Introduction

Fear of hypoglycaemia is a major barrier to physical activity among individuals with type 1 diabetes [1]. As such, most research to date has examined the acute glycaemic effects of different activity types to refine hypoglycaemia prevention guidelines. Most studies have focused on activity intensity and duration, in addition to its timing relative to meals, snacks and insulin adjustments.

Recently, interest in anaerobic (i.e. high intensity) exercise has increased due to its association with smaller declines in blood glucose and potentially fewer occurrences of hypoglycaemia. Anaerobic exercise is defined as intense, short-duration physical activity, where the fuels provided for the contracting muscles are independent of oxygen availability [2]. Weight lifting, sprinting and sustained high intensity (generally >70–80% of peak aerobic capacity [V̇O2peak) cycling, running, swimming, etc. are all considered ‘anaerobic’. As these activities cannot be sustained for long durations, they are often combined with periods of either low intensity aerobic activity or resting recovery. These exercise sessions are described as high intensity interval exercise (HIIE).

Since the 1980s, researchers examining exercise in the context of type 1 diabetes have generally believed that high intensity (anaerobic) activities increase blood glucose levels while aerobic exercise is associated with blood glucose declines. Indeed, even the most recent literature reviews and consensus statements around exercise/physical activity and type 1 diabetes make such statements [3, 4]. Many recommendations about starting blood glucose targets and insulin adjustments for exercise are also based on these foundations [5]. Is it possible, however, that these ideas are rooted in a decades-long misinterpretation of data?

Many studies of anaerobic exercise were undertaken in the morning, after an overnight fast. The decision to have participants perform these protocols while fasting was probably taken to avoid the complications of making insulin adjustments before exercise, and indeed to decrease the risk of hypoglycaemia from elevated insulin levels. Early morning, before the first meal of the day, is the only time that individuals are likely to be in a truly fasted, or post-absorptive, state. Most of the day is spent in a postprandial state, where remnants of the last meal may enter the blood stream as late as 4 to 6 h after it was consumed. In individuals with type 1 diabetes, it is also likely that some of the prandial insulin taken with that meal will still be circulating. In the fasted state, however, in addition to lower insulin levels, fuel selection both at rest and during exercise is very different, which may have had an unintended impact on blood glucose outcomes.

Sustained high intensity exercise

Several early studies of anaerobic activity in individuals with type 1 diabetes demonstrated clear blood glucose increases (and subsequent hyperglycaemia) following sustained (~10–12 min), high intensity (≥80% of V̇O2peak, or incremental tests to exhaustion) exercise. Mitchell et al [6], observed an increase (from 4.8 ± 0.2 to 7.1 ± 0.4 mmol/l) in blood glucose levels throughout exercise and up to 2 h following exercise when six (four female/two male) participants with type 1 diabetes performed sustained high intensity (80% V̇O2peak) cycling until exhaustion (10 ± 2 min). Similarly, when six adult male participants with type 1 diabetes performed an incremental workload test to exhaustion (~12–14 min), Purdon et al [7] measured an increase in blood glucose levels from 4.8 ± 0.2 mmol/l pre-exercise to a 2-min post-exercise peak of 7.9 ± 0.5 mmol/l. Blood glucose levels did not return to baseline after 2 h of recovery. Sigal et al [8] measured a comparable glucose increase (+2.65 ± 0.32 mmol/l) from the start of exercise to the glycaemic peak 4 min into recovery, when six young, fit male participants with type 1 diabetes performed sustained high intensity (89–98% V̇O2peak) cycle ergometer exercise (12.3 ± 0.7 min). In all three studies, participants had very good glycaemic management (close to normal HbA1c). Importantly, another design element common to all three studies (in addition to the sustained, anaerobic exercise) is that they were performed after an overnight fast.

Over the past 5 years, we have performed similar incremental workload tests to exhaustion for V̇O2peak tests in our lab on 34 individuals with type 1 diabetes (25 female; aged 29.2 ± 8.3 years; mean V̇O2peak=38.7 ± 7.7 ml kg-1 min-1; HbA1c=61 ± 3 mmol/mol [7.7 ± 1.0%]). The mean test duration was similar to those listed above (9.7 ± 3.0 min). All tests were performed within 3 to 5 h of consuming a meal. In stark contrast to the studies described above, the mean blood glucose change was a decrease (−1.0 ± 2.3 mmol/l, p=0.03) from the start to the end of exercise. More participants (n=21) experienced declining blood glucose than those who experienced an increase (n=13) (Fig. 1). While it can be argued that age, sex or physical fitness could have affected blood glucose outcomes [9], there was no consistent effect of these characteristics when assessed with regression analyses.

