Table 2 Criteria and benchmarks to assess certainty of the evidence using GRADE [10, 12, 16]
GRADE criterion | Meaning | Benchmark used in this review |
|---|---|---|
Risk of bias | Quality of the evidence | No risk of bias if at least one Level 1 study was present Serious risk of bias if only one Level 2 was present Very serious risk of bias if no Level 1 or 2 studies were present |
Inconsistency | Results for a given outcome not similar across studies | No inconsistency if improvements shown in at least: – Two thirds of Level 1 or 2 studies and half of Level 3 or 4 studies; or – Half of Level 1 or 2 studies and two third of Level 3 or 4 studies; or – Two thirds of Level 3 or 4 studies in absence of Level 1 or 2 studies |
Imprecision | Insufficient statistical power or wide confidence intervals | No imprecision if at least one study was sufficiently powered and at least one study showed narrow confidence intervals surrounding the estimate of effects |
Indirectness | Evidence differs from study eligibility criteria (PICO) | No indirectness if—across the studies—the following participant characteristics were represented: male/female, young and middle-aged adults (16–65 years) and older adults (> 65 years), time since injury > 1 year and > 1 year, and lesion characteristics (AIS and lesion level) with sufficient lower motor neuron capacity to respond to FES cycling |
Publication bias | Selective publication of studies | Publication bias present if unpublished studies added to the evidence summary would have changed assessment of any of the criteria shown above |
Reasons for upgrading level of certainty in the evidence | If lower-quality studies provide convincing evidence | – Consistent effects across a large number of Level 2, 3 or 4 studies – Plausible bias caused by including participants not responsive to FES cycling – Dose–response gradient present in one study or across all studies |