Associations of Iron Sucrose and Intradialytic Blood Pressure

Abstract

Rationale and Objective:

Intradialytic hypotension and intra-dialytic hypertension are associated with morbidity and mortality in hemodialysis (HD). Many factors can contribute to intra-HD blood pressure (BP) changes, such as drugs with vasoactive properties that could destabilize already tenuous BP. Intravenous iron sucrose (IS) is commonly administered to correct iron deficiency; however, reported associations with altered hemodynamics are not consistent.

Study Design:

Prospective cohort study.

Setting and participants:

950 outpatient HD participants.

Exposure:

Iron sucrose (IS) administered during HD.

Outcome(s):

Intradialytic hypotension, intradialytic hypertension, systolic blood pressure parameters.

Analytical Approach:

Unadjusted and adjusted Poisson and linear repeated measures regression models.

Results:

The mean age of patients included in the study was 53 ±22 years, 43% were females, and 38% were Black. Mean pre-HD SBP was 152 ±26 mmHg. At baseline, patients who received higher doses of IS tended to be diabetic, have longer HD sessions, and higher frequency of ESA use, compared with those who did not receive IS. In adjusted models, higher doses of IS were associated with an 11% lower rate (incidence rate ratio [IRR] for IS≥ 100mg vs. none 0.89; 95%CI 0.85 to 0.94) of intradialytic hypotension. In adjusted analyses, administration of higher doses IS during HD were associated with intradialytic hypertension (IRR for IS≥100mg vs. none 1.07; 95%CI 1.04 to 1.10).

Limitations:

Non-availability of the precise iron sucrose formulation (volume), laboratory data for each HD session and outpatient medications. Objective measures of volume status, home medications, and symptom data were not recorded in this study.

Conclusions:

We observed an independent association of intravenous IS administration during HD with a lower risk of intradialytic hypotension and higher risk of intradialytic hypertension. Future studies to better understand the mechanisms underlying these associations are warranted.

Plain Language Summary:

Intradialytic hypotension and intradialytic hypertension are common among patients on hemodialysis and they are associated with morbidity and mortality. While many factors may contribute to these risks, medications administered during hemodialysis play an important role. We studied the significance ofof intravenous iron sucrose (IS) and the impact it may have on blood pressure during dialysis. In our study of 950 outpatient hemodialysis (HD) patients, we observed that IS administration was associated with higher systolic blood pressure (pre-HD, intra-HD, and post-HD) as well as a lower risk of intradialytic hypotension We also observed that higher dose of IS are associated with development of intradialytic hypertension.

Introduction:

Abnormal blood pressure (BP) is a major health concern for the ~485,000 people in the United States receiving maintenance hemodialysis (HD).¹ Intradialytic hypotension (IDHypo) is one of the most common complications of HD and occurs in up to 68% of HD sessions, depending on the definition used.²IDHypo is associated with end-organ ischemic damage leading to increased risk of myocardial stunning, heart failure, limb ischemia, dementia, loss of renal reserve, and cardiac- and all-cause mortality.^3–6 On the other end of the spectrum, intradialytic hypertension (IDHyper) occurs when systolic BP rises or fails to lower over the course of a dialysis session and is estimated to affect up to 22% of HD sessions.^7,8 IDHyper is associated with higher rates of hospitalization, cardiac and all-cause mortality.^8–10

Intradialytic BP is affected by multiple patient and treatment-specific factors, including cardiovascular, neurohormonal, and autonomic dysfunction, in addition to blood, dialysate, and ultrafiltration rates.^9,11 Intradialytic medications must be considered carefully, as administration of drugs with vasoactive properties could destabilize an already tenuous BP. One such medication is intravenous IS, which is a nondialyzable, polynuclear iron (III)-hydroxide and sucrose mixture, widely used in the HD population to correct iron deficiency.^12,13 There have been concerns raised that IS infusions generate a bioactive, “labile” free iron fraction in the blood that may cause endothelial dysfunction through oxidative stress.^14–16 Data on the safety profile of intravenous IS is not entirely consistent, with some studies reporting an association with up to a threefold higher risk of IDHypo,¹⁷ while others do not.¹⁸

Based on prior data suggesting a dose-response relationship between IS and risk of transient hypotension (IDHypo), ¹⁹ we undertook this study with the a priori hypothesis that an association between IS administration and IDHypo would be seen in this large cohort.