Fig. 1
figure 1

Number of participants with type 1 diabetes in each category of glycaemic change during a non-fasting maximal aerobic capacity test

Full size image

Resistance exercise

Resistance exercise, generally consisting of short bouts of relatively forceful muscle contraction, is also an anaerobic activity. The acute glycaemic effects of this type of activity in individuals with type 1 diabetes has only been studied in the context of weight lifting, and has generally been limited to one type of protocol (a moderate/mass-building protocol involving three sets of eight repetitions) [10,11,12,13,14]. In 2013, Yardley et al [11], observed a blood glucose decrease from 8.4 ± 2.7 mmol/l to 6.8 ± 2.3 mmol/l (p=0.008) when 12 participants with type 1 diabetes (two female; aged 31.8 ± 15.3 years; V̇O2peak=51.2 ± 10.8; HbA1c=54 ± 0.3 mmol/mol [7.1 ± 1.1%]) performed three sets of eight repetitions of seven exercises targeting all major muscle groups. Exercises were performed at the participants’ eight repetition maximum (8RM—the maximum amount of weight that could be safely lifted with good form eight times). Using a very similar protocol (three sets of 8 to 12 repetitions at 60–80% of 1RM, targeting all major muscle groups), ten participants in a study by Reddy et al [10] [six female; age 33 ± 6 years; HbA1c=57 ± 0.3 mmol/mol [7.4 ± 1.0%]) experienced a similar blood glucose decline (−1.33 ± 1.78 mmol/l, p=0.007). Both protocols involved a 90 s rest between sets, were performed roughly 4 to 5 h after lunch, and incorporated insulin adjustments alongside a pre-exercise snack to prevent blood glucose declines.

In contrast, however, four separate publications by Turner et al [12,13,14,15] reported rising blood glucose levels with resistance exercise, despite altering the number of sets of exercise performed [15] and the intensity (i.e. resistance and number of repetitions) of the sets [14]. Importantly, an almost identical protocol to that used by both Reddy et al [10] and Yardley et al [11] (three sets of ten repetitions at ~64% of the individual’s 1RM) also increased blood glucose levels in eight participants (one female; aged 38 ± 6 years; HbA1c=72 ± 0.3 mmol/mol [8.7 ± 1.0%]) with type 1 diabetes [13]. The participants were notably similar across studies: there were slightly more female than male participants (except Yardley et al [11]), all of whom were recreationally active, having managed type 1 diabetes for more than a decade on average, and a similar mean age. As such, it is unlikely that age, sex or physical fitness affected the results [9]. It is more likely that one key aspect of study design was responsible for the differences observed: data collected by Turner et al [12,13,14,15] involved participants exercising after an overnight fast, while studies by Reddy et al [10] and Yardley et al [11] started their exercise protocols between 16:00 and 17:00 hours with participants having consumed a meal in the previous 4–5 h.

HIIE

If it can be believed that sustained high intensity exercise consistently increases blood glucose levels in individuals with type 1 diabetes, then logic would dictate that performing this type of activity intermittently with periods of rest or low intensity aerobic recovery between them should lead to relatively stable blood glucose levels during exercise. Indeed, HIIE has been associated with a lower risk of hypoglycaemia during exercise compared with aerobic exercise [16,17,18,19,20,21,22,23,24], although it may increase the risk of post-exercise hypoglycaemia [25]. However, glycaemic outcomes from HIIE studies involving participants with type 1 diabetes have been very inconsistent, with some showing no change [17, 18] or increasing blood glucose [16], with others showing non-significant [23, 24] or significant [19,20,21, 26] blood glucose declines by the end of exercise. Participant characteristics such as age, sex and physical fitness may play some role in the variable responses [9], as could discrepancies among the exercise protocols themselves (interval intensities between 85% V̇O2peak and supramaximal, interval length between 4 and 30 s). However, the only thing consistent about the studies where blood glucose declined is that they were all performed 3 to 5 h after a meal [19,20,21, 23, 24, 26].

Fasting vs fed repeated measures studies

A handful of small studies, using repeated measures study designs, have compared fasted vs fed exercise using a single group of participants with type 1 diabetes within each study. Where anaerobic exercise is concerned, two separate studies (both n=12) showed clear blood glucose decreases with afternoon resistance [27] and HIIE [28] (performed ~4–5 h after a meal), while the same protocols performed by the same participants after an overnight fast resulted in an increase or no change in blood glucose, respectively. Even aerobic exercise, thought to consistently decrease blood glucose levels [3, 4], can increase blood glucose when performed under these conditions [29, 30]. A small study by Ruegemer et al [29] (n=6; three female participants; age 30 ± 4 years; HbA1c=58 ± 0.2 mmol/mol [7.5 ± 0.6%]) measured an increase in blood glucose (6.7 ± 0.4 mmol/l to 9.1 ± 0.4 mmol/l) over 30 min of moderate (60% V̇O2peak) aerobic exercise performed after an overnight fast, which was absent when participants performed the same protocol at 16:00 hours (4 h after lunch) on a different day. Similarly, Yamanouchi et al [30] found a blood glucose decline (15.3 ± 3.0 to 11.0 ± 0.7 mmol/l) during a walking protocol 2 h after breakfast (09:00 hours), when the same protocol at roughly the same time of day (07:00 hours) was associated with no blood glucose decline when performed before breakfast. Overall, the trend from these studies is clear: on average, blood glucose does not decline, and will generally increase when exercise is performed after an overnight fast, regardless of exercise intensity.

