Document resource
WHAT IS ALREADY KNOWN ON THIS TOPIC National Institute for Health and Care Excellence guidelines recommend systematic anticancer therapy (SACT) for eligible patients with advanced non-small cell lung cancer (NSCLC) to improve survival and quality of life.Older age, poorer performance status, comorbidity and social deprivation have been associated with lower SACT uptake, but recent national-level data accounting for frailty are limited.Variation in SACT use across regions in England has been reported, but the magnitude and impact on survival in the modern treatment era are not well described.WHAT THIS STUDY ADDS In England (2019–2022), only 60% of patients with advanced NSCLC (stage 3B–4B) with good performance status received SACT.SACT use was strongly associated with younger age, lower frailty and lower social deprivation, independent of comorbidity and stage at diagnosis.There was a ∼20% absolute difference in adjusted SACT rates between cancer alliances, and higher treatment rates correlated with longer median survival.If all alliances matched the highest observed treatment rate, an estimated 448 additional patients could receive SACT annually, yielding 289 life-years gained.HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY Highlights persistent inequalities in SACT use by age and socio-economic deprivation, even after adjusting for performance status and frailty.Highlights unwarranted variation in SACT use across cancer alliances, even after adjusting for patient factors.Supports targeted interventions to reduce unwarranted variation in treatment, including optimisation of frailty assessment, earlier referral and regional best-practice sharing. Informs service improvements that could increase treatment uptake and extend survival.Introduction Clinical guidelines for people with advanced stage non-small cell lung cancer (NSCLC) recommend that systematic anticancer therapy (SACT) is offered to people with good performance status (PS). 1–3 Before 2014, chemotherapy was the mainstay of SACT, but since then there have been rapid developments in the range of systemic treatments available, and SACT now comprises three main therapeutic groups: cytotoxic chemotherapy, biological therapies and immunotherapies.4 5 The proportion of patients with advanced NSCLC who received SACT is reported to vary across high-income countries, with figures ranging from around 30% to almost 60% across 11 European countries before 2020.6 Figures published by the National Lung Cancer Audit (NLCA) show that, among the patients diagnosed with stage 3B–4 NSCLC in England and Wales during 2022, around 60% of people with good PS (0–1) received SACT.7 A study that explored patterns of chemotherapy use among people with advanced NSCLC in England suggested that patients aged over 75 years, compared with those under 75 years, tended to receive fewer cycles of chemotherapy and more treatment modifications and dose reductions, often due to comorbidities.8 However, these figures were based on treatments given in 2014–2017, and with the increasing availability of other SACT drug groups since then, it is not known whether these patterns have changed.The aim of this study is to investigate patterns of SACT use among patients with advanced NSCLC within England and describe how use varied by patient characteristics. Further, the study aimed to explore whether there is undertreatment among patients who appeared sufficiently fit to benefit from SACT.Methods Data source The study analysed a cohort of patients identified from the Rapid Cancer Registration Dataset (RCRD) provided by the National Cancer Registration and Analysis Service (NCRAS). 9 The RCRD dataset is released more quickly than the full cancer registration dataset, and the speed of production means that the range of data items is more limited and it does not have complete coverage of all patients diagnosed with lung cancer in England.10 The RCRD dataset was supplied by NCRAS with patient records linked to other national healthcare datasets, including Hospital Episode Statistics Admitted Patient Care (HES-APC) dataset, the National Radiotherapy Dataset and the SACT dataset.People with stage 3B–4B NSCLC diagnosed in England between January 2019 and December 2022 were identified from the RCRD, based on the assigned International Classification of Diseases (V.10) diagnosis code ‘C33’ or ‘C34’.11 The RCRD provided the data on patient demographics (age, sex, social deprivation) as well as the clinical variables disease stage at diagnosis, based on the