Surveillance after surgery for pancreatic cancer: a global scoping review of guidelines and a nordic Survey of contemporary practice.

OBJECTIVES: Most patients with pancreatic cancer who have undergone surgical resection eventually develop disease recurrence. |This study aimed to investigate whether there is evidence to support routine surveillance after pancreatic cancer surgery, with a secondary aim of analyzing the implementation of surveillance strategies in the Nordic countries. MATERIALS AND METHODS: A scoping review was conducted to identify clinical practice guidelines globally and research studies relating to surveillance after pancreatic cancer resection. This was followed by a survey among 20 pancreatic units from four Nordic countries to assess their current practice of follow-up for operated patients. RESULTS: Altogether 16 clinical practice guidelines and 17 research studies were included. The guidelines provided inconsistent recommendations regarding postoperative surveillance of pancreatic cancer. The clinical research data were mainly based on retrospective cohort studies with low level of evidence and lead-time bias was not addressed. Active surveillance was recommended in Sweden and Denmark, but not in Norway beyond the post-operative/adjuvant period. Finland had no national recommendations for surveillance. The Nordic survey revealed a wide variation in reported practice among the different units. About 75% (15 of 20 units) performed routine postoperative surveillance. Routine CA 19-9 testing was used by 80% and routine CT by 67% as part of surveillance. About 73% of centers continued follow-up until 5 years postoperatively. CONCLUSION: Evidence for routine long-term (i.e. 5 years) surveillance after pancreatic cancer surgery remains limited. Most pancreatic units in the Nordic countries conduct regular follow-up, but protocols vary.

Positive Impact of [18F]FDG-PET/CT on Mortality in Patients With Staphylococcus aureus Bacteremia Explained by Immortal Time Bias.

BACKGROUND: Several studies have suggested that in patients with Staphylococcus aureus bacteremia (SAB) [18F] fluorodeoxyglucose positron emission tomography/computed tomography ([18F]FDG-PET/CT) improves outcome. However, these studies often ignored possible immortal time bias. METHODS: Prospective multicenter cohort study in 2 university and 5 non-university hospitals, including all patients with SAB. [18F]FDG-PET/CT was performed on clinical indication as part of usual care. Primary outcome was 90-day all-cause mortality. Effect of [18F]FDG-PET/CT was modeled with a Cox proportional hazards model using [18F]FDG-PET/CT as a time-varying variable and corrected for confounders for mortality (age, Charlson score, positive follow-up cultures, septic shock, and endocarditis). Secondary outcome was 90-day infection-related mortality (assessed by adjudication committee) using the same analysis. In a subgroup-analysis, we determined the effect of [18F]FDG-PET/CT in patients with high risk of metastatic infection. RESULTS: Of 476 patients, 178 (37%) underwent [18F]FDG-PET/CT. Day-90 all-cause mortality was 31% (147 patients), and infection-related mortality was 17% (83 patients). The confounder adjusted hazard ratio (aHR) for all-cause mortality was 0.50 (95% confidence interval [CI]: .34-.74) in patients that underwent [18F]FDG-PET/CT. Adjustment for immortal time bias changed the aHR to 1.00 (95% CI .68-1.48). Likewise, after correction for immortal time bias, [18F]FDG-PET/CT had no effect on infection-related mortality (cause specific aHR 1.30 [95% CI .77-2.21]), on all-cause mortality in patients with high-risk SAB (aHR 1.07 (95% CI .63-1.83) or on infection-related mortality in high-risk SAB (aHR for 1.24 [95% CI .67-2.28]). CONCLUSIONS: After adjustment for immortal time bias [18F]FDG-PET/CT was not associated with day-90 all-cause or infection-related mortality in patients with SAB.

Immortal Time Bias With Time-Varying Exposures in Environmental Epidemiology: A Case Study in Lung Cancer Survival.

