Neonatal heart rate variability: a contemporary scoping review of analysis methods and clinical applications.

BACKGROUND: Neonatal heart rate variability (HRV) is widely used as a research tool. However, HRV calculation methods are highly variable making it difficult for comparisons between studies.
OBJECTIVES: To describe the different types of investigations where neonatal HRV was used, study characteristics, and types of analyses performed. ELIGIBILITY CRITERIA: Human neonates ≤1 month of corrected age. SOURCES OF EVIDENCE: A protocol and search strategy of the literature was developed in collaboration with the McGill University Health Center's librarians and articles were obtained from searches in the Biosis, Cochrane, Embase, Medline and Web of Science databases published between 1 January 2000 and 1 July 2020. CHARTING
METHODS: A single reviewer screened for eligibility and data were extracted from the included articles. Information collected included the study characteristics and population, type of HRV analysis used (time domain, frequency domain, non-linear, heart rate characteristics (HRC) parameters) and clinical applications (physiological and pathological conditions, responses to various stimuli and outcome prediction).
RESULTS: Of the 286 articles included, 171 (60%) were small single centre studies (sample size <50) performed on term infants (n=136). There were 138 different types of investigations reported: physiological investigations (n=162), responses to various stimuli (n=136), pathological conditions (n=109) and outcome predictor (n=30). Frequency domain analyses were used in 210 articles (73%), followed by time domain (n=139), non-linear methods (n=74) or HRC analyses (n=25). Additionally, over 60 different measures of HRV were reported; in the frequency domain analyses alone there were 29 different ranges used for the low frequency band and 46 for the high frequency band.
CONCLUSIONS: Neonatal HRV has been used in diverse types of investigations with significant lack of consistency in analysis methods applied. Specific guidelines for HRV analyses in neonates are needed to allow for comparisons between studies. © Author(s) (or their employer(s)) 2021. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ.

This is a comprehensive scoping review describing all study characteristics and population details, heart rate variability (HRV) analysis methods and clinical applications of the neonatal HRV literature.

The literature search strategy was developed in consultation with and performed by medical librarians.

This review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Scoping Review guidelines for reporting items.

This review does not include articles published prior to the year 2000.

A single reviewer performed both the screening (SL) and data collection (SL or JL), though all questions or conflicts were discussed among all reviewers.

Introduction

Heart rate variability (HRV) refers to the fluctuation of beat-to-beat intervals over time and is the result of the continuous counterbalancing input from the parasympathetic and sympathetic branches of the autonomic nervous system.1 In neonates, a diversity of studies using HRV have been published during the last 20 years. HRV analyses were used in investigations of brain injuries,2–5 response to pain,6 7 effects of prenatal drug exposure8 9 and to assess physiological maturation of preterm infants.10–14 More recently the gradual integration of heart rate characteristics (HRC) monitoring (HeRO, Medical Predictive Science Corporation) into the neonatal intensive care unit (NICU) setting has also increased the use of neonatal HRV in clinical practice.

HRV indices can be processed in many different ways following three analysis approaches: time domain, frequency domain or non-linear methods.1 Time domain methods use summary statistics of the normal-to-normal beat (NN) interval (or R-wave-to-R-wave (RR) intervals), such as the SD of NN intervals (SDNN). Frequency domain methods use spectral analyses to break down the RR interval time series into a distribution of the power as a function of frequency, which is then broken down into spectral ranges of interest. The third and more advanced type of HRV analysis is the non-linear method, which includes various measures of chaos, unpredictability and self-similarity, such as sample entropy (SampEn) and the detrended fluctuation analysis (DFA).1

In 1996, the Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology published specific guidelines for processing of HRV indices in adults that became the gold standard.15 Nevertheless, the majority of research projects did not strictly adhere to these standards.16 Such standards are important for consistency in the literature so that results of studies may be compared and combined accurately. In neonates, there are no specific guidelines for processing of HRV indices. Thus, in order to understand the full extent of neonatal HRV analysis and its various applications, a scoping review was conducted. The aim was to provide an overview of the current studies and population characteristics, analysis methods used, and clinical research applications in which HRV has been used in the neonatal population.

Methods

Protocol

A comprehensive protocol and search strategy of the literature was developed in collaboration with the McGill University Health Center’s librarians. The study is reported using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) extension for Scoping Reviews.

Eligibility criteria

All articles that analysed human neonatal HRV were included. Articles were limited to the last 20 years (ie, from the year 2000 onwards) to review more contemporary research in neonatal HRV as methodologies and clinical care have evolved significantly making difficult to ascertain validity of studies performed and published more than 20 years ago. Articles in which human neonates of any gestational age (GA) and ≤1 month corrected age had HRV measured in at least five neonates were eligible. Articles performed in fetuses, animals, infants and children above 1 month corrected age, or adults were excluded. Other types of articles that analysed or discussed human neonatal HRV (reviews, case reports, commentaries, methodological papers, etc) were excluded from data collection. A tally of these different types of articles was maintained in order to understand the full degree to which neonatal HRV has been investigated or used.

Information sources and search

The search was performed in Biosis, Cochrane, Embase, Medline and Web of Science databases and included articles of HRV in neonates in all languages, published between 1 January 2000 and 1 July 2020. The search was performed twice: first in October 2017, then updated again in July 2020. Details of the search strategy used for all databases is provided as online supplemental file 1.

Selection of articles

After duplicate removal, all titles and abstracts were screened for inclusion by one reviewer (SL) based on the eligibility criteria. Full-text articles were then obtained and reviewed by two reviewers (SL or JL) to confirm eligibility and extract data.

Data charting process

Two reviewers (SL or JL) extracted the data from the included articles using a predefined form which included the following sections and listed items below.

