Two runners start a race on a course with changing incline. One starts earlier, the other later, and the judge measures their speeds without knowing about the staggered starts. So it is with virus variants: data arrive with delays, and if those delays are unequal, estimates of competitive advantage can be badly wrong.
In evolutionary biology, advantage is often estimated from the slope of the logit-frequency curve of the invading type. The logit is the logarithm of the frequency ratio, and its slope shows how quickly one variant displaces another. But this estimate assumes that observations reflect the real process without distortion. In reality, days pass between infection and sample registration, and different variants can have different delays, for example due to testing peculiarities or prioritization.
The authors tested this in two-variant SIR models with complete cross-immunity and its extensions. They introduced fixed or random observation delays and examined how the slope estimate changed. In the basic model, the relative error ranged from -47.3% to +60.7%, depending on the delay difference and the timing of variant introduction. Strikingly, even when the logit trajectory was almost perfectly linear (R² = 0.9999), the bias was about 31%.
This means that a good logistic fit does not guarantee an unbiased estimate. In GISAID data for the UK, the logit slope for the Alpha→Delta replacement was 0.107 per day, and for Delta→Omicron BA.1 it was 0.237 per day. These numbers could have been distorted by delays, and without accounting for them, one cannot claim that Omicron truly grew 2.2 times faster than Delta.
Now that the runners have finished, the judge knows: without accounting for the staggered starts, medals may go to the wrong people. So too in science: before drawing conclusions about a variant's advantage, one must understand how observation delays have shifted the picture. Correction methods have yet to be developed, but it is already clear that ignoring this problem is not an option.
In GISAID data for the UK, the logit slope for the Alpha→Delta replacement was 0.107 per day, and for Delta→Omicron BA.1 it was 0.237 per day, corresponding to a doubling of the new variant's proportion roughly every 6.5 and 2.9 days, respectively.