Fasted exercise metabolism in type 1 diabetes

There are several explanations for the observed glycaemic trends. The first, and easiest, explanation is that circulating insulin levels are lower after an overnight fast than they are mid- to late afternoon. However, this on its own would not explain rising blood glucose levels regardless of exercise modality. An alternate explanation involves lipids being prioritised as a fuel source when the body is in a fasted state [31]. As triglycerides are metabolised, the resulting glycerol could act as a gluconeogenic precursor, thereby increasing blood glucose. The resulting increase in circulating NEFA from lipolysis (which are higher after fasted exercise) [32] could also temporarily increase insulin resistance [33], which would account for rising blood glucose levels during exercise, and persistent hyperglycaemia [16, 27] after exercise, when performed after an overnight fast.

High growth hormone levels have been associated with the ‘dawn phenomenon’, an early morning blood glucose surge experienced by individuals with type 1 diabetes [34, 35]. While this phenomenon itself may be responsible for some of the higher glucose levels experienced with fasted exercise, the underlying exercise-induced growth hormone elevations (which are higher with greater intensity [36], and enhanced by fasting [37]) may play a role in increasing blood glucose levels with exercise intensity and duration. In a previous study of individuals with type 1 diabetes, higher growth hormone was associated with glucose sparing [38] during exercise late in the day and may be playing a greater role when the activity is performed after an overnight fast.

Are these responses universal?

Anyone with type 1 diabetes who is physically active, and anyone who has involved individuals with type 1 diabetes in exercise studies, knows that blood glucose responses to exercise are extremely variable. Most of the trends discussed above are the mean changes in blood glucose in a group of participants. Data from non-fasted incremental V̇O2peak tests performed in our lab over the past 5 years (n=34) produced blood glucose changes ranging from −5.4 mmol/l to +3.5 mmol/l (Fig. 1). There were no clear differences between those using multiple daily injections of insulin vs those using insulin pumps, between male and female participants, between higher fitness (V̇O2peak >40 ml O2 kg-1 min-1) and lower fitness, or between younger (<30 years) and older individuals. It is likely that a combination of factors, including the timing of previous insulin injections, dictate whether glucose increases or decreases during one of these tests. However, it should be noted that the mean blood glucose change over all of these tests (performed 3–5 h after a meal) was −1.0 ± 2.3 mmol/l (p=0.03), rather than the increase one would expect if sustained high intensity exercise had a consistent hyperglycaemic response. These findings are consistent with a large observational study (n=5157) where self-reported data indicated a decrease in blood glucose levels in the majority of participants (75.8%) during exercise, regardless of the type or intensity [39].

As can be seen in Fig. 2, changes in glucose during fasted resistance exercise (Fig. 2a [methods described elsewhere [27]]) and fasted HIIE (Fig. 2b [methods described elsewhere [28]]), show that most participants experience rising blood glucose levels during fasted anaerobic exercise (9 out of 12 and 10 out of 12, respectively). Riddell et al [16] also noted this response among 16 participants with type 1 diabetes (four female; aged 34.7 ± 10.3 years; HbA1c=54 ± 0.3 mmol/mol [7.1 ± 0.8%]) performing a combination of interval cycling and multimodal training during a 25 min exercise session after an overnight fast. Where each participant performed the same session four times, increasing glucose was observed in 62 of the 64 exercise sessions.

Fig. 2
figure 2

Changes in blood glucose in individual participants with type 1 diabetes during 45 min of (a) fasted resistance exercise and (b) fasted high intensity exercise. N=12 for both datasets. Each solid line represents a different individual and mean glucose is represented by the dashed line

Full size image

So what?

Current exercise safety guidelines for individuals with type 1 diabetes suggest that those performing anaerobic activities can start with lower blood glucose levels [3]. Some even suggest that bolus insulin before exercise may be appropriate [5]. These recommendations are based on the assumption that anaerobic activities cause blood glucose levels to rise during exercise. However, from the literature available, this only seems to occur somewhat consistently when exercising after an overnight fast. While seasoned athletes with type 1 diabetes will generally have learned to manage their blood glucose levels accordingly, these recommendations could be misleading and may cause hypoglycaemia for less experienced exercisers. Conversely, the aggressive insulin reductions currently suggested for aerobic exercise may be unnecessary (and may actually cause hyperglycaemia) if the activity is performed after an overnight fast, provided that it is not prolonged (i.e. >45 min) [40]. In light of this information, future guidelines and consensus statements for exercise and physical activity in individuals with type 1 diabetes should consider providing different advice based on participant prandial status. Furthermore, for individuals whose fear of hypoglycaemia is a barrier to exercise and physical activity, the data suggest that exercising after an overnight fast may be a safer option.