Study characteristics and population

Study characteristics

(a) Name of the journal, (b) region where the study was performed, (c) country where the study was performed, (d) type of study design and (e) single or multicentre.

Population details

(a) Sample size, (b) GA and birth weight (BW) categories based on the averages (or medians) of infants enrolled within the study (ie, not based on the study’s inclusion criteria) and using the following definitions: term (≥37 weeks GA), preterm (<37 weeks GA), late preterm (LPT; 34–366 weeks GA), moderate preterm (MPT; 32–346 weeks GA), very preterm (28–316 weeks GA), extremely preterm (<28 weeks GA), normal BW (≥2500 g), low BW (<2500 g), very low BW (<1500 g) and extremely low BW (<1000 g).

HRV analysis methods

Time domain parameters evaluated

(a) SDNN: SD of normal-to-normal (NN) or R-wave-to-R-wave intervals, (b) CVNN: coefficient of variation of NN intervals, (c) SDANN: SD of the averages of NN intervals in all 5 min segments of the entire recording, (d) SDNNi: mean of the SDs of all NN intervals for all 5 min segments of the entire recording, (e) RMSSD: root mean square of the differences between adjacent NN intervals, (f) SDSD: SD of differences between adjacent NN intervals, (g) NN50/pNN50: pairs of adjacent NN intervals differing by more than 50 ms (count or per cent), (h) NNxx/pNNxx: pairs of adjacent NN intervals differing by a time other than 50 ms (count or per cent), (i) triangular index: total number of all NN intervals divided by the height of the histogram of all NN intervals, (j) TINN: baseline width of the minimum square difference triangular interpolation of the highest peak of the histogram of all NN intervals measured, (k) histogram analyses: includes measures of skew, kurtosis and so on and (l) other: any other time domain parameter not listed previously.15

Frequency domain

(a) Total power (TP), (b) ultra low frequency (ULF), (c) very low frequency (VLF), (d) low frequency (LF), (e) high frequency (HF), (f) LF/HF ratio and (g) other: any other frequency domain parameter not listed previously. Frequency ranges reported for each parameter were also collected.15

Non-linear

(a) Poincaré plot analyses (SD1 and SD2), (b) alpha (α): exponent of the 1/f pattern (power spectral density), (c) DFA α1 (short-term exponent), (d) DFA α2 (long-term exponent), (e) approximate entropy (ApEn), (f) sample entropy (SampEn) and (g) other: any other non-linear parameter not listed previously.17

Heart rate characteristics

Classified as its own category given its specific combination of HRV parameters, and the gradual integration of Heart Rate Observation System (HeRO) monitors (Medical Predictive Science Corporation, USA) providing these values into the NICU setting.

HRV application: studies were classified into four major groups

Physiological conditions

HRV measured under normal, healthy, physiological conditions. Includes: (a) normative: study provides data on normal or healthy infants either with the purpose of providing normative data or as controls, (b) sleep state: includes quiet sleep, active sleep, quiet awake and active awake statuses, (c) longitudinal: study provides longitudinal data (repeated measures, maturation, etc), (d) age or weight: study examines the effect of age or weight in a cross-sectional fashion, (e) feeding, (f) position: supine or prone, (g) tilt: study examines the effect of different degrees of bed inclination, (h) sex: male or female and (i) other: any other physiological application not listed previously.

Pathological conditions

HRV measured in infants with certain conditions or diseases, including: (a) sepsis, (b) necrotising enterocolitis, (c) infection, (d) respiratory distress syndrome, (e) apnoea of prematurity (AOP), (f) extubation failure: peri-extubation measurements in relation to extubation outcome, (g) hypoxic ischaemic encephalopathy (HIE), (h) other brain injuries (eg, intraventricular haemorrhage, periventricular leukomalacia, etc), (i) seizures, (j) patent ductus arteriosus (PDA), (k) congenital heart disease or defects (CHD) (l) in-hospital mortality and (m) other (ie, any other condition or disease not listed previously).

Response to certain exposures, stimuli or external factors

Includes: (a) pain, (b) medication, (c) sensory stimuli (light, sound or touch), (d) non-nutritive sucking (pacifiers), (e) kangaroo care (skin-to-skin), (f) sucrose, (g) procedures, (h) maternal factors and (i) other stimuli or responses not listed previously. Details regarding the types of medications, procedures and maternal factors were also collected.

Outcome prediction

HRV measured as predictor of outcomes during hospitalisation or beyond. Includes: (a) days on mechanical ventilation, (b) length of stay or postmenstrual age at discharge, (c) repeat hospitalisations, (d) neurodevelopmental impairment, (e) cerebral palsy, (f) behavioural or social assessments and (g) other: any other outcome not listed previously.

Synthesis of results

Data were summarised descriptively using counts (n) and percentages (%), and presented as pie charts, bar graphs or tables. Years of publication were presented in 5-year bins to observe the overall trend of number of publications over time and minimise year-to-year fluctuations. References for the articles of population characteristics, HRV analysis methods, and HRV application groups are provided.

Patient and public involvement

No patients were involved.

Results

Of the 3361 records identified, 286 articles were eligible and included for data collection; a PRISMA flow diagram of the screening process is provided as figure 1. A complete table (Microsoft Excel file) with each of the articles’ data collection items is provided in the public repository Figshare, along with an EndNote file (X7, Clarivate Analytics, Philadelphia, Pennsylvania, USA) with all references organised by each data collection item (figshare.com, DOI: 10.6084/m9.figshare.14750865).18

Figure 1

Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram for record screening and inclusion. HRV, heart rate variability.

