Leveraging a surrogate outcome to improve inference on a partially missing target outcome

Sample sizes vary substantially across tissues in the Genotype-Tissue Expression (GTEx) project, where considerably fewer samples are available from certain inaccessible tissues, such as the substantia nigra (SSN), than from accessible tissues, such as blood. This severely limits power for identifying tissue-specific expression quantitative trait loci (eQTL) in undersampled tissues. Here we propose Surrogate Phenotype Regression Analysis (Spray) for leveraging information from a correlated surrogate outcome (eg, expression in blood) to improve inference on a partially missing target outcome (eg, expression in SSN). Rather than regarding the surrogate outcome as a proxy for the target outcome, Spray jointly models the target and surrogate outcomes within a bivariate regression framework. Unobserved values of either outcome are treated as missing data. We describe and implement an expectation conditional maximization algorithm for performing estimation in the presence of bilateral outcome missingness. Spray estimates the same association parameter estimated by standard eQTL mapping and controls the type I error even when the target and surrogate outcomes are truly uncorrelated. We demonstrate analytically and empirically, using simulations and GTEx data, that in comparison with marginally modeling the target outcome, jointly modeling the target and surrogate outcomes increases estimation precision and improves power.

Changes in GFR and Albuminuria in Routine Clinical Practice and the Risk of Kidney Disease Progression

RATIONALE & OBJECTIVE

Changes in urinary albumin-creatinine ratio (UACR) and estimated glomerular filtration rate (eGFR) have been used separately as alternative kidney disease outcomes in randomized trials. We tested the hypothesis that combined changes in UACR and eGFR predict advanced kidney disease better than either alone.

STUDY DESIGN

Observational cohort study.

SETTING & PARTICIPANTS

91,319 primary care patients assembled from the Clinical Practice Research Datalink in the United Kingdom between 2000 and 2015.

EXPOSURES

Changes in UACR and eGFR (categorized as ≥30% increase, stable, or ≥30% decrease), alone and in combination, over a 3-year period.

OUTCOMES

The primary outcome was advanced CKD (sustained eGFR <30 mL/min/1.73 m2); secondary outcomes included kidney failure, cardiovascular disease, and all-cause mortality.

ANALYTICAL APPROACH

Multivariable Cox regression with bias from missing values assessed using multiple imputation; discrimination statistics compared across exposure groups.

RESULTS

91,319 individuals were studied, with a mean eGFR of 72.6 mL/min/1.73 m2 and median UACR of 9.7 mg/g; 70,957 (77.7%) had diabetes. During a median follow-up of 2.9 years, 2,541 people progressed to advanced CKD. Compared with stable values, hazard ratios for a ≥30% increase in UACR and ≥30% decrease in eGFR were 1.78 (95% CI, 1.59-1.98) and 7.53 (95% CI, 6.70-8.45), respectively, for the outcome of advanced CKD. Compared with stable values of both, the hazard ratio for the combination of an increase in UACR and a decrease in eGFR was 15.15 (95% CI, 12.43-18.46) for the outcome of advanced CKD. The combination of changes in UACR and eGFR predicted kidney outcomes better than either alone.

LIMITATIONS

Selection bias, relatively small proportion of individuals without diabetes, and very few kidney failure events.

CONCLUSIONS

In a large-scale general population, the combination of an increase in UACR and a decrease in eGFR was strongly associated with the risk of advanced CKD. Further assessment of combined changes in UACR and eGFR as an alternative outcome for kidney failure in trials of CKD progression is warranted.

