Rosuvastatin versus atorvastatin on CRP: a difference of 0.16 mg/L, and only after an outlier goes

Design

systematic review and meta-analysis of 35 randomised head-to-head trials, PROSPERO registered

Population

6,148 patients randomised to atorvastatin or rosuvastatin

Primary outcome

change in CRP, with IL-6, TNF-alpha and adiponectin secondary

Effect

CRP mean difference -0.53 mg/L (95% CI -1.10 to 0.05), p=0.07, I squared 82%; after outlier exclusion -0.16 mg/L (-0.28 to -0.03); no difference for IL-6, TNF-alpha or adiponectin

Thirty-five randomised head-to-head trials, 6,148 patients, comparing atorvastatin against rosuvastatin on inflammatory markers. The primary pooled comparison for C-reactive protein found no significant difference: mean difference 0.53 mg/L favouring rosuvastatin, 95% CI 1.10 lower to 0.05 higher, p=0.07, with I squared of 82%.

Significance appeared only after one outlying trial was excluded, at which point rosuvastatin was ahead by 0.16 mg/L (95% CI 0.03 to 0.28, p=0.01). Restricting to the nine trials at low risk of bias shrank the difference further, to 0.08 mg/L (95% CI 0.01 to 0.16). Nothing separated the two drugs on interleukin-6, TNF-alpha or adiponectin, all with intervals comfortably spanning no effect.

At these magnitudes the question is not which statin but whether the difference means anything. A CRP gap of 0.08 to 0.16 mg/L is far below the increments that separated risk groups in the trials which established CRP as a marker, and no clinical outcome was measured here. Statin choice in practice turns on potency needed, interactions, tolerability and cost - and on that last point atorvastatin's price advantage in India is not displaced by a tenth of a milligram per litre of CRP.

• Choose the statin on LDL target, interactions and cost, not on inflammatory markers
• Rosuvastatin remains the option where a high-intensity target is not met on atorvastatin
• Do not order CRP to choose between statins - no outcome data support the comparison
• Where baseline inflammation is genuinely high, look for the cause rather than switching statin
• Check for interacting drugs; that difference between the two agents is real and clinically larger than this one

Why it matters

It closes a comparison that generates a surprising amount of prescribing debate, and closes it as a non-difference.

Don't overread it

Inflammatory biomarkers are surrogates - no cardiovascular outcome was measured in this comparison.

The statistics, in plain English

The headline comparison was not significant, and the significant result exists only after removing a trial the authors judged an outlier - a legitimate sensitivity analysis, but one that turns a negative primary result into a positive secondary one, which is the weaker form of evidence. I squared of 82% falling to 63% after that exclusion tells you the trials disagreed substantially. And the effect shrinks as study quality rises, from 0.16 mg/L overall to 0.08 mg/L among the low risk-of-bias studies - the pattern you expect when a small effect is partly bias.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.