Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

Adverse events with placebo: representative of real life?


Bandolier 115 examined evidence about non-drug adverse events from two studies that collected adverse event information in young people without any disease and taking no medicine. The duration was three days, and symptoms were indicated from proffered lists.

Given that these were young men and women in their 20s, what was astonishing was the frequency with which adverse events were noted, with fatigue particularly common, but others also, like headache, joint and muscle pain, sleepiness, and inability to concentrate appearing frequently. Only 17% noted no symptoms, and many had multiple symptoms.

Clinical trials record adverse events in a number of ways, and even though adverse events are not usually the prime interest of clinical trials, we pore over the data because adverse events often drive therapy. The most obvious is that adverse events are a major driver of lack of adherence, or switching therapy, or both.

A review examines adverse events in placebo groups from statin trials, and makes comparison with another survey of adverse events in an adult German population [1], and leads us again to question the reliability of adverse event rates captured by clinical trials.


The review identified statin drug trials published between 1994 and 2003 with more than 100 subjects in a placebo arm. Adverse events in placebo were noted as all cause and drug related. Results from a representative sample of 2,500 German adults assessing 33 symptoms for the previous seven days were used for comparison.


Discontinuation rates for any cause in placebo groups of seven trials with 17,535 patients lasting 16 weeks to five years (mostly one year or longer) ranged from 3 to 40% on an annualised basis, averaging 10% a year. Discontinuation rates because of adverse events averaged 7% a year. In trials lasting at least three years there was less variability, with annualised rates of all cause discontinuations with placebo of 6% and adverse event discontinuations 4%.

Particular symptoms regarded as possibly drug related usually varied by factors of 4:1 or greater between large trials with at least 400 patients given placebo. Symptoms occurring with placebo and independent of presumed cause also varied between trials. Moreover, the rates at which these occurred with placebo in trials was different from that occurring in the absence of a trial (Figure 1).

Figure 1: Adverse events recorded with placebo in patients in placebo groups in statin trials, and over seven days in 2,552 German adults aged 16-99 years, mean age 48 years


While in part this paper simply rehearses what is already known about the problems with adverse event measurement and interpretation in clinical trials, it still gives us much to chew over. Usually variability between trials is due to small numbers, but not here. Different methods of capturing adverse events and different populations are theoretical reasons for differences.

But trials and clinical practice are not always the same, with relatively low discontinuation rates with statins in trials contrasting with high discontinuation rates in practice. And there's what happens without drugs. Again, it is instructive to see how many symptoms can be recorded without a trial, and why those symptoms are often so less frequent in trials.

There is much food for thought. Adverse event reporting, analysis, and interpretation is a topic that demands much more of our attention.


  1. W Rief et al. Medication-attributed adverse effects in placebo groups. Implications for assessment of adverse events. Archives of Internal Medicine 2006 166: 155-160.

previous story