The Unmeasured Variable: Why Cancer Burden Depends on More Than the Tumor A systems framework for…

The Unmeasured Variable: Why Cancer Burden Depends on More Than the Tumor A systems framework for cancer symptom burden Two patients can receive the same cancer diagnosis, the same chemotherapy, and the same treatment plan in the same clinic — yet go through treatment in very different ways.

One finishes treatment tired but functioning relatively well. The other struggles with persistent pain, cognitive slowing that is difficult to explain clinically, and fatigue severe enough to limit even simple daily activities.

We call this variability. We usually explain it through frailty, age, or comorbidities — and then move on. But over time, I started to feel that this explanation was incomplete. Not because it is wrong, exactly, but because it describes what we see at the end without fully explaining how patients arrived there.

I have been working on a framework for this. It started with a focus on environment — the air a patient breathes, the chemicals in her food and water, the neighborhood she comes home to after treatment. Those things matter, and the exposome framework introduced by Wild in 2005 provides a way to measure them systematically.

Then, a few weeks ago, my mentor read an early draft and pointed out something important that I had not fully considered. I was discussing the environment almost as if it acted on a neutral patient. But patients are not neutral starting points. They arrive with their own biological, psychological, and social histories — sex, mental health, socioeconomic conditions, comorbidities, social support, and many other host-related factors that shape both symptom burden and the way environmental exposures are experienced. He was right. This piece is my attempt to fix that gap.

It is written for clinicians and researchers who feel the distance between trial toxicity tables and what their patients describe in clinic.

Why Do Patients Suffer So Differently? Fatigue, pain, cognitive difficulties, sleep problems, and psychological distress tend to travel together in cancer patients. They share biological roots — mainly inflammatory and stress-hormone pathways, as symptom cluster research has consistently shown — and yet how severe they are varies enormously between people on identical treatment.

The standard explanation focuses on tumor biology and treatment toxicity. That framework has been genuinely useful. But it does not explain why a 52-year-old woman living near a busy road, working long hours, with limited social support and years of poorly managed anxiety, goes through breast cancer treatment so differently from a 54-year-old woman on the same drugs whose life circumstances are completely different.

The tumor and the treatment are the same. Something else is carrying a lot of the weight.

A Map for Thinking: S = f(H, T, E)

Recent work in cancer systems biology supports a shift away from strictly mutation-centric models toward frameworks that include adaptive cell states, environmental stress responses, and epigenetic memory. França and Yanai proposed a mechanism by which cancer cells adapt to stressful environments through adaptive genome regulation, while Oliveira et al. showed that drug resistance can be encoded through heritable epigenetic configurations that allow multiple phenotypic outputs from a single genetic background. Although these studies focus on tumor-cell adaptation rather than patient symptom burden, they support the broader principle that biological systems exposed to stress can acquire persistent, context-dependent states. This Review extends that logic to the host, asking whether immune, endocrine, mitochondrial, and microbial systems may also enter different baseline states before treatment begins, thereby shaping symptom vulnerability and treatment tolerance.

I want to offer a conceptual map — not a formal model, just a way of organising the problem (tumour biology is real and important; it is set aside here only to keep the argument focused on what usually goes unmeasured):

S = f(H, T, E)

— S is symptom burden — everything the patient feels and lives with — T is treatment — the biological effects of chemotherapy, radiotherapy, immunotherapy — E is environmental exposure — the cumulative inputs from the world the patient inhabits: air quality, chemicals, light, noise, neighborhood stress — H is the host — the full human being who receives both the treatment and the environment

H Is a Whole Person, Not Just a Biology In earlier drafts I called this variable B, for baseline biology — things like genetic variation, the microbiome, epigenetic marks from prior exposures. That part is real. But it is not the whole picture.

