TO THE EDITOR: The debate raised by Sitas and colleagues about whether smoking status should be recorded on death certificates represents data acquisition at the last point. What needs to be considered is not the merits of the proposal, but the systematic failure to record this crucial data in patients’ clinical records. The argument I present here concerns cancer, but may be extrapolated to other smokingrelated diseases. The absence of smoking history from the clinical records of patients with cancer up until now is understandable. Unlike the situation with infectious diseases, causative agents have had no relevance to the prognosis or the treatment of malignancy. Even if this pregenomic outlook persisted, the evidence we now have, showing that smoking affects response to therapy (eg, in prostate cancer), would warrant documentation of smoking history for every cancer patient. Among developed countries, the risk attributable to smoking in lung cancer is 70%–90%. In single-gene terms, polycycl ic a romatic hydrocarbons and tobacco-specific nitrosamines such as 4(methylni trosamino)-1-(3-pyridyl) -1butanone mediate malignant transformation by point mutations in onocogenes and tumour suppressor genes, typified by G-to-T transversions in codons 157, 158, 245, 248 and 273 of TP53; mutations in the KRAS gene and in the gene for epidermal growth-factor receptor (EGFR) are likewise relevant. Enough distinction can thus be made between smoking-related and non-smoking-related lung cancer to identify two entities. Patients with nonsmall-cell lung cancer, with exon 19 and 21 mutations in the EGFR gene are typically non-smokers, and their tumours have an 80% response rate to erlotinib. Indeed, these tumours respond better to chemotherapy than in patients lacking such mutations. As the single-gene era closes, smoking history will be required for all relevant biospecimens to avoid making genomic-wide data on pathways consequent on tobacco-related etiology — representing hundreds of tumours — inaccessible. Tobacco smoking causes cancer of the oral cavity, oropharynx, nasopharynx, and hypopharynx, oesophagus (adenocarcinoma and squamous-cell carcinoma), stomach, colorectum, liver, pancreas, nasal cavity and paranasal sinuses, larynx, lung, uterine cervix, ovary (mucinous), urinary bladder, kidney (body and pelvis), ureter and bone marrow (myeloid leukaemia), and possibly causes female breast cancer. The prospect of genomic-wide data for these tumours not being accompanied by data on smoking is daunting. The case for smoking data in relation to clinical trials has been made previously. Beyond this, the immediate goals for Australia are that smoking history be recorded for tobacco-related malignancies in hospital-based tumour-specific clinical cancer registries and for biospecimens. The need is for the adoption of uniform national vocabulary and coding methods. This will require leadership by an authority such as the National Health and Medical Research Council (NHMRC), the Cancer Council Australia, the Clinical Oncological Society of Australia, or Cancer Australia.