Flexible Parametric Survival Models Research Articles

Twenty years of melanoma in Victoria, Queensland, and South Australia (1997 – 2016)

BackgroundGiven the high incidence of melanoma in Australia alongside high mortality with later stage disease, we investigated the populations and locations most at risk, to optimise public health activities in areas where intervention is most needed. This study examines trends and identifies significant prognostic factors and potential disparities in incidence, mortality and survival between population groups in Victoria, Queensland and South Australia. MethodsThe analysis includes data from the population-based cancer registries of the three states over a twenty-year period (1997–2016). Age-standardized and age-specific incidence rates were calculated, and long-term trends analysed using Joinpoint Regression. Five-year relative survival estimates for the study population were calculated using the cohort method and multivariable flexible parametric survival models were applied for each jurisdiction to calculate adjusted excess mortality hazard ratios for the key characteristics. ResultsThere were more males with melanoma than females in all the three states. Over 60% of the cases occurred in the 40–74 years age group. Most melanomas had a Breslow thickness less than or equal to 1.0 mm. For males, Victoria and Queensland had a statistically significant increasing trend whereas in South Australia there was a decreasing trend. For females, the incidence rate trend was stable in Victoria but significantly decreasing in South Australia. In Queensland there was an increasing and statistically significant trend from 2006 to 2016. Across all three states there was a reducing incidence rate in the youngest cohort, stabilizing incidence in the 40–59-year-old age group, and increasing in the oldest cohorts. Five-year relative survival decreased with increasing age and with Breslow thickness across all three jurisdictions. Males had between 43%− 46% excess mortality compared to females in all the three states. There was higher risk with increasing age and Breslow thickness, with the largest risk among the 75 + age group and those with a Breslow thickness of > 4 mm. ConclusionIt is the first time that data from these three registries has been analysed together in a uniform way, covering more than half of the Australian population. This study compares the epidemiology of melanoma across three states and provides a better understanding of trends and factors affecting outcome for Australians with melanoma. While there has been some improvement in aspects of incidence and mortality, this has not been evenly achieved across Australia.

Treatment-Related Circulatory Diseases and Mortality in Hodgkin Lymphoma Patients Using Multi-State Modelling and Relative Survival

Introduction Diseases of the circulatory system (DCS) are some of the most common late effects of Hodgkin lymphoma (HL), primarily due to mediastinal radiotherapy and/or anthracycline-containing chemotherapy (e.g., ABVD and BEACOPP). As DCS is very common in the general population, modelling DCS attributable to HL therapy versus that expected in the general population is relevant to understand the magnitude of DCS reduction that could be achieved with less cardiotoxic therapies. This real-world data study used multi-state modelling and relative survival to estimate treatment-related DCS risk in HL survivors. Material and methods All HL patients registered in the Swedish Lymphoma Register 2000-2018, aged 18-80 years at diagnosis, and treated with anthracycline-containing therapy were included. DCS events were identified in the Swedish National Inpatient Register (ICD-10 codes I00-I99) and defined as the first admission for DCS after HL diagnosis. Sex-, year-, and age-specific DCS incidence rates in the general population were retrieved from public records. Patients were followed from treatment initiation through a series of states: alive and DCS-free, alive and with (excess or expected) DCS, dead without DCS, and dead after DCS. Follow-up ended on date of death, emigration, or December 31st, 2019, whichever came first. All state transitions were modelled separately using flexible parametric survival models, yielding hazard ratios (HR) with 95% confidence intervals (CI). A relative survival model incorporating the expected DCS rates was fitted to allow for estimation of excess DCS, which was then interpreted as treatment-related. Transition probabilities (i.e., the likelihood of being in a state) were predicted for specific patient groups, by sex and low/high (≤200/>200 mg/m2) cumulative anthracycline dose. Results In a total of 1,929 HL patients, the mean age at diagnosis was 42 years, 54% were males, and 49% were treated with high dose anthracycline (Table 1). The distribution of treatments was 66% ABVD, 20% CHOP-like, and 14% BEACOPP. During a median follow-up of 7.6 years (IQR: 4.5-11.0), 145 (7.5%) died without DCS, 377 patients (20%) were diagnosed with DCS, and 87 (5%) died after DCS. Figure 1a shows the sex-specific transition probabilities for a patient with advanced stage disease, diagnosed in 2000, at age 50, and treated with a cumulative anthracycline dose of >200mg/m2 in a ABVD regimen. The probability of being in the treatment-related (excess) DCS state at 10 years after first line treatment was 0.35 (95% CI: 0.09-0.75) for males and 0.41 (95% CI: 0.10-0.81) for females in this patient group. Figure 1b shows the probabilities for limited stage male patients diagnosed in 2000 at age 50, treated with ABVD, by low/high anthracycline dose. Here, the probability of being in the treatment-related (excess) DCS state at 10 years was 0.44 (95% CI: 0.16-0.76) and 0.53 (95% CI: 0.25-0.80) for low and high dose anthracycline patients, respectively. There were no significant differences in treatment-related (excess) DCS incidence rates or post DCS all-cause mortality rates (Table 1). Females had a significantly lower non-DCS mortality rate than males (adj. HR=0.69, 95% CI: 0.49-0.97), as did patients treated with high compared to low anthracycline dose (adj. HR=0.28, 95% CI: 0.18-0.44). Conclusion After 10 years, females had higher risks of treatment-related (excess) DCS compared to males, although not statistically significant and likely due to superior survival. The same was seen for high compared to low cumulative doses of anthracycline. Female sex and high dose anthracycline reduced the non-DCS mortality rate, but not the excess DCS incidence rate. Figure 1View largeDownload PPTFigure 1View largeDownload PPT Close modal