The person who sits across from me in the clinic is not just a set of lab values. Her symptom experience is also shaped by things we rarely measure and almost never enter into trial models:

Sex and gender: Sex affects how the immune system works, how pain is processed, how drugs are metabolised, and how the stress hormone system responds. Women generally mount stronger inflammatory responses to infection — which also means more inflammatory symptoms during treatment. Gender is different from sex: it shapes whether symptoms get reported, whether they get taken seriously, and what social roles the patient is trying to maintain while feeling terrible. A woman who is also the primary caregiver for her children and her elderly mother experiences fatigue differently from someone without those responsibilities. That is not a soft variable. It changes what the symptom costs her.

Mental health history: Depression and anxiety are not just consequences of having cancer — they are also biological states that change how the brain receives treatment. A patient who arrives with untreated depression already has raised inflammatory markers, a dysregulated stress-hormone axis, and altered pain sensitivity before chemotherapy begins. Her nervous system is not a neutral starting point.

Money and social position: Chronic financial insecurity activates the stress response in the body in the same way that other prolonged stressors do — raised cortisol, low-grade inflammation, disrupted sleep, faster biological ageing. Lower income is also linked to higher pollution exposure, less access to good food, noisier housing, and less time to rest. Socioeconomic position and environmental exposure are connected, but they are not the same thing.

Social connection — or the lack of it: We often treat loneliness as a psychological footnote, yet its measurable effects on inflammatory markers, immune function, sleep, and pain are, in some studies, as large as those from recognised medical risk factors. Patients with good social support consistently do better during cancer treatment. The reason is not mysterious: feeling safe and connected reduces the threat-detection systems in the brain that otherwise amplify how bad symptoms feel.

Other illnesses: Every significant comorbidity a patient brings to a cancer diagnosis is a prior disturbance of the same biological systems that treatment is about to hit again. Type 2 diabetes brings chronic inflammation and mitochondrial damage. Cardiovascular disease brings endothelial dysfunction and raised cytokines. Obesity carries a substantial inflammatory load on its own. A patient with two or three significant comorbidities is starting chemotherapy with biological reserves that were already low. What lands on that system — treatment, environment, stress — lands harder.

None of this is new in isolation. What a systems framework gives us is a way to see these things as interacting, not just adding up. The mental health history changes how the immune system responds to pollution. The financial situation shapes both environmental exposure and psychological state. The comorbidities reduce the biological capacity to absorb everything else. Sex biology runs through all of it.

The Environment Still Matters — But It Arrives Through the Host

My original argument was that environmental exposure is an under-measured variable in cancer symptom science. I still believe that. But my mentor’s point has changed how I think about the mechanism.

Environmental exposures do not act on an abstract biology. They act on a specific person, and that person’s H determines how much of the exposure reaches the relevant biological systems, how those systems respond, and how the response is felt.

A few examples:

Air pollution and brain inflammation: A patient returning home to a house along a major road breathes fine particles that can reach the brain and push microglial cells — the brain’s immune cells — into a state of heightened reactivity. In cancer patients whose blood-brain barrier is already weakened by chemotherapy, this may intensify cognitive symptoms and fatigue. But how much it does so depends on the host. Women may show stronger neuroinflammatory responses to the same particle load than men, partly because of how oestrogen interacts with immune cells. Patients with pre-existing depression already have more reactive microglia. Patients in lower-income areas carry higher pollution burdens and have less reserve to handle them. Same air, different outcome — because different H.

Endocrine-disrupting chemicals and treatment response: Chemicals like PFAS and BPA interfere with stress hormone regulation and thyroid function. But their effects are heavily shaped by the host’s existing hormonal environment — which is why the same chemical exposure has different effects before and after the menopause, or in a patient on aromatase inhibitors versus one on tamoxifen. The chemical lands in a landscape that the patient’s sex biology, treatment history, and comorbidities have already shaped.

