Gestation-Specific Vital Sign Reference Ranges in Pregnancy

OBJECTIVE: To estimate normal ranges for maternal vital signs throughout pregnancy, which have not been well defined in a large contemporary population. METHODS: We conducted a three-center, prospective, longitudinal cohort study in the United Kingdom from August 2012 to September 2017. We recruited women at less than 20 weeks of gestation without significant comorbidities with accurately dated singleton pregnan-cies. We measured participants’ blood pressure (BP), heart rate, respiratory rate, oxygen saturation and


CONCLUSION:
We present widely relevant, gestationspecific reference ranges for detecting abnormal BP, heart rate, respiratory rate, oxygen saturation and temperature during pregnancy. Our findings refute the existence of a clinically significant BP drop from 12 weeks of gestation. S uccessive inquiries into maternal deaths call for an evidence-based early warning score to facilitate earlier recognition and treatment of the unwell pregnant woman. [1][2][3][4] Recognition is complicated by the normal physiologic changes in vital signs during pregnancy. 1 Changes in normal ranges for blood pressure (BP) through pregnancy have been derived from routinely collected data-which may be biased by the reason for collection. 5 Blood pressure trends without normal ranges have been investigated in smaller studies. 6 Substantially fewer data exist for heart rate, [7][8][9][10][11] temperature, 12,13 respiratory rate 14,15 and oxygen saturation 14,16 patterns during pregnancy.
Parameters recommended internationally for normal ranges of vital signs in pregnancy ([United Kingdom, Confidential Enquiry into Maternal and Child Health 1 ; Ireland, Irish Maternity Early Warning System 17 ; United States, National Partnership for Maternal Safety 18 ]) and thresholds used for modified obstetric early warning scores [19][20][21][22] are not gestation-specific or evidence-based, but based on expert opinion with wide local, national and international variation.
Using robust estimates of the outer centiles of vital sign distributions to generate early warning scores is reliable in nonpregnant adults. 23,24 The approach may be particularly useful for pregnancy, where event rates are low. However, there are insufficient published data to generate evidence-based, gestation-specific reference ranges for vital signs. 25 A large-scale, contemporary prospective data set of vital signs in pregnancy is needed to provide gestation-specific normal ranges for use in clinical practice. We included a pragmatic population, representative of women who would be monitored using a modified obstetric early warning score system derived from the normal vital sign ranges obtained.
The primary objective of the Pregnancy Physiology Pattern Prediction (4P) study was to develop a database of prospective vital sign measurements using standardized measurement techniques throughout pregnancy. 26 Estimates of population distributions and associated centiles were derived from this database. Secondary objectives of developing a centilebased early warning score and investigating new patterns in vital sign data will be explored in further work.

METHODS
This work is reported following the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines. 27 We registered the study (https://doi.org/10.1186/ISRCTN10838017). Detailed methods are reported in the published protocol 26 and in brief here. We conducted a multicenter, longitudinal, observational, cohort study across three centers in the United Kingdom. We collected vital sign data during the antenatal, intrapartum, and postnatal periods. Here we present data from the antenatal period.
Recruitment commenced in August 2012, and collection of vital sign data was completed in August 2017. The study started in Oxford as a substudy of the INTERBIO-21 st Fetal Study, approved by Oxford South Central C Research Ethics Committee (REC):08/H0606/139), 28,29 and then expanded to include two additional centers (Newcastle and London, South East Coast-Brighton and Sussex REC:14/LO/1312) continuing after completion of   (December 2015).
We approached women when they attended the ultrasound department or antenatal clinic at three university hospitals before 20 weeks of gestation. Women were eligible if they were aged 16 years or older, with a singleton pregnancy, and fell within category 1 of the American Society of Anesthesiologists' classification of physical status before pregnancy ("a normal healthy patient without any clinically important comorbidity and without clinically significant past or present medical history" 30 ). Gestational age was determined by ultrasound measurement of crown-rump length before 14 weeks of gestation. Full eligibility criteria can be found in Appendix 1, available online at http://links.lww.com/AOG/B746.
The primary outcome was gestation-specific reference ranges comprising centile distributions for vital signs through pregnancy. We conducted sensitivity analyses to assess changes in performance of the sensors used to measure each vital sign over time. We assessed differences in methods for recording respiratory rate, agreement between two thermometers used and the effect of duplicate measurement.
We collected vital sign data at clinic visits every 4-6 weeks 31 throughout pregnancy for five physiologic parameters: BP, heart rate, SpO 2 , temperature, and respiratory rate. We measured all vital signs following study standard operating procedures. 26 Blood pressure was measured with an automated BP monitor validated for use in pregnancy (Microlife 3BT0-A (2)/WatchBP Home). Heart rate and oxygen saturation were measured with a Bluetooth-enabled pulse oximeter (Wris-tOx2 3150).