Robust and efficient semi-supervised estimation of average treatment effects with application to electronic health records data

We consider the problem of estimating the average treatment effect (ATE) in a semi-supervised learning setting, where a very small proportion of the entire set of observations are labeled with the true outcome but features predictive of the outcome are available among all observations. This problem arises, for example, when estimating treatment effects in electronic health records (EHR) data because gold-standard outcomes are often not directly observable from the records but are observed for a limited number of patients through small-scale manual chart review. We develop an imputation-based approach for estimating the ATE that is robust to misspecification of the imputation model. This effectively allows information from the predictive features to be safely leveraged to improve efficiency in estimating the ATE. The estimator is additionally doubly-robust in that it is consistent under correct specification of either an initial propensity score model or a baseline outcome model. It is also locally semiparametric efficient under an ideal semi-supervised model where the distribution of the unlabeled data is known. Simulations exhibit the efficiency and robustness of the proposed method compared to existing approaches in finite samples. We illustrate the method by comparing rates of treatment response to two biologic agents for treatment inflammatory bowel disease using EHR data from Partners' Healthcare.

How researchers should select the best outcomes for randomised clinical trials in paediatric dentistry?

Randomised clinical trial (RCT) is the best study design to evaluate the effect of the treatment and preventive healthcare procedures. The effects of the tested treatments on patient's health are compared in terms of outcomes, which are used to evaluate the participants' health changes. However, these outcomes should be relevant for the target population. In that way, RCTs represent the type of primary study design that provides the most reliable evidence to implement therapies into the clinical practice. In this review, an outline of some aspects related to the choice of RCTs' outcomes was presented, focusing on the conduction of relevant trials in Paediatric Dentistry. The importance and necessity of defining a primary outcome were addressed, preferentially a clinically relevant endpoint. The patients should perceive this outcome, and changes in this variable should reflect directly patient's health improvement or impairment. Moreover, considerations about the objective or subjective variables, use of surrogate outcomes, and the increasing tendency to develop core outcome sets were also presented in this review. The main idea of this manuscript is the RCTs must evaluate outcomes relevant to the children's oral health in order to contribute to the implementation of treatments in the evidence-based health practice.

Non-inferiority trials using a surrogate marker as the primary endpoint: An increasing phenotype in cardiovascular trials

BACKGROUND/AIMS

Non-inferiority trials are increasing in cardiovascular medicine, with approval of many drugs and devices on the basis of such studies. Surrogate markers as primary endpoints have been also more frequently used for efficient assessment of cardiovascular interventions. However, there is uncertainty about their concordance with clinical outcomes. Non-inferiority design using a surrogate marker as a primary endpoint may pose particular challenges in clinical interpretation. We sought to explore the publication trends, methodology, and reporting features of non-inferiority cardiovascular trials that used a primary surrogate marker as the primary endpoint.

METHODS

We searched six high-impact journals (The New England Journal of Medicine, The Journal of the American Medical Association, The Lancet, The Journal of the American College of Cardiology, Circulation, and European Heart Journal) from 1 January 1990 to 31 December 2018 and identified non-inferiority cardiovascular trials that used a surrogate marker as the primary endpoint. We assessed the non-inferiority margin reported in the manuscript and other publicly available platforms (e.g. protocol, clinicaltrials.gov). We also determined whether the included non-inferiority trials with surrogate markers as primary endpoints were followed by clinical outcome trials.

RESULTS

We screened 15,553 publications and identified 247 cardiovascular trials that used a non-inferiority design. Of these, 37 had a surrogate marker as a primary endpoint (18 drug trials, 13 device trials, 6 others). All of these non-inferiority trials with surrogate outcomes were published after 2000, mostly in cardiology journals (13 in The Journal of the American College of Cardiology, 9 in European Heart Journal, 8 in Circulation, 6 in The Lancet, 1 in The New England Journal of Medicine), and their publication rate increased over time (p < 0.001 for linear trend). The median number of patients in the primary analysis was 300 (interquartile range: 202-465). The study protocol or a methods paper was publicly available for only 13 (35.1%) trials, of which the non-inferiority margin was not reported in 4 trials. In 16 studies (43.2%), the manuscript did not acknowledge the limitations of using a surrogate endpoint or the need for a definitive clinical outcome trial. Thirty-four trials (91.9%) concluded that the tested intervention met non-inferiority criteria. However, only five (13.5%) were followed by clinical outcomes trials the results of which did not always confirm non-inferiority.

CONCLUSION

Non-inferiority trials that use a surrogate marker as the primary endpoint are being increasingly performed. However, these trials pose particular challenges with design, reporting, and interpretation, which are not systematically and consistently addressed or reported.

