"Probability results are from an ineffective treatment. The evidence of a negative publication bias allows us to perform an additional analysis. If treatment was not effective, and there is no publication bias, the results of studies should be randomly distributed. We can compute the probability that the observed percentage of positive results (or higher) could occur due to chance with an ineffective treatment, which is the probability of >= k heads in n coin tosses, or the one-sided sign test / binomial test. 79 of 95 studies report a positive result. The probability of this happening due to chance with an ineffective treatment is 1 in 58 billion. Publication bias will affect this estimate. Correcting for a negative bias will further decrease the chance that the results are from an ineffective treatment."

What?! 1 in 58 billion! Wow!

"Probability results are from an ineffective treatment. The evidence of a negative publication bias allows us to perform an additional analysis. If treatment was not effective, and there is no publication bias, the results of studies should be randomly distributed. We can compute the probability that the observed percentage of positive results (or higher) could occur due to chance with an ineffective treatment, which is the probability of >= k heads in n coin tosses, or the one-sided sign test / binomial test. 79 of 95 studies report a positive result. The probability of this happening due to chance with an ineffective treatment is 1 in 58 billion. Publication bias will affect this estimate. Correcting for a negative bias will further decrease the chance that the results are from an ineffective treatment."

What?! 1 in 58 billion! Wow!