Colorectal cancer: a slow-growing cancer we can catch before it starts

Start here: what colorectal cancer is

The colon and rectum are the last stretch of the digestive tract. Their inner lining renews itself constantly, and that constant division is where things go wrong. Most colorectal cancer begins as a polyp — a small benign outgrowth of the lining. One kind, the adenoma, carries mutations that let it keep growing; over years it can accumulate further mutations and become invasive cancer. That stepwise progression is the adenoma–carcinoma sequence, and it is the single most important fact about this disease ^{[1] [2]}.

It matters because it is slow, and because the intermediate step is physically removable. That combination is rare in oncology: colorectal cancer is one of the few common cancers that can be prevented rather than merely detected early.

Colorectal cancer is common. In the 2022 global estimates it was the third most frequently diagnosed cancer worldwide (9.6% of all cancers) and the second leading cause of cancer death (9.3% of cancer deaths) ^[3]. In 2020 there were roughly 1.9 million new cases and 0.9 million deaths, with incidence rising in middle- and low-income countries as diets and activity patterns change ^{[4] [5]}.

It is shifting younger. Incidence in adults under 50 is rising in a large fraction of countries with reliable registries — in one analysis of 36 countries, incidence in under-50s increased in 19, and in nine of those it rose while rates in older adults were flat or falling ^[6]. European registry data show the same pattern over 25 years ^[7]. Why is not settled, but obesity and weight gain since early adulthood are associated with early-onset disease in prospective cohort data ^{[8] [9]}.

It is heavily modifiable. Obesity, sedentary behaviour, red and processed meat, alcohol, and tobacco are the recognized drivers; fibre and overall dietary pattern run the other way ^{[5] [10]}. A minority of cases are hereditary — Lynch syndrome, caused by pathogenic variants in the mismatch-repair genes MLH1, MSH2, MSH6, and PMS2, produces gene-specific and sex-specific cancer risks large enough to change screening schedules for carriers ^{[11] [12]}.

And stage at diagnosis dominates outcome. Among people diagnosed with metastatic colorectal cancer, roughly 70–75% survive beyond 1 year, 30–35% beyond 3 years, and fewer than 20% beyond 5 years; 20% of new diagnoses are already metastatic, and another 25% of those presenting with localized disease will later develop metastases ^[13]. Against that, colonoscopic removal of adenomas produced 12 observed colorectal-cancer deaths where 25.4 were expected from population rates, over follow-up as long as 23 years ^[14]. Same disease, opposite ends of the same timeline.

This review follows three pillars — measurements, medicines, and progress — with a simple simulatable growth model in between that explains why the timeline is the whole game.

Pillar 1: measurements and diagnosis

Screening: the part that happens before there is a patient

Screening for colorectal cancer is unusual because it has two separate jobs. It finds cancers early, and it finds and removes the precursors, so incidence itself falls. Both effects are measurable.

Colonoscopy passes a camera through the whole colon; anything suspicious can be biopsied or removed in the same session. Its preventive effect is the National Polyp Study result above ^[14]. Flexible sigmoidoscopy examines only the lower colon; in a randomized trial of 154,900 people it reduced colorectal-cancer incidence by 21% over a median 11.9 years, with the effect concentrated in distal tumours — precisely the anatomy the scope reaches ^[15]. That asymmetry is itself informative: reach determines benefit. Systematic reviews and large community cohorts have quantified the same gradient between the two tests and between right- and left-sided disease ^{[16] [17] [18]}.

Stool tests trade sensitivity for reach and acceptability. The faecal immunochemical test (FIT) detects human haemoglobin in stool, is non-invasive, requires no bowel preparation, and can be repeated annually; a positive result triggers a colonoscopy. Its weakness is that a lesion has to bleed to be seen, which biases detection toward larger and more advanced lesions ^[19]. Head-to-head programme comparisons of once-only colonoscopy against repeated FIT are exactly the trial being run to settle the trade-off ^{[20] [21] [22]}. Guidelines therefore present a menu rather than a single test, on the reasoning that the best screening test is the one a person will actually do ^{[23] [24]}.

When to start moved. The American Cancer Society's 2018 update recommended beginning average-risk screening at 45 rather than 50, driven by microsimulation modelling that incorporated the rising incidence in younger adults ^[23] — a guideline change that follows directly from the epidemiology in the opening section.

