Filtered instead of unfiltered coffee
The paper filter retains cafestol, which raises LDL. French press, moka pot and boiled coffee let it through.
Order of magnitude
Filter coffee or espresso from a machine instead of French press, moka pot or boiled coffee, with the same number of cups.
Cai 2012: coffee versus none raises LDL by 5.4 mg/dL and TG by 12.6 mg/dL, more when unfiltered. Jee 2001: filtered coffee barely changes cholesterol. The swap uses the unfiltered rise as the avoidable amount (G-D8). No dedicated head-to-head pool. ApoB was not pooled.
Effects in detail
Each row is a directed edge from our database, with range, latency, evidence grade and the condition under which it applies at all.
| Marker | Effect | Latency | Mode of action | Outcome | Evidence | Condition | Sources |
|---|---|---|---|---|---|---|---|
| Triglycerides | helps. Lowers Direction known, size not established the source prints the interval 3.5 to 12.6 around a point estimate of 12.6, which cannot be right; the row stores the printed bounds as a plausible span and no point estimate Cai 2012 +12.6 mg/dL TG from coffee overall, larger unfiltered. Percentage at TG 150 mg/dL for switching off unfiltered coffee. Same G-D8 caveat as LDL: not a dedicated filter head-to-head pool. | 6–6 weeks | chronic | unclear | Evidence grade C | TG at or above 150 mg/dL | |
| Apolipoprotein B | helps. Lowers -5.40 mg/dL confidence interval of the estimate, not the range one person can expect Across 12 randomised trials with 1017 participants and a mean of 45 days, coffee raised LDL-C by 5.4 mg/dL (95 % CI 1.4 to 9.5) and total cholesterol by 8.1 mg/dL, and the total-cholesterol rise was larger with unfiltered than with filtered coffee. The number here is the coffee-versus-no-coffee estimate, used as an upper bound for the filtered-versus-unfiltered swap: the meta-analysis reports the filtered contrast only as a subgroup difference and gives no separate pooled estimate for filtered coffee (gaps.md G-D8). Trial proxy: ldl_c. | 6–6 weeks | chronic | unclear | Evidence grade B | applies regardless of the starting value |
What this measure costs
Counteracting effects of the same measure.
No counteracting effect is on file for this measure. That does not mean there is none, only that we found no defensible edge.
Sources
Full details of every study the rows above rely on.
The effect of coffee consumption on serum lipids: a meta-analysis of randomized controlled trials
Cai et al. · European Journal of Clinical Nutrition · 2012 · Meta-analysis