Temperature was measured with a tympanic thermometer (Genius 2). Temperature was also measured with a Bluetooth enabled tympanic thermometer (Fora IRb 20b) in the INTERBIO-21 st subgroup. Respiratory rate was recorded using two different methods. Trained midwives observed chest wall movement over 15 seconds. From October 2015 onward, midwives also tapped in time with observed respiratory rate for 1 minute using bespoke software on an Android tablet. Research midwives entered vital sign data (or in the case of Bluetooth-enabled devices, automatically transmitted) onto a studyspecific android tablet computer (Samsung Galaxy Tab 4.0 26 ).
We collected demographic information (age, height, weight, self-reported ethnicity, number of   previous pregnancies, smoking status), past medical and obstetric history, current health status, pregnancyrelated health and current medications at the initial assessment. At each follow-up appointment we collected assessments of smoking status, current health status, pregnancy related health and current medications. We extracted medical and obstetric history from the participants' notes. If participants did not respond to contact by research midwives during the study, they were treated as withdrawn after the third failed contact attempt. Data already collected from withdrawn participants were included in the final analysis, unless explicitly requested otherwise.
Trained research midwives undertook all antenatal observations and extracted all hospital data following study standard operating procedures. Research midwife coordinators performed frequent site visits to carry out midwife training and address any recruitment and equipment issues.
Vital sign measurements taken for the study were not included in the clinical record and were not communicated to the clinical team unless BP reached predefined values (systolic 140 mm Hg or higher or diastolic 90 mm Hg or higher) according to study standard operating procedures.
Duplicate measurements were recorded for a subset of participant visits to document the effect of using only the first recorded measurement for each vital sign parameter (as in clinical practice). To minimize participant inconvenience, duplicate measurements were reviewed by study statisticians to decide when an adequate number of duplicate measurements had been collected.
Our sample size determination may be found in Appendix 2 (available online at http://links.lww.com/ AOG/B747) and the published protocol. 26 In brief, a sample size of 1,000 women would achieve an SE of 0.05*SD at the 2.5 and 97.5 centiles, and even greater precision at the less extreme centiles. Adequate precision was also met for any subgroup analysis; for example, a sample size of 300 women would achieve an SE of 0.1*SD at the 2.5 and 97.5 centiles.
Vital sign data from all participants were included in the primary analysis. We constructed smoothed centiles for systolic and diastolic BP, heart rate, SpO 2 , temperature and respiratory rate by gestational age. We constructed gestation-specific reference ranges comprising smoothed centiles for vital sign distribution (3rd, 10th, 50th, 90th and 97th as used by the World Health Organization Multicentre Growth Reference Study, 32,33 with corresponding 95% CIs) for all women. We presented centiles from 12 to 40 weeks of gestation (because there were relatively few data at less than 12 and greater than 40 weeks of gestation) for each vital sign graphically and tabulated at fortnightly intervals, along with associated 95% CIs.
We used statistical methods to generate gestational age-specific centiles following those used in the INTERGROWTH-21 st Project for fetal growth. [34][35][36][37] .3) 0 (0) 3 (2.1) 5 (0.5) Termination of pregnancy at less than 12 wk 0 (0) 0 (0) 0 (0) 0 (0) Termination of pregnancy at more than 12 wk Once enrolled, we did not exclude women who developed conditions that might affect their vital signs (to generate a pragmatic, representative sample of pregnant women, maximizing the clinical applicability of centiles generated). We explored different statistical methods to achieve the best fit to the data (see Appendix 2, http://links.lww.com/AOG/B747). We expected some participants would become lost to follow-up or have missing measures. We included these participants in the analysis, unless all data were missing. We compared agreement between Fora and Genius thermometers using a Bland-Altman plot to assess whether it was possible to pool measures. We also compared data from the counting and tapping methods for recording respiratory rate before pooling.
We conducted sensitivity analyses to assess changes in sensor performance over time.
To explore whether limiting the population to those of optimal health would affect results, we defined a "restrictive" population of women aged younger than 40 years with body mass indexes (BMIs, calculated as weight in kilograms divided by height in meters squared) between 18.5 and 29.9, who did not smoke and did not have any medical comorbidities.
In this "restrictive" population, we excluded measures from women who developed a severe maternal condition during pregnancy (severe preeclampsia; hemolysis, elevated liver enzymes, and low platelet count [HELLP] syndrome; or eclampsia), from the time of diagnosis onward, in line with previous work. 38 We conducted a post hoc analysis using these definitions in comparison with the full "pragmatic" population.