Quality of the examination matters as much as the choice of test. Colonoscopy has a real miss rate, quantified by the adenoma detection rate, and a randomized trial of a real-time deep-learning polyp-detection system raised the adenoma detection rate from 20.3% to 29.1% — evidence that a meaningful share of the residual risk was operator attention rather than biology ^[25]. Post-polypectomy surveillance intervals are set separately, by the number and pathology of what was removed ^{[26] [27]}.

Staging: how far it has gone

Once cancer is found, treatment is chosen almost entirely by stage, summarized by the TNM system: T how deep the tumour has invaded through the bowel wall, N whether regional lymph nodes contain tumour, M whether it has spread to distant organs. These combine into stages I through IV. In plain terms: stage I is confined to the wall and usually cured by surgery alone; stage II has penetrated the wall but spared the nodes; stage III involves lymph nodes and is the classic indication for chemotherapy after surgery; stage IV is metastatic, most often to liver or lung, and is treated systemically ^{[28] [29] [13]}.

Molecular markers: which biology, not just how much

Two tumours at the same stage can need different drugs. Genome-scale characterization of 276 colorectal tumours found that 16% were hypermutated — three-quarters of those with high microsatellite instability, usually from hypermethylation and MLH1 silencing, and one-quarter from somatic mismatch-repair or POLE mutations — while the remainder shared recurrent mutations in APC, TP53, SMAD4, PIK3CA, and KRAS, plus targetable ERBB2 amplification ^[30]. Expression-based work then organized the disease into four consensus molecular subtypes, giving the field a shared vocabulary for its heterogeneity ^{[31] [32] [33]}.

Four markers change management directly:

• MSI-H / dMMR status. Microsatellite instability means the mismatch-repair machinery has failed, so the tumour accumulates enormous numbers of mutations — and therefore neoantigens the immune system can see. It occurs across many cancer types but is enriched in colorectal cancer ^[34], and testing for it is now a formal recommendation because it selects patients for immunotherapy ^[35].
• KRAS / NRAS mutation. RAS sits immediately downstream of EGFR. If it is mutated and constitutively active, blocking the receptor above it accomplishes nothing — a prediction confirmed in consortium analyses and in prospective-retrospective trial reanalysis ^{[36] [37]}. Anti-EGFR therapy is therefore restricted to RAS wild-type disease.
• BRAF V600E. Further down the same pathway, this marks a poor-prognosis subgroup with median overall survival of 4 to 6 months after failure of initial therapy ^[38].
• HER2 (ERBB2) amplification. A smaller subgroup, identified in the genomic characterization and now with its own targeted options ^{[30] [39]}.

Tumour sidedness turns out to carry some of the same information: in a pooled analysis of six randomized trials, primary tumour side was both prognostic and predictive of benefit from EGFR-directed antibodies in RAS wild-type metastatic disease ^[40].

Centerpiece: a simple simulatable growth model

Every argument for screening rests on a claim about time — that there is a usefully long interval during which a lesion is detectable but not yet dangerous. That claim can be written down.

Treat tumour volume as growing exponentially, the simplest of the classical growth laws:

V(t) = V₀ · 2^(t / T_d)

where T_d is the volume doubling time. Because a roughly spherical tumour's volume scales as the cube of its diameter, the diameter grows with a doubling time three times longer:

d(t) = d₀ · 2^(t / (3·T_d))

Inverting gives the time for a lesion to grow from one diameter to another:

t(d₁ → d₂) = 3 · T_d · log₂(d₂ / d₁)

The detection window — sometimes called the sojourn time in screening theory — is the interval between the size at which a test can find the lesion and the size at which it declares itself with symptoms. Under this model that window is proportional to the doubling time, and independent of where the tumour started:

W = 3 · T_d · log₂(d_symptomatic / d_detectable)

Grounding. The model form is standard and validated rather than invented here: the classical growth laws — exponential, exponential-linear, power law, Gompertz, logistic, von Bertalanffy — have been fitted head-to-head against experimental tumour-growth data and assessed for descriptive and predictive power ^[41], and catalogued systematically ^[42]. The precedent for reading growth rates out of screening programme data, rather than serial imaging of individuals, comes from mammography-screening work that jointly estimated tumour growth and size-dependent test sensitivity across 395,188 women — finding a mean 1.7 years for a tumour to grow from 10 mm to 20 mm and screen sensitivity rising from 26% at 5 mm to 91% at 10 mm ^[43]. That study is breast, not colorectal; it is cited here for the method and for the shape of the size-versus-detectability relationship, not for colorectal parameter values.