Methods:

Study Population

The current analyses were performed using a prospective cohort of anonymized samples and statistically de-identified clinical data from a biorepository assembled by DaVita Clinical Research (DCR) collected from January 2008 to December 2014 and made available to academic organizations through the Biospecimen Research Grant (BioReG) program. Patients who were <18 years old, with Hgb <8.0 g/dL, pregnant, or with any physical, mental, or medical condition which limited the ability to provide written informed consent were excluded from BioReG. The present study only included patients undergoing thrice-weekly HD. The sampling protocol was approved by an Institutional Review Board (Quorum IRB, Seattle, WA, USA) and patients provided written informed consent prior to the initiation of sample collection. All clinical and hemodialysis prescription data were collected from the electronic medical record. A randomly sampled subset of the total cohort was provided to each of four academic institutions by DCR in a deidentified format and all sessions were analyzed for our included population. Follow up was restricted to 1 year for the purpose of the present analyses.

Biospecimen collection and storage

Biospecimens were collected and processed according to a standardized protocol, including shipping on refrigerated packs on the same day as collection, processing, aliquoting, and storage at −80°C. Recollection was requested for any specimen with cause for rejection (e.g., unspun tubes, insufficient volume, or thawed specimens). Specimens received >48h from the time of collection were also rejected and re-collected. Samples were distributed frozen at −80°C across the four academic medical centers.

Exposure

The primary exposure for this study was the administration of various doses of intravenous IS (no IS (reference),<100mg, and ≥100mg), at each individual HD session. The prescribed dose and confirmation of IS administration were obtained from the electronic medical record as IS dosing can be dynamic and vary from session to session within an individual participant.

Outcome Ascertainment

The primary outcome was IDHypo, defined as either an absolute intradialytic nadir systolic BP<90 mmHg in patients with a pre-HD systolic BP of <160 mmHg or nadir systolic BP<100 mmHg in patients with pre-HD systolic BP ≥160 mmHg. This definition of IDHypo was selected over alternative definitions due to its association with mortality. ²

Secondary outcomes included other intra-HD blood pressure parameters (nadir intra-dialytic systolic BP and post-HD systolic BP). The development of intra-dialytic hypertension was also considered, defined as any increase in systolic BP from pre- to post-HD.²⁰ BP was measured at all study sessions per standard clinical guidelines and abstracted from the electronic medical record.

Assessment of Other Covariates

Demographic information including age, race, sex, dialysis access and comorbid conditions including diabetes, heart failure, coronary artery disease, cerebrovascular disease, peripheral vascular disease, liver disease, and lung disease, were recorded at baseline (taken as the first HD session during the observation period) and updated from the medical record (via ICD-9 codes) throughout the study. Additional information such as the HD prescription, erythropoiesis-stimulating agent (ESA) dose, vascular access, and laboratory data were collected at each session. Dialysis session length was categorized (≤180 mins; 181-209 mins; 210-239 mins; ≥240 mins). Ultrafiltration volume was calculated by subtracting the post-dialysis weight from the pre-dialysis weight.

Statistical Analysis

Baseline characteristics were recorded according to tertile of IS doses (0mg, <100mg, ≥100mg) and groups were compared using tests for trend based on linear regression, χ2 trend test, and the Cuzick non-parametric trend test, as appropriate.

Poisson and linear random effects models, including a random intercept for subject-wise variability to account for within patient correlation among sessions, were fit to estimate the association of tertiles of IS administration with outcomes of interest on a per session basis. Model 0 was unadjusted. Subsequently, multivariable adjusted linear random effects and Poisson regression models were fit, including a random intercept for subject-wise variability. Assessment for the presence of effect modification for pre-specified variables (sex, diabetes, heart failure) was performed via the inclusion of cross-product terms in Model 2. Model 1 was adjusted for age, sex, race, access, and pre-HD systolic BP; Model 2 was adjusted for the same variables as model 1 with additional adjustment for vintage, ultrafiltration rate, diabetes, heart failure, ischemic heart disease (history of coronary artery disease, myocardial infarction, or angina), peripheral vascular disease, liver disease, lung disease, erythropoietin dose. Model 3 adjusted for the same variables as model 2, with additional adjustment for hemoglobin (Hb); Model 4 adjusted for the same variables as model 3, with additional adjustment for endothelin-1. The latter two models were considered exploratory, as Hb was generally only available on a monthly basis and ET-1 was only measured at baseline. Thus, missing data was imputed for these analyses on a last non-missing basis. Model covariates were selected for inclusion based on clinical and biologic plausibility.