eighth edition of tumour, node, metastases staging12 and Eastern Cooperative Oncology Group (ECOG) PS.13 Variables for comorbidity burden and frailty were derived from information in the national hospital dataset (HES-APC).14 Comorbidity was measured using the Charlson comorbidity score and was calculated using comorbidities recorded in HES records in the year prior to the cancer diagnosis. Frailty was measured using the Secondary Care Administrative Records Frailty (SCARF) index.15 The SCARF index is based on a cumulative deficit model of frailty and was originally developed using data from patients with breast cancer. The scores are based on the number of frailty deficits present in patient records in the 2 years prior to diagnosis, with the higher number of deficits identified indicating greater levels of frailty.15 Social deprivation was measured using the Index of Multiple Deprivation (IMD) 2019, based on the patient’s postcode at diagnosis.16 This allocates a deprivation score to people in the same area of residence. The areas (known as lower super output areas (LSOAs)) have an average population of approximately 1700 people,17 and there are 33 755 in England currently. Patients were allocated the quintile of the ranked LSOA IMD scores, with one indicating people living in the least deprived areas and five the most deprived areas.Study outcome The primary outcome was the receipt of SACT, including chemotherapy, targeted therapies and immunotherapy. The outcome was primarily identified from records in the SACT dataset. 18 A person was flagged as receiving SACT when the first recorded cycle was within 6 months of the date of diagnosis. Because a few National Health Service (NHS) hospitals did not submit information to the SACT dataset during the study period, records in the HES-APC were used to supplement these outcome data. Patients who had an inpatient spell where chemotherapy was given were identified by Classification of Surgical Operations and Procedures (V.4) codes (OPCS-4 codes: X70–X74) within 6 months of diagnosis.Statistical methods All statistical analyses were performed using Stata software. 19 Patients missing a recorded PS were excluded from the analyses. Descriptive statistics were produced to summarise patient characteristics and their cancer. Patients were allocated to the NHS trust of diagnosis for the analysis of geographical patterns of care. Each NHS trust is located within one of 21 cancer alliances, which are organisations responsible for coordinating cancer pathways within a region.20 In secondary descriptive analyses, SACT was further classified into chemotherapy, immunotherapy, targeted therapy or other, based on drug-level information recorded in the SACT dataset. Patients were assigned to a treatment category according to the first SACT regimen received within 6 months of diagnosis.To examine the association between the use of SACT and patient characteristics, we used a mixed-effects multivariable logistic regression model. Models were constructed separately for patients with PS 0–1 and 2. The models included fixed effects for age group (≤54, 55–64, 65–74, 75–84, ≥85), sex, year of diagnosis (2019–2022), disease stage (3B/C vs 4), socio-economic deprivation (IMD quintile), Charlson comorbidity score (0, 1, 2, ≥3) and frailty status (SCARF index: none/mild, moderate, severe) (online supplemental table 1). The cancer alliance of diagnosis was included in the model as a random effect, with differences between alliances modelled as a random intercept. A complete-case analysis was performed; patients with missing data for any included variable were excluded from the multivariable models. The statistical significance of covariates in the model was assessed using likelihood ratio tests on nested models. Because the baseline rate of SACT use was far from 0%, ORs from logistic models were presented as relative risks using the transformation of Zhang and Yu.21 Empirical-Bayes estimates of treatment use in each cancer alliance were estimated from the mixed-effects multivariable model and are presented with 99.8% credible intervals. We examined whether the association between frailty and receipt of SACT varied by age by adding an interaction term between age group and frailty category to the regression model.SP110.1136/bmjonc-2025-000957.supp1Supplementary data We further explored the association between age at diagnosis and use of SACT within each cancer alliance. Similar multivariable logistic regression models for SACT treatment were fitted separately for each cancer