This paper addresses an important topic.

It is well established that insulin degludec (IDeg) shows lower rates of symptomatic, nocturnal and severe hypoglycaemia than insulin glargine U100 in patients with type 1 diabetes, however, studies in patients with high hypoglycaemia risk are scarce and data on patients with high risk of nocturnal hypoglycaemia are missing.

The SWITCH-1 trial included patients at some risk of hypoglycaemia, but has been criticised for very tight glycaemic targets potentially inducing hypoglycaemic events. Pedersen-Bjergaard et al. compared degludec and glargine (in combination with insulin aspart) in a cross-over trial with treatment periods of one year for each insulin in patients with recurrent severe nocturnal hypoglycaemia (median 1 event in the past 2 years).

Individual treatment targets were defined for the study participants, end of study HbA1c-levels were considerably higher than in SWITCH-1 (7.7-8.0% vs. 6.8-6.9%) and not different between treatment groups.

Details on the study design were published previously (reference 12 of the manuscript). Despite the relatively small sample size (n=149) the investigators found significant reductions in level 1 and level 2 symptomatic hypoglycaemia as well as in all-day severe hypoglycaemia.

The primary endpoint of the study was nocturnal severe hypoglycaemia which was not different between treatments (p=0.08).

The study was well executed and the results are presented nicely.

This reviewer only has minor comments.

1. Abstract: It should be clearly mentioned that the primary endpoint was not different between treatments. This comment also applies to the Results Section on severe hypoglycaemia (pages 11 and 12 according to the authors’ page numbers).

2. Statistical methods: Sample size considerations were given in the “design paper” (reference 12), but not in the current manuscript.

This is of course acceptable, however, as the primary endpoint did not reach statistical significance it would be important to present the statistical power for differences in this endpoint based on the incidence observed in the study.

This is of particular importance as the “design paper” did not include a formal power (or sample size) analysis stating that this is difficult for hypoglycaemia rates because of the skewed distribution.

1. Neither the “design paper” nor the current manuscript states how the authors corrected for multiple testing (if they did). This should be clarified.

2. Level 2 hypoglycaemia was defined as blood glucose concentrations below or equal to 3.0 mmol/l rather than below 3.0 mmol/l as recommended by International Hypoglycaemia Study Group. These recommendations were not yet available at study start, so the small difference in definition is understandable. It would still be interesting to see the incidence of level 2 hypoglycaemia as now defined (below 3.0 mmol/l) in a post-hoc analysis.

3. Insulin dose: The presentation on insulin doses is not quite clear to this reviewer. The authors report slightly higher mean total insulin doses for degludec versus glargine at the end of the two maintenance periods (42.7 U vs. 42.5 U for the first period, 49.2 U vs. 48.1 U for the second period).

The statistical analysis, however, showed significantly lower doses for degludec (p=0.04). The same applies to mean basal insulin doses. This should be clarified.

1. The authors did not discuss the time distribution of severe hypoglycaemia with insulin glargine (Suppl. Figure 1b). It might be worth pointing out that the incidence of hypoglycaemic events declined at about 12 hours post-dosing (in particular at midday and afternoon) which is very much in line with the PK/PD-characteristics of insulin glargine.

It might be worth to shortly discuss this point as one could hypothesise that BG-levels might have slightly increased with glargine during this time (to be investigated with CGM in other trials).

1. Table 1: Insulin dose should be specified as (I)U/day (rather than IU/day). International units only apply to human insulin formulations (e.g. NPH-insulin), whereas all analogue insulins are expressed in units only.

2. Figures 2a and 2b: It should be indicated which side of the line of unity (RRR=0) favours degludec or glargine.

All my comments were well addressed. I think this is a nice paper for publication in DOM.