All analyses were performed using Stata MP version 16 (StataCorp LP). A two-sided P-value <0.05 was considered to be statistically significant, without adjustment for multiple testing.

Results

We examined data from 950 subjects and 135,412 HD sessions from the BioReG cohort (Figure 1). The median number of sessions per patient was 154 [148, 157]. Those included were more likely to have a lower baseline pre-HD systolic BP and were more likely to have had an ESA administered during HD, compared with those excluded from the final cohort (Table S1).

Figure 1.

Consort Diagram

The mean age of patients included in the study was 56 ±20 years, 43% were females, and 38% were Black. Mean pre-HD systolic BP was 152 ±26 mmHg. At baseline, patients who received higher doses of IS tended to be diabetic, have longer HD session lengths, higher hemoglobin, and greater ESA use, compared with those who did not receive IS (Table 1).

Table 1.

Baseline characteristics according to dosage of iron sucrose (IS)

Baseline Characteristics*	IS (−) (n=828)	IS (<100mg) (n=42)	IS ≥100 mg (n=80)	P-trend
Age, yrs	53 ± 22	50 ± 20	52 ± 20	0.6
Female, n(%)	363 (44%)	17 (40%)	34 (42%)	0.7
Pre-Dialysis weight, kg	90.3 ±24.1	92.4 ± 25.1	88.6 ± 23.8	0.7
Ultrafiltration volume, L	2.1 ± 1.5	2.5 ± 1.6	2.1 ± 1.5	0.6
Ultrafiltration rate, ml/kg/hour	6 [4, 10]	7 [4, 11]	6 [3, 10]	0.7
Race, n(%)				0.8
White	323 (39%)	15 (36%)	35 (44%)
Black	317 (38%)	21 (50%)	23 (29%)
Other	188 (23%)	6 (14%)	22 (28%)
Pre-HD Systolic BP, mmHg	152 ± 27	152 ± 28	153 ± 30	0.8
Nadir intra-HD SBP, mmHg	120 ±25	120 ±23	123 ±28	0.4
Post-HD SBP, mmHg	147 ±27	143 ±26	149 ±28	0.6
Access, n (%)				0.3
AVF	567 (67%)	30 (71%)	47 (59%)
AVG	108 (13%)	9 (21%)	20 (25%)
Tunneled Catheter	153 (18%)	3 (7%)	13 (16%)
Dialysis Vintage, months	0 [0, 18]	16 [4, 53]	2 [0, 26]	0.01
Dialysis Session Length†, mins				0.03
<180	203 (25%)	4 (10%)	11 (14%)
>180	134 (16%)	10 (24%)	17 (21%)
≥210	271 (33%)	14 (34%)	24 (30%)
≥240	217 (26%)	13 (32%)	28 (35%)
Diabetes, n(%)	355 (43%)	20 (48%)	45 (56%)	0.02
Hypertension, n(%)	247 (30%)	15 (36%)	31 (39%)	0.07
Ischemic Heart Disease, n(%)	73 (9%)	3 (7%)	6 (8%)	0.6
Heart Failure, n(%)	70 (8%)	6 (14%)	8 (10%)	0.4
Peripheral Vascular Disease, n(%)	16 (2%)	3 (7%)	3 (4%)	0.1
Liver Disease, n(%)	34 (4%)	2 (5%)	4 (5%)	0.7
Lung Disease, n(%)	27 (3%)	1 (2%)	1 (1%)	0.3
Serum Albumin, g/dL	3.5 ± 0.5	3.6 ± 0.5	3.3 ± 0.5	0.09
Hemoglobin, g/dL	10.5 ±1.5	11.2 ±1.8	11.0 ±1.8	0.02
Endothelin-1, pg/mL	2.0 [1.5, 2.7]	2.3 [1.8, 3.0]	2.1 [1.6, 2.6]	0.2
ESA Dose, units per HD	0 [0, 6600] 3944 ±5969	3850 [0, 8800] 5657 ±7528	6600 [3300, 11000] 8118 ±6930	<0.001

Abbreviations: BP, blood pressure; AVF, arteriovenous fistula; AVG arteriovenous graft; ESA, erythropoiesis stimulating agent

Normally distributed continuous variables are presented as means standard deviation; non-normally distributed continuous variables are presented as median [25^th-75^th percentiles].

