The objective of this study was to investigate whether glycated hemoglobin (HbA1c) is a valid surrogate for evaluating the effectiveness of antihyperglycemic drugs in diabetes mellitus (DM) trials.

We conducted a systematic review of placebo-controlled randomized clinical trials (RCTs) evaluating the effect of a treatment on HbA1c (mean difference between groups) and clinical outcomes (relative risk of mortality, myocardial infarction, stroke, heart failure, and/or kidney injury) in patients with DM. Then, we investigated the association between treatment effects on HbA1c and clinical outcomes using regression analysis at the trial level. Lastly, we interpreted the correlation coefficients (R) using the cut-off points suggested by the Institute for Quality and Efficiency in Healthcare (IQWiG). HbA1c was considered a valid surrogate if it demonstrated a strong association: lower limit of the 95 percent confidence interval (95 percent CI) of R greater than or equal to .85.

Nineteen RCTs were identified. All studies included adults with type 2 DM. None of the associations evaluated was strong enough to validate HbA1c as a surrogate for any clinical outcome: mortality (R = .34; 95 percent CI −.14 to .69), myocardial infarction (R = .20; −.30 to .61), heart failure (R = .08; −.40 to .53), kidney injury (R = −.04; −.52 to .47), and stroke (R = .81; .54 to .93).

The evidence from multiple placebo-controlled RCTs does not support the use of HbA1c as a surrogate to measure the effectiveness of antihyperglycemic drugs in DM studies.

A primary outcome in oncology trials is overall survival (OS). However, to estimate OS accurately requires a sufficient number of patients to have died, which may take a long time. If an alternative end point is sufficiently highly correlated with OS, it can be used as a surrogate. Progression-free survival (PFS) is the surrogate most often used in oncology, but does not always satisfy the correlation conditions for surrogacy. We analyze the methodologies used when extrapolating from PFS to OS.

Davis et al. previously reviewed the use of surrogate end points in oncology, using papers published between 2001 and 2011. We extend this, reviewing papers published between 2012 and 2016. We also examine the reporting of statistical methods to assess the strength of surrogacy.

The findings from 2012 to 2016 do not differ substantially from those of 2001 to 2011: the same factors are shown to affect the relationship between PFS and OS. The proportion of papers reporting individual patient data (IPD), strongly recommended for full assessment of surrogacy, remains low: 33 percent. A wide range of methods has been used to determine the appropriateness of surrogates. While usually adhering to reporting standards, the standard of scholarship appears sometimes to be questionable and the reporting of results often haphazard.

Standards of analysis and reporting PFS to OS surrogate studies should be improved by increasing the rigor of statistical reporting and by agreeing to a minimum set of reporting guidelines. Moreover, the use of IPD to assess surrogacy should increase.