Table 2.
Summary of Findings.
| Quality assessment | Summary of findings | |||||||
|---|---|---|---|---|---|---|---|---|
| No of studies (participants) | Risk of bias | Consistency | Directness | Precision | Publication bias | Relative effect (95% CI) a |
Absolute effect (95% CI) b |
Quality, GRADE |
| Skin cancers: 5 (552) |
No serious limitation c | No inconsistency d | Serious indirectness (−1) e |
No serious imprecision | Not detected f | Rate ratio = .50 (0.29, 0.85) |
−1.22 per person-year (-1.83, −0.62) |
⊕⊕⊕⊝, moderate |
| BCC: 5 (552) |
No serious limitation c | Serious inconsistency (−1) g |
Serious indirectness (−1) h |
No serious imprecision | Not detected f | Rate ratio = .46 (0.22, 0.95) |
−0.74 per person-year (-1.13, −0.35) |
⊕⊕⊝⊝, low |
| cSCC: 5 (552) |
No serious limitation c | No inconsistency i | Serious indirectness (−1) e |
No serious imprecision | Not detected f | Rate ratio = .48 (0.26, 0.88) |
−0.53 per person-year (-1.03, −0.04) |
⊕⊕⊕⊝, moderate |
| AK: 3 (492) |
No serious limitation | Serious inconsistency (−1) j |
Serious indirectness (−1) h |
Serious imprecision (−1) i | Not detected f | — | −4.48 (-12.68, 3.73) |
⊕⊝⊝⊝, very low |
| Melanoma: 2 (416) |
No serious limitation c | No inconsistency | Serious indirectness (−1) e |
Serious imprecision (−1) l | Not detected f | RR = .89 (0.29, 2.79) |
0.43% fewer melanoma (3.51 fewer to 2.65 more) |
⊕⊕⊝⊝, low |
| GI AE: 21 (1859) |
Serious limitations (−1) m |
No inconsistency | Serious indirectness (−1) o |
Serious imprecision (−1) l | Unlikely | RR = 1.78 (1.30, 2.45) |
5.5% more GI AEs (3.1%, 8.0% more) |
⊕⊝⊝⊝, very low |
| Cutaneous AE: 19 (1805) |
Serious limitations (−1) n |
No inconsistency | Serious indirectness (−1) o |
No serious imprecision | Unlikely | RR = 1.13 (0.87, 1.47) |
1.6% more cutaneous AEs (1.2% fewer to 4.3% more) | ⊕⊕⊝⊝, low |
| Biochemical AE: 9 (1491) | No serious limitation | No inconsistency | Serious indirectness (−1) o |
Serious imprecision (−1) l | Not detected f | RR = 1.57 (0.67, 3.66) |
2.0% more biochemical AEs (0.2%, 3.8% more) | ⊕⊕⊝⊝, low |
Abbreviations: AE, adverse effect;GI, gastrointestinal; RR, relative risk.
aRelative risk (RR) and rate ratio based on random effects models.
bAbsolute risk could only be calculated for 4/5 studies related to skin cancers, BCCs and cSCCs (follow-up per individual not available for rate difference calculation in one trial).
cNo serious risk of bias limitation. Only 2/5 studies reporting BCCs and cSCCs were designed to evaluate skin cancers, and 0/2 trial reporting melanoma was designed to evaluate melanoma (theorical increased risk of selective reporting). However, not downgraded because cancer numbers are not numerical results subject to selection from multiple measurements or analyzes.
dNo inconsistency. Not downgraded because variability in effect estimates (global I 2 = 64%) can be explained by risk of bias between trials (subgroup I 2 = 0% or not applicable).
eSerious indirectness. Downgraded from high to moderate because all relevant trials were restricted to tertiary prevention of skin cancers.
fPossibility of publication bias not excluded but not considered sufficient to downgrade quality of evidence.
gSerious inconsistency. Downgraded from high to moderate because variability in effect estimates (global I 2 = 53%) not explained in subgroup analyses.
hSerious indirectness. Downgraded from moderate to low because all relevant trials were restricted to tertiary prevention.
iNo inconsistency. Heterogeneity between trials (global I 2 = 67%) could be explained by variations in daily doses of nicotinamide and risk of bias.
jSerious inconsistency. Downgraded from high to moderate because variability in effect estimates (global I 2 = 61%) not explained in subgroup analyses.
kSerious imprecision. Downgraded because null value (MD = 0) is included in 95% CI, and both arms are greater than 25% of relative effect.
lSerious imprecision. Downgraded because total number of events < 300, and both arms are greater than 25% of relative effect.
mSerious limitations due to 3/21 trials with per-protocol analyzes; 5/21 open label or single blind trials; and inability to judge the risk of selective reporting of adverse effects in 8/21 studies.
nSerious limitations due to 4/19 trials with per-protocol analyzes; 6/19 open label or single blind trials; and inability to judge the risk of selective reporting of adverse effects in 6/19 studies.
oSerious indirectness. Downgraded from moderate to low because all trials relevant to adverse effects were conducted for other indications than skin cancer chemoprophylaxis.