^*the values used to construct this table are from the first recorded hemodialysis session for each patient

^†For Dialysis Session Length, min: 4 participants lacked a documented session length at baseline.

Iron Sucrose and intra-dialytic hypotension

During the course of the observation period, IS was administered at 35,766 of the 135,892 sessions (20,001 (15%) sessions with IS < 100 mg and 15765 (12%) session with IS ≥100 mg). In unadjusted analyses, higher doses of IS were associated with a 16% lower rate of experiencing IDHypo (incidence rate ratio [IRR] IS ≥100mg: 0.84 (0.80 to 0.89). In adjusted models 1 and 2 this association persisted, with 14% lower rate (IRR for IS ≥100mg 0.86; 95%CI 0.81 to 0.90) and 11% lower rate (IRR for IS≥ 100mg 0.89; 95%CI 0.85 to 0.94), respectively. In exploratory models that were adjusted for hemoglobin and hemoglobin and endothelin, respectively, higher IS administration was associated with a 11% (IRR for IS ≥100mg 0.89 (0.80 to 0.99) and 13% lower rate of IDHypo (IRR for IS ≥100mg 0.87 (0.77 to 0.98). When compared with the 0mg IS group, there were no significant associations with the <100mg group and outcomes of interest.

Iron Sucrose and intradialytic hypertension

In unadjusted and fully adjusted analyses, administration of higher doses of IS during HD was associated with intradialytic hypertension (IRR for IS≥100mg 1.07; 95%CI 1.04 to 1.10 in model 2). This pattern persisted in exploratory models that adjusted for hemoglobin and endothelin (Table 2). There was no association between the lower dose IS group (<100mg) and outcomes of interest when compared with the reference group.

Table 2.

Association of iron sucrose (IS) administration by dosage with intradialytic hypotension and hypertension.

Incident Rate Ratio (95% CI) for IS administration by dosage versus not
	No IS	IS < 100mg	IS ≥100mg
IRR (95%CI) for IDH
Unadjusted	Ref	0.99 (0.95 to 1.04)	0.84 (0.80 to 0.89)
Model 1	Ref	1.01 (0.96 to 1.05)	0.86 (0.81 to 0.90)
Model 2	Ref	0.98 (0.94 to 1.03)	0.89 (0.85 to 0.94)
Model 2 + Hb	Ref	0.96 (0.88 to 1.06)	0.89 (0.80 to 0.99)
Model 2 + Hb + ET-1	Ref	0.92 (0.83 to 1.03)	0.87 (0.77 to 0.98)
IRR (95%CI) for IDHyper
Unadjusted	Ref	0.93 (0.90 to 0.95)	1.10 (1.07 to 1.13)
Model 1	Ref	0.97 (0.94 to 1.00)	1.12 (1.09 to 1.15)
Model 2	Ref	1.00 (0.97 to 1.03)	1.07 (1.04 to 1.10)
Model 2 + Hb	Ref	0.98 (0.93 to 1.04)	1.05 (0.99 to 1.11)
Model 2 + Hb + ET-1	Ref	1.00 (0.94 to 1.07)	1.04 (0.97 to 1.11)

Abbreviations: IDH, intradialytic hypotension; IDHyper, intradialytic hypertension; Hb, hemoglobin; ET-1, endothelin-1

Iron Sucrose and HD-related systolic BP parameters

At baseline, the mean pre-HD systolic BP was 150 ±26 and 152 ±26, the mean nadir systolic BP was 114 ±14 and 116 ±22, and the mean post-HD systolic BP was 139 ±25 and 141 ±25 in the IS (−) and IS(+) groups, respectively (P<0.001 for all; Table 1).

In unadjusted analyses, IS administration was associated with 1.4 (95%CI 1.2. to 1.7) mmHg higher pre-HD systolic BP (Table 3). Overall, in the fully adjusted model, IS ≥100 mg administration was associated with 1.2 (95%CI 1.0 to 1.5) mmHg higher nadir systolic BP and 1.3 (95%CI 0.9 to 1.6) mmHg higher post-HD systolic BP, compared with no IS administration.

