Scientific foundation for ithlete


Simon Wegerif

Founder Simon Wegerif explains how ithlete was designed in response to an athlete’s genuine need to train smart.

‘Like many endurance athletes, I had used a heart rate monitor for years. It helped me measure how hard my body was working during training as well as keeping me just to the right side of the red line during competition. In search of more effective ways to develop my limited talents, I began investigating smarter training methods that would improve my performance whilst avoiding injury due to overtraining. This is how I found out about heart rate variability (HRV).

I was quickly convinced that I should incorporate HRV in my own athletic training. I looked around for a commercial product but could find nothing that included important measurement criteria such as colour-coded readiness indicators, the ability to visualize trends graphically, and the simplicity of a quick morning test that anyone can fit into his or her daily routine.

I concluded that there was no easy-to-use, affordable product that provided a convenient daily measure of HRV. As a professional engineer, with a long career in consumer electronics & signal processing, building such an HRV product seemed a worthwhile challenge.

During 2009, I read over 500 research papers on HRV and consulted cardiologists, coaches and trainers. My device had to be scientifically valid, practical, and uncomplicated to use. During this research, and following patent applications in the US, UK and EU, I was invited to establish sports and medical research collaborations on HRV application in cardiac rehabilitation and elite sports training.  By mid 2009, I had completed the first working ithlete smartphone app and a prototype receiver which was the first able to record heart rate on an iPhone.’

The following material is based on extracts from US patent 8666482 and links to relevant supporting research, and explains how ithlete measures HRV and creates training indications for the user.

Scientific basis for the ithlete measurement

Choice of HRV parameter

Many parameters have been for measuring HRV over the last 30 years. They fall into three types: time domain, frequency domain and entropy (chaos) measures.  Very few of these are suitable for use outside the laboratory by untrained users with limited time available for measurement.  We chose RMSSD (Root Mean Square of the Successive Differences), a time domain parameter that correlates very highly with more complex frequency domain measure HF (High Frequency), without requiring a breathing rate greater than 9 breaths per minute or stationarity of underlying heart rate. The raw RMSSD measure has poor statistical properties, so I decided to apply natural Log (Ln) transformation, which allows common statistical measures such as standard deviation (SD) and coefficient of variation (CV) to be used. This article provides a more in-depth rationale for this choice.

Measurement duration

Users need to fit the measurement into their daily morning routine, so it is important that it should not take very long.  We established that 1 minute is sufficient for a very good level of scientific validity whilst being short enough to maximize user compliance. Although just 30 seconds is sufficient for RMSSD from a signal processing perspective, the 1 minute measure of LnRMSSD has now been validated in a peer reviewed journal paper, which reports an intraclass correlation of 0.98 (0.93, 0.99) and 0.0 bias (LoA 0.22) for the 1 minute measure compared to the criterion measure of 5 minutes.

Measurement time of day

Morning wake was chosen as the optimum time of day for the following reasons:

  • Measurement first thing in the morning provides an indication of recovery following sleep without influence from food and drink (including caffeine), and daily physical or mental stresses.
  • Knowledge of the HRV value first thing in the morning allows the user to alter training or activity plans for the day ahead.
  • Relevant, commonly used subjective recovery parameters such as sleep quality, general fatigue, stress and muscle soreness can be recorded at the same time as the HRV measure, giving a more complete picture of the impact of lifestyle factors impacting recovery and readiness to train.

Sensor types

Sensors need to be carefully selected and well cared for, as the precision measurement of every interbeat interval is a demanding task, and much more so than giving an indication of average heart rate at rest or during exercise. Traditionally, medical grade ECG has been used for HRV studies and clinical practice.  (See this article’s charts for more on the evolution of heart rate sensors suitable for HRV use) Polar™ type chest band sensors have been validated for HRV use, and these are now available with Bluetooth Smart™ (BLE, 4.0) transmission, but not all manufacturers’ products transmit the required R-R intervals over Bluetooth, and even worse, some transmit inaccurate data. Certain pulse sensors have also recently been validated for HRV measurement in smartphone applications.

Researchers at the University of Alabama have now performed a comprehensive validation of the ithlete Finger Sensor in a young athletic population in seated, standing and supine positions. Full text is available here.