Characteristics of included articles

Of the 286 articles included, 275 (96%) were published in English and the rest were in Russian (n=3), German (n=2), Chinese (n=2), Polish (n=1), French (n=1) and Japanese (n=1). Articles spanned 127 different journals, with the top 10 journals that most commonly published about HRV listed in table 1. Most articles were from North America (47%), primarily from the USA (41%) and Europe (32%) (figure 2). Since 2000, the number of publications in neonatal HRV has increased over time, most markedly in the last 5 years (figure 3). The type of study design was primarily observational (n=261; 91%), including case–control, prospective cohort and cross-sectional articles. There were 14 (4.9%) randomised controlled trials or crossover trials, and 11 (3.8%) retrospective or database articles. Most of the articles were single centre (n=251; 88%).

Table 1

Top 10 journals publishing articles concerning neonatal heart rate variability

Journal name	n (%)
Early Human Development	26 (9.1)
Pediatric Research	24 (8.4)
Biology of the Neonate/Neonatology	11 (3.8)
Acta Paediatrica	10 (3.5)
Institute of Electrical and Electronics Engineers (IEEE)	10 (3.5)
Journal of Perinatology	10 (3.5)
Developmental Psychobiology	9 (3.1)
Pediatrics	8 (2.8)
Journal of Pediatrics	7 (2.4)
Physiological Measurement	6 (2.1)

Figure 2

Distribution of neonatal heart rate variability literature by region and nation. (A) Distribution of articles published by region. (B) Distribution of articles published by nation. Other countries included Brazil (n=6), Italy (n=6), Belgium (n=5), Russia (n=5), Turkey (n=4), Switzerland (n=3), Croatia (n=3), Taiwan (n=2), Sweden (n=2), South Africa (n=2), Solvenia (n=2), Slovakia (n=2), Poland (n=2), China (n=2), Austria (n=2), Vietnam (n=1), Serbia (n=1), New Zealand (n=1), India (n=1), Greece (n=1) and Finland (n=1). HRV, heart rate variability.

Figure 3

Distribution of studies published by 5-year intervals.

Study population

Most articles enrolled a small number of infants (table 2). For the largest cohorts, n=6 articles enrolled >1000 patients, with n=3 (1%) between 1000 and 1999 and n=3 (1%) between 2000 and 2999. However, it should be noted that two of the three published articles that had between 2000 and 2999 infants recruited were analyses of the same large cohort study. Term infants with BW ≥2500 were the most common GA and BW categories studied (table 2). BW was more frequently omitted in articles than GA (n=46 vs n=11 not reported) and 18 articles (6.3%) reported including preterm patients, without specifying the ages (table 2).

Table 2

Study population characteristics

	n (%)	References
Sample size
≤20	63 (22.0)	23 26 28 29 31 32 34 42–47 49 51–54 60 64 65 70 72 78 84 91 94 98 99 108–110 112 113 134 143 148 151 154 168 169 174–195
21–49	108 (37.8)	9 20–22 24 25 35–39 41 50 57 59 63 66 67 71 73–75 81 82 85 87 90 92 93 101 103–106 111 118 120 125–128 138 142 147 149 150 152 153 155–158 160 161 167 196–248
50–99	57 (19.9)	2 4 10 13 19 33 40 56 61 62 68 69 76 79 83 97 100 114 115 117 122 129 131 132 136 137 139 141 146 159 163 249–274
100–199	32 (11.2)	48 55 77 80 88 89 95 96 102 107 116 121 123 130 133 144 162 165 275–288
200–499	12 (4.2)	27 30 86 119 124 289–295
≥500	14 (4.9)	58 135 140 145 164 166 296–303
Gestational age category
Term (≥37 weeks)	136 (47.6)	2 4 9 10 13 19 20 23 26–28 30 37 38 40 46 49 50 52 53 55–57 59 60 62 65 69 75–79 81–83 85 86 88–92 94–96 99 102–104 106–108 111 115–118 121–123 127 128 130 136 137 139 141 142 144 151 152 158 162 168 169 174 176 186 188 189 193 194 196–198 202–206 208 209 213–215 219 221–223 228–230 232 234 237 240 241 244 247 249 253–255 258–260 265 266 268 270 274 275 279–281 285–291 294 295 303
Preterm (<37 weeks)	18 (6.3)	10 13 22 30 47 49 102 113 114 141 187 203 208 218 234 241 252 257
Late preterm (34–366 weeks)	15 (5.2)	62 67 71 80 89 95 110 129 134 202 245 266 275 294 295
Moderate preterm (32–346 weeks)	21 (7.3)	29 41 54 64 80 84 101 109 115 133 157 200 201 204 216 239 271 273 296 297 299
Very preterm (28–316 weeks)	75 (26.2)	21 24 25 31–34 36 39 48 51 61 70 80 87 93 98 108 120 131 132 135 136 143 145 146 153–156 159–161 163 165 167 178 182 185 192 195 202 204 207 210–212 220 224 226 227 231 233 235 236 238 242 243 246 248 250 258 261 264 265 269 272 276–278 282 283 292 298 300
Extremely preterm (<28 weeks)	42 (14.7)	35 45 62 66 68 73 74 76 100 105 112 119 124–126 132 138 140 147–150 164 175 177 179–181 183 184 191 199 202 217 225 251 262 263 267 284 293 302
Not reported	11 (3.8)	42–44 58 63 72 97 166 190 256 301
Birth weight category
Normal (≥2500 g)	117 (40.9)	2 4 9 10 13 19 20 23 26 27 29 30 37 38 40 52 53 55–60 62 65 67 69 75 77–79 81–83 85 86 90–92 94–96 99 102–104 106–108 111 115–118 121–123 127–130 134 136 137 139 141 142 144 152 158 162 168 176 186 188 189 193 196–198 202 204–206 209 213 215 219 221–223 230 232 237 240 244 247 249 253 254 259 260 265 266 268 270 274 275 279–281 285–287 289 290 295
LBW (<2500 g)	49 (17.1)	10 41 54 62 64 71 80 95 98 108–110 114 115 120 131–133 141 157 160 163 165 185 195 201 202 204 207 211 216 227 239 243 245 248 258 259 265 266 271 273 275 282 296–300
VLBW (<1500 g)	52 (18.2)	13 32–36 39 45 48 51 61 70 80 84 93 102 113 124 135 136 143 146 153 154 156 159 161 166 178 180 182 192 200 204 212 218 220 224 231 235 246 250 261 264 267 269 272 276–278 283 292
ELBW (<1000 g)	42 (14.7)	21 31 62 66 68 73 74 87 100 112 119 125 126 132 138 140 145 147–150 164 167 175 177 179 181 183 184 191 199 202 217 225 226 236 251 262 263 284 293 302
Not reported	46 (16.1)	22 24 25 28 42–44 46 47 49 50 63 72 76 88 89 97 101 105 151 155 169 174 187 190 194 203 208 210 214 228 229 233 234 238 241 242 252 255–257 288 291 294 301 303

