Today's meta-analysis is a worked example of a pattern that recurs across gastroenterology: a class comparison, a point estimate that looks alarming or reassuring, and an interval nobody quotes. Anti-TNF agents came out at an odds ratio of 3.04 for cardiovascular events in randomised data - a figure that, read alone, would move prescribing. Its interval runs from 0.31 to 29.47.

The practice change is a reading habit applied at the point of decision. When a safety signal prompts you to consider switching a patient who is in remission, find the interval before you find the alternative drug. If it spans benefit and harm, the signal is not evidence and the switch carries its own known cost: loss of response on re-exposure, a new immunogenicity risk, and a period of active disease during the changeover. Those harms are certain and the signal prompting them is not.

This is not an argument for ignoring safety data. It is an argument for distinguishing a signal that has been measured from one that has merely been estimated, and for recognising that in IBD, where events are rare and trials are short, most class safety comparisons currently fall into the second category.

Read the confidence interval before acting on any class safety signal.
Weigh the certain harms of switching - loss of response, immunogenicity, disease flare - against the uncertain harm you are avoiding.
Distinguish regulatory label warnings, which reflect assessed evidence, from meta-analytic trends, which may not.
Where a patient is in stable remission, the bar for switching on a non-significant signal should be very high.
Document the reasoning when you decide not to switch, so the decision is visible if the evidence later changes.

Why it matters

Switching a patient in remission on the strength of an unmeasured signal trades a certain harm for an uncertain one, and it happens routinely.

The statistics, in plain English

A point estimate is the single most likely value given the data; the interval is the range the data cannot rule out. When rare events drive the analysis, the estimate is unstable and the interval is enormous, so the headline number carries far less information than its precision implies. An interval from 0.31 to 29.47 is the statistical equivalent of 'we do not know'.