Measurement accuracy

HRV was first analysed in clinical settings using hand measurements of ECG rhythm strips, and then by recorders which digitized at 128Hz, giving an accuracy of approx. 8 milliseconds.  This has steadily improved over the years to 1-2 milliseconds and this is the range chosen also for ithlete, with the strap based measurements aiming for 1-1.5 milliseconds and the finger pulse sensor for 2 milliseconds precision.  Insufficient accuracy leads to bias in the results, which is not constant, but dependent on the measured value.

Validation of the ithlete measurement accuracy when used with a chest strap can be found here.

And for the finger sensor here.

There has been interest recently in the use of smartphone cameras for HR & HRV measurement.  Reasonable estimates of resting HR can be obtained using this technology, but the time resolution of the camera at 30-60 frames per second is far too low for reliable HRV measurement.

Measurement repeatability & breathing rate

Many studies have shown that HRV is affected by respiratory activity – both breathing rate and depth. HRV measurement using LnRMSSD under free breathing resting conditions is not reliable (Coefficient of Variation = 12% reported by Al Haddad et al 2011). However, researchers have found that controlled (paced) breathing overcomes the problems experienced with free breath and improves reliability and repeatability of HRV results.

When implementing paced breathing, it is important to avoid stressing the user by making them breathe at a significantly different rate to that which they are comfortable with.  The rate of 7.5 breaths per minute chosen for ithlete back in 2009 has recently been endorsed by researchers who found that this was the average free breathing rate of a group of club runners at rest (Saboul 2013).

HRV result scale

The raw transformed LnRMSSD measure of HRV occupies a range of 2.5 to 4.5 for healthy individuals and as such is not very friendly or intuitive for non scientific users. A scaling factor of 20 was introduced in order to improve intelligibility so that the fittest athletes would tend to be in the range 90-100.  The scale does not have a hard stop, and occasionally ithlete readings as high as 115 have been reported.  This represents very marked respiratory sinus arrhythmia where the highest heart rate during inspiration is about twice that observed at the end of expiration.

Another beneficial property of the 20x scaling is that one unit represents about 0.25 standard deviations for a user who is careful with their measurement practice, and is therefore the lowest value likely for the smallest worthwhile change (SWC, W. Hopkins). This property means that whole numbers can be reported and eliminates the need for values after the decimal point.

Baseline comparison

Since the first prototypes of ithlete, a 7 day weekly moving HRV average has been used as an individual baseline from which to identify significant changes.  This has also been validated as best practice in this review by Plews (2013).

Identifying normal & abnormal daily readings

Daily readings are compared with the baseline described above to determine whether they are within the normal range.  The Smallest Worthwhile Change (SWC, Hopkins) is set at 1 standard deviation (SD) for the purposes of traffic light user indications. If a reading is more than 1 SD below baseline, it is marked amber as a warning to the user.  If the following day’s reading also meets the same criteria, it is marked red as a strong warning to the user that their HRV is below normal range.

As well as flagging HRV values below normal for the user, the ithlete algorithms also indicate HRV values very significantly above baseline as signs of possible parasympathetic dominance/ adrenal fatigue, especially when combined with abnormally low resting HR.

Identifying normal & abnormal trends

HRV response to progressive overload in runners was studied by V. Pichot et al (2002), and their findings are embedded in the ithlete weekly and monthly change indications to flag when overload may become maladaptive (i.e. excessive and difficult to recover from). Indications are progressive with amber and red flags. Conversely, during taper (the period at the end of a training block when loading is reduced) and rest periods, beneficial upwards trends are flagged green. Rapid rises in HRV over successive days which are unlikely to be the result of tapering but rather onset of adrenal fatigue are also flagged.

Additional data recording

As well as HRV and resting HR, ithlete also allows users to capture the following objective & subjective variables:

  • Training load (arbitrary units)
  • Sleep quality (Visual Analog Scale, VAS 1 -9)
  • Fatigue (VAS 1 -9)
  • Muscle soreness (VAS 1 -9)
  • Stress level (VAS 1 -9)
  • Mood (VAS 1 -9)
  • Diet (VAS 1 -9)

These are presented graphically so that users can visually identify relationships between HRV and lifestyle factors, allowing them to experiment with changes.

Pin It on Pinterest