To explore whether the small reduction in heart rate near term resulted from women with higher heart rates giving birth earlier, we conducted a post hoc analysis with all women who delivered at less than 37 weeks of gestation excluded. We undertook a prespecified subgroup analysis including only nulliparous women. Participants provided informed written consent and could withdraw from the study at any time.

RESULTS
We screened 4,279 women between August 1, 2012, and December 28, 2016, of whom 1,054 agreed to take part. Thirteen women subsequently withdrew consent for any data to be used and were excluded completely. This provided a total pragmatic study cohort of 1,041 women who contributed antenatal vital sign data (Fig. 1). Delivery information was unavailable for 37 women, who were either lost to follow-up (15 women) or discontinued participation (22 women) during the study period. Within this group, 36 women provided data at 12 weeks of gestation, 24 at 24 weeks, 12 at 28 weeks, and two at 32 weeks, beyond which they no longer participated.
Delivery information was available for 1,004 women. Table 2 details pregnancy outcomes and perinatal events within the study cohort.
Temperature measurements with both Fora and Genius devices were recorded at 2,990 visits. The Bland-Altman plot showed wide limits of agreement, around 2°C (Appendix 5, available online at http:// links.lww.com/AOG/B750). Therefore, it was not appropriate to pool data from the two thermometers. For the Genius device, median (3rd-97th centile) temperature decreased from its maximum at 12 weeks of gestation of 36.7 (35.6-37.5)°C to a minimum of 36. For each vital sign measured, gestation-specific values for the smoothed centiles and a smoothed gestation-specific centile plot with associated 95% CI can be found in Appendix 4 (http://links.lww. com/AOG/B749). Duplicate measurements were recorded for a subset of 273 participant visits (4.6%). There were no clinically meaningful differences between first and second measures of vital signs in pregnancy, so duplicate readings were not continued (Appendix 6, available online at http:// links.lww.com/AOG/B751). We found no evidence of systematic sensor performance alteration over time (see Appendix 7, available online at http:// links.lww.com/AOG/B752).
Applying the "restrictive" population definitions reduced the cohort to 595/1,041 women (Appendix 8, available online at http://links.lww.com/AOG/B753). Although systolic and diastolic BP centiles were numerically slightly lower in the restrictive cohort (higher BPs having been excluded) the relatively narrow CIs mainly overlapped. Centiles for other vital signs were similar to the "pragmatic" cohort.
Blood pressures from 460 nulliparous women were slightly higher than those from the 581 parous women (Appendix 9, available online at http://links.lww.com/ AOG/B754). Excluding women who delivered before 37 weeks of gestation did not change the shape of the heart rate centiles (Appendix 10, available online at http://links.lww.com/AOG/B755).

DISCUSSION
Our multicenter study includes longitudinal data from more than 1,041 women. From these data we produced evidence-based, gestation-specific centiles for vital signs during pregnancy. Our findings refute a large drop in BP during pregnancy after week 12, when women are usually first measured. Textbooks suggest a 10-15-mm Hg drop in diastolic BP 39-41 during pregnancy. Evidence supporting this BP drop came from small single center studies 6,[42][43][44][45] or larger birth cohorts from one geographical region 46,47 commonly using routinely collected, 6,44 nonstandardized measurements 42,45 with equipment not ratified for use in pregnancy. 6,[42][43][44][46][47][48] Our study concurs with recent publications, including a recently published systematic review 7,25,49 that challenge this teaching, and strongly supports the resulting call to abandon this notion. 50 Several studies suggest a drop occurs from prepregnancy levels in the first few weeks of pregnancy. 8,48,51,52 However, preconception BPs are not normally available in clinical practice. Our findings allow clinicians to recognize that relative hypotension from the first BP in pregnancy may warrant investigation. The third centile for systolic BP was never less than 94 mm Hg and was greater than 96 mm Hg after 30 weeks of gestation in all groups. These thresholds are above the less than 90 mm Hg used to recognize sepsis in pregnancy 53,54 and modified obstetric early warning score charts trigger for escalation. 1,17,18,55 The BP rise toward the end of pregnancy suggests gestational threshold adjustment could improve detection of deteriorating mothers.