alliance, but with age treated as a continuous variable. To allow for a non-linear relationship, age was modelled as a cubic spline with three knots, placed at default percentiles of the distribution of a winsorised age variable (values below 50 were set to 50 and values above 90 were set to 90), in order to retain all patients in the analysis while limiting the influence of sparse extreme values at the tails of the age distribution.To investigate the association between treatment and survival at the alliance level, we initially calculated the median survival in each cancer alliance and plotted this against the adjusted (marginal) rate of patients who received SACT as estimated from the mixed-effects regression model described above. The strength of this association was quantified using the Pearson correlation coefficient.We developed a flexible parametric survival model to examine the relationship between patient, tumour and treatment factors and survival. A 2-month landmark was incorporated to reduce immortal time bias and because patients who die shortly after diagnosis are unlikely to be candidates for SACT. Survival time was therefore measured from 2 months after diagnosis, and patients who died within the first 2 months were excluded from the analysis. Treatment status was defined consistently with the main analysis: patients were classified as ‘treated’ if they had a recorded SACT cycle within 6 months of diagnosis. We also defined survival as the restricted mean survival over 36 months because less than 10% of patients had follow-up beyond 36 months. From this model, standardised survival probabilities were estimated for patients who did and did not receive SACT. These probabilities were produced from restricted mean survival predictions for each individual which were then averaged across the study population.22 Finally, we estimated how many extra patients would be treated each year and how many life-years gained, if the proportion of patients who had SACT within this population of patients with advanced SACT was at the highest level observed across the cancer alliances.Patient and public involvement The NLCA Patient and Public Involvement forum were consulted about the design and reporting of this study.Results Demographics of patients diagnosed with advanced NSCLC Overall 66 052 patients were diagnosed with advanced NSCLC in England between 1 January 2019 and 31 December 2022. The analysed cohort contained 60 210 patients with a diagnosis of advanced NSCLC and a recorded PS. Median age at diagnosis was 73 years (IQR 66–80); 83% had stage 4 disease at diagnosis and 25% of patients lived in the most socially deprived areas ( table 1). ECOG PS was good in 43% of patients (PS=0/1) and poor in 30% (PS=3/4). Among patients with good PS, 37% had a Charlson comorbidity score of 0, compared with 20% (n=3649) of patients with PS=3/4. A similar gradient was observed across the frailty SCARF index scores, with 72% of patients with PS=0/1 having a frailty index of none/mild compared with 31% of patients with PS=3/4.Table 1Characteristics of patients diagnosed with advanced NSCLC, stage 3B–4B, England 2019–2022PS*All (n=60 210)PS 0–1 (n=26 055)PS 2 (n=10 352)PS 3–4 (n=17 961)Year of diagnosis 201915 047 (25%)6895 (26%)2571 (25%)4413 (25%) 202014 501 (24%)6291 (24%)2544 (25%)4391 (24%) 202115 099 (25%)6393 (25%)2578 (25%)4654 (26%) 202215 563 (26%)6476 (25%)2659 (26%)4503 (25%)Age group (years) <543744 (6%)2576 (10%)380 (4%)399 (2%) 55–649742 (16%)5746 (22%)1469 (14%)1587 (9%) 65–7419 823 (33%)9915 (38%)3388 (33%)4684 (26%) 75–8419 610 (33%)6715 (26%)3786 (37%)7240 (40%) ≥857291 (12%)1103 (4%)1329 (13%)4051 (23%)IMD quintile 1 (least deprived)9129 (15%)4348 (17%)1442 (14%)2448 (14%) 210 998 (18%)5019 (19%)1845 (18%)3056 (17%) 311 914 (20%)5308 (20%)1939 (19%)3451 (19%) 412 874 (21%)5438 (21%)2230 (22%)3861 (22%) 5 (most deprived)15 295 (25%)5942 (23%)2896 (28%)5145 (29%)Stage 3B/C10 011 (17%)5219 (20%)1739 (17%)2232 (12%) 450 199 (83%)20 836 (80%)8613 (83%)15 729 (88%)Charlson score 017 581 (29%)9607 (37%)2729 (26%)3649 (20%) 118 987 (32%)8742 (34%)3188 (31%)5147 (29%) 214 528 (24%)5362 (21%)2723 (26%)5018 (28%) ≥39114 (15%)2344 (9%)1712 (17%)4147 (23%)Frailty Index None/mild32 244 (54%)18 738 (72%)5123 (49%)5529 (31%) Moderate15 459 (26%)5489 (21%)3189 (31%)5245 (29%) Severe12 507 (21%)1828 (7%)2040 (20%)7187 (40%)*Missing=5842.IMD, Index of Multiple Deprivation; NSCLC, non-small cell lung cancer; PS, performance status.Patients with advanced NSCLC receiving SACT Among patients with good PS, 