for routine clinical use yet extensible enough to incorporate real‑world data, multiple harmonics, and confounders. No single mathematical tool can capture the full complexity of every patient’s experience, but the cosine framework remains a cornerstone of chronobiological research and a practical aid for improving outcomes in SAD. As data collection technologies and analytical methods advance, these models will become increasingly precise, helping to bring relief to those who suffer from this predictable but debilitating condition.

For further reading, consult the Mayo Clinic overview of SAD, the National Institute of Mental Health fact sheet, and the meta‑analysis of light therapy efficacy published in JAMA Psychiatry.