A Quantitative Systematic Review of Normal Values for Short-term Heart Rate Variability in Healthy Adults

David Nunan, Ph.D.; Gavin R. H. Sandercock, Ph.D.; David A. Brodie, Ph.D.


Pacing Clin Electrophysiol. 2010;33(11):1407-1417. 

In This Article


Results of Literature Retrieval for Normal Values of Short-term HRV

From over some 3,100 citations, only 44 reported short-term measures of HRV in healthy adult participants (n ≥ 30) and were in accordance with Task Force methodological standards/recommendations. The number of studies was limited by the following factors:

  • Many studies of HRV assessed longer term 24-hour monitoring;

  • Studies were powered for the use of small sample sizes;

  • Studies often include clinical populations without the inclusion of a healthy cohort and/or reference to healthy values;

  • Adherence to the Task Force methodological recommendations was poor.

Some of the factors pertaining to the above findings can be more easily explained than others. A preferred use of 24-hour measurements to that of short-term measurements could lie in their greater prognostic power,[3,52–54] or the additional information such as night:day ratios that can only be determined from 24-hour monitoring. A more plausible explanation lies in the fact that many studies of HRV are retrospective in nature, reporting data from 24-hour Holter monitoring carried out as part of standard cardiac assessment.

The fact that studies utilize only a small sample size may be explained by the nature of the study, limitations in resources, and/or the calculations of statistical power.[55] Other factors, such as the failure to report the actual values for measures of HRV, were found to occur when studies were interested in change scores[56] or preferred to present results graphically.[57]

The failure to report mean RR interval by 54% of the studies is a concern. Because of the reciprocal nature of HR and mean RR interval, studies reporting measures of HRV often choose to report only mean HR[36,43] or in some cases, neither.[25,41] This error can be likened to assessing the suspension behavior of a car without acknowledging the car's speed. Such errors also reflect, on the part of both author and publishing editor, failures in understanding of the fundaments of HRV data and their analysis.

Thirty-six percent of included studies reported TP and VLF which are not recommended from short RR recordings due to their ambiguous physiological meaning under such conditions.[1] The use of units that differ from standard units (e.g., beats per minute/√Hz[58]) further limited the number of eligible studies. When such studies are published, they reflect a weakness in adherence to Task Force recommendations. This also demonstrates a lack of coherence between authors and editors as to how and what to present when reporting short-term measures of HRV.

Comparisons Between Literature and Task Force Values

The Task Force does not provide norm values for short-term time-domain measures of HRV and therefore comparisons can only be made between spectral measures. The Task Force figures are as follows: 1,170 ms2 for LF power, 975 ms2 for HF power, 54 and 29 for normalized LF and HF, and 1.5–2.0 for the LF:HF ratio. The Task Force LF value is more than 1.5 SD above the mean literature value (519 ms2). The Task Force HF value is also higher compared with that from the literature (657 ms2). Task Force and literature-normalized measures of LF and HF power are more homogenous but the Task Force value for LF:HF (1.5–2.0) is considerably lower than the value gained from the literature (2.8).

Reasons for these discrepancies could be due a number of factors including differing characteristics of participants and differences in spectral decomposition methods. The studies from which the norms were obtained were not cited by the Task Force authors so comparisons in terms of participants are not possible. The Task Force report does provide details as to the frequency bandwidths used for determining LF and HF power distributions. Oscillations in RR intervals occurring at LF were assessed between 0.04 and 0.15 Hz and at HF between 0.15 and 0.4 Hz. Forty-seven percent of the studies presented here report values for LF and HF power obtained at frequency bandwidths differing from those recommended by the Task Force. Some considered oscillations in heart periods at frequencies as low as zero to 0.003 as part of the LF component.[19,40] Others utilized much lower cutoff values (0.3 Hz) for the HF component.[21] Discrepancies in LF and HF frequency bands could lead to the inclusion and/or exclusion of oscillations of differing physiological origins and would certainly result in varying values for LF, HF, and/or both. It is both interesting and somewhat telling then that these studies report some of the largest discrepancies for spectral measures of HRV.