Parameters are illustrative and flagged. The doubling times and the two size thresholds below are teaching values chosen to span a plausible range — the substrate assembled for this review does not contain a measured colorectal volume-doubling-time distribution, so no number here should be read as a colorectal-specific estimate. The caution is not merely formal: comparing seven ordinary-differential-equation growth models on the same data produced a 12-fold difference in predicted doubling times depending on which model was chosen ^[44]. Exponential growth is the most optimistic-looking and least realistic of the family — real tumours decelerate as they enlarge, which is why Gompertz-type models often fit better ^{[41] [45]}.

What the model explains. Three things, all of which are otherwise counterintuitive. First, why screening intervals of several years can work at all: if the detection window is measured in years, a test repeated every two to ten years still lands inside it for most lesions. Second, why screening preferentially catches slow-growing tumours and preferentially misses fast ones — the window is proportional to T_d, so aggressive tumours are the ones most likely to appear as symptomatic interval cancers between screens. Third, why the adenoma–carcinoma sequence is such a good target: intervening at the polyp stage removes the lesion before the clock in this model even starts ^{[14] [2]}.

Pillar 2: medicines

Treatment in colorectal cancer is stage-directed first and marker-directed second.

Surgery is the curative modality in early disease. For localized colon cancer, removing the tumour with its draining lymph nodes is the treatment, and everything else is an adjunct. The main change over two decades has been how the operation is delivered rather than what it accomplishes: randomized trials established that laparoscopic-assisted resection achieves oncologic outcomes comparable to open surgery in colon cancer, with the rectal-cancer question tested separately and more cautiously because the pelvis makes complete mesorectal excision harder to guarantee ^{[46] [47] [48] [49] [50] [51]}. For rectal cancer specifically, the sequencing of radiotherapy, chemotherapy, and surgery has shifted toward total neoadjuvant therapy — delivering systemic treatment before the operation ^{[52] [53]}.

Adjuvant chemotherapy after surgery aims at micrometastatic disease left behind — cells too few to see but enough to recur. The backbone is a fluoropyrimidine (fluorouracil or oral capecitabine), an antimetabolite that blocks thymidylate synthase and starves DNA synthesis, combined with oxaliplatin, a platinum agent that forms DNA crosslinks. The two standard regimens are FOLFOX (infusional fluorouracil, leucovorin, oxaliplatin) and CAPOX (capecitabine plus oxaliplatin); adding oxaliplatin to a fluoropyrimidine improved disease-free survival in stage III colon cancer ^[54]. The live question is how much, because oxaliplatin causes cumulative and often permanent peripheral neuropathy. The IDEA pooled analysis of six concurrent randomized trials in 12,834 patients tested 3 months against 6; non-inferiority was not confirmed overall, but held for CAPOX — which is why duration is now chosen by regimen and recurrence risk rather than applied uniformly ^[55]. Stage II is the genuinely contested zone, since the absolute benefit is small enough that many patients are treated to help few ^[56].

Metastatic disease is treated systemically, with a fluoropyrimidine backbone plus irinotecan or oxaliplatin, intensified to the three-drug FOLFOXIRI regimen when the goal is maximum response ^{[57] [13]}. Two antibody classes are added on top:

• Anti-VEGF (bevacizumab) targets vascular endothelial growth factor rather than the tumour cell, starving the tumour's blood supply; it is combined with chemotherapy irrespective of RAS status ^[57].
• Anti-EGFR (cetuximab, panitumumab) blocks the epidermal growth factor receptor at the top of the MAPK cascade ^[58]. It improved outcomes when added to first-line chemotherapy ^[59], but only in RAS wild-type tumours — in the panitumumab–FOLFOX4 reanalysis, patients without RAS mutations had median progression-free survival of 10.1 versus 7.9 months and overall survival of 26.0 versus 20.2 months, while RAS-mutant patients derived no benefit ^{[37] [60] [36]}. Resistance, when it comes, is molecularly legible: it emerges through RAS and other MAPK-pathway alterations that can be tracked in blood ^[61].

