Analysis of Clinical Cohort Data Using Nested Case-control and Case-cohort Sampling Designs. A Powerful and Economical Tool

Clonal hematopoiesis of indeterminate potential and diabetic kidney disease: a nested case-control study

Introduction

The disease trajectory of diabetic kidney disease (DKD) shows a high interindividual variability not sufficiently explained by conventional risk factors. Clonal hematopoiesis of indeterminate potential (CHIP) is a proposed novel cardiovascular risk factor. Increased kidney fibrosis and glomerulosclerosis were described in mouse models of CHIP. Here, we aim to analyze whether CHIP affects the incidence or progression of DKD.

Methods

A total of 1419 eligible participants of the PROVALID Study were the basis for a nested case-control (NCC) design. A total of 64 participants who reached a prespecified composite endpoint within the observation period (initiation of kidney replacement therapy, death from kidney failure, sustained 40% decline in estimated glomerular filtration rate or sustained progression to macroalbuminuria) were identified and matched to 4 controls resulting in an NCC sample of 294 individuals. CHIP was assessed via targeted amplicon sequencing of 46 genes in peripheral blood. Furthermore, inflammatory cytokines were analyzed in plasma via a multiplex assay.

Results

The estimated prevalence of CHIP was 28.91% (95% CI 22.91%–34.91%). In contrast to other known risk factors (albuminuria, hemoglobin A1c, heart failure, and smoking) and elevated microinflammation, CHIP was not associated with incident or progressive DKD (hazard ratio [HR] 1.06 [95% CI 0.57–1.96]).

Conclusions

In this NCC study, common risk factors as well as elevated microinflammation but not CHIP were associated with kidney function decline in type 2 diabetes mellitus.

Inverse Probability Weighting Enhances Absolute Risk Estimation in Three Common Study Designs of Nosocomial Infections

Purpose

When studying nosocomial infections, resource-efficient sampling designs such as nested case-control, case-cohort, and point prevalence studies are preferred. However, standard analyses of these study designs can introduce selection bias, especially when interested in absolute rates and risks. Moreover, nosocomial infection studies are often subject to competing risks. We aim to demonstrate in this tutorial how to address these challenges for all three study designs using simple weighting techniques.

Patients and Methods

We discuss the study designs and explain how inverse probability weights (IPW) are applied to obtain unbiased hazard ratios (HR), odds ratios and cumulative incidences. We illustrate these methods in a multi-state framework using a dataset from a nosocomial infections study (n = 2286) in Moscow, Russia.

Results

Including IPW in the analysis corrects the unweighted naïve analyses and enables the estimation of absolute risks. Resulting estimates are close to the full cohort estimates using substantially smaller numbers of patients.

Conclusion

IPW is a powerful tool to account for the unequal selection of controls in case-cohort, nested case-control and point prevalence studies. Findings can be generalized to the full population and absolute risks can be estimated. When applied to a multi-state model, competing risks are also taken into account.

Which patients to sample in clinical cohort studies when the number of events is high and measurement of additional markers is constrained by limited resources

PURPOSE

We consider an existing clinical cohort with events but limited resources for the investigation of a further potentially expensive marker. Biological material of the patients is stored in a biobank, but only a limited number of samples can be analyzed with respect to the marker. The question arises as to which patients to sample, if the number of events preclude standard sampling designs.

METHODS

Modifications of the nested case-control and the case-cohort design for the proportional hazards model are applied, that allow efficient sampling in situations where standard nested case-control and case-cohort are not feasible. These sampling designs are compared to simple random sampling and extreme group sampling, the latter including only patients with extreme outcomes, ie either with an event early in time or without an event until at least a point later in time.

RESULTS

The modified nested case-control design and the modified case-cohort design provide powerful methods for sampling in a clinical cohort with many events. The simple random sampling usually is less efficient. If focus is on precise estimation of a potential effect in terms of a hazard ratio, extreme group sampling is not competitive. If focus is on screening for important biomarkers, extreme group sampling markedly outperforms the other sampling designs.

CONCLUSIONS

When it is not feasible to sample all events, a modified nested case-control design or case-cohort design leads to efficient effect estimates in the proportional hazards model. If screening for important biomarkers is the primary objective, extreme group sampling is preferable.