From Table SI, it can be seen that the following population-based studies report values for short-term HRV measurements from large samples (~1,000): Rennie et al.,[6] Kuo et al.[18] Dekker et al.,[20] Liao et al.,[32] Hemingway et al.,[36] Britton et al.[43] On closer examination, a number of these studies were based upon ongoing longitudinal and/or cross-sectional assessments of the same participant populations. While these studies present different sized samples and were testing different hypotheses, there is a potential for significant overlap between their respective samples. This may explain the similarity in values between Dekker et al.[20] and Liao et al.[32] and among Rennie et al.,[6] Hemingway et al.,[36] and Britton et al.[43] (Table SII). For these reasons, it could be argued that only three large populations have been assessed since the 1996 Task Force report.[6,18,43] Moreover, the lowest participant age across these three populations was 40 years. This means that there are currently no published data for short-term HRV measures obtained in a large population including adults aged less than 40. The negative relationship between HRV and age may also explain the relatively low values for HRV measures observed by these studies. The impact these large samples have on the mean publication values presented here is also noteworthy.

Studies Reporting Discrepant Absolute HRV Values

Approximately 85% of studies demonstrated values within 1.5 SD of the mean publications value for one or more short-term HRV measure. Closer scrutiny of the 15% of studies demonstrating values greater than 1.5 SD can help identify conditions leading to disparate values for short-term measures of HRV. Discussion of the following studies demonstrating discrepant values will adopt a measure-by-measure approach: Melanson[21] (mRR, SDNN, rMSSD, LF, HF), Sandercock et al.[34] (LF), Evrengul et al.[40] (SDNN), Mehlsen et al.[48] (SDNN), Sandercock et al.[50] (SDNN, rMSSD), Nunan et al.[51] (LF).

A closer look at the characteristics of the above studies revealed a number of similarities and differences related to study participants, RR interval data recording, artifact identification, and interpolation and spectral decomposition protocols. As these factors can have differing effects depending on the measure, they will be discussed separately for time- and frequency-domain measures, respectively.

Time-domain Measures

The high RR values reported by Melanson[21] and the high SDNN values reported by both Melanson[21] and Sandercock et al.[50] might be explained by their use of young and moderate-to-well trained participants. There is a well-established link between age and HRV, with a decrease in HR for increasing age with younger individuals demonstrating higher values.[1,16,18,59] SDNN is also a function of the recording length, with longer analyzed recordings producing larger values.[60] For this reason, the Task Force recommends a standardized duration of 5 minutes for short-term SDNN (and other measures of HRV). These factors most likely explains the larger values observed by Evrengul and colleagues[40] who determined the SDNN of RR interval data recorded over a 1-hour period. No justification for such a recording length was given by the authors.

Parasympathetic nerve traffic enacts its effects at a much faster (<1 second) rate than sympathetic outflow (>5 seconds); therefore, beat-to-beat changes in RR intervals (rMSSD) are considered a reflection of vagal outlfow.[52,54] Measures of rMSSD are highly variable under conditions of enhanced vagal outflow.[61] One such condition is paced breathing, particularly in the supine position. In addition, the bradycardia observed for more highly trained individuals is commonly accompanied by augmented markers of cardiac vagal modulation,[62,63] although this relationship is not always observed.[64] The discrepant values for rMSSD reported by Melanson[21] and Sandercock et al.[50] are likely to result from the combined effect of young, trained individuals with higher baseline vagal tone and the use of supine and paced breathing protocols.

Frequency-domain Measures

A number of human and animal studies have demonstrated findings of both sympathetic[65–67] and parasympathetic[68] origins for LF oscillations and spectral power. An augmented and diminished LF power under parasympathetic blockade has implications for studies where vagal conditions are enhanced, such as during paced breathing conditions.[68] The higher values observed by Melanson[21] may be the consequence of a vagally mediated augmentation of LF power resulting from the paced breathing condition.

