Preoperative and intraoperative blood pressure variability independently correlate with outcomes

Abstract

Background:

Blood pressure variability (BPV) describes visit-to-visit blood pressure (BP) changes independent of hypertension. Preoperative BPV and intraoperative BPV are associated with increased postoperative outcomes. We investigated the impact of both preoperative BPV and intraoperative BPV on elective surgical outcomes, specifically whether preoperative BPV and intraoperative BPV were independent risk factors for surgical complications.

Materials and Methods:

We investigated 600 patients undergoing elective surgery lasting more than two hours and who had ≥8 outpatient BP recordings over three preoperative years. Age, sex, ethnicity, BMI, current medical problems, and medications at time of surgery were recorded. BPV was calculated as the standard deviation (SD) of systolic or diastolic BP for the 369 valid patients. Average BPVs were compared between adverse outcomes of readmission, wound infection, AKI, death, MI, and CVA.

Results:

Three-hundred-sixty-nine (52.6% male, 47.4% female, 98.1% non-Hispanic) patients (mean age 62.5) were included in the study. Preoperative systolic (p=.043) and diastolic (p=.009) BPV were higher for patients with the combined endpoint of all adverse events. Preoperative systolic BPV was correlated with intraoperative BPV (p=.010). Both systolic and diastolic preoperative BPV was found to be independent from intraoperative BPV. Otolaryngology procedures were associated with less adverse outcomes (p=.034), while antimicrobials (p=.022), autonomic drugs (p<.001), or respiratory drugs (p=.032) was associated with an increased likelihood of adverse outcome.

Conclusions:

Preoperative DBPV is associated with increased risk of readmission, wound infection and the combined endpoint of all adverse events. Intraoperative SPBV is associated with increased risk of AKI and the combined endpoint of all adverse events. Preoperative DBPV and intraoperative SBPV are independent risk factors for ninety-day postoperative outcomes. BPV should be considered in individualized risk assessment when assessing patient eligibility for elective procedures.

Introduction

Blood pressure variability (BPV) is described as the degree of instability in a person’s ambulatory outpatient blood pressure over years.¹ Distinct from hypertension, BPV can be calculated using the standard deviation (SD), average change, or largest change in systolic and diastolic blood pressure.^1–3 The etiology of BPV is likely to be multifactorial, including intrinsic patient characteristics such as autonomic instability⁴ and changes in the biology of the vascular wall^5–7, as well as extrinsic factors such as patient anxiety and treatment compliance^8–12 Regardless of the cause of the BPV, recent reports suggest that surgical patients with chronically high preoperative BPV (specifically excluding acute preoperative changes in blood pressure) are at increased risk for myocardial infarction (MI), cerebral vascular accident (CVA), renal failure, readmission, and mortality after elective surgery.^1,2 Because of this associated increase in adverse events¹, preoperative BPV is therefore of interest for the prediction of adverse outcomes in surgical patients.

Variability in intraoperative blood pressure is also correlated with adverse postoperative outcomes. Intraoperative hypotension has poor negative implications for postoperative acute kidney injury (AKI), MI, organ dysfunction, and mortality.^13,14 Intraoperative hypertension has been associated with factors which contribute to myocardial ischemia such as increased inflammatory responses, reperfusion injury, and platelet activation which may contribute to worse postoperative outcomes.¹⁵ Other findings suggest that intraoperative BPV is associated with 30-day mortality in patients undergoing coronary revascularization and/or coronary valve replacement.^15,16 BPV in non-cardiac procedures (gastrointestinal, gynecological, and neurosurgical) correlated significantly with increased risk of postoperative mortality, with a stronger correlation to DBPV.¹⁷ Some intraoperative BPV reflects untoward surgical or anesthetic events, such as hemorrhage, intraoperative MI, release of cytokines or vasoactive mediators by tumors or bacteremia, or misregulation of anesthesia. However, as for outpatient BPV, it seems some intraoperative BPV may reflect intrinsic patient biology. The relationship between preoperative BPV and intraoperative BPV has not previously been studied.

We therefore sought to investigate the impact of both preoperative BPV and intraoperative BPV on elective surgical outcomes. Specifically, we hypothesized that preoperative BPV and intraoperative BPV would independently increase the risk of surgical complications. In particular, we chose to study 90-day postoperative incidence of MI, CVA, AKI, wound infection, readmission, and mortality as adverse surgical outcomes. We chose these outcomes because they have previously been recorded to be associated with high preoperative BPV and represent common and serious sequelae that are often studied.^{2–3,18–20} We investigated a consecutive series of patients undergoing elective surgery lasting more than two hours (to allow for adequate numbers of recorded intraoperative blood pressure measurements), recording preoperative and intraoperative BPV and postoperative outcomes.