Estimated GFR Trajectories in Pediatric and Adult Nephrotic Syndrome: Results From the Nephrotic Syndrome Study Network (NEPTUNE)

Rationale & Objective

Surrogate outcomes for end-stage kidney disease often assume linear changes, which may not reflect true estimated glomerular filtration rate (eGFR) trajectories. This study’s objective was to characterize nonlinear eGFR trajectories in nephrotic syndrome.

Study Design

Observational cohort study.

Setting & Participants

Nephrotic Syndrome Study Network (NEPTUNE) is a multicenter study of adult and pediatric patients with proteinuria enrolled at clinically indicated kidney biopsy or initial presentation of disease (pediatric only).

Predictors

Patient demographic, clinical, and pathology variables at study enrollment and follow-up time.

Outcome

eGFR was calculated using the Chronic Kidney Disease Epidemiology Collaboration (patients ≥ 18 years old) or modified Chronic Kidney Disease in Children Study–Schwartz (patients < 18 years) formulas. The probability of nonlinearity (PNL) was calculated for individual eGFR trajectories.

Analytical Approach

Associations between predictors and PNL were assessed using multivariable linear regression.

Results

453 patients with ≥3 eGFR measurements and 1 or more year of follow-up were included (median follow-up, 3.6 years). Median PNL was 0.052; 56% and 16% had PNL < 10% and >50%, respectively. In both adults and pediatric patients, higher baseline eGFR was associated with higher PNL, whereas longer follow-up time was associated with lower PNL. Higher urine protein-creatinine ratio and steroid use were also associated with higher PNL in adults. Higher percentages of tubular atrophy and foot-process effacement were associated with lower and higher PNLs, respectively, in adults.

Limitations

Relatively short follow-up time, inability to assess acute kidney injury events, and variable eGFR measurement frequency across patients.

Conclusions

Although increasing follow-up time resulted in more linear trajectories, nonlinear eGFR trajectories were common in this cohort. Future studies in nephrotic syndrome should consider novel outcomes that do not rely on linearity assumptions.

Surrogate Biomarkers for Monitoring Healthcare Quality for Chronic Diseases Such as Diabetes Care

Some laboratory tests or biomarkers are used as surrogate outcomes for health care effectiveness. HbA1c is defined as a surrogate biomarker since HbA1c values have been approved to be used in predictions of clinically important complications of diabetes mellitus. With the advance of information technology (IT) the real life data are aggregating as electronic health records (EHRs). About 70-85% of individuals admitted to hospitals have laboratory test results. As such, medical laboratories are the data centers in the hospitals. The test results can be used for assessment of health care delivered, especially for chronic diseases. This information provides insights of healthcare services, and can be used to enhance for individual and population well-being, research, and education. This article focuses on the importance of using laboratory tests results as outcome measures for specific population health status that are important in assessing the quality of health care services. The findings from our studies on the diabetic care quality is presented.

Prognostic factors versus markers of response to treatment versus surrogate endpoints: Three different concepts

Multiple sclerosis is a highly heterogeneous disease; the quantitative assessment of disease progression is problematic for many reasons, including the lack of objective methods to measure disability and the long follow-up times needed to detect relevant and stable changes. For these reasons, the importance of prognostic markers, markers of response to treatments and of surrogate endpoints, is crucial in multiple sclerosis research. Aim of this report is to clarify some basic definitions and methodological issues about baseline factors to be considered prognostic markers or markers of response to treatment; to define the dynamic role that variables must have to be considered surrogate markers in relation to specific treatments.

