This observational study reports blood glucose values measured by insulin-treated pilots within two hours before or during the flight. Blood glucose values were classified according to the traffic light system specified in the regulations by the UK Civil Aviation Authority.

Only 2.3% of the 38,621 values reported were outside the green range and only 0.12% in the low red range. Nearly all values were immediately (within 30 minutes) corrected to the green range, and no case of pilot incapacitation was reported.

The authors therefore conclude that the traffic light system seems to be effective in identifying and reducing the frequency and severity of out-of-range values.

The paper is well written and reports many additional analyses all confirming the appropriateness of the traffic light system.

As the main results are all given in the two tables and the figures it should be feasible to shorten the paper to a research letter or a brief report.

This reviewer only has minor comments.

1. Methods: The rather long description of the aviation regulations could be shortened. Classification of BG values and associated actions could be summarised in a supplementary table.

2. Results: The rather lengthy analysis of the demographics of pilots in each specified blood glucose range should be deleted.

The authors do not present the demographics of the whole group of pilots (which would actually be helpful), but it seems that this is a rather homogenous population (all but two are male, most are patients with type 1 diabetes, age range is relatively narrow (34-57 years) and HbA1c seems to be in a tight range around 55 mmol/l). It is therefore questionable that the comparison of demographics across traffic light colours will give meaningful results (indeed, the authors did not find significant differences).

1. Results in general could be shortened – all relevant results are given in Table 2 and the figures.

2. The statistical analysis for differences in rate of BG changes before and after low red and low amber values is a self-fulfilling prophecy that should be deleted. The rate of decline must have been steeper with low red values, otherwise BG would not have been fallen into the red range (BG is checked very hour by pilots).

The rate of increase is higher because pilots reacted appropriately to low BG values as acknowledged by the authors.

1. Figures: The colour-coded data points/lines in Figures 1B and 2C should be removed or explained.

2. Figure 3: The choice of polynomial trend lines should be justified. This reviewer would not expect that BG changes follow a polynomial pattern. Without the first two data points in the red curve of Figure 3A (showing increasing rather than decreasing BG values) and the first data points in Figure 3B (measured before carbohydrate intake) a polynomial trend might no longer be visible.

Rate of change should be calculated from the source values rather than trend lines (which was probably done, but then there is no need for trend lines).