Methods

Our study was approved by the IRB’s of the University of North Dakota and Altru Health System in Grand Forks. The requirement for written informed consent was waived. This manuscript adheres to the applicable STROBE guidelines. We tested our hypothesis by retrospectively examining 600 patients who had an operation between January 2019 through August 2019 at Altru Health System in Grand Forks and who underwent general anesthesia for greater than 2 hours for elective surgical procedures, but eliminated 231 for not meeting the requirement of having had 8 or more recorded outpatient blood pressures within the past three years prior to the surgery date. To test for potential bias in the included patients, they were compared to the excluded patients by sex, age, and BMI.

Inpatient non-intraoperative blood pressure (BP) readings the day of surgery were not included due to the possibility of being affected by an underlying illness, stress of surgery, or pre-surgical preparation. BPV was calculated as the standard deviation (SD) of either systolic or diastolic BP, using 8 or more recorded outpatient BPs over the previous three years. We then reviewed clinical records for post-operative surgical follow-ups, as well as any other medical records from the Grand Forks area.

Outcomes were identified through manual chart abstraction and were defined as ninety-day postoperative incidence of MI, CVA, AKI, [pre- to postoperative creatinine increase greater than 0.3 or 50%]²¹ wound infection, readmission, and mortality. Age, sex, ethnicity, BMI, type of surgery, and medications at time of surgery were recorded as patient attributes for the 369 patients included in the study. As potential confounders, their association to the outcomes was estimated. In particular, preoperative outpatient usage of blood pressure medications of beta blockers, ACEI, ARB, calcium channel blocker (CCB) and combinations were also recorded in order to control for the use of these drugs due to recent suggestions that long acting CCBs and other blood pressure medications may protect against effects of high BPV.^22–26

Statistical Analysis

Chi-Square statistics and independent t-tests were used to test for association between included and excluded patients with regard to sex, ethnicity, age, surgery procedure, and BMI. Significant results were considered indicators of potential bias between the populations. Chi-Square statistics were also used to reveal association between patient attributes (sex, ethnicity, age, BMI, surgery, and medication type) and having an adverse outcome. Independent t-tests were used to detect significant average mean differences in age, BMI, total medications and number of blood pressure readings, systolic, systolic SD, diastolic, and diastolic SD between those with and without adverse outcomes. One-way Analysis of Variance (ANOVA) was used to test differences in average BPV between levels of age and number of medications. Bonferroni multiple comparison adjustments were used with alpha=.05. Pearson’s correlation coefficient along with the root mean square error (RMSE) based on residuals from the regression line was used to test relationships between pre- and intraoperative systolic and diastolic BPVs. Logistic regressions of pre- and intraoperative BPV predicting the more prevalent outcomes of readmissions (27.1%), wound infection (20.1%), AKI (19.5%), and any adverse effect (46.9%) were ran including covariates of age, gender, BMI, on blood pressure medication, total number of medications, mean intraoperative BP, and mean preoperative BP. Analysis of Covariance was used to find the adjusted mean difference in intraoperative BPV between adverse outcomes while controlling for preoperative BPV. Any changes in significance from adjustment or significant interaction might be indicators of statistical dependence. One-tailed tests were used in the estimations of odds ratio and means as prior research has shown that increased BPV increases likelihood of adverse events at many levels and subgroups of people. Directional hypotheses will allow us to specifically see if preoperative BPV follows the same patterns as other BPVs, including intraoperative. SAS v. 9.4 was used for the analysis.

Results

Of the 600 medical records identified as corresponding to patients undergoing consecutive elective surgical procedures, we identified 369 (61.5%) patients who underwent elective surgery and for whom at least eight BP readings were documented both preoperatively (not counting the day of surgery) and intraoperatively. Table 1 compares these 369 patients with the 231 (38.5%) patients who were excluded because they did not have at least eight BP readings documented both preoperatively and intraoperatively. Supplemental Table 1 lists the variety of surgical procedures the included patients underwent. There was no significant difference in sex (p=.945) or ethnicity (p=.700) between the two groups. However, the excluded group of patients were younger (p<.001), and less likely to be overweight or obese (p=.009). They also tended to have different types of surgical procedures.