Biomarkers in Pediatric ARDS: Future Directions

Acute respiratory distress syndrome (ARDS) is common among mechanically ventilated children and accompanies up to 30% of all pediatric intensive care unit deaths. Though ARDS diagnosis is based on clinical criteria, biological markers of acute lung damage have been extensively studied in adults and children. Biomarkers of inflammation, alveolar epithelial and capillary endothelial disruption, disordered coagulation, and associated derangements measured in the circulation and other body fluids, such as bronchoalveolar lavage, have improved our understanding of pathobiology of ARDS. The biochemical signature of ARDS has been increasingly well described in adult populations, and this has led to the identification of molecular phenotypes to augment clinical classifications. However, there is a paucity of data from pediatric ARDS (pARDS) patients. Biomarkers and molecular phenotypes have the potential to identify patients at high risk of poor outcomes, and perhaps inform the development of targeted therapies for specific groups of patients. Additionally, because of the lower incidence of and mortality from ARDS in pediatric patients relative to adults and lack of robust clinical predictors of outcome, there is an ongoing interest in biological markers as surrogate outcome measures. The recent definition of pARDS provides additional impetus for the measurement of established and novel biomarkers in future pediatric studies in order to further characterize this disease process. This chapter will review the currently available literature and discuss potential future directions for investigation into biomarkers in ARDS among children.

Complete and Partial Remission as Surrogate End Points in Membranous Nephropathy

Absent a remission of proteinuria, primary membranous nephropathy (MN) can lead to ESRD over many years. Therefore, use of an earlier end point could facilitate the conduct of clinical trials. This manuscript evaluates complete remission (CR) and partial remission (PR) of proteinuria as surrogate end points for a treatment effect on ESRD in patients with primary MN with heavy proteinuria. CR is associated with a low relapse rate and excellent long-term renal survival, and it plausibly reflects remission of the disease process that leads to ESRD. Patients who achieve PR have better renal outcomes than those who do not but may have elevated relapse rates. How long PR must be maintained to yield a benefit on renal outcomes is also unknown. Hence, available data suggest that CR could be used as a surrogate end point in primary MN, whereas PR seems reasonably likely to predict clinical benefit. In the United States, surrogate end points that are reasonably likely to predict clinical benefit can be used as a basis for accelerated approval; treatments approved under this program must verify the clinical benefit in postmarketing trials. Additional analyses of the relationship between treatment effects on CR and PR and subsequent renal outcomes would inform the design of future clinical trials in primary MN.

Assessing treatment effects with surrogate survival outcomes using an internal validation subsample

BACKGROUND

In studies with surrogate outcomes available for all subjects and true outcomes available for only a subsample, survival analysis methods are needed that incorporate both endpoints in order to assess treatment effects.

METHODS

We develop a semiparametric estimated likelihood method for the proportional hazards model with discrete time data and a binary covariate of interest. Our proposed method allows for real-time validation of surrogate outcomes and flexible censoring mechanisms.

RESULTS

Our proposed estimator is consistent and asymptotically normal. Through numerical studies, we showed that our proposed method for estimating a covariate effect is unbiased compared to the naïve estimator that uses only surrogate endpoints and is more efficient with moderate missingness compared to the complete-case estimator that uses only true endpoints. We further demonstrated the advantages of our proposed method in comparison with existing approaches when there is real-time validation. We also illustrated the use of our proposed method by estimating the effect of gender on time to detection of Alzheimer's disease using data from the Alzheimer's Disease Neuroimaging Initiative.

CONCLUSION

The proposed method is able to account for the uncertainty of surrogate outcomes using a validation subsample of true outcomes in estimating a binary covariate effect. The proposed estimator can outperform standard semiparametric survival analysis methods and can therefore save on costs of a trial or improve power in detecting treatment effects.

[Patient-relevant outcomes and surrogates in the early benefit assessment of drugs: first experiences]