Dynamic inference in general nested case-control designs

Nested case-control designs are attractive in studies with a time-to-event endpoint if the outcome is rare or if interest lies in evaluating expensive covariates. The appeal is that these designs restrict to small subsets of all patients at risk just prior to the observed event times. Only these small subsets need to be evaluated. Typically, the controls are selected at random and methods for time-simultaneous inference have been proposed in the literature. However, the martingale structure behind nested case-control designs allows for more powerful and flexible non-standard sampling designs. We exploit that structure to find simultaneous confidence bands based on wild bootstrap resampling procedures within this general class of designs. We show in a simulation study that the intended coverage probability is obtained for confidence bands for cumulative baseline hazard functions. We apply our methods to observational data about hospital-acquired infections.

Nested exposure case-control sampling: a sampling scheme to analyze rare time-dependent exposures

For large cohort studies with rare outcomes, the nested case-control design only requires data collection of small subsets of the individuals at risk. These are typically randomly sampled at the observed event times and a weighted, stratified analysis takes over the role of the full cohort analysis. Motivated by observational studies on the impact of hospital-acquired infection on hospital stay outcome, we are interested in situations, where not necessarily the outcome is rare, but time-dependent exposure such as the occurrence of an adverse event or disease progression is. Using the counting process formulation of general nested case-control designs, we propose three sampling schemes where not all commonly observed outcomes need to be included in the analysis. Rather, inclusion probabilities may be time-dependent and may even depend on the past sampling and exposure history. A bootstrap analysis of a full cohort data set from hospital epidemiology allows us to investigate the practical utility of the proposed sampling schemes in comparison to a full cohort analysis and a too simple application of the nested case-control design, if the outcome is not rare.

Serial Fibroblast Growth Factor 23 Measurements and Risk of Requirement for Kidney Replacement Therapy: The CRIC (Chronic Renal Insufficiency Cohort) Study

RATIONALE & OBJECTIVE

Studies using a single measurement of fibroblast growth factor 23 (FGF-23) suggest that elevated FGF-23 levels are associated with increased risk for requirement for kidney replacement therapy (KRT) in patients with chronic kidney disease. However, the data do not account for changes in FGF-23 levels as kidney disease progresses.

STUDY DESIGN

Case-cohort study.

SETTING & PARTICIPANTS

To evaluate the association between serial FGF-23 levels and risk for requiring KRT, our primary analysis included 1,597 individuals in the Chronic Renal Insufficiency Cohort Study who had up to 5 annual measurements of carboxy-terminal FGF-23. There were 1,135 randomly selected individuals, of whom 266 initiated KRT, and 462 individuals who initiated KRT outside the random subcohort.

EXPOSURE

Serial FGF-23 measurements and FGF-23 trajectory group membership.

OUTCOMES

Incident KRT.

ANALYTICAL APPROACH

To handle time-dependent confounding, our primary analysis of time-updated FGF-23 levels used time-varying inverse probability weighting in a discrete time failure model. To compare our results with prior data, we used baseline and time-updated FGF-23 values in weighted Cox regression models. To examine the association of FGF-23 trajectory subgroups with risk for incident KRT, we used weighted Cox models with FGF-23 trajectory groups derived from group-based trajectory modeling as the exposure.

RESULTS

In our primary analysis, the HR for the KRT outcome per 1 SD increase in the mean of natural log-transformed (ln)FGF-23 in the past was 1.94 (95% CI, 1.51-2.49). In weighted Cox models using baseline and time-updated values, elevated FGF-23 level was associated with increased risk for incident KRT (HRs per 1 SD ln[FGF-23] of 1.18 [95% CI, 1.02-1.37] for baseline and 1.66 [95% CI, 1.49-1.86] for time-updated). Membership in the slowly and rapidly increasing FGF-23 trajectory groups was associated with ∼3- and ∼21-fold higher risk for incident KRT compared to membership in the stable FGF-23 trajectory group.

LIMITATIONS

Residual confounding and lack of intact FGF-23 values.

CONCLUSIONS

Increasing FGF-23 levels are independently associated with increased risk for incident KRT.