In healthy normotensive controls, a value of 82 ms2 was reported by Piccirillo et al.[33] Moreover, this value was used to determine "abnormal" HF power in chronic heart failure (CHF) patients. Inclusion of these values in the present study may explain the lower overall mean value for HF power. An important observation is that these values are considerably lower than the Task Force norm value for HF and the mean studies value presented here. As is common throughout the literature, consideration as to the "normality" of the so-called "healthy" values is ignored.

Spectral measures are highly sensitive to technical errors within RR data such as artifacts, misplacement of missing data, poor pre-processing, and nonstationarity. Information regarding error detection methods for 1-hour Holter RR interval data was not provided by Evrengul et al.[40] and no indication as to the number of errors observed and/or removed was given. The fact that Mehlsen et al.[48] do not report the performance of any error identification, removal, and/or correction procedures suggest a failure to understand the importance of correct RR interval data in the analysis of its variation. RR intervals were also considered to be "within normal range," yet the authors provide no reference for this so-called "normal" range.

The Task Force[1] recommendations stress the need for manual editing of RR interval data. Evidence of a strong prognostic value for fully automated measures of HRV[12] and their accurate and reliable determination compared to traditional methods[34,51] suggests that the Task Force[1] recommendations may be outdated. At the very least, they require updating to account for the computational power of current automated RR recording and HRV analysis devices.

Studies Reporting Discrepant Log-transformed HRV Values

Of the studies reporting log-transformed measures of HRV, only one demonstrated discrepant values for HRV measures.[12] In the study by Ho et al.[12] data for spectral measures of HRV were obtained in a healthy control group matched for age and sex to a group of patients suffering from CHF. The participants in the control group were 44% female, with a mean age of 72 years and a resting HR of 76 beats/min. There is a well-known age-related decline in HRV that particularly affects measures related to vagal modulations of HR in females.[18] Data presented elsewhere demonstrate a negative correlation between HR and spectral measures of HRV.[69] These two factors alone may explain the low values for LF (2.05 ln ms2) and particularly for HF power (0.08 ln ms2) observed by Ho et al.[12] As with the majority of studies utilizing a control "reference" group, the values presented in the control group are not questioned by the authors as to their normality/abnormality.

Summary of Main Factors Underlining Discrepant Values in Short-term HRV from Healthy Individuals

The measure-by-measure analysis performed for those studies reporting discrepant values revealed a number of underlying factors including:

  1. Moderate to high level of participant habitual physical activity;

  2. The use of paced breathing protocols, particularly when performed in participants with moderate to high physical activity levels;

  3. Where younger participants are measured, values for HRV are typically higher;

  4. Poor reporting and/or performance of RR interval error recognition, removal, and/or correction procedures;

  5. The use of differing frequency bandwidths and normalization methods for LF and HF spectral measures;

  6. Wide variation in HRV measures between healthy participants of the same study;

  7. The misclassification of participants as healthy;

  8. A failure of studies to recognize the normality/abnormality of values obtained in healthy participants.

Some of the points above (1, 2, 3, and 6) were not unexpected. Of some surprise was the failure to perform error correction procedures by a number of studies and the poor reporting of these procedures by others. The last three summary points are particularly important and highlight the inherent problem of defining a so-called "normal" HRV.

These points are also inter-related in that the failure to question the normality of data when obtained in healthy participants possibly stems from the fact that even in homogonous healthy groups, measures of HRV can display wide interindividual variations (as high as 260,000%, Fagard et al.;[11] Table IV ).

It is important, however, to recognize other factors could influence discrepancies between studies. Measures of HRV are influenced by diet (caffeine and alcohol intake) and physical and mental stress. Very few of the studies included here include information on these factors and their impact on values presented cannot be determined. When assessing studies reporting so-called normal HRV, readers should employ close scrutiny of the factors outlined above as well as potential other factors (e.g., diet, stress) related to the individual aspects of each study. With consideration of these factors, the data presented in this study may provide users of HRV with reference ranges by which to determine disparate values for common measures of short-term HRV.