Before our study, the best evidence for longitudinal reference ranges for BP in normal pregnancy probably came from the Avon Longitudinal Study of Parents and Children, a large, single-center cohort study using data from 20 years ago. 5 The Avon Longitudinal Study of Parents and Children reported a small (1.7 mm Hg) nadir of systolic BP at 17-18 weeks of gestation, similar to the 1-mm Hg drop at 18.6 weeks of gestation in our multicenter cohort. The subtle nadir in diastolic BP was also similar in both studies. The Avon Longitudinal Study of Parents and Children relied on BPs documented in women's routine obstetric records, rather than measuring BP with validated equipment following a standard operating procedure. Despite this, the median and ranges of our systolic BPs are similar, providing reassurance that our BP ranges will be valid in routine clinical practice.
It is commonly taught that heart rate increases by 10-15 bpm from the first trimester onward. 40,56 An increase in heart rate of 20-25% is reported by studies which compare heart rate in pregnancy to prepregnancy baseline. 8,51,57 However, prepregnancy vital sign data are not usually available. We demonstrated a smaller increase in heart rate of 9 bpm between 12 weeks of gestation and the third trimester. This is consistent with our recent systematic review, that also highlighted that the outer centiles of heart rate in pregnancy have not been demonstrated in large modern studies. 25 From 18 weeks of gestation, heart rates of more than 100 bpm (more than 105 bpm from 28 weeks of gestation) occurred in more than 10% of observations taken in healthy pregnancy. However the 97th centile for heart rate increased from 105 bpm at 12 weeks of gestation to 115 at 36 weeks, suggesting that a 120 bpm threshold for the highrisk threshold for modified obstetric early warning score escalation 1,17,55 may be too high.
Centiles for respiratory rate in pregnancy have not been determined in a large modern cohort, Green. Vital Sign Reference Ranges for Pregnancy. Obstet Gynecol 2020. though our finding that respiratory rate does not change with gestation is supported by a longitudinal study of 20 women. 14 Our work shows that a respiratory rate of more than 22 breaths per minute (as used in the quick Sepsis Related Organ Failure Assessment tool 58 ) occurs in fewer than 3% of observations in normal pregnancy, suggesting this threshold could translate from other medical practice. Current moderate and high-risk thresholds for respiratory rate of 21-24 and 25 breaths per minute or more (as advocated by the UK Sepsis Trust, 54 the Irish Maternity Early Warning System 17 and the Scottish Patient Safety Programme 55 ) may more accurately detect women at risk of sepsis than moderate 1 and high-risk 1,18 thresholds of more than 21-30 and more than 30 breaths per minute.
We identified a small drop in SpO 2 during pregnancy. Previous small studies suggest a modest reduction in SpO 2 between the second and third trimester 59 and oxygen partial pressure (PaO 2 ) across gestation. 14 The lower thresholds of oxygen saturation have not been determined at scale. We showed that SpO 2 less than 93% is an abnormal finding in pregnancy. However, from 16 weeks of gestation onward, SpO 2 of 94% is within the normal range. These findings suggest that a lower alerting threshold of less than 95% 1,17,18,55 may be too high.
Longitudinal temperature studies during pregnancy providing data on outer centiles are lacking, with only small previous studies. 13 We found a clinically insignificant drop in temperature from 12 to 40 weeks of gestation (less than 0.2°C). Many modified obstetric early warning score systems use a high-risk escalation threshold of 38°C or higher. 1,17,55 However, temperatures of 37.5°C were uncommon (fewer than 3% of observations), suggesting this may be an appropriate threshold to further investigate women.
Although higher BPs were excluded from the restrictive cohort, BP differences from the pragmatic cohort were marginal. In the nulliparous group, BPs were higher than in the parous cohort, particularly at 40 weeks of gestation. All other vital sign centiles were similar between subgroups.
In our systematic literature review 25,60 we could not identify a study using World Health Organization recommended centiles developed from longitudinal vital sign data to provide reliable data for clinicians in clinical practice. By adopting a prescriptive approach to gestational age calculation and using standardized, pregnancy-specific equipment to collect data prospectively from three sites, we are confident our reference ranges are robust. Our study population is of similar age to the U.K. national average for pregnancy 61 and of similar BMI to previous U.K. findings. 62 Women were predominantly of Caucasian ethnicity (85.2%), equivalent to the most recent England and Wales census data (86%). 63 Our population therefore appears representative and applicable to clinical practice. Extending to an international population would improve external validity for other settings.
We present widely relevant, gestation-specific reference ranges for systolic and diastolic BP, heart rate, respiratory rate, oxygen saturation and temperature. Our findings refute the idea that a clinically significant first-to-mid trimester drop in BP is normal and suggest evidence-based ranges for other vital signs. These reference ranges can be used to facilitate early recognition of unwell pregnant women. The differences from current accepted values used to detect deterioration in pregnant women emphasize the need for robustly derived ranges for a modified obstetric early warning score system.