Immortal time bias is a well-recognized bias in clinical epidemiology but is rarely discussed in environmental epidemiology. Under the target trial framework, this bias is formally conceptualized as a misalignment between the start of study follow-up (time 0) and treatment assignment. This misalignment can occur when attained duration of follow-up is encoded into treatment assignment using minimums, maximums, or averages. The bias can be exacerbated in the presence of time trends commonly found in environmental exposures. Using lung cancer cases from the California Cancer Registry (2000-2010) linked with estimated concentrations of particulate matter less than or equal to 2.5 µm in aerodynamic diameter (PM2.5), we replicated previous studies that averaged PM2.5 exposure over follow-up in a time-to-event model. We compared this approach with one that ensures alignment between time 0 and treatment assignment, a discrete-time approach. In the former approach, the estimated overall hazard ratio for a 5-µg/m3 increase in PM2.5 was 1.38 (95% confidence interval: 1.36, 1.40). Under the discrete-time approach, the estimated pooled odds ratio was 0.99 (95% confidence interval: 0.98, 1.00). We conclude that the strong estimated effect in the former approach was likely driven by immortal time bias, due to misalignment at time 0. Our findings highlight the importance of appropriately conceptualizing a time-varying environmental exposure under the target trial framework to avoid introducing preventable systematic errors.

Prognostic impact of severe neutropenia in colorectal cancer patients treated with TAS-102 and bevacizumab, addressing immortal-time bias.

BACKGROUND: Several studies have reported an association between severe neutropenia and long-term survival in patients treated with trifluridine-tipiracil (TAS-102). Because some of these studies failed to address immortality time bias, however, their findings should be interpreted with caution. Additionally, the association between severe neutropenia and survival in patients receiving TAS-102 in combination with bevacizumab (Bmab) remains unclear. PATIENTS AND METHODS: We conducted a single-center retrospective cohort study in patients with colorectal cancer who received Bmab + TAS-102. We compared overall survival (OS) between patients who developed grade ≥ 3 neutropenia during the treatment period and those who did not. To account for immortal time bias, we used two approaches, time-varying Cox regression and landmark analysis. RESULTS: Median OS was 15.3 months [95% CI: 14.1-NA] in patients with grade ≥ 3 neutropenia and 10.0 months [95% CI: 8.1-NA] in those without. In time-varying Cox regression, onset grade ≥ 3 neutropenia was significantly related to longer survival after adjustment for age and modified Glasgow Prognostic Score. Additionally, 30-, 60-, 90-, and 120-day landmark analysis showed that grade ≥ 3 neutropenia was associated with longer survival after adjustment for age and modified Glasgow Prognostic Score, with respective HRs of 0.30 [0.10-0.90], 0.65 [0.30-1.42], 0.39 [0.17-0.90], and 0.41 [0.18-0.95]. CONCLUSION: We identified an association between long-term survival and the development of severe neutropenia during the early cycle of Bmab + TAS-102 using an approach that addressed immortality time bias.

Metanalyses on metformin's role in pancreatic cancer suffer from severe bias and low data quality - An umbrella review.

AIMS: Discrepancies in the preclinical evidence, retrospective studies, and randomized trials evaluating metformin's role in pancreatic cancer are difficult to disentangle. We aimed to critically and systematically examine the quality and sources of heterogeneity between meta-analyses investigating the association between metformin intake and the prognosis of patients with pancreatic cancer. MATERIALS AND METHODS: We performed a literature search on PubMed, MEDLINE, Embase, and Scopus on October 31, 2021 to identify meta-analyses investigating the impact of metformin treatment on the prognosis of patients with pancreatic cancer. Meta-analyses quality was assessed using according to the AMSTAR 2 criteria. We assessed bias in individual studies included in the meta-analyses, with particular attention to immortal time bias and quality of reporting. RESULTS: Eleven meta-analyses describing 24 individual studies were included. All meta-analyses were rated low (n = 5) or critically low (n = 6) quality. Only 4 followed PRISMA reporting guidelines and only in 5 presented data were sufficient to replicate the analyses. Most meta-analyses combined results from clinical trials and retrospective studies (n = 6); patients with different cancer stages (resectable vs advanced) and from studies with different control group definitions. Immortal time bias was present and not accounted for in most (65.2%) of the individual retrospective studies; almost all (n = 9) meta-analyses failed to identify and correct for this source of bias. CONCLUSIONS: Meta-analyses describing the association of metformin use in patients with pancreatic cancer are plagued by various types of bias inherent in retrospective studies. The quality of evidence linking metformin to decreased pancreatic cancer mortality is generally low.

Incidental Findings and Low-Value Care.