The Act on the Reform of the Market for Medicinal Products (AMNOG) became effective in Germany on January 1, 2011. Since then, the assessment of the added benefit of new drugs versus a therapeutic standard on the basis of dossiers submitted by pharmaceutical companies has been required by law. The Federal Joint Committee (G-BA) generally commissions the Institute for Quality and Efficiency in Health Care (IQWiG) with this task. The added benefit is primarily to be demonstrated on the basis of patient-relevant outcomes. The aim of this paper is to describe the feasibility of the early benefit assessment on the basis of patient-relevant outcomes by systematically characterising the outcomes available in company dossiers and comparing the companies' and IQWiG's evaluations regarding patient relevance and surrogate validity. Dossier assessments published between October 2011 and June 2012 were used for this purpose. The outcomes available and the respective evaluations were extracted and compared. 12 out of 22 submitted dossiers contained sufficient data to assess outcomes; all 12 assessable dossiers provided data on patient-relevant outcomes. Data on mortality and adverse events were available in all dossiers, except that one dossier did not contain adverse event data on the relevant subpopulation. In contrast, data on morbidity and health-related quality of life were available in 8 and 7 dossiers, respectively. Of a total of 214 outcomes extracted by IQWiG, 124 patient-relevant and 3 surrogate outcomes were included in IQWiG's assessment (companies: a total of 183 outcomes included, of which 172 were patient-relevant and 11 were surrogates). The first experiences with AMNOG have shown that in principle an early benefit assessment of drugs based on patient-relevant outcomes is feasible. The companies' and IQWiG's evaluations regarding patient relevance and surrogate validity of outcomes partly deviated from each other. By increasingly considering patient-relevant outcomes in approval studies, pharmaceutical companies can create the necessary data basis for the early benefit assessment.

Surrogate outcomes: experiences at the Common Drug Review

BACKGROUND

Surrogate outcomes are a significant challenge in drug evaluation for health technology assessment (HTA) agencies. The research objectives were to: identify factors associated with surrogate use and acceptability in Canada's Common Drug Review (CDR) recommendations, and compare the CDR with other HTA or regulatory agencies regarding surrogate concerns.

METHODS

Final recommendations were identified from CDR inception (September 2003) to December 31, 2010. Recommendations were classified by type of outcome (surrogate, final, other) and acceptability of surrogates (determined by the presence/absence of statements of concern regarding surrogates). Descriptive and statistical analyses examined factors related to surrogate use and acceptability. For thirteen surrogate-based submissions, recommendations from international HTA and regulatory agencies were reviewed for statements about surrogate acceptability.

RESULTS

Of 156 final recommendations, 68 (44%) involved surrogates. The overall 'do not list' (DNL) rate was 48%; the DNL rate for surrogates was 41% (p = 0.175). The DNL rate was 64% for non-accepted surrogates (n = 28) versus 25% for accepted surrogates (odds ratio 5.4, p = 0.002). Clinical uncertainty, use of economic evidence over price alone, and a premium price were significantly associated with non-accepted surrogates. Surrogates were used most commonly for HIV, diabetes, rare diseases, cardiovascular disease and cancer. For the subset of drugs studied, other HTA agencies did not express concerns for most recommendations, while regulatory agencies frequently stated surrogate acceptance.

CONCLUSIONS

The majority of surrogates were accepted at the CDR. Non-accepted surrogates were significantly associated with clinical uncertainty and a DNL recommendation. There was inconsistency of surrogate acceptability across several international agencies. Stakeholders should consider collaboratively establishing guidelines on the use, validation, and acceptability of surrogates.

Debate: The slippery slope of surrogate outcomes

Surrogate outcomes are frequently used in cardiovascular disease research. A concern is that changes in surrogate markers may not reflect changes in disease outcomes. Two recent clinical trials (Heart and Estrogen/Progestin Replacement Study [HERS], and the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial [ALLHAT]) underscore this problem since their results contradicted what was expected based on the surrogate outcomes. The current regulatory policy to allow new therapies to be introduced onto the market based solely on surrogate outcomes may need to be reviewed.

Current Controlled Trials: an opportunity to help improve the quality of clinical research

Some problems with the quality of controlled clinical trials can be addressed by following these procedures: registering all trials at inception; using systematic reviews to inform the design of new studies; posting and obtaining feedback on preprints; reporting all well conducted trials, regardless of their results; reducing biased and inefficient assessment of reports submitted for publication; publishing sufficiently detailed reports; linking trial reports to relevant external information; providing readier access to reports; and reviewing and amending reports after initial publication. The launch of a new range of electronic journals by Current Controlled Trials offers an opportunity to contribute to progress in these ways.