Association of Maternal Preeclampsia With Offspring Risks of Ischemic Heart Disease and Stroke in Nordic Countries

An association between maternal preeclampsia and an increased risk of cardiovascular disease in the offspring is plausible, but evidence in this area is limited. To investigate (1) the association between maternal preeclampsia and risks of ischemic heart disease (IHD) and stroke in the offspring, (2) whether the association varies by severity or timing of onset of preeclampsia, and (3) the role of preterm birth and small for gestational age (SGA) birth, both of which are related to preeclampsia and cardiovascular diseases, in this association. This multinational population-based cohort study obtained data from Danish, Finnish, and Swedish national registries. Live singleton births from Denmark (1973-2016), Finland (1987-2014), and Sweden (1973-2014) were followed up until December 31, 2016, in Denmark and December 31, 2014, in Finland and Sweden. Data analyses were performed between September 2020 and September 2022. Preeclampsia and its subtypes, including early onset (<34 gestational weeks) and late onset (≥34 gestational weeks), severe and mild or moderate, and with and without SGA birth. Diagnoses of IHD and stroke were extracted from patient and cause-of-death registers. Cox proportional hazards regression models and flexible parametric survival models were used to analyze the associations. Sibling analyses were conducted to control for unmeasured familial factors. The cohort included of 8 475 819 births (2 668 697 [31.5%] from Denmark, 1 636 116 [19.3%] from Finland, and 4 171 006 [49.2%] from Sweden, comprising 4 350 546 boys [51.3%]). Of these offspring, 188 670 (2.2%) were exposed to maternal preeclampsia, 7446 (0.1%) were diagnosed with IHD, and 10 918 (0.1%) were diagnosed with stroke during the median (IQR) follow-up of 19.3 (9.0-28.1) years. Offspring of individuals with preeclampsia had increased risks of IHD (adjusted hazard ratio [HR], 1.33; 95% CI, 1.12-1.58) and stroke (adjusted HR, 1.34; 95% CI, 1.17-1.52). These associations were largely independent of preterm or SGA birth. Severe forms of preeclampsia were associated with a higher stroke risk than less severe forms (severe vs mild or moderate: adjusted HR, 1.81 [95% CI, 1.41-2.32] vs 1.22 [95% CI, 1.05-1.42]; early vs late onset: adjusted HR, 2.55 [95% CI, 1.97-3.28] vs 1.18 [95% CI, 1.01-1.39]; with vs without SGA birth: adjusted HR, 1.84 [95% CI, 1.44-2.34] vs 1.25 [95% CI, 1.07-1.48]). Sibling analyses suggested that the associations were partially explained by unmeasured familial factors. Results of this study suggest that offspring born to individuals with preeclampsia had increased IHD and stroke risk that were not fully explained by preterm or SGA birth, and that the associated risks for stroke were higher for severe forms of preeclampsia.

Predictors of Adverse Clinical Outcome in Diffuse Large B-Cell Lymphoma Richter Transformation of CLL/SLL: A Consortium of Richter Transformation Investigators (CORTI) Multicenter Study across 8 US, UK, and Australian Academic Centers