ELBW, extremely low birth weight; LBW, low birth weight; VLBW, very low birth weight.

HRV analysis methods

The list of all HRV analysis methods and their references is provided as table 3.

Table 3

Heart rate variability analysis methods

	N (%)	References
Time domain	139 (48.6)
SDNN	125 (43.7)	4 9 19 21 22 24–28 30–37 40 41 43 44 51 53 73 75 82 84 86–93 97 101–105 107 108 114 116 121 122 127 128 130 134 137 139 141 143 146 148 149 152 153 156 158 161 167 169 175 179 182 187 189–191 193 195 196 198 203 205–207 209 211–213 215 221 222 224 226 229 231 234 235 239 243 244 246–248 250 251 253 255 259 261 266 267 270 272 273 275 279–281 284 285 287–289 294–296 298 303
CVNN	13 (4.5)	26 28 30 40 43 44 46 109 110 149 176 213 242
SDANN	20 (7.0)	9 24 25 44 86 90 91 101 114 123 130 141 175 187 190 193 253 259 261 275
SDNNi	14 (4.9)	9 86 90 91 130 175 187 190 193 248 253 259 261 275
RMSSD	83 (29.0)	9 21–28 30–36 40 53 86–93 101–105 107 108 114 116 121 122 128 130 134 137 141 148 149 152 158 161 175 182 187 189 190 195 203 206 212 213 215 222 224 226 229 231 235 238 239 243 247 248 250 253 259 261 267 273 275 279–281 289 294 295 303
SDSD	14 (4.9)	19 22 28 107 116 122 127 148 149 152 187 190 209 224
NN50	33 (11.5)	9 30–32 40 86 87 90 91 95 103 107 128 134 139 141 148 149 191 195 203 205 206 215 239 243 247 253 259 261 270 273 275
NNxx	15 (5.2)	9 22 28 36 40 84 86 107 161 190 195 206 213 235 259
TINN	7 (2.4)	4 21 86 105 107 169 285
Triangular index	13 (4.5)	9 24 25 36 86 95 101 107 190 253 259 261 290
Histogram	17 (5.9)	21 35 43 44 46 77 86 92 97 105 167 212 226 251 278 296 298
Other*	18 (6.3)	19–36
Frequency domain	210 (73.4)
Total power	78 (27.3)	9 13 20 26 28 30 36 38–41 43 44 46 53 54 74 79 86 87 95 96 99 107 111 116 120 123 125 126 129–131 134 136 141 144 148 149 151 152 155–157 176 180 186 188 195 201 202 204 206 208–210 212 213 216 218 223–226 232 233 241 247 249 253 258–260 265 275 281 286 289
Ultra low frequency	5 (1.7)	86 141 253 259 275
Very low frequency	42 (14.7)	4 9 21 28 45 48 86 102 105 111 120 130 134 141 148 149 152 155 161 169 176 186 190 191 195 202 208 213 222 225 231 235 244 248 253 258–260 275 281 285 291
Low frequency	168 (58.7)	4 9 10 13 20 21 23–26 28 30 31 36–48 51 53 54 57 65 73–84 86–88 92 93 95 96 99 102–105 107 111–113 115 116 118 120 122 123 125–127 129–134 136–138 141 146–157 159 161 169 174 176 178–188 190–192 195 197 199–204 206–210 212 213 216 220–226 229–233 235 237 241 244 245 247–250 253 254 256 258–262 264 265 267 271 273 275–277 281 285 286 288 289 291
High frequency	188 (65.7)	4 9 10 13 20 21 23–26 28 30 31 36–38 40–44 46–48 50 51 53 54 57 63–65 73–84 86–89 92 93 95 96 99 102–105 107 111–113 115–118 120 122 123 125–127 129–134 136–138 141 146–157 159 161 163 165 168 169 174 176–188 190–192 194 195 197 199–204 206–210 212–214 216 218–226 228–233 235–238 241 244 245 247–250 253 254 256–262 264 265 267 268 271 273 275–277 280–282 285 286 288 289 291 294 303
LF/HF ratio	117 (40.9)	4 9 13 21 23 26 28 30 31 36–38 43 44 46 48 51 54 57 78 84 86 87 92 93 95 102 103 105 111 115 116 118 120 122 125 126 129–131 133 134 136–138 141 146 148 149 151–153 155–157 159–162 169 174 176 178 179 181 184–186 191 192 195 197 199 202 203 206 208 209 212 213 216 218–227 229–233 235 245 247 253 254 256 258 259 261 262 264 265 267 271 273 275–277 281 285 286
Other*	41 (14.3)	22 26 28 30 36–71
Non-linear	74 (25.9)
Poincaré	29 (10.1)	28 30 36 84 86 87 92 93 95 102 103 105 107 149 152 158 161 190 191 194 213 222 231 236 243 246 247 267 273
Alpha (α)	4 (1.4)	86 93 98 113
DFA α1	28 (9.8)	30 36 51 72–84 86 87 93 103 212 217 222 226 231 235 246 267
DFA α2	25 (8.7)	30 36 51 72–74 76–83 86 87 93 103 212 217 222 231 235 246 267
ApEn	19 (6.6)	21 24 25 30 36 37 86 87 92 103 105 212 222 226 231 235 243 267 273
SampEn	22 (7.7)	24 25 36 48 86–88 92 93 97 100 103 113 212 222 235 243 252 273 274 296 298
Other*	48 (16.8)	21 24 25 28 30 36 37 43 44 46 48 72–107 240
Heart rate characteristics	25 (8.7)	2 100 119 124 135 140 142 145 164 166 251 263 269 272 283 284 292 293 296–302