Incidental imaging findings are common and analogous to the results of screening tests when screening is performed of unselected, low-risk patients. Approximately 15-30% of all diagnostic imaging and 20-40% of CT examinations contain at least one incidental finding. Patients with incidental findings but low risk for disease are likely to experience length bias, lead-time bias, overdiagnosis, and overtreatment that create an illusion of benefit while conferring harm. This includes incidental detection of many types of cancers that, although malignant, would have been unlikely to affect a patient's health had the cancer not been detected. Detection of some incidental findings can improve health, but most do not. Greater patient- and disease-related risk increase the likelihood an incidental finding is important. Clinical guidelines for incidental findings should more deeply integrate patient risk factors and disease aggressiveness to inform management. Lack of outcome and cost-effectiveness data has led to reflexive management strategies for incidental findings that promote low-value and sometimes harmful care.

Core concepts in pharmacoepidemiology: Key biases arising in pharmacoepidemiologic studies.

Pharmacoepidemiology has an increasingly important role in informing and improving clinical practice, drug regulation, and health policy. Therefore, unrecognized biases in pharmacoepidemiologic studies can have major implications when study findings are translated to the real world. We propose a simple taxonomy for researchers to use as a starting point when thinking through some of the most pervasive biases in pharmacoepidemiology. We organize this discussion according to biases best assessed with respect to the study population (including confounding by indication, channeling bias, healthy user bias, and protopathic bias), the study design (including prevalent user bias and immortal time bias), and the data source (including misclassification bias and missing data/loss to follow up). This tutorial defines, provides a curated list of recommended references, and illustrates through relevant case examples these key biases to consider when planning, conducting, or evaluating pharmacoepidemiologic studies.

Immeasurable Time Bias in Self-controlled Designs: Case-crossover, Case-time-control, and Case-case-time-control Analyses.

BACKGROUND: Impact of immeasurable time bias (IMTB) is yet to be examined in self-controlled designs. METHODS: We conducted case-crossover, case-time-control, and case-case-time-control analyses using Korea's healthcare database. Two empirical examples among elderly patients were used: 1) benzodiazepines-hip fracture; 2) benzodiazepines-mortality. For cases, the date of hip fracture diagnosis or death was defined as the index date, and the inherited date of their matched cases for controls or future cases. Exposure was assessed in the 1-30 day (hazard) and 61-90 day (control) windows preceding the index date. A non-missing exposure setting included in- and outpatient prescriptions and the pseudo-outpatient setting included only the outpatients. Conditional logistic regression was done to estimate odds ratios (ORs) with 95% confidence intervals (CIs), where the relative difference in OR among the two settings was calculated to quantify the IMTB. RESULTS: The IMTB had negligible impacts in the hip fracture example in the case-crossover (non-missing exposure setting OR 1.27; 95% CI, 1.12-1.44; pseudo-outpatient setting OR 1.21; 95% CI, 1.06-1.39; magnitude 0.05), case-time-control (OR 1.18; 95% CI, 0.98-1.44; OR 1.13; 95% CI, 0.92-1.38; 0.04, respectively), and case-case-time-control analyses (OR 0.99; 95% CI, 0.80-1.23; OR 0.94; 95% CI, 0.75-1.18; 0.05, respectively). In the mortality example, IMTB had significant impacts in the case-crossover (non-missing exposure setting OR 1.44; 95% CI, 1.36-1.52; pseudo-outpatient setting OR 0.72; 95% CI, 0.67-0.78; magnitude 1.00), case-time-control (OR 1.38; 95% CI, 1.26-1.51; OR 0.68; 95% CI, 0.61-0.76; 1.03, respectively), and case-case-time-control analyses (OR 1.27; 95% CI, 1.15-1.40; OR 0.62; 95% CI, 0.55-0.69; 1.05, respectively). CONCLUSION: Although IMTB had negligible impacts on the drug's effect on acute events, as these are unlikely to be accompanied with hospitalizations, it negatively biased the drug's effect on mortality, an outcome with prodromal phases, in the three self-controlled designs.

ICU triage decisions and biases about time and identity.