Table 1.

Attributes of 369 included and 231 excluded patients.

	369 Included		231 Excluded		Chi-Square	p
	N	%	N	%
Sex: Female	175	47.43	109	46.78	.0050	.945
Male	194	52.57	124	53.22
Ethnicity: Hispanic	7	1.90	5	2.15		.700
Non-Hispanic	362	98.10	228	97.85
Surgery:					537.5	<.001
Neurosurgery	84	22.76	1	0.43
Cardiothoracic	47	12.47	3	1.29
Abdominal surgery	54	14.63	3	1.29
Otolaryngology surgery	13	3.52	1	0.43
Musculoskeletal	26	7.05	0	0
Endocrine surgery	10	2.71	3	1.29
Pelvic surgery	13	3.52	0	0
Urologic surgery	19	5.15	0	0
Vascular surgery	69	18.70	5	2.15
Plastic surgery	35	9.49	1	0.43
Other			216	92.70
BMI: Underweight	7	1.90	1	0.43	15.443	.009
Normal	54	14.63	43	18.61
Overweight	106	28.73	85	36.80
Obese I	89	24.12	39	16.88
Obese II	75	20.33	31	13.42
Ext Obese	38	10.30	32	13.85
	Mean	S.D.	Mean	S.D.	t	p
Age	62.53	14.97	54.13	62.53	6.06	<.001
BMI	31.41	6.86	31.06	7.38	0.59	.552

Adverse outcomes were defined as having been readmitted within 90 days after surgery (100, 27.1%), developing a wound infection (74, 20.1%), pre- to postoperative creatinine increase greater than 0.3 or 50% (72, 19.5%), death (13, 3.5%), MI (6, 1.6%), or a CVA (5, 1.4%). A total of 173 patients (46.9%) had at least one adverse outcome. Sex, age, ethnicity, and BMI were not associated with having adverse outcomes (Table 2). Although the number of cases is low (N=13), compared to other surgical procedures, otolaryngology procedures were associated with less adverse outcomes (p=.034) though the surgery type itself was not significant (p=.065). Being on antimicrobials (p=.022), autonomic drugs (p<.001), or respiratory drugs (p=.032) was associated with an increased likelihood of adverse outcome. Five classifications of blood pressure drugs were investigated, and none were significantly associated with having adverse outcomes (separately and in combination: beta blockers, ACEI, ARB, CCB).

Table 2.

Attributes of 369 patients associated with at least one adverse outcome.