*Others are described in detail within the manuscript text (with references).

ApEn, approximate entropy; CVNN, coefficient of variation of NN intervals; DFA, detrended fluctuation analysis; LF/HF ratio, low frequency/high frequency ratio; NN, normal-to-normal beat intervals; NN50, pairs of adjacent NN intervals differing by more than 50 ms (count or per cent); NNxx, pairs of adjacent NN intervals differing by a time other than 50 ms (count or per cent); SampEn, sample entropy; SDANN, SD of the averages of NN intervals in all 5 min segments of the entire recording; SDNN, standard deviation of NN intervals; SDNNi, mean of the SDs of all NN intervals for all 5 min segments of the entire recording; SDSD, SD of differences between adjacent NN intervals; TINN, triangular interpolation of NN intervals.

Time domain HRV

Time domain was the second most commonly employed type of HRV analysis (n=139; 48.6%), mostly by calculating SDNN (n=125; 43.7%) and RMSSD (n=83; 29.0%) (table 3). Some articles reported other time domain parameters (n=18), including: SDNN/SD delta NN,19 log MSSD,20 SDNN/RMSSD,21 variance NN,22 long-term variability,23–25 short-term variability,24 25 interval index,24 25 differential index,24 25 long-term irregularity,24 25 expiration/inspiration ratio,26 percentage of successive changes in NN intervals sustained over two consecutive increases and decreases,27 temporally scaled SDNN,28 temporally scaled CVNN,28 log variance of heart rate,29 IQRs,30 per cent of NN intervals of transient decelerations,31–35 SD of NN intervals of transient decelerations,31–35 average acceleration and deceleration response35 and an undefined variable ‘X’.36

Frequency domain HRV

Frequency domain was the most common type of HRV analysis used (n=210; 73.4%) (table 3). The most frequently used parameters were the LF (n=168; 58.7%), HF (n=188; 65.7%) and LF/HF ratio (n=117; 40.9%) (table 3). Many articles also reported other frequency domain parameters (n=41), including: mid-frequency (MF, reported at various ranges),10 26 37–46 super HF,47 ultra HF,47 VLF/HF ratio,28 48 MF/HF ratio,26 40 41 LF/LF+VLF ratio,48 peak frequencies,41 49 power in other sub-bands,28 HF variability index,50 beta (slope of VLF),51 respiratory sinus arrhythmia,22 47 52–60 vagal tone index,61 62 area under the curve50 63–65 and the Newborn Infant Parasympathetic Evaluation index.36 64 66–71

Among the frequency domain parameters, there were two different reported ranges used for ULF frequency band (upper border limit 0.003 (n=1), 0.0033 (n=2), not reported (n=2)), 13 different ranges for the VLF band, 29 different ranges used for the LF band and 46 different ranges used for the HF band (figure 4 and online supplemental file 2). The most common VLF and LF ranges were 0.003–0.04 Hz (n=8) and 0.04–0.15 Hz (n=60) (figure 4A and B), respectively, which follow the Task Force guidelines for adults.15 For the HF range, the most common was 0.2–2 Hz (n=20), which is different from the Task Force guidelines range of 0.15–0.4 Hz (n=19) recommended for adults (figure 4C). Of the 210 articles using frequency domain analyses, only 17 (8.1%) articles had both their LF and HF ranges following the Task Force guidelines. Nine of these articles also included VLF measures, of which seven also matched the Task Force guidelines for VLF ranges, while the other two used the slightly altered range of 0.0033–0.04 Hz.15 A summary of the number of articles adhering to the frequency ranges from the Task Force guidelines is provided as table 4. A few articles reported using an individualised approach to determine the range for each patient, often by taking into consideration the breathing frequency: fully individualised range (n=8) and upper limit individualised (n=3)(figure 4C).

Figure 4

Reported very low frequency, low frequency and high frequency ranges for the articles using frequency domain analysis. (A) Reported very low frequency (VLF) ranges with n≥3 articles using each range. (B) Reported low frequency (LF) ranges with n≥3 articles using each range. (C) Reported high frequency (HF) ranges with n≥3 articles using each range. All ranges excluded from the figure (n≤2 articles) can be found in online supplemental file 2. Ind, individualised; NR, not reported.