The immunotherapy shift is the largest change of the last decade, and it applies to a minority. MSI-H/dMMR tumours carry so many mutations that they present abundant neoantigens; checkpoint inhibitors release the brake that lets those tumours evade T cells ^[62]. In previously treated dMMR/MSI-H metastatic disease, nivolumab plus ipilimumab produced a 55% objective response rate with 94% of responses ongoing at data cutoff ^[63]. KEYNOTE-177 then moved it to first line: in 307 patients with untreated MSI-H/dMMR metastatic colorectal cancer, pembrolizumab beat chemotherapy on progression-free survival, 16.5 versus 8.2 months (HR 0.60) ^{[64] [65]}. The corollary is just as important — the same drugs do little in the microsatellite-stable majority, which is why MSI testing is not optional ^{[35] [66]}.

Targeted therapy for the hard subgroups arrived by treating pathway biology seriously. Single-agent BRAF inhibition failed in colorectal cancer because blocking BRAF triggers feedback reactivation through EGFR — a mechanism worked out at the bench ^{[67] [68]} and confirmed clinically ^[69]. BEACON tested the combination that mechanism implies, in previously treated BRAF V600E-mutant metastatic disease, and reported the two arms separately. The triplet — encorafenib, binimetinib and cetuximab — raised median overall survival from 5.4 months on control to 9.0 months (HR 0.52). The doublet of encorafenib plus cetuximab, without the MEK inhibitor, reached 8.4 months (HR 0.60 versus control) — most of the benefit from two drugs rather than three ^{[38] [70] [71]}. The same logic recurs with KRAS G12C: adagrasib alone is limited by EGFR-driven adaptive resistance, and adding cetuximab improves response ^{[72] [73]}. HER2-directed therapy extends the breast- and gastric-cancer playbook to HER2-amplified colorectal tumours, including antibody-drug conjugates ^{[39] [74]}.

Pillar 3: progress

Early-onset disease is the open epidemiological problem. Incidence under 50 is rising across many high-income countries while falling in older adults, which rules out any explanation that acts uniformly across ages ^{[6] [7]}. Obesity and adult weight gain are implicated in prospective cohorts ^[8], and the gut microbiome is an active candidate — faecal metagenomic signatures discriminate colorectal cancer across cohorts, shift along the adenoma–carcinoma sequence, and can improve FIT sensitivity when combined with it ^{[75] [1] [76] [77] [78]}. Causality is not established ^[9].

ctDNA and minimal residual disease. Circulating tumour DNA — fragments shed by tumour cells into blood — can be measured after surgery, when imaging shows nothing. Ultradeep sequencing of plasma in stages I–III colorectal cancer showed post-operative ctDNA strongly associated with recurrence ^[79], and the DYNAMIC trial then used it to make a decision: in 455 patients with stage II colon cancer, ctDNA-guided management reduced adjuvant chemotherapy use without compromising 2-year recurrence-free survival ^[56]. This is the most concrete answer yet to the stage II over-treatment problem, and a task-force whitepaper sets out the remaining requirements before broader adoption ^[80]. Detection sensitivity is still the limiting technical factor ^{[81] [82]}.

Neoadjuvant immunotherapy may be the biggest near-term shift. The NICHE study showed pathological responses to neoadjuvant checkpoint blockade in early-stage colon cancer, striking in dMMR tumours ^{[83] [84]}. Then a prospective phase 2 study gave single-agent dostarlimab to patients with dMMR locally advanced rectal cancer, with chemoradiotherapy and surgery held in reserve for anyone who did not achieve a clinical complete response ^[85]. For a disease whose standard treatment can mean permanent colostomy, an organ-preserving path is a different category of benefit from a survival hazard ratio.

Screening keeps improving on two axes. Computationally, real-time AI polyp detection has already shown a randomized increase in adenoma detection ^{[25] [86]}. Biologically, stool- and blood-based molecular tests aim to raise sensitivity for the pre-cancerous lesions FIT tends to miss ^{[87] [88] [89]}. Both target the same quantity in the model above: widening the effective detection window by lowering the size at which a lesion becomes findable.