Background Richter transformation (RT) of chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL) is a known but uncommon complication with dismal prognosis despite improvements in CLL outcomes with novel agents. Previously published studies are characterized by cohort heterogeneity with respect to CLL treatments (predating the rituximab era and novel agents) and RT histology [diffuse large B-cell lymphoma (DLBCL) and classic Hodgkin lymphoma (cHL)], limiting their generalizability in the modern era. To this end, we examined characteristics and outcomes in 222 DLBCL-RT cases, enriching for patients treated with novel agents. Methods We collected 247 RT cases diagnosed between 1995-2022 across eight academic institutions, excluding 23 with cHL-RT and 2 with "pseudo-transformation" after BTKi withdrawal (final n = 222; 41% of cases post-2010). We analyzed > 65 CLL/RT biologic, pathologic, and clinical parameters. Primary outcomes included complete response (CR) to first line of therapy [in CLL (iwCLL 2008/2018) and RT (Lugano 2014)] and 5-year overall survival (5yr-OS) after RT, estimated using Cox proportional hazards [yielding hazard ratios (HRs) and 95% confidence intervals (CIs)] and flexible parametric survival models (for prediction of standardized survival). Results The antecedent CLL included: 11% cases with atypical CLL, 84% IGHVUM, and 70% ZAP70+ with 46% and 43% of patients harboring del(13q) and del(17p), respectively. CLL/RT was synchronous (CLL diagnosis within 2 months of RT diagnosis) in 13% of cases. Only del(17p) correlated with C/HC (≥ 3 / ≥ 5 chromosome abnormalities) karyotype (p < 0.0001), and ZAP70/CD38 had no association with IGHV status. RT was primarily nodal (47.3%), non-GCB (90.3%), and EBV-negative (95%) with p53 IHC-positive large cells in 60% of cases and MYC translocation in 10%. Preceding CLL was treated with a median of 2 lines of therapy (range: 1-7) with 53% heavily-treated (≥ 2 lines). The most common first lines (LOT1) were chemotherapy (fludarabine-based or bendamustine/CHOP equivalents; 73.6%) and BTKi/novel agents (17%); 37% received BTKi as any LOT, while 44% received any novel agent (BTKi, BCL2i, PI3Ki, or SYKi). CLL CRLOT1 was achieved in 41.2% with overall response of 77%. Median time to RT from CLL diagnosis (sequential group) was 47.6 months pre-2010 (IQR: 87.8) vs. 21.8 months post-2010 (IQR: 34.6, p < 0.0001). During a median follow up of 10.2 months (IQR: 27.1 months) post-RT, 59% of patients died. Median OS was 17.4 months (range: 11.3-24.2) with 28% 5yr-OS. In the pre-rituximab era (i.e. prior to 2010), median OS was significantly shorter compared to the present era (8 vs. 19.9 months). Indicators of inferior 5yr-OS on univariable analysis included: sequential transformation (HR: 2.34 [95% CI: 1.25-4.35]), not achieving CRLOT1 for CLL (HR: 1.62 [95% CI: 1.01-2.58]), heavily-treated CLL (HR: 1.98 [95% CI: 1.36-2.88]), and C/HC karyotype (HR: 1.59 [95% CI: 1.02-2.46]). While trisomy 12 predicted superior outcome (HR: 0.54 [95% CI: 0.32-0.94]), del(13q) and del(17p) had no impact. Adverse baseline covariates at RT diagnosis were low (< 3.5 g/dL) serum albumin (HR: 3.13 [95% CI: 2.09-4.69]), low (< 150x103/uL) platelet count (HR: 2.38 [95% CI: 1.64-3.47]), LDH > 1.5x ULN (HR: 1.63 [95% CI: 1.12-1.38]), and ECOG-PS ≥ 2 (HR: 2.4 [95% CI: 1.51-3.81]). Notably, p53 IHC positivity in RT predicted increased all-cause mortality (HR: 2.13 [95% CI: 1.27-3.5]), retaining significance in both minimally- and heavily-treated CLL subsets. RT was treated using CHOP-R or EPOCH-R in 62%. Not achieving CRLOT1 for RT predicted worse 5yr-OS (HR: 3.90 [95% CI: 2.09-7.29]). A multivariable, main effects Cox model was constructed using five variables (age at transformation, CLL treatment number, LDH > 1.5x ULN, albumin, and p53 IHC). In this age-adjusted model, all variables retained adverse prognostic significance (Figure 1). Marginal OS was predicted by p53 IHC, standardizing over remaining covariates, and with a trend towards inferior 5yr-OS in p53-positive vs. -negative RT (13% vs. 38%; Figure 2). Conclusion This multicenter, international study is the largest cohort of RT in the era of novel therapies with solely DLBCL histology. We confirmed the utility of several traditional predictors. In addition, we identified p53 IHC as a promising predictor of adverse outcome (in univariable and multivariable analyses) based on these preliminary analyses. Figure 1View largeDownload PPTFigure 1View largeDownload PPT Close modal

Modelling multiple time-scales with flexible parametric survival models

BackgroundThere are situations when we need to model multiple time-scales in survival analysis. A usual approach in this setting would involve fitting Cox or Poisson models to a time-split dataset. However, this leads to large datasets and can be computationally intensive when model fitting, especially if interest lies in displaying how the estimated hazard rate or survival change along multiple time-scales continuously.MethodsWe propose to use flexible parametric survival models on the log hazard scale as an alternative method when modelling data with multiple time-scales. By choosing one of the time-scales as reference, and rewriting other time-scales as a function of this reference time-scale, users can avoid time-splitting of the data.ResultThrough case-studies we demonstrate the usefulness of this method and provide examples of graphical representations of estimated hazard rates and survival proportions. The model gives nearly identical results to using a Poisson model, without requiring time-splitting.ConclusionFlexible parametric survival models are a powerful tool for modelling multiple time-scales. This method does not require splitting the data into small time-intervals, and therefore saves time, helps avoid technological limitations and reduces room for error.