Table 3.

Association of iron sucrose (IS) by dosage with dialysis-related systolic blood pressure parameters.

SBP Parameter	Difference in Systolic BP (95%CI)			P-trend
	No IS	IS < 100 mg	IS ≥100 mg
Nadir SBP
Unadjusted	Ref	0.2 (−0.1 to 0.5)	2.5 (2.2 to 2.8)	<0.001
Model 1	Ref	−0.4 (−0.6 to −0.1)	2.0 (1.7 to 2.3)	<0.001
Model 2	Ref	0.1 (−0.1 to 0.4)	1.2 (1.0 to 1.5)	<0.001
Model 3	Ref	−0.3 (−0.9 to 0.3)	1.3 (0.7 to 1.8)	<0.001
Model 4	Ref	−0.1 (−0.7 to 0.6)	1.3 (0.7 to 2.0)	<0.001
Post-HD SBP
Unadjusted	Ref	0.0 (−0.3 to 0.3)	2.8 (2.5 to 3.2)	<0.001
Model 1	Ref	−0.5 (−0.8 to −0.2)	2.3 (2.0 to 2.7)	<0.001
Model 2	Ref	0.2 (−0.1 to 0.5)	1.3 (0.9 to 1.6)	<0.001
Model 3	Ref	−0.6 (−1.2 to 0.1)	1.0 (0.4 to 1.7)	0.03
Model 4	Ref	−0.3 (−1.0 to 0.4)	1.1 (0.3 to 1.8)	0.02

Abbreviations: BP, blood pressure; mmHg, millimeters of Mercury; CI, confidence interval

Effect modification and sub-group analyses

There was no evidence of effect modification by sex, diabetes, and heart failure for the association of categories of IS with IDHypo (P-interaction>0.2 for all).

Discussion

In this large cohort of patients receiving maintenance HD, we found that higher doses of intravenous IS during HD were associated with a lower rate of IDHypo. Further, we observed and an association between higher doses with development of IDHyper. Similarly, IS administration was associated with higher nadir intra-dialytic systolic BP and post-HD systolic BP.

Iron deficiency is common among patients receiving maintenance HD and is thought to be related to impaired nutrition, chronic blood loss, and decreased intestinal absorption of iron.^21–23 As an integral component in erythropoiesis, administration of iron has been promoted as a means to correct anemia, reduce the need for transfusions, and to reduce the requirement for erythropoiesis-stimulating agents.²⁴ In the setting of HD, iron deficiency is associated with adverse symptoms and important clinical outcomes, such as hospitalization and mortality.^25,26 Indeed, proactive administration of IV iron (400 mg per month) was reported to be superior to a reactive IV iron strategy (0-400 mg per month) in the PIVOTAL trial in terms of reducing the risk for cardiovascular outcomes (composite of non-fatal myocardial infarction, non-fatal stroke, heart failure, or all-cause death).²⁷

The development of anaphylaxis is perhaps the most feared complication from administration of intravenous iron. In the non-CKD/ESKD population, the reported frequency of serious reactions related to intravenous iron administration appears to be relatively rare. For example, a meta-analysis by Wang et al comparing the safety profiles of different IV iron formulations found that use of IS carried a very low risk of anaphylaxis (21 per 100,000 persons), relative to the use of iron dextran.²⁸ Another study by Baile et al. that used a large United States Food and Drug Administration surveillance database found zero episodes of anaphylaxis/million mg of IS administered.²⁹ IS is frequently used to correct iron deficiency anemia in ESKD and is widely considered to be safe in HD patients with virtually no risk of causing anaphylaxis¹⁶. Still, transient hypotension, believed to be caused by labile free iron toxicity,^15–17,30 has been observed in patients receiving larger doses of IS¹⁹ and represents a potential etiology for increased risk of hypotension in patients with CKD or HD dependence. This was seen in a trial comparing IS to oral iron in patients with CKD (n=188) which described transient hypotension in two patients in the context of 500 mg dose administration.³¹ The Iron Sucrose Clinical Trial of repletion versus maintenance (n=665 patients on maintenance HD) reported IS-related nonserious IDHypo in 0.0004% of exposures and 0.004% of patients.³² Other studies, including the single-arm North American Clinical Trial (n=77), which looked at the effect of 10 consecutive 100mg doses of IS, and the Proactive IV Iron Therapy in Haemodialysis Patients (PIVOTAL; n=2141), which compared proactive (frequent) vs reactive (less frequent) doses of IS, did not raise concerns about hypotension.²⁷ However, it should be noted that these studies did not specifically examine prespecified definitions of IDHypo.