62% (16 155/26 055) received SACT ( table 2). The association between use of SACT and age was particularly strong, with 76% of patients aged under 55 having treatment compared with 22% of those aged 85 or over (online supplemental figure 1). A strong association was also observed with the frailty index; approximately two-thirds of the ‘none/mild’ group were treated compared with one third of the ‘severe’ group. Patients with moderate PS (PS=2), 25% received SACT (2,601/10,352) and significant associations were again observed with age, disease stage (3B/C/4), Charlson score and SCARF index.Table 2Use of systemic treatment by performance status for patients with advanced NSCLC, stage 3B–4B, England 2019–2022Performance status 0–1Performance status 2NTreatedNTreatedAll26 05516 155 (62.0%)10 3522601 (25.1%)Year of201968954354 (63.2%)2571615 (23.9%)diagnosis202062913781 (60.1%)2544624 (24.5%)202163934010 (62.7%)2578680 (26.4%)202264764010 (61.9%)2659682 (25.7%)Age group≤5425761952 (75.8%)380189 (49.7%)(years)55–6457464000 (69.6%)1469536 (36.5%)65–7499156461 (65.2%)3388971 (28.7%)75–8467153497 (52.1%)3786811 (21.4%)≥851103245 (22.2%)132994 (7.1%)Stage3B/C52193637 (69.7%)1739490 (28.2%)4A–B20 83612 518 (60.1%)86132111 (24.5%)IMD quintile1 (least)43482814 (64.7%)1442379 (26.3%)250193172 (63.2%)1845472 (25.6%)353083328 (62.7%)1939485 (25.0%)454383310 (60.9%)2230551 (24.7%)5 (most)59423531 (59.4%)2896714 (24.7%)Charlson096076400 (66.6%)2729730 (26.8%)score187425575 (63.8%)3188871 (27.3%)253623015 (56.2%)2723663 (24.4%)≥323441165 (49.7%)1712337 (19.7%)FrailtyNone/mild18 73812 593 (67.2%)51231535 (30.0%)Moderate54892896 (52.8%)3189757 (23.7%)Severe1828666 (36.4%)2040309 (15.2%)IMD, Index of Multiple Deprivation; NSCLC, non-small cell lung cancer.An exploratory analysis of the interaction between age and frailty in relation to receipt of SACT suggested an association between frailty and treatment uptake varied by age, with a steeper gradient observed in younger patients (online supplemental table 2 and online supplemental figure 2) but the strength of the evidence was weak (p=0.046).The adjusted relative risks, estimated by the mixed-effects regression analysis, are presented in table 3. Among patients with good PS (0–1), receipt of SACT varied by age at diagnosis, and there were clear gradients in the use of SACT by degree of frailty, socio-economic deprivation and disease stage categories. Variation over calendar time was minimal. Among patients with PS 2, receipt of SACT similarly varied by age, disease stage, socio-economic deprivation and frailty.Table 3Multivariable logistic regression of factors associated with receipt of SACT*PS 0–1PS 2RR (95% CI)†P valueRR (95% CI)†P valueYear20191.00 (reference)0.00161.000.124020200.95 (0.92 to 0.98)1.01 (0.89 to 1.15)20211.00 (0.96 to 1.04)1.10 (0.97 to 1.24)20220.99 (0.96 to 1.03)1.09 (0.98 to 1.22)Age group≤541.00 (reference)<0.00011.00<0.000155–640.93 (0.90 to 0.96)0.73 (0.61 to 0.86)65–740.88 (0.85 to 0.91)0.58 (0.48 to 0.70)75–840.72 (0.68 to 0.75)0.43 (0.35 to 0.53)≥850.31 (0.28 to 0.34)0.14 (0.11 to 0.18)Stage3B/C1.00 (reference)<0.00011.000.00014A–B0.86 (0.82 to 0.89)0.83 (0.75 to 0.92)IMD quintile1 (least)1.00 (reference)<0.00011.000.000220.96 (0.93 to 0.99)0.93 (0.83 to 1.03)30.93 (0.89 to 0.98)0.87 (0.76 to 0.98)40.88 (0.85 to 0.92)0.77 (0.69 to 0.87)5 (most)0.85 (0.80 to 0.89)0.77 (0.69 to 0.85)Comorbidity score01.00 (reference)0.00191.000.410810.99 (0.96 to 1.02)1.07 (0.99 to 1.16)20.95 (0.92 to 0.98)1.05 (0.94 to 1.17)≥30.98 (0.94 to 1.02)1.04 (0.90 to 1.19)FrailtyNone/mild1.00 (reference)<0.00011.00<0.0001Moderate0.83 (0.81 to 0.85)0.79 (0.73 to 0.85)Severe0.61 (0.56 to 0.67)0.55 (0.47 to 0.65)Variance estimateVariance estimateCancer alliance‡0.06 (0.04 to 0.12)<0.00010.10 (0.04 to 0.21)<0.0001*Mixed effects multivariable logistic regression, separate models for PS=0/1 and PS=2. †Relative risk calculated from OR.