	All (N=369)		Adverse Outcomes (N= 173)		Normal Outcomes (N=196)		p
	N	%	N	%	N	%
Sex: Female	175	47.43	78	45.09	97	49.49	.398
Male	194	52.57	95	54.91	99	50.51
Ethnicity: Hispanic	7	1.90	5	2.89	2	1.02	.957
Non-Hispanic	362	98.10	168	97.11	194	98.98
Age: <=50	75	20.33	29	16.76	46	23.47	.225
51–65	124	33.60	63	36.42	61	31.12
66–70	43	11.65	17	9.83	26	13.27
>=71	127	34.42	64	36.99	63	32.14
BMI: Underweight	7	1.90	5	2.89	2	1.02	.381
Normal	54	14.63	25	14.45	29	14.80
Overweight	106	28.73	53	30.64	53	27.04
Obese I	89	24.12	35	20.23	54	27.55
Obese II	75	20.33	39	22.54	36	18.37
Ext Obese	38	10.30	16	9.25	22	11.22
Surgery:							.065
Neurosurgery	84	22.76	32	18.50	52	26.53	.231
Cardiothoracic surgery	47	12.74	24	13.87	23	11.73	.900
Abdominal surgery	54	14.63	27	15.61	27	13.78	.941
Otolaryngology surgery	13	3.52	1	0.58	12	6.12	.034
Musculoskeletal	26	7.05	11	6.36	14	7.14	.996
Endocrine surgery	10	2.71	4	2.31	6	3.06	.993
Pelvic surgery	13	3.52	5	2.89	8	4.08	.945
Urologic surgery	19	5.15	11	6.36	8	4.08	.184
Vascular surgery	69	18.70	38	21.87	31	15.82	.387
Plastic surgery	35	9.49	20	11.56	15	7.65	.546
Medication Type:
Antimicrobials	74	20.05	44	25.43	30	15.31	.022
Antineoplastics	10	2.71	4	2.31	6	3.06	.775
Autonomic drugs	84	22.76	54	31.21	30	15.31	<.001
Anticoagulants	196	53.12	97	56.07	99	50.51	.335
Anti-inflammatory	213	57.72	99	57.23	114	58.16	.856
Antihistamine drugs	53	14.36	21	12.14	32	16.33	.319
Analgesics/Anesthetics	134	36.31	71	41.04	63	32.14	.096
Cardiovascular drugs	184	49.86	92	53.18	92	46.94	.275
Diabetic drugs	92	24.93	50	28.90	42	21.43	.125
Dyslipidemia drugs	176	47.70	82	47.40	94	47.96	.997
Endocrine drugs	80	21.68	35	20.23	45	22.96	.611
Gastrointestinal drugs	166	44.99	81	46.82	85	43.37	.575
Immunomodulatory	11	2.98	5	2.89	6	3.06	.999
Neuro/Psychiatric	169	45.80	81	46.82	88	44.90	.791
Renal drugs	216	58.54	103	59.54	113	57.65	.794
Respiratory drugs	81	21.95	47	27.17	34	17.35	.032
Topicals	17	4.61	9	5.20	8	4.08	.792
Vitamins/Minerals	179	48.51	90	52.02	89	45.41	.244
Miscellaneous Other	69	18.70	38	21.97	31	15.82	.168
Blood pressure drugs:
Any	219	59.35	105	60.69	114	58.16	.698
Beta blockers	131	35.50	70	40.46	61	31.12	.078
ACEI	70	18.97	34	19.65	36	18.37	.856
ARB	45	12.20	18	10.40	27	13.78	.408
CCB	70	18.97	31	17.92	39	19.90	.726
Combination	12	3.25	4	2.31	8	4.08	.508
Total Medications:
0 – 3	74	20.05	28	16.18	46	23.47	.118
4 – 5	92	24.93	41	23.70	51	26.02
6 – 7	91	24.66	42	24.28	49	25
8 – 13	112	30.35	62	35.84	50	25.51
	Mean	S.D.	Mean	S.D.	Mean	S.D.	p
Age	62.53	14.97	64.03	14.87	61.19	14.96	.069
BMI	31.41	6.86	31.15	6.77	31.64	6.94	.497
Total Medications	5.97	2.90	6.38	2.86	5.62	2.89	.012
Pre-Operative
Number of BPs	18.02	9.80	19.59	10.75	16.64	8.66	.004
Systolic	127.42	10.39	128.27	10.69	126.66	10.08	.139
Systolic SD	12.48	4.31	12.89	4.37	12.11	4.24	.043
Diastolic	74.25	7.21	73.8	8.03	74.65	6.4	.26
Diastolic SD	7.95	2.32	8.26	2.42	7.68	2.21	.009
Intra-Operative
Number of BPs	122.25	76.41	127.77	86.34	117.38	66.27	.193
Systolic	112.44	12.28	113.73	12.16	111.3	12.3	.058
Systolic SD	17.30	6.43	10.63	3.85	10.25	3.39	.010
Diastolic	65.41	8.91	65.04	9.02	65.73	8.82	.453
Diastolic SD	10.43	3.62	18.12	6.74	16.58	6.06	.158

Chi-Square test used for association of categorical variables with having at least one adverse outcome. Independent t-test used to test mean difference of continuous variable between having and not having at least one adverse outcome. Independent t-tests for BPVs are directional hypothesizing that the adverse event group will have higher BPV.

Table 2 also shows the relationships between average values of continuous variables and having an adverse outcome. Those with adverse outcomes had slightly less than one medication on average (p=.012). The number of preoperative BPs were significantly different (16.64 to 19.59, p=.004). The systolic SD was significantly higher for those with adverse events (12.89 to 12.11, p=.043) and the diastolic SD was also significantly higher (8.26 to 7.68, p=.009). There was no significant differences in any other mean preoperative variable. Intraoperative means were only significantly higher for systolic SD (10.63 to 10.25, p=.010). No other intraoperative means were significantly different.