We often show a greater inclination to assist and avoid harming people identified as those at high risk of great harm than to assist and avoid harming people who will suffer similar harm but are not identified (as yet). Call this the identified person bias. Some ethicists think such bias is justified; others disagree and claim that the bias is discriminatory against statistical people. While the issue is present in public policy and politics, perhaps the most notable examples can be found in medical ethics such as in ICU triage decisions made during the COVID-19 pandemic. The Rule of Rescue, as the application of the identified person bias is sometimes called, states that spending large amounts of resources rescuing identifiable individuals who are in imminent danger is justified. In this paper, I show that our distorted attitudes toward time play a role in identified person bias. I claim that ICU triage decisions are better explained by a preference to treat people sooner rather than later, which is at least partially due to near bias (positive events are to be preferred to be near rather than distant), than by a preference to treat identified lives over statistical lives. Thus, another bias, near bias, is involved in the reasoning behind the identified person bias and the Rule of Rescue.

Statins and Mortality in COPD: A Methodological Review of Observational Studies.

Randomized controlled trials and observational studies have reported conflicting results on the potential beneficial effects of statins on mortality in patients with chronic obstructive pulmonary disease (COPD). We performed a systematic search of the literature to review all observational studies reporting relative risks of death with statin use in COPD, focusing on potential sources of bias. We identified 15 observational studies, out of 2835, of which 12 were affected by time-related and other biases and the remaining 3 by confounding bias. All 15 studies were also subject to confounding bias due to lack of adjustment for important COPD-related factors. The risk of death associated with statin use was reduced across all 15 studies (pooled relative risk (PRR) 0.66; 95% CI: 0.59-0.74). The reduction was observed in 7 studies with immortal time bias (PRR 0.62; 95%: 0.53-0.72), two with collider-stratification bias (PRR 0.60; 95% CI: 0.45-0.80), one with time-window bias (RR 0.61; 95% CI: 0.38-0.98), one with immeasurable time bias (RR 0.50; 95% CI: 0.40-0.62), and one with exposure misclassification (RR 0.86; 95% CI: 0.72-1.03). The three studies that avoided these biases were, however, affected by confounding bias resulting in a PRR of 0.77 (95% CI: 0.61-0.98). In conclusion, the observational studies investigating statin use and mortality in COPD are affected by major biases, many of which can result in spurious protective effects. Well-designed observational studies that carefully emulate randomized trials are needed to resolve this uncertainty regarding the potential beneficial benefits of statins on mortality in patients with COPD.

Left truncation in linked data: A practical guide to understanding left truncation and applying it using SAS and R.

Time-to-event data such as time to death are broadly used in medical research and drug development to understand the efficacy of a therapeutic. For time-to-event data, right censoring (data only observed up to a certain point of time) is common and easy to recognize. Methods that use right censored data, such as the Kaplan-Meier estimator and the Cox proportional hazard model, are well established. Time-to-event data can also be left truncated, which arises when patients are excluded from the sample because their events occur before a specific milestone, potentially resulting in an immortal time bias. For example, in a study evaluating the association between biomarker status and overall survival, patients who did not live long enough to receive a genomic test were not observed in the study. Left truncation causes selection bias and often leads to an overestimate of survival time. In this tutorial, we used a nationwide electronic health record-derived de-identified database to demonstrate how to analyze left truncated and right censored data without bias using example code from SAS and R.

Assessing lead time bias due to mammography screening on estimates of loss in life expectancy.

BACKGROUND: An increasingly popular measure for summarising cancer prognosis is the loss in life expectancy (LLE), i.e. the reduction in life expectancy following a cancer diagnosis. The proportion of life lost (PLL) can also be derived, improving comparability across age groups as LLE is highly age-dependent. LLE and PLL are often used to assess the impact of cancer over the remaining lifespan and across groups (e.g. socioeconomic groups). However, in the presence of screening, it is unclear whether part of the differences across population groups could be attributed to lead time bias. Lead time is the extra time added due to early diagnosis, that is, the time from tumour detection through screening to the time that cancer would have been diagnosed symptomatically. It leads to artificially inflated survival estimates even when there are no real survival improvements. METHODS: In this paper, we used a simulation-based approach to assess the impact of lead time due to mammography screening on the estimation of LLE and PLL in breast cancer patients. A natural history model developed in a Swedish setting was used to simulate the growth of breast cancer tumours and age at symptomatic detection. Then, a screening programme similar to current guidelines in Sweden was imposed, with individuals aged 40-74 invited to participate every second year; different scenarios were considered for screening sensitivity and attendance. To isolate the lead time bias of screening, we assumed that screening does not affect the actual time of death. Finally, estimates of LLE and PLL were obtained in the absence and presence of screening, and their difference was used to derive the lead time bias. RESULTS: The largest absolute bias for LLE was 0.61 years for a high screening sensitivity scenario and assuming perfect screening attendance. The absolute bias was reduced to 0.46 years when the perfect attendance assumption was relaxed to allow for imperfect attendance across screening visits. Bias was also present for the PLL estimates. CONCLUSIONS: The results of the analysis suggested that lead time bias influences LLE and PLL metrics, thus requiring special consideration when interpreting comparisons across calendar time or population groups.