‡Cancer alliance included in models as random effect.IMD, Index of Multiple Deprivation; PS, performance status; SACT, systematic anticancer therapy.Use of SACT by treatment modality Among patients with stage 3B–4B NSCLC and good PS (PS 0–1) who received SACT, the type of treatment administered changed over time ( online supplemental table 3). The proportion of patients receiving immunotherapy rose from 40.6% in 2019 to 48.2% in 2020 and remained above 50% in subsequent years. In contrast, chemotherapy use declined from 46.5% in 2019 to 35.2% in 2020 and stabilised at around 34% thereafter. Targeted therapies accounted for approximately 13–15% of treatments each year.Variation across England Figure 1 describes the extent of the variation in the use of SACT among patients with advanced NSCLC (PS=0/1) between cancer alliances, after adjustment for patient characteristics (For patients PS=2, online supplemental figure 3). The proportion of patients who received SACT varied from 51% in Humber, Coast and Vale to around 72% in North East London. We investigated the association between age at diagnosis and use of SACT across the cancer alliances using the mixed-effects regression with age as a continuous variable. The change in use by age was broadly similar across cancer alliances, with use of SACT in most alliances falling sharply among patients aged over 70 years (figure 2; for patients PS=2, online supplemental figure 4).Figure 1Systemic treatment use in patients with good PS (PS=0/1), by cancer alliance (adjusted for age, disease stage, social deprivation, comorbidity and frailty). PS, performance status.Figure 2Use of systemic anticancer therapies by age in patients with advanced NSCLC (PS=0/1), by cancer alliance. Marginal rates for age are adjusted for disease stage, comorbidity and frailty. NSCLC, non-small cell lung cancer; PS, performance status.Association between use of SACT and survival In patients with advanced NSCLC (PS=0/1), median survival was longer in cancer alliances with a greater proportion receiving SACT, with a correlation coefficient of 0.62 (p=0.003, figure 3). The association was significant among patients with moderate PS (PS=2, r=0.45, p=0.043, online supplemental figure 5).Figure 3Median survival and adjusted rates of SACT use by cancer alliance for patients with advanced NSCLC and good PS (PS=0/1). mths, months; NSCLC, non-small cell lung cancer; PS, performance status; SACT, systematic anticancer therapy.Without the 2-month landmark incorporated into the model, the restricted mean survival over 36 months was 7.6 months in patients who did not have SACT, and 18.1 months in patients that did. Introducing the 2-month landmark into the model reduced the cohort to 21 740 patients with advanced NSCLC (PS=0/1), of whom 15 647 (72%) had SACT. From this adjusted model, the estimated restricted mean survival time was 9.5 months among those who did not receive SACT, and 17.3 months among those who did. The removal of patients who died within 2 months increased the highest treatment rate across the alliances from 72% to 80%. Assuming that each cancer alliance was able to achieve the highest rate, we estimated that a further 448 patients would be treated each year, with a total of 289 life years gained.Discussion This study investigated the use of SACT among patients with advanced NSCLC and explored how the demographic and clinical characteristics were associated with the receipt of treatment. Our study revealed notable differences between patients with good PS who received SACT and those who did not. Age at diagnosis was a critical factor, with older patients less likely to receive SACT compared with their younger counterparts. This observation aligns with existing literature, where age-related comorbidities, frailty and concerns about treatment tolerance often result in more conservative treatment approaches for older patients. 8 23–25 The observed shift from chemotherapy to immunotherapy between 2019 and 2020 is likely to reflect services responding to changes in national treatment guidelines. The National Institute for Health and Care Excellence (NICE) guideline for NSCLC,1 published in 2019, incorporated immunotherapy as a standard treatment option for patients with stage 3B–4B NSCLC and suitable biomarker profiles.We also observed significant differences in the use of SACT across the cancer alliances that persisted after accounting for differences in patient characteristics. If every alliance achieved the highest treatment rate, we estimated that a further 448 patients would be treated each year, with a gain of 289 life years overall.Comparison with other studies Various studies have examined the use of SACT among patients with advanced NSCLC. A study from Portugal in the preimmunotherapy era reported that approximately 76% of patient with advanced NSCLC received SACT; they found increasing age and stage 4 disease were the key predictors for patients not receiving SACT but unfortunately data on PS was not available. 