Science, Medicine, and the Anesthesiologist

Science, Medicine, and the Anesthesiologist

Preexisting Neuropsychiatric Conditions and Associated Risk of Severe COVID-19 Infection and Other Acute Respiratory Infections

Evidence indicates that preexisting neuropsychiatric conditions confer increased risks of severe outcomes from COVID-19 infection. It is unclear how this increased risk compares with risks associated with other severe acute respiratory infections (SARIs). To determine whether preexisting diagnosis of and/or treatment for a neuropsychiatric condition is associated with severe outcomes from COVID-19 infection and other SARIs and whether any observed association is similar between the 2 outcomes. Prepandemic (2015-2020) and contemporary (2020-2021) longitudinal cohorts were derived from the QResearch database of English primary care records. Adjusted hazard ratios (HRs) with 99% CIs were estimated in April 2022 using flexible parametric survival models clustered by primary care clinic. This study included a population-based sample, including all adults in the database who had been registered with a primary care clinic for at least 1 year. Analysis of routinely collected primary care electronic medical records was performed. Diagnosis of and/or medication for anxiety, mood, or psychotic disorders and diagnosis of dementia, depression, schizophrenia, or bipolar disorder. COVID-19-related mortality, or hospital or intensive care unit admission; SARI-related mortality, or hospital or intensive care unit admission. The prepandemic cohort comprised 11 134 789 adults (223 569 SARI cases [2.0%]) with a median (IQR) age of 42 (29-58) years, of which 5 644 525 (50.7%) were female. The contemporary cohort comprised 8 388 956 adults (58 203 severe COVID-19 cases [0.7%]) with a median (IQR) age of 48 (34-63) years, of which 4 207 192 were male (50.2%). Diagnosis and/or treatment for neuropsychiatric conditions other than dementia was associated with an increased likelihood of a severe outcome from SARI (anxiety diagnosis: HR, 1.16; 99% CI, 1.13-1.18; psychotic disorder diagnosis and treatment: HR, 2.56; 99% CI, 2.40-2.72) and COVID-19 (anxiety diagnosis: HR, 1.16; 99% CI, 1.12-1.20; psychotic disorder treatment: HR, 2.37; 99% CI, 2.20-2.55). The effect estimate for severe outcome with dementia was higher for those with COVID-19 than SARI (HR, 2.85; 99% CI, 2.71-3.00 vs HR, 2.13; 99% CI, 2.07-2.19). In this longitudinal cohort study, UK patients with preexisting neuropsychiatric conditions and treatments were associated with similarly increased risks of severe outcome from COVID-19 infection and SARIs, except for dementia.

The effect of vitamin D supplementation on risk of keratinocyte cancer: an exploratory analysis of the D-Health randomized controlled trial.

Vitamin D may play a role in prevention of keratinocyte cancer (KC), but observational studies examining the association between serum 25-hydroxy vitamin D concentration and KC are largely uninformative because sun exposure causes both KC and vitamin D production. There is scant evidence from clinical trials of supplementary vitamin D. To examine the effect of vitamin D supplementation on the risk of developing KC. We used data from the D-Health Trial, a randomized placebo-controlled trial of vitamin D supplementation (60 000 international units monthly for 5 years) among Australians aged ≥60 years. KC outcomes were captured through linkage to a national administrative dataset for those who consented (N = 20 334; 95%). We used negative binomial regression to analyse the incidence of KC excisions and the incidence of actinic lesions treated using cryotherapy or serial curettage, and flexible parametric survival models for analysis of time to first KC excision. Randomization to vitamin D supplementation did not reduce the incidence of KC lesions treated by excision [incidence rate ratio (IRR) 1·04; 95% confidence interval (CI) 0·98-1·11], the incidence of actinic lesions treated using other methods (IRR 1·01; 95% CI 0·95-1·08) or time to first histologically confirmed KC excision (hazard ratio 1·02; 95% CI 0·97-1·08). However, in subgroup analysis vitamin D increased the incidence of KC excisions in adults aged ≥ 70 years (IRR 1·13, 95% CI 1·04-1·23; P-value for interaction = 0·01). Vitamin D supplementation did not reduce the incidence of KC or other actinic lesions. What is already known about this topic? Laboratory studies have suggested possible protective effects of vitamin D on skin cancer. Observational studies investigating the association between vitamin D and risk of keratinocyte cancer are largely uninformative as ultraviolet radiation both causes skin cancer and is the primary source of vitamin D. The evidence from randomized controlled trials of vitamin D is limited and inconclusive. What does this study add? This population-based, randomized controlled trial suggests that supplementing older adults with a high monthly dose of vitamin D for 5 years does not affect the incidence of keratinocyte cancer.