Immune-related adverse events in urothelial cancer patients: Adjustment for immortal time bias.

To investigate the association between the onset, severity, and type of immune-related adverse events (irAEs) and the efficacy of pembrolizumab in patients with platinum-pretreated advanced urothelial carcinoma (UC), we retrospectively collected clinical datasets of 755 patients and conducted landmark analysis. Patients who survived for fewer than 3 months were excluded from the evaluation to reduce the immortal time bias. In total, 620 patients were evaluated, of whom 220 patients (35.5%) experienced grade ≥2 irAEs, including 134 patients with grade 2 irAEs and 86 with grade ≥3 irAEs. Propensity score matching extracted 198 patients with and without grade ≥2 irAEs. The onset of grade ≥2 irAEs was associated with longer median progression-free survival (PFS) (8.3 months vs. 4.5 months, p = 0.003) and overall survival (OS) (20.4 months vs. 14.3 months, p = 0.031) and a higher objective response rate (ORR) (44.8% vs. 30.2%, p = 0.004). Patients with grade 2 irAEs had significantly better oncological outcomes (PFS, OS, and ORR) than grade ≤1 and ≥3 irAEs. Patients with grade ≥3 irAEs had worse outcomes than grade 2 irAEs. Endocrine and skin irAEs were related with better survival outcomes, and the rate of severities was lower in these categories. In conclusion, the occurrence of irAEs, particularly low-grade irAEs, was predictive of pembrolizumab efficacy in patients with platinum-pretreated advanced UC.

Development and evaluation of safety and effectiveness of novel cancer screening tests for routine clinical use with applications to multicancer detection technologies.

Multicancer screening is a promising approach to improving the detection of preclinical disease, but current technologies have limited ability to identify precursor or early stage lesions, and approaches for developing the evidentiary chain are unclear. Frameworks to enable development and evaluation from discovery through evidence of clinical effectiveness are discussed.

Immortal time bias for life-long conditions in retrospective observational studies using electronic health records.

BACKGROUND: Immortal time bias is common in observational studies but is typically described for pharmacoepidemiology studies where there is a delay between cohort entry and treatment initiation. METHODS: This study used the Clinical Practice Research Datalink (CPRD) and linked national mortality data in England from 2000 to 2019 to investigate immortal time bias for a specific life-long condition, intellectual disability. Life expectancy (Chiang's abridged life table approach) was compared for 33,867 exposed and 980,586 unexposed individuals aged 10+ years using five methods: (1) treating immortal time as observation time; (2) excluding time before date of first exposure diagnosis; (3) matching cohort entry to first exposure diagnosis; (4) excluding time before proxy date of inputting first exposure diagnosis (by the physician); and (5) treating exposure as a time-dependent measure. RESULTS: When not considered in the design or analysis (Method 1), immortal time bias led to disproportionately high life expectancy for the exposed population during the first calendar period (additional years expected to live: 2000-2004: 65.6 [95% CI: 63.6,67.6]) compared to the later calendar periods (2005-2009: 59.9 [58.8,60.9]; 2010-2014: 58.0 [57.1,58.9]; 2015-2019: 58.2 [56.8,59.7]). Date of entry of diagnosis (Method 4) was unreliable in this CPRD cohort. The final methods (Method 2, 3 and 5) appeared to solve the main theoretical problem but residual bias may have remained. CONCLUSIONS: We conclude that immortal time bias is a significant issue for studies of life-long conditions that use electronic health record data and requires careful consideration of how clinical diagnoses are entered onto electronic health record systems.