Exposure density sampling: Dynamic matching with respect to a time-dependent exposure

Estimating the potential risk associated with an exposure occurring over time requires complex statistical techniques, since ignoring the time from study entry until the exposure leads to potentially seriously biased effect estimates. A prominent example is estimating the effect of hospital-acquired infections on adverse outcomes in patients admitted to the intensive care unit. Exposure density sampling has been proposed as an approach to dynamic matching with respect to a time-dependent exposure. Firstly, exposure density sampling can be useful to reduce the workload of study follow up, as it includes all exposed but only a subset of the not yet exposed individuals. Secondly, it can help to obtain a comparable control group by including propensity score matching. In the present article, we provide the theoretical justification that data obtained by exposure density sampling can be analyzed as a left-truncated cohort. It is shown that exposure density sampling allows estimation of the effect of a time-dependent exposure as well as further baseline covariates on a subsequent event, with only minor loss in precision as compared with a full cohort analysis. The sampling is applied to a real data example (hospital-acquired infections in intensive care units) and in a simulation study. We also provide an estimate of the loss in precision in terms of an increased standard error in the reduced data set after exposure density sampling as compared with the full cohort.

Improving nested case-control studies to conduct a full competing-risks analysis for nosocomial infections

OBJECTIVE

Competing risks are a necessary consideration when analyzing risk factors for nosocomial infections (NIs). In this article, we identify additional information that a competing risks analysis provides in a hospital setting. Furthermore, we improve on established methods for nested case-control designs to acquire this information.

METHODS

Using data from 2 Spanish intensive care units and model simulations, we show how controls selected by time-dynamic sampling for NI can be weighted to perform risk-factor analysis for death or discharge without infection. This extension not only enables hazard rate analysis for the competing risk, it also enables prediction analysis for NI.

RESULTS

The estimates acquired from the extension were in good agreement with the results from the full (real and simulated) cohort dataset. The reduced dataset results averted any false interpretation common in a competing-risks setting.

CONCLUSIONS

Using additional information that is routinely collected in a hospital setting, a nested case-control design can be successfully adapted to avoid a competing risks bias. Furthermore, this adapted method can be used to reanalyze past nested case-control studies to enhance their findings.

Administration of oral fluoroquinolone and the risk of rhegmatogenous retinal detachment: A nationwide population-based study in Korea

PURPOSE

To investigate the association between oral fluoroquinolones (FQ) and the risk of rhegmatogenous retinal detachment (RRD) using a nationwide population-based study in Korea, designed to control for time-related bias.

METHODS

As a nested case-control study within a cohort, the KNHIS-NSC 2002-2013 (Korean National Health Insurance Service National Sample Cohort) data used for the investigation. The subjects who visited an ophthalmologist were included in a cohort. Subjects with infectious ocular diseases, severe ocular trauma, and congenital diseases were excluded. Within the cohort, subjects who underwent surgery for RRD were defined as cases, and controls were matched by age group, sex, and cohort entry date using an incidence density sampling method. After investigating the exposure to oral FQ, the odds ratio was calculated by the FQ exposure rate of both groups and adjusted by the confounding factors of demography, health service utilization, and comorbidities.

RESULTS

A total of 1,151 subjects in the case group and 11,470 subjects in the control group were included. There were intergroup differences in household income, numbers of ophthalmologic visits and drug prescriptions, events of intraocular surgeries, and prevalence of diabetes and degenerative myopia (all P's<0.05). The crude odds ratio of the total group was 1.06 (P = 0.53, 95% CI 0.88-1.28), and the odds ratio adjusted for all pre-defined confounders was 1.00 (P = 0.99, 95% CI 0.81-1.24). The crude and adjusted odds ratios were not showed statistical significance (all P's>0.05).

CONCLUSIONS

By the nested case-control design, this study showed that oral administration of FQ was not associated with the increased risk of development of RRD.

Valid and efficient subgroup analyses using nested case-control data

BACKGROUND

It is not uncommon for investigators to conduct further analyses of subgroups, using data collected in a nested case-control design. Since the sampling of the participants is related to the outcome of interest, the data at hand are not a representative sample of the population, and subgroup analyses need to be carefully considered for their validity and interpretation.

METHODS

We performed simulation studies, generating cohorts within the proportional hazards model framework and with covariate coefficients chosen to mimic realistic data and more extreme situations. From the cohorts we sampled nested case-control data and analysed the effect of a binary exposure on a time-to-event outcome in subgroups defined by a covariate (an independent risk factor, a confounder or an effect modifier) and compared the estimates with the corresponding subcohort estimates. Cohort analyses were performed with Cox regression, and nested case-control samples or restricted subsamples were analysed with both conditional logistic regression and weighted Cox regression.