Association of cardiovascular health with diabetic complications, all-cause mortality, and life expectancy among people with type 2 diabetes

BackgroundWe aimed to assess the impact of healthy cardiovascular health (CVH) on diabetic complications, mortality, and life expectancy among people with type 2 diabetes and to explore whether inflammation marker mediate these associations.MethodsThis prospective cohort study included 33,236 participants (aged 40–72) with type 2 diabetes from the UK Biobank with annual follow-up from 2006 to 2010 to 2020. Type 2 diabetes was ascertained from self-report, glycated hemoglobin ≥ 6.5%, hospital inpatient registry, or glucose-lowering medication use. Information on mortality was derived from the national death registry. Favorable CVH metrics consisted of non-smoker, regular physical activity, a healthy diet, non-overweight, untreated resting blood pressure < 120/<80 mm Hg, and untreated total cholesterol < 200 mg/dL. Participants were categorized into three groups according to the number of favorable CVH metrics: unfavorable (0 or 1); intermediate (any 2 or 3); and favorable (4 or more). Inflammation marker, as measured by C-reactive protein (CRP), was assessed at baseline and categorized as low (≤ 3 mg/L) and high (> 3 mg/L). Data were analyzed using Cox regression models, flexible parametric survival models, and mediation models.ResultsDuring the follow-up (median: 11.7 years), 3133 (9.4%) cases of diabetes complications and 4701 (14.1%) deaths occurred. Compared to unfavorable CVH, favorable CVH was associated with a reduced risk of diabetes complications (HR, 0.35; 95% CI, 0.26–0.47) and all-cause mortality (HR, 0.53; 95% CI, 0.43–0.65). In participants with unfavorable CVH, life expectancy at age 45 had a significantly reduction of 7.20 (95% CI, 5.48–8.92) years compared to those with a favorable CVH. Among people with type 2 diabetes, the proportions of diabetes complications and all-cause mortality that would be reduced by promoting the favorable CVH was 61.5% and 39.1%, respectively. CRP level mediated 14.3% and 29.7% of the associations between CVH and diabetic complication and all-cause mortality, respectively.ConclusionA favorable CVH was associated with lower risk of diabetes complications and mortality risk, and was associated with a longer life expectancy among people with type 2 diabetes. This association may be in part accounted for by inflammatory processes. Our findings highlight the importance of favorable CVH for the prevention of diabetic complications and all-cause mortality among people with type 2 diabetes, and underscores the need to monitor inflammation among people with unfavorable CVH.

Cancer survival differentials for Aboriginal and Torres Strait Islander peoples in Queensland: the impact of remoteness

PurposeIn Australia, Aboriginal and Torres Strait Islander peoples (First Nations population) often have low overall cancer survival, as do all residents of geographically remote areas. This study aimed to quantify the survival disparity between First Nations and other Queenslanders for 12 common cancer types by remoteness areas.MethodsFor all Queensland residents aged 20–89 years diagnosed with a primary invasive cancer during 1997–2016, we ran flexible parametric survival models incorporating age, First Nations status, sex, diagnosis time period, area-level socioeconomic status, remoteness categories and where appropriate, broad cancer type. Three survival measures were predicted: cause-specific survival, survival differences and the comparative survival ratio, each standardised to First Nations peoples’ covariate distributions.ResultsThe standardised five-year cause-specific cancer survival was 60% for urban First Nations and 65% for other Queenslanders, while remote residents were 54% (First Nations) and 58% (other). The absolute survival differential between First Nations and other Queenslanders was often similar, regardless of remoteness of residence. The greatest absolute difference in five-year standardised cancer survival was for head and neck cancers, followed by cervical cancer.The five-year comparative survival ratio (First Nations: other Queenslanders) for urban cancer patients was 0.91 (95% CI 0.90–0.93), similar to outer regional, inner regional and remote areas. The greatest comparative survival differential was for oesophageal cancer.ConclusionFirst Nations’ survival inequalities are largely independent of geographical remoteness. It remains a priority to determine the contribution of other potential factors such as the availability of culturally acceptable diagnostic, management and/or support services.

The duration of protection against clinical malaria provided by the combination of seasonal RTS,S/AS01E vaccination and seasonal malaria chemoprevention versus either intervention given alone

BackgroundA recent trial of 5920 children in Burkina Faso and Mali showed that the combination of seasonal vaccination with the RTS,S/AS01E malaria vaccine (primary series and two seasonal boosters) and seasonal malaria chemoprevention (four monthly cycles per year) was markedly more effective than either intervention given alone in preventing clinical malaria, severe malaria, and deaths from malaria.MethodsIn order to help optimise the timing of these two interventions, trial data were reanalysed to estimate the duration of protection against clinical malaria provided by RTS,S/AS01E when deployed seasonally, by comparing the group who received the combination of SMC and RTS,S/AS01E with the group who received SMC alone. The duration of protection from SMC was also estimated comparing the combined intervention group with the group who received RTS,S/AS01E alone. Three methods were used: Piecewise Cox regression, Flexible parametric survival models and Smoothed Schoenfeld residuals from Cox models, stratifying on the study area and using robust standard errors to control for within-child clustering of multiple episodes.ResultsThe overall protective efficacy from RTS,S/AS01E over 6 months was at least 60% following the primary series and the two seasonal booster doses and remained at a high level over the full malaria transmission season. Beyond 6 months, protective efficacy appeared to wane more rapidly, but the uncertainty around the estimates increases due to the lower number of cases during this period (coinciding with the onset of the dry season). Protection from SMC exceeded 90% in the first 2–3 weeks post-administration after several cycles, but was not 100%, even immediately post-administration. Efficacy begins to decline from approximately day 21 and then declines more sharply after day 28, indicating the importance of preserving the delivery interval for SMC cycles at a maximum of four weeks.ConclusionsThe efficacy of both interventions was highest immediately post-administration. Understanding differences between these interventions in their peak efficacy and how rapidly efficacy declines over time will help to optimise the scheduling of SMC, malaria vaccination and the combination in areas of seasonal transmission with differing epidemiology, and using different vaccine delivery systems.Trial registrationThe RTS,S-SMC trial in which these data were collected was registered at clinicaltrials.gov: NCT03143218