Although FDA labeling for Venofer (IS) references a 36% prevalence of hypotension among patients in the three, single-arm clinical trials used in the drug approval process, this risk may be overstated. None of the trials used standardized definitions of IDHypo and instead, events were recorded if a decline in blood pressure was felt to be significant in the investigator’s opinion. In the two published trials, all episodes of IDHypo occurred at least 2 hours after IS administration and were felt to be unrelated to the drug.³³ Van Wyck even observed a higher frequency of IDHypo among patients during the observation period rather than during sessions when IS was infused.³³

Conversely, one study examining different doses of IS in pediatric patients with CKD observed an adverse event of increased BP in 6.4% of patients in the higher dose arm, but granular information was not available on what definition of IDHypo was used or when this occurred in relation to the HD session.³⁴ In our study, we found that the administration of higher doses of IS were associated with a lower risk of IDHypo when compared with no IS administration. This contrasts with the common clinical concerns related to potential hypotension. As such, this analysis cannot rule out the possibility that IS administration may be a surrogate marker for another factor associated with higher blood pressures, such as more intense/attentive medical care or another unmeasured confounding variable. In this respect, intravenous IS is hyperosmolar (1200 mOsm/L) and is often administered in 100mL of isotonic saline (combined 404 mOsm/L). The addition of even small volumes of hypertonic fluid has been reported to elevate intradialytic BPs, likely through both expansion of the intravascular volume and enhanced vasopressin activity.³⁵ Due to data limitations, we were not able to determine the volume of diluent at each administration in the present analyses. Several other plausible but speculative mechanisms for a hypertensive effect exist as well. IS administration, through generation of labile free iron, can cause reactive oxygen species formation³⁶ and stimulate the release of the potent vasoconstrictive endothelial peptide endothelin-1³⁷. However, the fact that the association between IS and IDHyper persisted in our models after adjustment for endothelin-1 may argue against an endothelin-1-mediated etiology or represent a limitation from imputation of endothelin-1 values in our dataset. The treatment of iron deficiency anemia itself may play a role in augmenting blood pressure through a) reduction of pathologically elevated levels of serum nitric oxide³⁸ thereby counteracting its vasodilatory effect, and b) improvement in cardiovascular outcomes²⁷ which may act through improved cardiac function.³⁹ However, it is not clear if this would explain improved hemodynamic parameters at an individual session level.

Our study has several strengths such as a relatively large sample size, duration of follow-up, and availability of detailed, per-HD session intravenous medication and hemodynamic data. However, several limitations are present, including the non-availability of the precise IS formulation and volume, nonavailability of laboratory data for each HD session and outpatient medications (including recent administration of blood pressure medications), limited data on the dialysate temperature, exact timing of intra-dialytic BP readings, and potential misclassification of covariables secondary to the use of ICD-9 codes. No data is available for other intravenous iron formulations and we did not have access to the exact timing of IS administration during the HD session. Although other formulations are generally considered to be better tolerated than IS, we are unable to address their precise hemodynamic effects in this dataset.^14,40 Other potentially relevant information relating to objective measures of volume status, home medications, and symptom data were not recorded in this study, and it is possible that transient episodes of hypotension or hypertension may have been missed. We used data imputation in our exploratory models to account for a lack of available laboratory values, which carries the risk of introducing bias into the analysis. For all analyses, the potential for residual confounding remains. Although, the patient population and data collection methodology strongly reflect the US in-center hemodialysis population, it is impossible to account for hidden sources of bias and confounding.

In summary, we observed an independent association of intravenous IS administration with a lower risk of intra-dialytic hypotension and higher intra-dialytic systolic BP parameters. While these differences in blood pressure between sessions where IS is versus is not administered may appear clinically modest, on a population level, these may reflect important clinical outcomes and provide reassurance that risks of hypotension with IS were not observed. Future studies to better understand the mechanisms underlying this data are warranted.

Supplementary Material

1

Table S1. Comparison of included and excluded participants