RESULTS

For all studied scenarios, the subgroup analyses provided unbiased estimates of the exposure coefficients, with conditional logistic regression being less efficient than the weighted Cox regression.

CONCLUSIONS

For the study of a subpopulation, analysis of the corresponding subgroup of individuals sampled in a nested case-control design provides an unbiased estimate of the effect of exposure, regardless of whether the variable used to define the subgroup is a confounder, effect modifier or independent risk factor. Weighted Cox regression provides more efficient estimates than conditional logistic regression.

Levels of interleukin (IL)-6 and IL-8 are elevated in serum and bronchoalveolar lavage fluid of haematological patients with invasive pulmonary aspergillosis

Aspergillus spp. have been shown to induce T-helper cell (Th) 1 and Th17 subsets resulting in elevated levels of several cytokines. The objective of this study was to analyse a bundle of cytokines in serum and bronchoalveolar lavage fluid (BALF) in patients with and without invasive pulmonary aspergillosis (IPA). This nested case-control analysis included 10 patients with probable/proven IPA and 20 matched controls without evidence of IPA, out of a pool of prospectively enrolled (2014-2017) adult cases with underlying haematological malignancies and suspected pulmonary infection. Serum samples were collected within 24 hours of BALF sampling. All samples were stored at -70°C for retrospective determination of cytokines. IL-6 and IL-8 were significantly associated with IPA in both serum (P = .011 and P = .028) and BALF (P = .006 and P = .012, respectively), and a trend was observed for serum IL-10 (P = .059). In multivariate conditional logistic regression analysis, IL-10 remained a significant predictor of IPA in serum and IL-8 among BALF cytokines. In conclusion, levels of IL-6 and IL-8 were significantly associated with probable/proven IPA, and a similar trend was observed for serum IL-10. Future cohort studies should determine the diagnostic potential of these cytokines for IPA, and evaluate combinations with other IPA biomarkers/diagnostic tests.

A case-cohort approach for multi-state models in hospital epidemiology

Analysing the determinants and consequences of hospital-acquired infections involves the evaluation of large cohorts. Infected patients in the cohort are often rare for specific pathogens, because most of the patients admitted to the hospital are discharged or die without such an infection. Death and discharge are competing events to acquiring an infection, because these individuals are no longer at risk of getting a hospital-acquired infection. Therefore, the data is best analysed with an extended survival model - the extended illness-death model. A common problem in cohort studies is the costly collection of covariate values. In order to provide efficient use of data from infected as well as uninfected patients, we propose a tailored case-cohort approach for the extended illness-death model. The basic idea of the case-cohort design is to only use a random sample of the full cohort, referred to as subcohort, and all cases, namely the infected patients. Thus, covariate values are only obtained for a small part of the full cohort. The method is based on existing and established methods and is used to perform regression analysis in adapted Cox proportional hazards models. We propose estimation of all cause-specific cumulative hazards and transition probabilities in an extended illness-death model based on case-cohort sampling. As an example, we apply the methodology to infection with a specific pathogen using a large cohort from Spanish hospital data. The obtained results of the case-cohort design are compared with the results in the full cohort to investigate the performance of the proposed method. Copyright © 2016 John Wiley & Sons, Ltd.

A full competing risk analysis of hospital-acquired infections can easily be performed by a case-cohort approach

OBJECTIVES

We provide a case-cohort approach and show that a full competing risk analysis is feasible even in a reduced data set. Competing events for hospital-acquired infections are death or discharge from the hospital because they preclude the observation of such infections.

STUDY DESIGN AND SETTING

Using surveillance data of 6,568 patient admissions (full cohort) from two Spanish intensive care units, we propose a case-cohort approach which uses only data from a random sample of the full cohort and all infected patients (the cases). We combine established methodology to study following measures: event-specific as well as subdistribution hazard ratios for all three events (infection, death, and discharge), cumulative hazards as well as incidence functions by risk factor, and also for all three events.

RESULTS

Compared with the values from the full cohort, all measures are well approximated with the case-cohort design. For the event of interest (infection), event-specific and subdistribution hazards can be estimated with the full efficiency of the case-cohort design. So, standard errors are only slightly increased, whereas the precision of estimated hazards of the competing events is inflated according to the size of the subcohort.

CONCLUSION

The case-cohort design provides an appropriate sampling design for studying hospital-acquired infections in a reduced data set. Potential effects of risk factors on the competing events (death and discharge) can be evaluated.