Preterm birth, abnormal foetal growth and the risk of atrial fibrillation later in life: a cohort study in Denmark, Finland and Sweden

Abstract Background Increasing evidence suggests that adverse birth outcomes, including preterm birth, small for gestational age (SGA) and large for gestational age (LGA), are associated with increased risks of hypertension, ischemic heart disease, stroke and heart failure. However, knowledge regarding their associations with atrial fibrillation (AF) is limited and inconsistent. Objective To investigate whether preterm birth, SGA and LGA are associated with increased risks of AF later in life. Methods We conducted a population-based cohort study involving 8,012,433 live singleton births based on the nation-wide Danish (1978–2016), Swedish (1973–2014) and Finnish (1987–2014) Medical Birth Registers. Information on birth outcomes, atrial fibrillation and covariates was obtained from nationwide health and socioeconomic registers. We estimated hazard ratios (HR) and 95% confidence intervals (CI) for AF according to preterm birth (&amp;lt;37 weeks), SGA and LGA (&amp;lt;10th and &amp;gt;90th birth weight for gestational age percentiles, respectively) with multivariable Cox proportional hazard models and flexible parametric survival models. We conducted sibling analyses to control for unmeasured familial risk factors. Results Preterm birth and LGA were associated with increased AF risks in both the full population cohort and in the sibling analyses; the multivariate HRs (95% CI) from the cohort analyses were 1.53 (1.37–1.71) and 1.55 (1.44–1.65), respectively. The associations were stronger with AF in childhood than in adulthood. Children born SGA had an increased risk of AF in the first 18 years of life, but not later. Conclusions Individuals born preterm or LGA had an increased AF risk independently of familial confounding factors. Individuals born SGA had an increased AF risk only during childhood. Persons born prematurely or with abnormal foetal growth may benefit from early screening and prevention to reduce the risk of AF later in life. Funding Acknowledgement Type of funding sources: Public grant(s) – National budget only. Main funding source(s): Swedish Council for Working Life and Social Research; China Scholarship Council

T065: Reproduction patterns among Classical Hodgkin Lymphoma Survivors Treated with BEACOPP and ABVD in Sweden, Denmark, and Norway

Figure 1: Aalen-Johansen estimates of the cause-specific cumulative incidence (CIF) of childbirth estimated in the presence of competing risks of death. Background: Reproduction in classical Hodgkin lymphoma (cHL) survivors has been shown to be reduced compared to the general population, likely due to multi-agent chemotherapy. Understanding if contemporary treatment protocols are associated with reduced reproduction is important as treatment guidelines shift towards more liberal use of more intensive chemotherapy. Methods: We identified 2,937 individuals aged 18–40 years with cHL in Swedish and Danish lymphoma registers, and in the clinical database at Oslo University Hospital (OUH) between 1995 and 2019, who were linked to national medical birth registers in each country. Cox regression adjusted for stage, performance status, year and age at diagnosis was used to estimate hazard ratios (HRs) contrasting time to first childbirth by treatment groups (ABVD, 2–4 BEACOPP, 6–8 BEACOPP) up to ten years after diagnosis. Cause-specific cumulative incidence (CIF) of childbirths was further estimated using flexible parametric survival models. All analyses included an interaction between cHL treatment and sex. Results: Overall, 71.4% of patients were treated with ABVD, 3.6% with 2–4 BEACOPP, and 10.3% with 6–8 BEACOPP. The reproduction rates (per 1,000 person-years) were 45.5 (males) and 48.2 (females) in the ABVD group, and 23.8 (males) and 39.9 (females) in the 6–8 BEACOPP group. The adjusted HR comparing reproduction rates in individuals treated with 6–8 BEACOPP to ABVD was 0.51 (95% CI 0.35–0.73) for males and 0.82 (95% CI 0.57–1.19) for females. The CIF after 10 years was 20.0% (CI: 14.7%–27.2%) for males and 33.0% (CI: 24.6%–44.3%) for females treated with 6–8 BEACOPP. Conclusion: We found that BEACOPP treatment is associated with decreased reproduction rates compared to ABVD in male cHL patients. Infertility counselling should be prioritized for this group. This abstract has been accepted and presented at the EHA 2022 congress.

Lung Cancer Absolute Risk Models for Mortality in an Asian Population using the China Kadoorie Biobank.