Table 4

Articles using frequency domain analyses that follow the Task Force guideline’s frequency ranges

Frequency range	Task Force ranges15	Number of articles following Task Force ranges
VLF	0.003–0.04 Hz	8/210 (3.8%)
LF	0.04–0.15 Hz	60/210 (28.6%)
HF	0.15–0.4 Hz	18/210 (8.5%)
LF+HF	0.04–0.15 and 0.15–0.4 Hz	17/210 (8.1%)
VLF+LF+HF	0.003–0.04, 0.04–0.15 and 0.15–0.4 Hz	7/210 (3.3%)

Data reported as n/N (%), where N=number of articles using frequency domain analyses (N=210).

HF, high frequency; LF, low frequency; VLF, very low frequency.

Non-linear HRV

A total of 74 articles reported using non-linear HRV analysis (table 3). The parameters used were evenly distributed between the Poincaré plot analyses (n=29; 10.1%), DFA α1 (n=28; 9.8%) and α2 (n=25; 8.7%), approximate entropy (n=19; 6.6%) and sample entropy (n=22; 7.7%). Most frequently, however, articles reported using some other non-linear parameters (n=48; 16.8%) (table 3) such as: DFA RMS1 and/or RMS2,72–83 Shannon entropy,28 36 43 44 46 84–87 other entropies,25 28 48 84 86–89 deceleration/acceleration capacity24 25 86 90–93 and numerous (n=26) other combinations of measures, indices, exponents and dimensions of various chaos, fractality, sample asymmetry, correlations and graphical analyses.21 28 30 36 37 48 72 84–87 93–107

Heart rate characteristics

N=25 articles reported using HRC (table 3). Most of them (n=20) come from within the same group of researchers. However, since 2016, other groups have begun using HRC.

HRV applications

The list of all HRV applications and their references is provided as table 5.

Table 5

Neonatal heart rate variability applications

	n (%)	References
Physiological conditions	162 (56.6)
Normative	48 (16.8)	13 30 33 40 41 47 60 76 86 88–93 107 113 117 118 130 131 141 160 163 178 193 195 208 210 213 215 218 223 229 232 233 243 247 248 253 264–266 271 274 294 295 303
Sleep state	49 (17.1)	10 11 13 19 24 27 30 37 38 48 53 62 84 88 94 99 101 115 125 126 134 136 158 160 161 163 174 176 181 188 189 195 205 211 213 220 223 232–235 250 264 265 271 275 288 294 295 303
Longitudinal	80 (28.0)	2 13 19 20 25 30 33 35 39 48 51 53 54 56 60–62 65 72 80 86 88 90 91 93 94 96 98 104 107 114 117–119 125 126 128 129 136 138 140 142 145 147 150 152 158 160 163 168 177 192–194 198 200 207 208 218 221 223 227–229 232 233 243 244 247 248 256 257 260 265 279 291 293 295 296 303
Age or weight	49 (17.1)	10 13 19 21 25 33 40 41 62 73 76 89 93 95 98 102 107 111 113 115 116 118 122 125 132 134 137 141 181 183 202–204 208 241 243 250 258 259 265 266 271 274 275 280 294–296 303
Feeding	16 (5.6)	117 120 127 159 168 182 183 194 207 214 221 228 236 280–282
Position	13 (4.5)	19 23–25 111 146 182 201 205 211 232 233 273
Tilt	7 (2.4)	96 111 197 200 209 223 303
Sex	6 (2.1)	33 55 73 86 111 151
Other*	14 (4.9)	33 47 62 108–118
Pathological conditions	109 (38.1)
Sepsis	20 (7.0)	35 92 97 100 119 126 135 164 166 191 212 251 252 278 292 296 298–300 302
NEC	7 (2.4)	74 76 119 131 135 263 302
Infection	8 (2.8)	76 119 137 145 186 283 298 301
RDS	2 (0.7)	39 139
AOP	1 (0.3)	231
Extubation failure	7 (2.4)	87 140 148 149 225 269 284
HIE	19 (6.6)	2 4 28 52 65 72 75 81–83 142 151 169 190 219 230 237 244 285
Other brain injury	7 (2.4)	135 175 199 210 217 242 293
Seizures	6 (2.1)	28 42–44 46 241
PDA	2 (0.7)	73 224
CHD	8 (2.8)	79 194 206 214 221 228 256 275
In-hospital mortality	9 (3.1)	26 65 72 135 164 166 293 297 298
Other*	37 (12.9)	21 29 30 45 56 78 81 101 104 105 107 119–144
Responses	136 (47.6)
Pain	32 (11.2)	36 48 50 57 59 63 64 67 69 71 85 103 106 132 133 150 153 154 162 185 199 220 222 240 249 254 260 262 268 276 277 281
Medication	18 (6.3)	2 4 70 73 78 125 128 132 146 180 192 196 203 231 246 249 267 284
Light, sound or touch	22 (7.7)	22 29 64 68 146 153–155 160 184 187 227 238 239 245 255 257 262 264 273 281 282
NNS	5 (1.7)	49 127 232 233 281
Kangaroo care	13 (4.5)	31–34 61 66 165 179 185 188 194 220 264
Sucrose	3 (1.0)	59 162 281
Procedure	22 (7.7)	2 26 50 65 67 72 82 117 119 125 146 168 169 196 214 221 226 260 261 267 272 285
Maternal	30 (10.5)	9 27 37–39 55–58 60 117 122 126–128 137 193 197 209 216 229 235 254 255 270 282 286 287 289 290
Other*	38 (13.3)	20 23 32 34 48 55 59 64 69 77 104 112 118 121–123 128 129 137 145–163
Outcome prediction	30 (10.5)
NDI	12 (4.2)	4 57 58 65 72 100 114 126 167 175 190 293
Cerebral palsy	5 (1.7)	4 100 114 126 175
LOS or GA at discharge	5 (1.7)	129 164 165 175 256
Repeat hospitalisations	1 (0.3)	175
Behaviour or social	6 (2.1)	62 177 207 228 279 282
Days on MV	4 (1.4)	129 132 164 165
Other	11 (3.8)	4 101 119 137 157 164–169

*Others are described in detail within the manuscript text (with references).

