Could someone help me understand how to interpret this line:
> Higher mushroom consumption was associated with lower risk of total cancer (pooled RR for the highest compared with the lowest consumption groups: 0.66; 95% CI: 0.55, 0.78; n = 17)
Specifically the “0.66; 95% CI: 0.55, 0.78; n = 17” line?
ChatGPT told me:
> The “pooled RR” of 0.66 indicates that the risk of cancer was 34% lower in the group with the highest mushroom consumption compared to the group with the lowest consumption. The “95% CI” represents the confidence interval, indicating the range within which the true risk reduction is likely to fall. In this case, it’s between 0.55 and 0.78. “n = 17” likely refers to the number of studies or datasets included in the analysis.
Is that accurate? The top line conclusion of this study was that eating more mushrooms was associated with a 34% reduction of risk of all cancers?
IANA statistician, but yes I believe that's correct.
The confidence interval means that it might not actually be 0.66, sample size n = only 17 after all, but (with 95% certainty) the true value for the population sampled lies within 0.55-0.78.
(threading along for a minor nitpick on above post)
Ya, correct enough.
> > Higher mushroom consumption was associated with lower risk of total cancer (pooled RR for the highest compared with the lowest consumption groups: 0.66; 95% CI: 0.55, 0.78; n = 17)
> Specifically the “0.66; 95% CI: 0.55, 0.78; n = 17” line?
For a given study with a cohort of A entities, and a count of events A_event, and a cohort of B entities, with count of events B_event; the A_risk will be A_event/A, B_risk will be B_event/B; and the risk ratio (RR) between the cohorts will be RR = A_risk/B_risk. [wiki gives a cursory math explainer: https://en.wikipedia.org/wiki/Relative_risk#Numerical_exampl...] If the RR here is lower than 1, whatever distinguishes cohort A from B is interpreted as reducing the risk.
The linked study does a systematic review of the literature. From that review, they ultimately find 17 studies which are suitably comparable under their criteria. Each of these 17 studies are observational studies over different sizes of cohorts (table 1 in linked paper characterizes the n for each study) that report their own risk ratios (figure 3 in the paper plots each of those RR), then they pooled those risk ratios with a random effects model to synthesize each of the different risk ratios between those studies into the single pooled RR within some confidence interval quoted above. ("pooled" because it's pooling the results from multiple studies.)
>sample size n = only 17 after all
The nitpick is that n=17 is referring to the number of papers pooled, not the population sampled in the underlying observational studies. (that number is higher than 17; lee et al in table 1/ fig 3 is n=112,991).
> Higher mushroom consumption was associated with lower risk of total cancer (pooled RR for the highest compared with the lowest consumption groups: 0.66; 95% CI: 0.55, 0.78; n = 17)
Specifically the “0.66; 95% CI: 0.55, 0.78; n = 17” line?
ChatGPT told me:
> The “pooled RR” of 0.66 indicates that the risk of cancer was 34% lower in the group with the highest mushroom consumption compared to the group with the lowest consumption. The “95% CI” represents the confidence interval, indicating the range within which the true risk reduction is likely to fall. In this case, it’s between 0.55 and 0.78. “n = 17” likely refers to the number of studies or datasets included in the analysis.
Is that accurate? The top line conclusion of this study was that eating more mushrooms was associated with a 34% reduction of risk of all cancers?