Cross-cutting teaching notes for critical appraisal and clinical reasoning: 33 concepts in 6 themes. Open "Clinical reflection" on each card for a question to discuss with your team.
Every module follows one way of thinking
The vocabulary every cancer decision is built on.
Stage describes disease extent; grade describes selected microscopic features; molecular subtype can guide treatment. They answer different questions.
Which of these findings is prognostic, and which predicts drug benefit?
Explain treatment intent explicitly. The same medicine may have different goals and endpoints depending on the setting.
What outcome is this regimen trying to improve?
A 30% relative reduction can represent very different absolute benefits depending on baseline risk. Use absolute numbers over a defined time horizon.
How many out of 100 patients benefit, and how many experience serious harm?
ECOG or Karnofsky performance status is one of the strongest predictors of benefit and toxicity, and most trials enrolled only ECOG 0–1 patients.
Is this patient’s poor performance status caused by the cancer, and could it recover with response?
RECIST 1.1 defines partial response as at least a 30% fall in the sum of target-lesion diameters, and progression as at least a 20% rise from the smallest sum (and at least 5 mm) or any new lesion.
Would the scan result change management if the patient is clinically better?
Most serum tumour markers are useful for following a known cancer, not for finding one. A raised value needs context; a normal value does not exclude cancer.
What will I do differently if this marker comes back high or normal?
How to judge whether a result should change practice.
Start with the population, intervention, comparator and primary endpoint. Then inspect absolute effect size, follow-up, adverse events and crossover.
Is the result clinically important, or only statistically significant?
A longer progression-free interval can matter, but does not prove the same improvement in overall survival. Subsequent therapies and crossover may affect OS.
What was the absolute PFS gain, and was quality of life measured?
Exploratory subgroup signals are less reliable when many comparisons are made. Look for prespecification, interaction tests and consistency.
Was the subgroup finding prespecified and adequately powered?
A hazard ratio of 0.75 means a 25% lower rate of the event at any moment during follow-up; it does not mean patients live 25% longer.
At 2 or 3 years, how many more patients out of 100 are alive or free of disease?
Non-inferiority trials test whether a simpler, shorter or cheaper treatment is “not unacceptably worse”. Everything depends on the margin chosen.
How much efficacy is the patient willing to trade for less toxicity or cost?
Pathological complete response, disease-free survival and response rate are surrogates. They are validated in some settings and not in others.
Has this surrogate been shown to predict survival benefit in this disease and setting?
What a biomarker report can and cannot tell you.
An alteration may be predictive in one tumour and not another. Variant class, assay quality, line of treatment and local approval all matter.
What evidence supports this drug for this precise variant and tumour type?
These biomarkers capture different biological features and have different predictive performance across tumour types. None guarantees response.
Which test and cutoff were validated for the proposed therapy?
A tumour-only BRCA result does not establish hereditary cancer. Confirmatory germline testing and counselling may be indicated in appropriate settings.
Does the finding have implications for family members?
Plasma circulating tumour DNA can find actionable mutations quickly and without a new biopsy. A positive result is usually reliable; a negative result does not rule a mutation out.
Would a negative plasma result change the decision to rebiopsy?
Some drugs are approved by biomarker regardless of where the cancer started: for example MSI-H/dMMR, NTRK fusions, high TMB, BRAF V600E, RET fusions and HER2 IHC 3+.
How many patients with this tumour type were in the trials behind the approval?
Situations where the first step decides the outcome.
Subtype, nodal status, menopausal status and selected genomic assays influence decisions. Separate prognosis from predicted chemotherapy benefit.
Which patients were represented in the supporting trials, and what is the absolute benefit?
Advanced nonsquamous NSCLC often requires comprehensive molecular profiling and PD-L1 assessment before first-line systemic therapy decisions when clinically feasible.
Could starting treatment before results return limit the best first-line option?
RAS, BRAF V600E and MSI/MMR status, tumour sidedness, performance status and goals of care affect systemic therapy choice.
Which biomarkers actually change the first-line regimen?
Some presentations need action within hours, often before a cancer diagnosis is confirmed. Recognising them is the job of whoever sees the patient first.
Which of these can my team recognise and start treating before the oncologist is reached?
Chemotherapy, anti-CD20 antibodies and long steroid courses can reactivate hepatitis B; reactivation can cause liver failure and is largely preventable.
Was hepatitis B status documented before the first cycle?
Age alone does not predict tolerance. A brief geriatric screen (for example G8) followed by geriatric assessment finds problems that change treatment.
Would the treatment plan change if we knew this patient falls, lives alone or has early memory loss?
Fitting the evidence to the person in front of you.
Frailty, organ function, treatment burden, travel, affordability and patient priorities affect whether trial results apply.
What is the most important benefit to this patient, and what is the burden?
More treatment is not always better. Consider absolute recurrence risk, competing mortality, toxicity and de-escalation evidence where applicable.
What happens if we omit or shorten this component?
Check study design, endpoints, denominators, conflicts of interest, publication status and whether comparisons were randomized.
Is the claimed advantage based on a direct head-to-head comparison?
Cost is a clinical issue. Treatment that families cannot sustain leads to missed cycles, debt and abandoned care, which can be worse than a simpler regimen completed.
Can this family complete the whole plan, and if not, what is the best plan they can complete?
Palliative care is not only end-of-life care. In a randomised trial in metastatic lung cancer (Temel 2010), early palliative care alongside oncology improved quality of life and mood, with less aggressive end-of-life treatment.
Has this patient with incurable cancer met the palliative care team yet?
How information is given affects understanding, trust and choices. Structured approaches such as SPIKES help: set up, find out what the patient knows, ask how much they want to know, give knowledge, respond to emotions, summarise.
Does the patient, not only the family, know the aim of treatment?
Short case discussions that test the principles.
When disease progresses, consider adherence, radiological confirmation, resistant clones, sanctuary sites and whether repeat testing is informative.
Would local therapy, a different systemic drug or repeat biopsy change management?
New diarrhoea on checkpoint treatment requires assessment for infection, immune-mediated colitis, severity and dehydration; do not assume a benign drug effect.
Which symptoms trigger same-day evaluation and specialist input?
FDG uptake is not cancer-specific. Correlate with morphology, prior imaging, inflammatory conditions and whether tissue confirmation would change treatment.
Will confirming this lesion change stage or management?
A woman on an EGFR inhibitor for lung cancer is diagnosed with pulmonary TB. Rifampicin is a strong CYP3A4 inducer and can sharply lower the levels of many oral targeted drugs.
Who checks drug interactions when another doctor starts a new medicine?
Medically reviewed by Dr. Haritha Chiramana on 8 October 2026. This page is general education and cannot replace an individual consultation.