Lung cancer is the leading cause of cancer mortality globally. Early detection through risk-based screening can markedly improve prognosis. However, most risk models were developed in North American cohorts of smokers, whereas less is known about risk profiles for never-smokers, which represent a growing proportion of lung cancers, particularly in Asian populations. Based on the China Kadoorie Biobank, a population-based prospective cohort of 512 639 adults with up to 12 years of follow-up, we built Asian Lung Cancer Absolute Risk Models (ALARM) for lung cancer mortality using flexible parametric survival models, separately for never and ever-smokers, accounting for competing risks of mortality. Model performance was evaluated in a 25% hold-out test set using the time-dependent area under the curve and by comparing model-predicted and observed risks for calibration. Predictors assessed in the never-smoker lung cancer mortality model were demographics, body mass index, lung function, history of emphysema or bronchitis, personal or family history of cancer, passive smoking, and indoor air pollution. The ever-smoker model additionally assessed smoking history. The 5-year areas under the curve in the test set were 0.77 (95% confidence interval = 0.73 to 0.80) and 0.81 (95% confidence interval = 0.79 to 0.84) for ALARM-never-smokers and ALARM-ever smokers, respectively. The maximum 5-year risk for never and ever-smokers was 2.6% and 12.7%, respectively. This study is among the first to develop risk models specifically for Asian populations separately for never and ever-smokers. Our models accurately identify Asians at high risk of lung cancer death and may identify those with risks exceeding common eligibility thresholds who may benefit from screening.

Flexible parametric survival analysis with multiple timescales: Estimation and implementation using stmt

In this article, we describe methodology that allows for multiple timescales using flexible parametric survival models without the need for time splitting. When one fits flexible parametric survival models on the log-hazard scale, numerical integration is required in the log likelihood to fit the model. The use of numerical integration allows incorporation of arbitrary functions of time into the model and hence lends itself to the inclusion of multiple timescales in an appealing way. We describe and exemplify these methods and show how to use the command stmt , which implements these methods, alongside its postestimation commands.

Perioperative Beta-Blocker Supply and Survival in Women With Epithelial Ovarian Cancer and a History of Cardiovascular Conditions.

Surgery for epithelial ovarian cancer (EOC) may activate stress-inflammatory responses that stimulate tumor growth and increase metastatic growth. Animal and in vitro studies have shown that inhibition of the catecholamine-induced inflammatory response via beta-adrenergic receptor blockade has antitumor potential in EOC. However, observational studies have reported mixed results. We assessed whether beta-blocker (BB) use at the time of primary ovarian cancer surgery was associated with improved survival in a large population-based study. Using linked administrative data, a population-based cohort of 3,844 Australian women age 50 years or older with a history of cardiovascular conditions who underwent surgery for EOC was followed for survival outcomes. The average treatment effect of selective BB (SBB) and nonselective BB (NSBB) supply at the time of surgery on survival was estimated from a causal inference perspective using covariate-balanced inverse probability of treatment weights with flexible parametric survival models that allowed for time-varying survival effects. Around the time of surgery, 560 (14.5%) women were supplied a SBB and 67 (1.7%) were supplied a NSBB. At 2 years postsurgery, the survival proportion was 80% (95% CI, 68 to 88) for women dispensed NSBBs at surgery compared with 69% (95% CI, 67 to 70) for women not supplied NSBBs. The survival advantage appeared to extend to at least 8 years postsurgery. No association was observed for women dispensed a SBB around the time of surgery. Perioperative supply of NSBBs appeared to confer a survival advantage for women age over 50 years with a history of cardiovascular conditions. Long-term clinical trials are required to confirm these findings.

Long-term survival and life expectancy following an acute heart failure hospitalization in Australia and New Zealand.

Contemporary long-term survival following a heart failure (HF) hospitalization is uncertain. We evaluated survival up to 10 years after a HF hospitalization using national data from Australia and New Zealand, identified predictors of survival, and estimated the attributable loss in life expectancy. Patients hospitalized with a primary diagnosis of HF from 2008-2017 were identified and all-cause mortality assessed by linking with Death Registries. Flexible parametric survival models were used to estimate survival, predictors of survival and loss in life expectancy. A total of 283 048 patients with HF were included (mean age 78.2 ± 12.3 years, 50.8% male). Of these, 48.3% (48.1-48.5) were surviving by 3 years, 34.1% (33.9-34.3) by 5 years and 17.1% (16.8-17.4) by 10 years (median survival 2.8 years). Survival declined with age with 53.4% of patients aged 18-54 years and 6.2% aged ≥85 years alive by 10 years (adjusted hazard ratio [aHR] for mortality 4.84, 95% confidence interval [CI] 4.65-5.04 for ≥85 years vs. 18-54 years) and was worse in male patients (aHR 1.14, 95% CI 1.13-1.15). Prior HF (aHR 1.20, 95% CI 1.18-1.22), valvular and rheumatic heart disease (aHR 1.11, 95% CI 1.10-1.13) and vascular disease (aHR 1.07, 95% CI 1.04-1.09) were cardiovascular comorbidities most strongly associated with long-term death. Non-cardiovascular comorbidities and geriatric syndromes were common and associated with higher mortality. Compared with the general population, HF was associated with a loss of 7.3 years in life expectancy (or 56.6% of the expected life expectancy) and reached 20.5 years for those aged 18-54 years. Less than one in five patients hospitalized for HF were surviving by 10 years with patients experiencing almost 60% loss in life expectancy compared with the general population, highlighting the considerable persisting societal burden of HF. Concerted multidisciplinary efforts are needed to improve post-hospitalization outcomes of HF.