As age and total number of medications showed potential to be related to the outcomes, average BPV was compared between levels of these variables to further test them for potential confounding using one-way ANOVA (Table 3). Mean systolic BPV differed between age groups at both pre-op (BPV=11.3 for under 50 and BPV=13.3 for 51 to 65) and intra-op (BPV=13.9 under 50 differed from other three age groups and 51 to 65 BPV=16.9 differed from 70 and older BPV=19.0). For diastolic BPV, only the intra-op values in the under 50 age group (BPV=11.4) were significantly different from the 70 and older age group (BPV=9.9). When considering total medications, patients taking 0 to 3 had systolic pre-op BPV=10.9 significantly lower than those taking 4–5 (BPV=12.7) or 8 or more (BPV=13.1). For intra-op, the lowest medication group 0 to 3 had BPV of 14.5, significantly lower than any other medication group (BPV 17.2 to 18.6). For diastolic BPV, only pre-op for people with 0 to 3 medications (BPV=7.4) was significantly lower than those taking 8 or more (BPV=8.3).

Table 3.

Differences in mean BPV according to age groups and number of medication groups.

Age	N	Pre-Op			Intra-Op
		Mean	SD	Differed	Mn	SD	Differed
Systolic
<50	75	11.302	3.761	51–65	13.875	5.828	51–65, 66–70, >70
51–65	124	13.307	5.031	<50	16.916	6.353	<50, >70
66–70	43	12.232	3.665		19.405	6.333	<50
>70	127	12.440	3.909		18.991	6.032	<50, 51–65
Diastolic
<50	75	7.761	2.709		11.374	4.088	>70
51–65	124	8.193	2.589		10.396	3.474
66–70	43	7.954	1.980		10.508	3.986
>70	127	7.832	1.875		9.886	3.236	<50
Total Medications
Systolic
0–3	74	10.902	3.622	4–5, 8–13	14.508	4.649	4–5, 6–7, 8–13
4–5	92	12.746	4.653	0–3	18.106	6.568	0–3
6–7	91	12.667	4.515		17.206	6.354	0–3
8–13	112	13.139	4.067	0–3	18.565	6.873	0–3
Diastolic
0–3	74	7.363	2.270	8–13	10.476	3.490
4–5	92	8.090	2.721		10.486	3.553
6–7	91	7.812	2.264		10.178	3.530
8–13	112	8.345	1.967	0–3	10.566	3.849

BPV measured as standard deviation three years pre-op or intra-op. Groups significantly different (p<.05 after Bonferroni adjustment) are listed in the Differed column.

Both pre- and intraoperative BPV were entered into logistic regressions predicting the four more common adverse events, readmissions, wound infections, AKI, and the combined endpoint of all adverse events (Table 4). Potential covariates in the model are shown on the table if significant. With both BPVs and covariates in the model, systolic preoperative BPV was not a significant predictor of any adverse event. Systolic intraoperative BPV may be associated with increased risk of AKI (OR=1.040, 95% CI 0.998–1.0783) and the combined endpoint of all adverse events (OR=1.036, 95% CI 1.002 – 1.070). Diastolic preoperative BPV was associated with increased risk of readmission (OR=1.108), wound infection (OR=1.116) and the combined endpoint of all adverse event (OR=1.100). Diastolic intraoperative had some association with increased risk of AKI (OR=1.040).

Table 4.

Logistic regressions predicting adverse effects from pre- and intra-operative BPV.

Adverse Effect	Variable	OR	95% CI	p
Systolic:
Readmissions	Pre-Operative BPV	1.035	0.982 – 1.091	.198
	Intra-Operative BPV	1.011	0.976 – 1.048	.537
Wound Infection	Pre-Operative BPV	1.005	0.946 – 1.068	.863
	Intra-Operative BPV	0.996	0.957 – 1.037	.858
	Total Meds	1.118	1.021 – 1.225	.016**
AKI	Pre-Operative BPV	1.047	0.987 – 1.112	.126
	Intra-Operative BPV	1.040	0.998 – 1.083	.059*
	Age	1.030	1.009 – 1.051	.005***
Any Adverse Effect	Pre-Operative BPV	1.036	0.987 – 1.087	.154
	Intra-Operative BPV	1.036	1.002 – 1.070	.036**
Diastolic:
Readmissions	Pre-Operative BPV	1.108	1.006 – 1.221	.038**
	Intra-Operative BPV	1.033	0.970 – 1.100	.308
Wound Infection	Pre-Operative BPV	1.116	1.004 – 1.241	.042**
	Intra-Operative BPV	1.014	0.945 – 1.087	.701
	Total Meds	1.109	1.013 – 1.214	.025**
AKI	Pre-Operative BPV	1.047	0.897 – 1.112	.126
	Intra-Operative BPV	1.040	0.998 – 1.083	.059*
	Age	1.030	1.009 – 1.050	.005***
Any Adverse Effect	Pre-Operative BPV	1.100	1.003 – 1.206	.043**
	Intra-Operative BPV	1.030	0.972 – 1.091	.313
	Total Meds	1.087	1.011 – 1.170	.025**

Regression equations were based on pre- and intra-operative BPV predicting adverse effect.