23 A study from the USA, focusing on patients with stage 4 SCLC from 2012 to 2015, found 80% received chemotherapy based SACT.24 They reported that older patients had SACT regimens that were modified more often, however survival for older patients was similar to younger patients. The study did not report PS, the demographics of patients who did not receive chemotherapy and it is possible that older patients who underwent chemotherapy had fewer comorbidities and were less frail compared with those who did not receive SACT.A UK-based retrospective cohort study of advanced NSCLC diagnosed between 2007 and 2017 similarly reported that SACT uptake was low (30.6%), with older age and poorer PS associated with non-treatment.25 However, that study preceded the widespread use of immunotherapy and did not include data on frailty or more recent treatment trends.In our study, receipt of SACT varied substantially across age groups, even after accounting for age-related patient characteristics such as PS, frailty and comorbidity burden. These findings are in keeping with a previous study conducted in England which considered PS at first oncology clinic appointment but lacked an objective measure of frailty.8 The national guidelines for lung cancer recommend treatment decisions are based on overall fitness rather than age alone. However, there is often an underrepresentation of older adults within randomised control trials (RCTs) for cancer treatments, including SACT for lung cancer26 27; this is often because of inclusion criteria for trials indirectly and disproportionally excluding older patients.28 The lower rates of SACT use in older populations may reflect the lack of evidence on older people from RCTs evaluating SACT.Social deprivation was also associated with patients receiving SACT, with patients who lived in the most deprived areas being least likely to have treatment. This finding is consistent with concerns that socio-economic factors may influence access to treatment, adherence to guidelines and overall outcomes.29 A UK-based study focusing on socio-economic inequalities in patients with stage 4 NSCLC (diagnosed 2012–2017) reported that twice as many patients in the most affluent areas had a molecular targeted therapy or other novel SACT drug than patients in the most deprived areas (9.56% vs 5.56%).30 Across cancer alliances in England, after adjustment for differences in patient demographics, the lowest and highest rates of SACT use differed by around 20%. It is suggested that centres that actively participate in research are more up to date with medical developments, have more clinical resources available to them and overall have improved patient outcomes.31 This may well be the case with adoption of newer SACT agents. A challenge then remains to improve dissemination of changing practices to reach clinical teams and their patients in all trusts. If all cancer alliances achieved the treatment rate of the highest, approximately 450 more patients would be treated each year, an increase of approximately 10%.Survival for patients with advanced NSCLC is known to be poor, with median survival times after initiation of SACT reported to be around 10–12 months.23 24 Our estimates use data on patients diagnosed between 2019 and 2022, and provide a more recent picture of outcomes. In the whole cohort of patients with stage 3B–4 (PS 0–1), we estimated the restricted mean survival time over 36 months between treated and untreated patients to be 18.1 and 7.6 months, respectively. However, it is likely that some patients who did not have SACT would not be candidates due to their poor prognosis. Using the 2-month landmark analysis, the estimated restricted mean survival over 36 months was 17.3 months among patients who received SACT and 9.5 months among those who did not.Strengths and limitations The study has several strengths. It used a large, population-based cohort of patients recently diagnosed with lung cancer using national cancer registration data. The patient data reflect modern clinical practice and the data were collected using nationally agreed data definitions.There was detailed information on the systemic therapies delivered to patients in the SACT dataset, and these data were augmented where necessary using national hospital admission data. The use of a landmark period was intended to reduce immortal-time bias in the survival analysis, and estimates of survival probabilities for receiving or not receiving SACT were derived using standardised survival curves. Nevertheless, some residual immortal-time bias cannot be fully excluded, as treatment was defined based on receipt within 6 months of diagnosis in a population with poor overall prognosis.The study has some