AOP, apnoea of prematurity; CHD, congenital heart defects or disease; GA, gestational age; HIE, hypoxic ischaemic encephalopathy; LOS, length of stay; MV, mechanical ventilation; NDI, neurodevelopmental impairment; NEC, necrotising enterocolitis; NNS, non-nutritive sucking (eg, pacifier); PDA, patent ductus arteriosus; RDS, respiratory distress syndrome; SIDS, sudden infant death syndrome.

Physiological condition

N=162 articles (56.6%) included some form of physiological application of HRV (table 5). The most common form was the use of longitudinal design or repeated measures (n=80). There were 14 articles that examined HRV in relation to 13 other types of normal physiological application or parameters: QT interval,108 EEG bursts,109 110 respiration,33 47 111 112 heart rate,33 blood pressure,111 112 oxygen saturation,111 fetal HRV,113 movements,114 crying frequency,115 body fat percentage,116 ethnicity,117 height-for-age ratio118 and the Neonatal Behavior Assessment Scale.62

Pathological conditions

N=109 articles included an investigation of a specific condition or disease. The list of the conditions studied and their references is provided as table 5. The most commonly studied conditions were sepsis (n=20) and HIE (n=19) (table 5). There were 37 articles that examined 27 other conditions: cardiorespiratory events,30 119 120 abnormal fetal status,107 121–123 unplanned intubation,124 hypotension,21 105 125 abnormal polysomnography,101 126 neonatal abstinence syndrome,56 127 128 Transport Risk Index of Physiologic Stability score,29 129 hyperbilirubinaemia,104 128 130 Score for Neonatal Acute Physiology score,131 132 delivery complications,107 123 clinical course score,21 105 risk of neurodevelopmental impairment,133 gastro-oesophageal reflux disease,134 bronchopulmonary dysplasia,135 inflammatory cytokines,81 fetal growth restriction,136 cardiovascular development,137 APGAR score (newborn health assessment of Appearance, Pulse, Grimace, Activity, and Respiration),137 retinopathy of prematurity,138 intrathoracic masses,139 stroke,78 antibiotic use,140 non-benign tachyarrhythmia,141 death/disability score,142 T-wave alternans,143 epigenetic ageing144 and Clinical Risk Index for Babies (CRIB) score.45 125

Response

N=136 articles investigated HRV in response to certain stimuli. The list of the responses studied and their references is shown in table 5. The most common stimuli studied were response to pain (n=32), followed by light, sound or touch (n=22) and procedures (n=22; table 5). There were 11 different medication or medication groups, 9 different procedures, and 21 different maternal factors studied; a list of these 3 subgroups and their references is provided as table 6. There were 38 articles examining 17 other stimuli: respiratory support,112 145–151 delivery mode,20 69 77 121–123 137 phototherapy,104 128 146 152 consoling devices,32 34 59 153 154 electromagnetic field,155 156 stress,48 118 157 cocoon/swaddle,64 158 transport,129 hypoxic hypercapnia,23 feeding protocol,159 environmental tobacco smoke,55 caregiving,160 incubator temperature,161 non-invasive electrical stimulation at acupuncture points,162 CRIB versus incubator,156 self-consoling behaviour157 and family nurture intervention.163

Table 6

Detailed types of medications, procedures and maternal factors within the responses category of heart rate variability applications

Response subgroup	n	References	Response subgroup	n	References
Medications (n=18)			Maternal factors (n=30)
Vasoactives	3	73 125 180	Smoke	9	27 55 56 58 126 137 197 216 235
Phenobarbital	3	2 4 78	Cocaine	8	37 38 56 58 128 209 289 290
Morphine	3	4 128 132	Opioids	4	56 58 127 128
Caffeine	3	146 231 246	Depression	2	117 255
Thiopental	1	196	SSRI	2	56 254
EMLA	1	249	Alcohol	3	27 57 58
Cisapride	1	203	Other drugs	4	56 58 289 290
Aminophylline	1	192	Diabetes	3	122 270 287
Corticosteroids	2	73 284	Glycaemic index	1	122
Ibuprofen	1	267	Fatty acid status	1	193
Surfactant	1	70	Low SES	2	117 282
Procedures (n=22)			Benzodiazepines	1	254
Therapeutic hypothermia	6	2 65 72 82 169 285	Magnesium sulfate	1	39
Intubation	2	146 196	Ritodrine	1	39
Surgery	7	50 117 119 168 214 221 260	Life stressors	1	117
Immunisation	2	226 267	Antenatal steroids	2	9 137
ECMO	1	26	Hypertension	1	137
UVC	1	261	PET	1	137
IAC	1	125	Cardioresp. Phys.	1	60
Chest tube insertion	1	67	Aerobic exercise	1	229
Transfusion	1	272	Maternal diet	1	286

Cardioresp Phys, cardiorespiratory physiolog; ECMO, extracorporeal membrane oxygenation; IAC, indwelling arterial catheter; PET, pre-eclampsia toxaemia; SES, socioeconomic status; SSRI, selective serotonin reuptake inhibitor; UVC, umbilical venous catheter.