Significant covariates in the models are shown.

^*p<.1;

^**p<.05;

^***p<.01

Potential covariates were age, gender, BMI, on blood pressure medications, total medications, average intra-operative BP, pre-operative BPV, and average pre-operative BP.

Figure 1A shows the distribution of preoperative and intraoperative systolic BPV. The r value for this correlation was 0.135 (p=.010), with a RMSE of 6.377, statistically significant but not strong. Intraoperative systolic BPV was 5 mmHg higher than preoperative BPV (17.30 to 12.48, t=12.80, p<.001). Figure 1B shows diastolic preoperative and intraoperative BPVs. These BPVs did not correlate significantly (r=.021, p=.691 with RMSE of 3.620). Again, the average intraoperative diastolic BPV was 2.5 mmHg higher than the preoperative diastolic BPV. (10.43 to 7.95, t=11.19, p<.001).

Figure 1A.

Systolic pre- and intra-operative BPV

Figure 1B. Diastolic pre- and intra-operative BPV

The average intraoperative BPV between the four adverse events are shown in Figure 2. These are adjusted for the preoperative BPV. The unadjusted intraoperative BPV is significantly higher for those with adverse outcomes of any event (p=.011) and AKI (p=.002). When adjusting for systolic preoperative BPV, (figure 2a) the average difference declines somewhat but the significance does not change. All of the interaction terms were not significant (all p>.05 and not shown). Figure 2b shows the results for diastolic BPV. None of the unadjusted intraoperative BPVs were significant between adverse outcomes and again, there is no change in significance when adjusted for preoperative BPV. None of the interaction terms were significant (all p>.05 and not shown).

Figure 2A.

Average increase in intra-operative systolic BPV from having an adverse event unadjusted, and adjusted for presence of pre-operative systolic BPV

Figure 2B. Average increase in intra-operative diastolic BPV from having an adverse event unadjusted, and adjusted for presence of pre-operative diastolic BPV

Discussion

BPV, distinct from hypertension, affects prognosis in medical patients.^1–3 Both outpatient preoperative BPV^2,3 and intraoperative BPV^13–17 are associated with adverse outcomes. We found preoperative SBPV to be associated with the combined endpoint of all adverse events in variation with intraoperative SBPV; while preoperative DBPV was associated with increased risk of readmissions, wound infection, CVA, and the combined endpoint of all adverse events. Intraoperative SBPV was associated with increased risk of AKI and the combined endpoint of all adverse events, while intraoperative DBPV was associated with increased risk of AKI. Antimicrobials, autonomic, and respiratory medications were associated with more frequent adverse outcomes, whereas otolaryngology surgery seemed less so. Finally, preoperative BPV and intraoperative BPV were independent risk factors.

The cause of BPV is likely to be multifactorial. Much of BPV may autonomic instability,⁴ while other contributing factors include medication non-compliance,^9,11,12 “white coat hypertension,”^8,10 and inaccurate BP measurement.²⁷ BPV has been found to correlate with prolonged hospitalization, stroke, other vascular events, and mortality in medical patients.^1,28,29 Methods to control SBPV may be inadequate as lowering ambulatory SBPV with antihypertensives did not significantly reduce mortality risk in one large study.³⁰

We found preoperative SBPV to be associated with increased risk of the combined endpoint of all adverse events in variation with intraoperative SBPV. When intraoperative SBPV was included in preoperative SBPV analysis, preoperative SBPV was no longer an independent factor. Logically, intraoperative SBPV could not cause variation in preoperative SPBV, suggesting that the pathophysiology of preoperative SPBV is likely acting to destabilize intraoperative SBPV or is causing other pathophysiologic changes not yet explored which destabilize intraoperative SBPV. Preoperative DBPV was associated with increased risk of wound infection, readmission, CVA and the combined endpoint of all adverse events. Conversely, preoperative DBPV remained an independent factor even when intraoperative DBPV was included in analysis. Further exploration is needed to better understand these differences between SBPV and DBPV and how they might affect adverse outcomes. The findings of preoperative DBPV within this study is consistent with observations in a surgical cohort from the US Veterans’ Administration (VA) system that preoperative DBPV was associated with MI, CVA, AKI, readmission and mortality.² The VA study included 28,439 patients, with greater statistical power.² That study relied on computer abstraction from a large database to identify outcomes and could not access information from other hospitals that veterans might have visited with complications, while charts and outlying hospital records were reviewed individually in this study. Interestingly, wound infection was not significantly associated with preoperative DBPV in the larger VA study. The VA population was male- dominant; our patient sample was more gender-balanced. VA patients averaged ten years older, with a mean age of 68, and more commonly smokers. Finally, different criteria could have been used in these health care systems for diagnosing or recording postoperative wound infections. The potential contribution of these differences to the association of DBPV with wound infection awaits further exploration.