A prognostic survival model for women diagnosed with invasive breast cancer in Queensland, Australia

PurposePrognostic models can help inform patients on the future course of their cancer and assist the decision making of clinicians and patients in respect to management and treatment of the cancer. In contrast to previous studies considering survival following treatment, this study aimed to develop a prognostic model to quantify breast cancer-specific survival at the time of diagnosis.MethodsA large (n = 3323), population-based prospective cohort of women were diagnosed with invasive breast cancer in Queensland, Australia between 2010 and 2013, and followed up to December 2018. Data were collected through a validated semi-structured telephone interview and a self-administered questionnaire, along with data linkage to the Queensland Cancer Register and additional extraction from medical records. Flexible parametric survival models, with multiple imputation to deal with missing data, were used.ResultsKey factors identified as being predictive of poorer survival included more advanced stage at diagnosis, higher tumour grade, “triple negative” breast cancers, and being symptom-detected rather than screen detected. The Harrell’s C-statistic for the final predictive model was 0.84 (95% CI 0.82, 0.87), while the area under the ROC curve for 5-year mortality was 0.87. The final model explained about 36% of the variation in survival, with stage at diagnosis alone explaining 26% of the variation.ConclusionsIn addition to confirming the prognostic importance of stage, grade and clinical subtype, these results highlighted the independent survival benefit of breast cancers diagnosed through screening, although lead and length time bias should be considered. Understanding what additional factors contribute to the substantial unexplained variation in survival outcomes remains an important objective.

Associations between frailty trajectories and cardiovascular, renal, and mortality outcomes in chronic kidney disease.

BackgroundFrailty is characterized by the loss of biological reserves and vulnerability to adverse outcomes. In individuals with chronic kidney disease (CKD), numerous pathophysiological factors may be responsible for frailty development including inflammation, physical inactivity, reduced energy intake, and metabolic acidosis. Given that both CKD and frailty incur a significant healthcare burden, it is important to understand the relationship of CKD and frailty in real‐world routine clinical practice, and how simple frailty assessment methods (e.g. frailty indexes) may be useful. We investigated the risk of frailty development in CKD and the impact of frailty status on mortality and end‐stage kidney disease (ESKD).MethodsA retrospective cohort study using primary care records from the Clinical Practice Research Datalink linked to Hospital Episode Statistics and the UK Office for National Statistics was undertaken in 819 893 participants aged ≥40 years, of which 140 674 had CKD. Frailty was defined using an electronic frailty index, generated electronically from primary care records. Cox proportional hazard and flexible parametric survival models were used to investigate the risk of developing frailty and the effect of frailty on risk of all‐cause and cardiovascular mortality, and ESKD.ResultsThe mean age of those with CKD was 77.5 (SD 9.7) years [61.0 (SD 12.1) years in no‐CKD group]; 62.0% of the CKD group were female (compared with 53.3% in no‐CKD group). The mean estimated glomerular filtration rate of those with CKD was 46.1 (SD 9.9) mL/min/1.73 m2. The majority of those with CKD (75.3%) were frail [vs. 45.4% in those without CKD (no‐CKD)]. Over 3 years (median), 69.5% of those with CKD developed frailty. Compared with no‐CKD, those with CKD had increased rates of developing mild (hazard ratio: 1.02; 95% confidence interval: 1.01–1.04), moderate (1.30; 1.26–1.34), and severe (1.50; 1.37–1.65) frailty. Mild (1.22; 1.19–1.24), moderate (1.60; 1.56–1.63), and severe (2.16; 2.11–2.22) frailty was associated with increased rates of all‐cause and cardiovascular‐related mortality (mild 1.35; 1.31–1.39; moderate 1.96; 1.90–2.02; and severe 2.91; 2.81–3.02). All stages of frailty significantly increased ESKD rates.ConclusionsFrailty is highly prevalent and associated with adverse outcomes in people with CKD, including mortality and risk of ESKD. Preventative interventions should be initiated to mitigate the development of frailty. The use of a simple frailty index, generated electronically from health records, can predict outcomes and may aid prioritization for management of people with frailty.

MSR62 Estimating Life Expectancy By Extrapolating Relative Survival: Comparison of the Flexible Parametric Model and the Rolling Extrapolation Algorithm Using Real-World Cancer Data

In health technology assessments, life expectancy estimates are often used for evaluating the effectiveness of interventions. Due to limited follow-up data, parametric models are typically used to fit patient-level data but yield inaccurate extrapolation until lifetime. We aim to compare two relatively new approaches, the flexible parametric relative survival model (FPM) by Andersson et al. (2013) and the rolling extrapolation algorithm (REA) by Hwang et al. (2017), which incorporate relative survival framework into their extrapolation processes with the intention to provide more robust extrapolated survival for estimating life expectancy.