Outcome prediction

N=30 articles examined HRV in relation to long-term outcomes at discharge and beyond. The list of outcomes studied and their references is provided as table 5. The most common outcome examined was neurodevelopmental impairment (n=12) (table 5). There were 11 articles that included 9 other outcome measures: days on antibiotics,119 164 days on oxygen,137 165 days off mechanical ventilation,166 days alive,166 death,4 167 feeding skill or type of feeding at discharge,157 168 abnormal polysomnography,101 ‘unfavourable’ outcome (multifactorial)169 and cardiovascular development (echocardiography and blood pressure measurements).137

Discussion

Study and population characteristics

In the last 20 years, there has been an increasing interest in neonatal HRV, with a notable increase from 2017 onward. Indeed, in this scoping review, we also found 118 review articles that mentioned neonatal HRV, further pointing to its potential usefulness as a clinical tool. Interestingly, the majority of published articles were single-centre investigations that enrolled a small number of infants (<50). It is possible that sample sizes may increase over time as a consequence of advances in monitoring technology and in-hospital data storage systems, which will allow access to ECG signals recorded during hospital stay rather than performing separate signal collection for specific research projects.

There were notably fewer articles in MPT and LPT infants than in the other GA groups despite the fact that these infants account for the majority of preterm infants admitted to the NICUs.170 These patients still encounter a variety of clinical issues such as respiratory problems requiring different types of respiratory support, feeding and thermoregulation difficulties, and prolonged hospitalisation.171 172 This is a population where further HRV studies may show clinical benefit.

A large variety of HRV analyses methods were noted. This review identified 37 additional parameters used for HRV analysis, primarily within the non-linear analyses. Furthermore, the 46 different frequency ranges reported for the HF frequency domain parameter demonstrates how a single HRV parameter can have multiple definitions. Few articles fully adhered to the Task Force guidelines for the frequency domain ranges; however, guidelines for the adult population may not be appropriate for use in neonates given the differences in respiratory and heart rates.173 Moreover, this review did not collect certain aspects and details concerning ECG signal or HRV analysis, such as the duration of the ECG segment and handling of artefacts, which can also vary between studies but is not often clearly reported. As a result, the methodological variation and lack of consensus in neonatal HRV analysis makes synthesis and comparisons between investigations very difficult, if not impossible. As a result, analysis of neonatal HRV data have been limited to simplified interpretations of trends or shifts in HRV before and after any intervention or condition.

Neonatal HRV was used in numerous clinical applications well over the 34 different applications prespecified in this review. Including the applications listed within the ‘other’ categories, as well as the subgroups of responses to medications, procedures and maternal factors, a total of 138 clinical applications have been identified in this review. The articles evaluating physiological conditions and responses to stimuli articles should be particularly useful in highlighting certain considerations when planning a research study. It is encouraging that many studies used a longitudinal or repeated measures design, as this allows each patient to act as their own control which is important in the absence of established normative data. The context within these designs is key in interpreting, for example, in which state or condition is the newborn stressed versus relaxed, or the system’s balanced versus disorganised and dysfunctional. It is important that future research also consider performing similar and multiple HRV measures, with the goal of quantitatively combining the results of several investigations. Furthermore, this review highlights areas where more in-depth reviews and larger scale studies could be worth pursuing such as AOP and various long-term outcomes.

Harmonising neonatal HRV

Until a consensus is reached and a guideline is made available, neonatal HRV will continue to have varied methodologies, with simplified interpretability and limited meta-analytic potential. Without the ability to combine results from multiple studies, establishing normative data within this population is not possible. While waiting for an established neonatal HRV methodology to be recommended, we provide a table of suggested reporting items for all studies of neonatal HRV (table 7). Consistency in reporting will ensure, at least, that methodologies can be adequately compared, which will in turn allow certain studies with matching methodologies to be compared and combined. Indeed, with the ever-expanding capabilities of storing large amounts of data, such as the ECGs used for HRV analysis, it may be possible to re-analyse past datasets while applying a newly standardised methodology.

Table 7

Suggested reporting items for neonatal HRV

	Reporting item	Details
Methodology
ECG	Acquisition	Describe the acquisition set-up, including devices and software used, sampling frequency, sleep state and positioning of the newborn, and details of the period during which ECG is recorded (eg, 1 hour prior to extubation, immediately after feeding, etc).
	Segment	Describe the segment length used to calculate HRV and methods used to select the segment (if any).
	Handling	Describe any filtering applied to the ECG or artefact removal methods, including any software used.
HRV	R-wave identification	Describe the algorithms or software used to identify beats (R-waves).
	Handling	Describe any filtering, selection or artefact removal methods applied to the RR intervals, including any software used.
	Parameters	Describe all HRV parameters calculated with definitions and methods for obtaining the values, with references where appropriate and including any software used. Report all frequency domain ranges.
Results	Demographics	Report the baseline demographics of all patients and subgroups of patients where HRV was investigated, including weights, gestational ages and day of life.
	Reporting	Report all values for all HRV parameters calculated, regardless of statistical significance.

HRV, heart rate variability.

Limitations

This scoping review has some limitations. First, with the intent of keeping the review contemporary, we did not include articles published before 2000. Second, a single reviewer performed both the screening and data collection, but any questions or conflicts were discussed among all reviewers. Third, as previously mentioned, additional details concerning ECG signal or HRV analysis (eg, duration of the ECG segment and handling of artefacts) were not collected since they are not clearly reported in a number of publications. However, this review provides an updated and comprehensive view on neonatal HRV that can be used as reference for future investigations.

Conclusion

This scoping review highlights the growing interest in neonatal HRV, with numerous applications being investigated. Most importantly, it reveals the lack of consistency in calculating and reporting HRV, likely due to the lack of consensus and published guidelines for neonatal HRV. Future investigations will benefit from a consensus guideline for neonatal HRV, and this review may help in the planning of that.