Although our main focus was preoperative BPV, intraoperative SBPV also correlated with the combined endpoint of all adverse events, consistent with reports.^13–17 Our findings differ from a previous report that intraoperative DBPV correlated more strongly than intraoperative SBPV with postoperative problems.¹⁷ Intraoperative BPV can reflect intra-procedural autonomic instability⁴ or adverse surgical or anesthetic events.^17,31 It seems intuitive that adverse intraoperative events would risk poor postoperative outcomes. Furthermore, surgical stress under anesthesia releases immunomodulatory cytokines that may affect postoperative outcomes.^32–36

The relationship between preoperative and intraoperative BPV had not previously been explored. Patient compliance^9,11,12 and measurement error²⁷ exogenously influence preoperative BPV, in addition to intrinsic biology such as autonomic instability⁴ and vascular wall stiffness.^5–7 Since we excluded BPs on the day of surgery from preoperative BPV, acute dehydration or other preoperative conditions would not have contributed. Conversely, intraoperative BPV is exogenously influenced by hydration, blood loss, decreased venous return from pneumoperitoneum, and anesthetic management^13–15 and is intrinsically influenced by the patient’s illness. Regardless of the underlying mechanism, we found preoperative and intraoperative BPV to affect prognosis independently.

Due to the incidental finding that otolaryngology patients seemed to have more favorable outcomes, we retrospectively compared the characteristics of these patients and their procedures to the rest of the study sample. Other than the nature of the procedure itself, we found the only difference in the patients undergoing otolaryngology surgery was that they were younger, used fewer medications, and were less likely to use any BP medication, particularly beta-blockers, than the other patients studied. These variables were included in our multivariate analysis, and otolaryngology surgery still fell out as an independent predictor in that analysis, suggesting that there may have been other distinguishing characteristics either of the patients or the surgical procedures that may have contributed to these favorable outcomes. One possibility is that the procedures may have been less invasive, caused less blood loss, or caused less tissue damage and inflammation than the other surgical procedures studied here. Such speculation exceeds the scope of the current study and would require a much more detailed analysis of a much larger group of patients for further exploration. In contrast, preoperative use of antimicrobials, autonomic drugs, and respiratory drugs increased the likelihood of postoperative adverse outcomes. Perhaps ongoing illnesses or comorbidities predisposed these patients to worse postoperative outcomes.

One potential limitation to this study is selection bias. We eliminated 231 of 600 patients who did not have at least eight recorded preoperative BP readings in the previous three years. One consideration in interpreting these results would be whether results had been biased by the exclusion of the patients who did not have enough BP measurements for accurate BPV calculation. The chief differences between the two groups were that the excluded patients were younger, somewhat less likely to be overweight, and most fell into the “other” surgical category. Neither age nor BMI affected outcomes in our multivariate analysis, so it does not seem likely that differences between the included and excluded patients in age and BMI would have substantially skewed our results. Additionally, the majority of the excluded patients fell into the “other” surgical category, indicating they were not undergoing the typical types of surgical procedures that were examined in this study in our included cohort of patients. Age and total medication were related to both outcomes (Table 2) and BPV (Table 3). As such, they had some effect as confounders when looking at the BPV to outcome relationship (Table 4). Both may be a reflection of severity or the healthiness of the individual as they increased the risk of adverse effects. At times, such as systolic BPV predicting wound infection, they were the only significant marker. They are factors that should be included in models as they may have key effects on the relationship of BPV to adverse effects. While there is always the possibility that exclusion might have also biased the study because of some other difference between the two groups that we were unable to capture, the fundamental observation remains that, at least in the patients we could study, both preoperative BPV and intraoperative BPV were independent risk factors with regard to adverse outcomes. Thus, this is likely true for at least some patients, even if not certainly so for all patients. We required eight BP values as a study qualification because this provides adequate accuracy for BPV calculation.^{1–3,37–39} The NHANES study by Mutner analyzed mortality with systolic BPV in 956 patients who had 3 values over six years.³⁸ Three BP readings would not be as reliable in our smaller study population. The mean systolic BPV among our patient population was 12.48 mmHg which is consistent with previous analyses by our own group and others in unselected data sets.^1–3,37 Another possible limitation is that we might have missed complications managed elsewhere. This is geographically unlikely. Altru is the only hospital in Grand Forks, and surgeons’ access outside records through shared data user agreements for follow-up appointments at neighboring facilities.

