Guideline Concordance Can Hide Poor Cancer Care

Cancer services often use guideline concordance as shorthand for quality. In Germany, for example, cancer centers certified by the German Cancer Society report annually against quality indicators derived from strong guideline recommendations and must justify results below target at audit (Stuebs et al. 2023).

It does not show whether the recommendation fitted the patient. Nor does it show whether a departure reflected thoughtful individualisation, poor access, or bias.

That blind spot matters more as the population treated for cancer grows older. People aged 65 and over make up about a fifth of the population of the EU and EFTA countries but account for an estimated 62% of new cancer cases; on current rates, ageing alone will raise new cases in this group by around 38% by 2040 (Joint Research Centre 2023).

Older patients were also over-represented in the high-risk group that, in Orcutt et al. (2025), gained the least from standard treatment. The next step in quality improvement is not to weaken guidelines. It is to make the reasoning behind their application visible and auditable.

The patient outside the trial

Guidelines compress large bodies of evidence into usable recommendations. That is their strength and their limit. Much of the evidence comes from trial populations selected by eligibility criteria, while routine practice includes patients with frailty, multiple conditions, cognitive impairment, limited support, and competing health risks.

Orcutt et al. (2025) used electronic health records from around 280 US cancer clinics to emulate 11 landmark phase 3 trials in advanced lung, breast, prostate, and colorectal cancer. Patients at low or medium prognostic risk had survival and treatment benefits broadly similar to those reported in the trials. High-risk patients—who tended to be older, with poorer performance status, and adverse laboratory values—had markedly lower survival and smaller benefits.

Across the emulated trials, the survival gain from treatment was, on average, 3 months smaller than in the published trials. The shortfall grew with risk: a median of 2.3 months for low-risk, 4.2 months for medium-risk, and 5.7 months for high-risk patients.

These are observational estimates from US data, and the authors acknowledge possible unmeasured confounding. Even so, the findings suggest that a patient's prognostic risk shapes how well trial results translate into practice. A service, therefore, needs a reliable way to identify when that translation is uncertain—and to record how clinicians handled it.

A useful metric—and an incomplete one

Consider two illustrative guideline-discordant decisions. In the first, an oncologist reduces treatment intensity after documenting frailty, renal impairment, toxicity risk, and the patient's priorities. In the second, treatment is withheld because an older patient is assumed to be unable to tolerate it, without formal assessment.

A concordance dashboard records both as the same event. One reflects careful judgement; the other may reflect age-based undertreatment.

The reverse problem can also occur. In our view, high concordance can conceal overtreatment when a standard regimen is applied to a patient who differs materially from the population that established its benefit—the high-risk group in which Orcutt et al. (2025) found the smallest gains.

Geriatric oncology illustrates the implementation gap. ASCO recommends geriatric assessment for patients aged 65 and over receiving systemic cancer therapy (Dale et al. 2023); an earlier version of the recommendation had been in place since 2018. Yet in a 2019–20 survey of 349 US oncologists, 74% of them were community-based; only 13% used a formal assessment for all older patients, 60% used none for any, and 19% were unaware of any validated instrument (Gajra et al. 2022).

The commonest reasons for non-use were that assessment was too cumbersome for routine practice (44%) and that it added nothing to a history and examination (36%). The barrier was partly workflow and partly perceived value—and neither shows up in a concordance rate.

De-escalation creates a similar challenge. PRIME II randomised 1,326 women aged 65 or over—most treated in the UK—to whole-breast radiotherapy or none (Kunkler et al. 2023). All had node-negative, hormone-receptor-positive tumours up to 3 cm, treated with breast-conserving surgery and adjuvant endocrine therapy.

At 10 years, local recurrence was 9.5% without radiotherapy and 0.9% with it, while overall survival was almost identical (80.8% vs 80.7%). A simple 'radiotherapy given' indicator cannot show whether omission was right for a particular woman. That depends on absolute benefit, life expectancy, adherence to endocrine therapy, treatment burden and her preferences.

Make clinical judgement auditable

Cancer services do not need a new score for every exception. They need a small number of workflow changes that reveal why a decision was made.

1. Record applicability, not a defensive essay. A short structured note, ideally completed at the multidisciplinary team (MDT) meeting, captures four items: whether the patient resembles the evidence population, the expected absolute benefit, major competing risks, and the patient's stated priorities. The aim is not more documentation but more informative documentation.
2. Move assessment upstream. If geriatric screening, medication review, or functional assessment happens before the treatment consultation, the results are available when the decision is made. Asking the oncologist to discover and complete every assessment in a time-limited visit makes non-use predictable; 'too cumbersome' was the reason oncologists most often gave for not using geriatric assessment (Gajra et al. 2022).
3. Treat discordance as a review trigger, not an automatic failure. Audit a sample of departures for documented reasoning. Examine patterns by age, sex, socioeconomic status, treatment site, and, where lawful and recorded, ethnicity. This helps separate appropriate individualisation from systematic undertreatment or unequal access.
4. Build feedback on reasoning into existing forums. Educational meetings probably improve practice slightly, but whether interactive formats outperform lectures remains uncertain (Forsetlund et al. 2021). Exploratory analyses in the same review associate feedback and goal-setting with greater change. For guideline application, that argues for periodic case review in MDT or audit meetings, where clinicians commit to a decision and receive feedback on both the choice and the reasoning.

An audit matrix for guideline concordance

Combining concordance with whether the reasoning is documented gives four categories, each needing a different governance response.

Table 1. Illustrative framework only

What it takes – and where it can fail

The approach is simple to describe but not free to run. Leaders should plan for four resource demands:

• Clinician time: the structured applicability note adds time to each MDT case.
• Record build: structured fields for applicability, assessment completion and patient priorities, so audits can draw on data rather than free-text review.
• Audit capacity: a clinically informed reviewer to examine a regular sample of discordant and undocumented cases.
• Upstream assessment: nursing or allied health time for pre-consultation screening and a pathway to act on what it finds.

Three failure points are foreseeable:

• Box-ticking: structured notes can become templated text that records nothing. Audit the content, not only the completion.
• Blame: if discordance review feels punitive, clinicians may retreat to defensive concordance – the opposite of the goal. Frame review as learning, and report patterns rather than individuals where possible.
• Small numbers: subgroup patterns by age or deprivation in a single service may be too small to interpret. Pool data over time or across a network.

Most of the evidence cited here comes from the US; PRIME II is the exception. In many European systems, treatment decisions already pass through multidisciplinary tumour boards, which makes them a natural place to record applicability and review departures. Equity auditing needs local adaptation: in several European countries ethnicity is not routinely recorded in health records or its collection is restricted, so deprivation indices, language needs or other lawful proxies may be required. Any new data collection must comply with GDPR.

The guardrail: audit in both directions

'Individualised care' can become a respectable label for inconsistent care. Any programme that encourages thoughtful departures must also check who is being offered less treatment, whether validated assessments were used, and whether similar patients receive different recommendations across clinicians or sites.

The governance question is not 'Did clinicians follow the guideline?' or 'Did clinicians individualise?' It is: 'Was the decision supported by the evidence, the patient's condition and the patient's goals – and can we see that in the record?'

A better definition of high-quality cancer care

Guidelines remain the foundation of oncology care, and concordance should remain part of quality measurement. But concordance alone cannot distinguish good adherence from mechanical adherence, or thoughtful deviation from biased undertreatment.

The more useful standard combines both: follow evidence-based recommendations when they fit and document the reasoning when they do not. Services that make this reasoning explicit can learn from variation rather than merely count it – and can judge whether care was appropriate for the person, not only compliant with the pathway.