Time-related biases in perinatal pharmacoepidemiology: A systematic review of observational studies.

BACKGROUND: Time-related biases, such as immortal time and time-window bias, frequently occur in pharmacoepidemiologic research. However, the prevalence of these biases in perinatal pharmacoepidemiology is not well understood. OBJECTIVE: To describe the frequency of time-related biases in observational studies of medications commonly used during pregnancy (antibiotic, antifungal, and antiemetic drugs) via systematic review. METHOD: We searched Medline and EMBASE for observational studies published between January 2013 and September 2020 and examining the association between antibiotic, antifungal, or antiemetic drugs and adverse pregnancy outcomes, including spontaneous abortion, stillbirth, preterm delivery, small-for-gestational age, pre-eclampsia, and gestational diabetes. The proportion of studies with time-related biases was estimated overall and by type (immortal time bias, time-window bias). RESULTS: Our systematic review included 20 studies (16 cohort studies, 3 nested case-control studies, and 1 case-control study), of which 12 examined antibiotic, 6 antiemetic, and 2 anti-fungal drugs. Eleven studies (55%) had immortal time bias due to the misclassification of unexposed, event-free person-time between cohort entry and exposure initiation as exposed. No included study had time-window bias. The direction of effect varied for both studies with and without time-related bias, with many studies reporting very wide confidence intervals around the effect estimates, thus making the direction of effect less interpretable. However, studies with time-related bias were more likely to show protective or null associations compared with studies without time-related bias. CONCLUSION: Time-related biases occur frequently in observational studies of drug effects during pregnancy. The use of appropriate study design and analytical approaches is needed to prevent time-related biases and ensure study validity.

Timing and severity of COVID-19 during pregnancy and risk of preterm birth in the International Registry of Coronavirus Exposure in Pregnancy.

BACKGROUND: Studies of preterm delivery after COVID-19 are often subject to selection bias and do not distinguish between early vs. late infection in pregnancy, nor between spontaneous vs. medically indicated preterm delivery. This study aimed to estimate the risk of preterm birth (overall, spontaneous, and indicated) after COVID-19 during pregnancy, while considering different levels of disease severity and timing. METHODS: Pregnant and recently pregnant people who were tested for or clinically diagnosed with COVID-19 during pregnancy enrolled in an international internet-based cohort study between June 2020 and July 2021. We used several analytic approaches to minimize confounding and immortal time bias, including multivariable regression, time-to-delivery models, and a case-time-control design. RESULTS: Among 14,264 eligible participants from 70 countries who did not report a pregnancy loss before 20 gestational weeks, 5893 had completed their pregnancies and reported delivery information; others were censored at time of their last follow-up. Participants with symptomatic COVID-19 before 20 weeks' gestation had no increased risk of preterm delivery compared to those testing negative, with adjusted risks of 10.0% (95% CI 7.8, 12.0) vs. 9.8% (9.1, 10.5). Mild COVID-19 later in pregnancy was not clearly associated with preterm delivery. In contrast, severe COVID-19 after 20 weeks' gestation led to an increase in preterm delivery compared to milder disease. For example, the risk ratio for preterm delivery comparing severe to mild/moderate COVID-19 at 35 weeks was 2.8 (2.0, 4.0); corresponding risk ratios for indicated and spontaneous preterm delivery were 3.7 (2.0, 7.0) and 2.3 (1.2, 3.9), respectively. CONCLUSIONS: Severe COVID-19 late in pregnancy sharply increased the risk of preterm delivery compared to no COVID-19. This elevated risk was primarily due to an increase in medically indicated preterm deliveries, included preterm cesarean sections, although an increase in spontaneous preterm delivery was also observed. In contrast, mild or moderate COVID-19 conferred minimal risk, as did severe disease early in pregnancy.

Nonaspirin nonsteroidal anti-inflammatory drugs and gastric cancer risk after Helicobacter pylori eradication: A territory-wide study.