One key reason we chose SD as a measure of variability over other measures was its near normal distribution which would then yield to use the most efficient form of BPV and ease of interpretation. Blood pressure variability can be expressed in different ways, including SD, variance, standard error, greatest change between consecutive readings, coefficient of variation, and variability independent of the mean (VIM)^{1–3, 18,37–42} The use of SD tends to yield a higher magnitude of the expressed variability than some other calculations of variability. High BPV tends to be high and low BPV tends to be low regardless of which system is used, however students of BPV should be aware of the potential difference in magnitude of BPV expression when comparing studies that use different calculations. The various definitions of BPV generally yield comparable correlations with risk factors,^{1–3, 37–39} and in general tend to parallel each other for most measurements of variability in other data sets as well, as long as the data set does not range far from normality. We chose to express BPV using SD in this study rather than other measures because BPV was nearly normally distributed, yielding the most efficient form of BPV and ease of interpretation, as well as ready comparability with other studies that have used this method of calculating BPV^{1–3, 18, 37–39}Additionally, there may be differences between outpatient and short-term inpatient BP measurement because of differences in the setting, how the measurement is done, the acute illness or event that has led to hospitalization, and the time interval between measurements.

The overall adverse event rate in our patient sample appears to be somewhat higher than what one might expect to observe in a consecutive surgical patient roster. This likely reflects three different factors. First, we excluded by our study criteria patients who had shorter operative procedures (less than two hours) who would also be less likely to experience postoperative adverse events. Second, it is possible that the requirement for at least 8 preoperative and 8 intraoperative BP measurements may have selected for somewhat sicker patients. We know that the included patients tended to be older, more obese, and fell into standard surgical categories (Table 1). Third, adverse events here were defined by careful specific review of the full patient record by unbiased observers. This may have tended to identify some possible adverse events (such as an increase in postoperative creatinine) that others at the time would not have considered “complications” worthy of reporting.

This study included a broad range of procedures (Supplemental Table 1), while previous studies preselected patients for specific procedures. Therefore, our data can be generalized over a broad range of procedures. However, we did preselect procedures that lasted at least two hours to assure that we would have enough intraoperative BP readings to assess intraoperative BPV. Therefore, these results may not generalize to short minor procedures. Emergency procedures were excluded to avoid outcomes dominated by acute illness.

Conclusion

This study demonstrates that preoperative DBPV is associated with increased risk of readmission, wound infection, and the combined endpoint of all adverse events. Intraoperative SBPV is associated with increased AKI and the combined endpoint of all adverse events. These results can be applied in practice. Preoperative BPV should be factored into decision making when evaluating risk for procedures. In addition, although patients cannot modify their BPV, information about BPV risk might motivate patients to address modifiable risk factors before elective procedures. Intraoperative BP is not known until after a surgical procedure and so would not be useful for preoperative prognostication, but patients with both preoperative DBPV and intraoperative SBPV should be monitored closely after surgery, as they have two independent risk factors for adverse outcomes. Calculating BPV appears daunting, but a physician can use the greatest change between two consecutive BP readings as a surrogate for BPV¹ until the electronic medical record (EMR) can be set to pre-calculate BPV as it does for BMI. BPV should be also considered in risk stratification in large population studies or clinical trials and should be included in surgical risk adjustment systems as an additional patient risk factor in calculating observed/expected complication ratios to more accurately assess system quality.

Glossary of Terms:

AKI

acute kidney injury

BMI

body mass index

BP

blood pressure

BPV

blood pressure variability

CCB

calcium channel blocker

CVA

cerebral vascular accident

DBPV

diastolic blood pressure variability

EMR

electronic medical record

MI

myocardial infarction

OR

odds ratio

SBPV

systolic blood pressure variability

SD

standard deviation

VA

Veterans’ Administration