limitations. The use of a complete-case analysis resulted in the exclusion of a small proportion of patients. Although missing PS data accounted for fewer than 10% of cases, multiple imputation was not undertaken because PS was a primary stratifying variable and a key clinical determinant of treatment eligibility. The RCRD dataset, while enabling timely national analyses, has some limitations related to data completeness for key variables such as PS and cancer stage.10 Second, our study could not account for certain variables, such as specific comorbidities that may be absolute or relative contraindications for SACT or patient preferences, which are crucial factors in treatment planning. For example, smoking status, autoimmune disorders and route of presentation may all influence SACT uptake. The regression models described associations between treatment use and patient characteristics while accounting for differences in comorbidity burden and frailty. The conditions used to derive these indices are generally accurately coded in HES.32 The frailty measure used in this study (SCARF index) relies on diagnostic coding from hospital data and therefore may underdetect frailty in patients with limited secondary care contact or incomplete coding.15 The SCARF index was developed and validated in a breast cancer population15 and has subsequently been validated in a colorectal cancer population,33 but it has not been validated specifically in lung cancer; at present, no frailty tool validated for patients with lung cancer using routine administrative data exists. Time from diagnosis to treatment is likely to be an important factor in patients successfully starting their recommended treatment; in 2023, the median time between diagnosis of stage 4 NSCLC and initiation of SACT was 43 days.7 A further limitation is the lack of information on biomarkers in our NSCLC population. Biomarker testing is essential in SACT treatment planning to determine the optimal regime.34 Our study did not report on biomarker testing results and furthermore did not appreciate if patients received appropriate SACT regimes in accordance with their biomarker results. A recent study in England was the first national study of biomarker testing in NSCLC; they found that when a targetable biomarker was found, only 68% of those patients received SACT as per recommendations from NICE.35 We used the Zhang and Yu method21 to approximate relative risks from ORs to improve interpretability. When outcomes are common, ORs tend to overestimate the magnitude of association compared with relative risks; the Zhang and Yu transformation reduces this overestimation. However, this approach has limitations, including assumptions about baseline risk and the absence of effect modification, which may not hold in all settings. These limitations should be considered when interpreting the adjusted estimates.Finally, the COVID-19 pandemic during 2020 and 2021 are clearly atypical in the provision of cancer care. The study observed the lowest number of lung cancer diagnoses and the lowest use of SACT in 2020 compared with any of the other years. This is consistent with reports on practice in England during the pandemic.36 37 In the years following 2020, the lung cancer diagnoses have increased, but SACT treatment rates have been slower to return to prepandemic levels.Conclusion Among NHS England patients with stage 3B–4B NSCLC and good PS (PS 0/1), SACT was received by 60% of patients. The receipt of SACT was influenced by patient age at diagnosis and socio-economic deprivation after accounting for comorbidities and frailty. These findings suggest decisions about treatment may deviate from NICE guideline recommendations, although it is unclear to what extent patient preferences contribute to these differences.There is considerable variation in use of SACT between cancer alliances in England. We estimated that if all cancer alliances treated the same proportion of patients as the cancer alliance with the highest treatment rate, 289 life-years could be gained each year. This suggests the potential improvement in outcomes if higher treatment rates among patients with stage 3B–4B NSCLC (PS 0–1) were achieved. The extent to which these can be realised depends on various factors, not least patient preferences. Nonetheless, improving the survival of patients with advanced NSCLC remains a priority. Addressing inequalities in SACT utilisation and ensuring adherence to NICE guidelines are essential steps towards improving outcomes for all patients with advanced NSCLC.