Soil zinc deficiency and mucositis: In parts of the Middle East, South Asia, and sub-Saharan Africa, zinc-poor soils mean that staple crops contain low zinc, leaving many people in a state of marginal deficiency. Zinc matters for the lining of the gut and mouth; deficiency makes chemotherapy-induced mucositis worse. But how severe that is for a particular patient depends on her broader nutritional state, her income and dietary flexibility, and whether she has the social support that makes it possible to eat adequately during treatment. The geochemical fact and the host life situation are inseparable in the clinical outcome.

In each case: environment creates a pressure. The host — her biology, her mental state, her social world, her economic situation — determines what happens to that pressure inside the body.

What This Changes in the Equation The updated framework means that H is not a single biological variable. Some of it is biological — genetics, microbiome, prior disease, organ function, the accumulated damage of other illnesses. Some of it is how far the body’s reserves have already been drawn down before treatment begins. And some of it is the kind of information we almost never enter into a clinical record: mental health history, who is at home, whether the bills are paid, what the air outside is like. Patients do not experience these as separate layers. They arrive as one person, and what happens to that person during treatment is shaped by all of it at once.

H interacts with both T and E. A difficult H amplifies the biological impact of treatment. It also amplifies the impact of environmental exposure — raising vulnerability to pollution, circadian disruption, nutritional gaps. The patient who “should” be tolerating treatment but is not is often the patient whose H was never assessed, whose E was never measured, and whose T was calibrated in trials that enrolled neither.

The variance we keep calling frailty is, in fact, unmeasured H and E. This is not an argument against frailty scores — they capture something real about biological reserve. It is an argument that reserve is not random, and that measuring it without asking where it came from limits what we can do with it.

Why Trials Keep Missing This

Clinical trials control for T very carefully. They are beginning to account for some tumor biology within H. But they measure almost nothing of the H I have described above, and almost nothing of E.

Real-world patients differ from trial participants not just in age and comorbidity score, but in the full texture of their host profile. The woman going through treatment while caring for two children and an ill parent, working a physically demanding job, living on a busy road, carrying ten years of undertreated anxiety — she is not simply “frailer” than the trial participant. She is a different H, in a different E, meeting the same T.

Frailty scores catch some of this downstream, after the damage is already visible. They do not tell us where it came from. And without understanding where it came from, we cannot change it.

Three Things That Would Actually Change If the framework holds, here is what it demands in practice:

Expand what we assess: Supportive care built only around tumor and treatment variables is working with half the picture. Psychological screening, social needs assessment, basic environmental data, and patient-reported outcomes that go beyond toxicity grading would begin to locate patients in H × E space — which is where their actual experience lives.

Recognise that “soft” interventions are biological: If psychological state modulates how the immune system responds to environmental stress, then psychosocial support is partly a biological intervention, not just a comfort measure. If social connection reduces the threat response that amplifies symptom perception, then care navigation and social prescribing are mechanistic, not decorative. The framing matters because it changes what gets funded and what gets taken seriously.

Measure H and E in trials: Studies of symptom burden, treatment toxicity, and long-term survivorship should be recording — at minimum — mental health history, social support, household income, and basic environmental exposure proxies. Not as background variables to adjust away, but as biologically active components of the system being studied.

If you are designing a trial or building a dataset, S = f(H, T, E) is one place to start putting the patient back into the analysis.

The Patient Was Always the Centre Systems biology has given oncology powerful tools for understanding tumor evolution, drug resistance, and clonal heterogeneity. We have used them mostly to understand the cancer.

The patient’s experience of having cancer — which is, after all, the reason any of this work matters — has received far less of that intellectual effort.

The argument here is simple: symptom burden is what happens when a specific treatment meets a specific host inside a specific environment. Leave out the host — really leave it out, not just mention it in a limitations paragraph — and the model will always underperform in the real world. It will keep producing “unexplained variability” and attributing it to frailty, because it was never built to see what frailty is made of.

Patients do not experience cancer as a tumor alone. They experience it as a treatment landing inside a life.

Some wording and structural edits were refined with the help of an AI-based language model; the author reviewed and approved all content.