BACKGROUND: Despite Helicobacter pylori (HP) eradication, individuals can still develop gastric cancer (GC). Prior studies have demonstrated that nonaspirin nonsteroidal anti-inflammatory drugs (NA-NSAIDs) reduce the risk of GC, but this may be caused by immortal time bias and failure to adjust for HP status. The objective of this study was to investigate whether NA-NSAIDs reduced the risk of GC in patients who undergo H. pylori eradication. METHODS: Adult patients who had received clarithromycin-based triple therapy between 2003 and 2016 were identified from a territory-wide health care database. Exclusion criteria included prior GC or GC diagnosed <6 months after HP eradication, prior gastrectomy, gastric ulcer after HP eradication, and failure of triple therapy. Covariates included age, sex, prior peptic ulcer disease, other comorbidities, and concurrent medications (aspirin, proton pump inhibitors, statins, and metformin). To avoid immortal time bias, NA-NSAID use (≥90 days) was treated as a time-dependent variable in a multivariable Cox model (time-dependent analysis). Time-independent analysis was also performed. RESULTS: During a median follow-up of 8.9 years (interquartile range, 5.4-12.6 years), 364 of 92,017 patients (0.4%) who underwent HP eradication developed GC. NA-NSAID use was associated with a significant reduction in the risk of GC in time-fixed analysis (adjusted hazard ratio [aHR], 0.65; 95% CI, 0.47-0.90), but not in time-dependent multivariable analysis (aHR, 1.35; 95% CI, 0.97-1.87). Time-dependent subgroup analyses also did not indicate any significant association between NA-NSAID use and either cardia GC (aHR, 0.75; 95% CI, 0.27-2.06) or noncardia GC (aHR, 1.28; 95% CI, 0.83-1.98). CONCLUSIONS: NA-NSAID use was not associated with a reduced risk of GC among patients who underwent HP eradication. The chemopreventive effect of NA-NSAIDs observed in prior studies may have been confounded by immortal time bias.

Immortal time bias exaggerates the effect of metformin on the risk of gastric cancer: A meta-analysis.

High heterogeneity has been reported among epidemiological studies exploring the relationship between metformin and the risk of gastric cancer. Immortal time bias might be one of the vital factors causing heterogeneity because of its widespread existence in pharmacological observational studies and it could severely exaggerate the drug's effectiveness. Immortal time bias could occur in an observational study if exposure status is determined based on a measurement or event that occurs after baseline. In this study, we aimed to assess whether immortal time bias is responsible for the false assumption that metformin reduces the risk of gastric cancer. We searched PubMed, Embase, Web of Science and Cochrane Library databases for relevant studies from the inception to August 9, 2020. The strength of the relationship was assessed using pooled relative risks (RRs) with corresponding 95% confidence intervals (95% CIs). Statistical analyses were carried out using a random-effects model. Pooled RR from 6 cohort studies with immortal time bias found a clear 33% reduced risk associated with metformin use (RR = 0.67, 95% CI = 0.59, 0.77; P < 0.001; I2 = 48.5%). However, pooled RR from 8 cohort studies without immortal time bias indicated no association between the use of metformin and gastric cancer risk (RR = 0.95, 95% CI = 0.85, 1.05; P = 0.317; I2 = 64.5%). From a univariate meta-regression model, the presence of immortal time bias was associated with a significant reduction of 29% in the effect estimate of metformin on gastric cancer risk (ratio of RR = 0.71, 95% CI = 0.58, 0.86; P = 0.002). This meta-analysis indicates that metformin use has no protective effect on gastric cancer risk. The relationship between metformin use and gastric cancer risk has been exaggerated as a result of the presence of immortal time bias. Further studies are required to confirm the results by controlling for immortal time bias based on appropriate study designs and statistical methods.

Analysis of time-to-event for observational studies: Guidance to the use of intensity models.

This paper provides guidance for researchers with some mathematical background on the conduct of time-to-event analysis in observational studies based on intensity (hazard) models. Discussions of basic concepts like time axis, event definition and censoring are given. Hazard models are introduced, with special emphasis on the Cox proportional hazards regression model. We provide check lists that may be useful both when fitting the model and assessing its goodness of fit and when interpreting the results. Special attention is paid to how to avoid problems with immortal time bias by introducing time-dependent covariates. We discuss prediction based on hazard models and difficulties when attempting to draw proper causal conclusions from such models. Finally, we present a series of examples where the methods and check lists are exemplified. Computational details and implementation using the freely available R software are documented in Supplementary Material. The paper was prepared as part of the STRATOS initiative.