Repeat Adrenal Vein Sampling in Aldosteronism: Reproducibility and Interpretation of Persistently Discordant Results

Abstract

Context

The reproducibility of adrenal vein sampling (AVS) is unknown.

Objective

This work aimed to determine reproducibility of biochemical results and diagnostic lateralization in patients undergoing repeat AVS.

Methods

A retrospective chart review was conducted of single-center, single-operator AVS procedures at a tertiary care center. Included were patients with confirmed primary aldosteronism (PA) undergoing repeat AVS because of concerns about technical success or discordant diagnostic results. Simultaneous AVS was performed by an experienced operator using a consistent protocol of precosyntropin and postcosyntropin infusion. Among successfully catheterized adrenal veins (selectivity index ≥ 2), the correlation of the adrenal_{aldosterone/cortisol} (A/C) ratio was measured between the first and second AVS. The secondary outcome measure was diagnostic agreement on repeat AVS lateralization (lateralization index ≥ 3).

Results

There were 46 sets of AVS from 23 patients at a median of 3 months apart. There was moderate correlation in A/C ratios in the adrenal veins and inferior vena cava (Spearman r = 0.49-0.59, P < .05) pre cosyntropin. Post cosyntropin, the correlation was better (Spearman r = 0.67-0.76, P < .05). In technically successful AVS, there was moderate correlation between the repeated lateralization indices (Spearman r = 0.53, P < .05). In 15 patients in whom repeat AVS was performed because of apparent lateralization discordance with computed tomography imaging, the final diagnosis was the same in the second AVS procedure. Initial failed AVS was successful 75% of the time on repeat attempt.

Conclusion

Repeat AVS was feasible and usually successful when an initial attempt failed. There was modest correlation between individual repeat adrenal A/C ratios and lateralization indices when AVS was performed twice. The final lateralization diagnosis was identical in all cases. This demonstrates that AVS is a reliable and reproducible localizing test in PA.

Primary aldosteronism (PA) is the most common form of secondary and remediable hypertension (1, 2). After biochemical diagnosis, adrenal vein sampling (AVS) is an evidence-based and strongly recommended step for subtyping to identify patients suitable for surgery (3-5). It is important to ensure that every step along the diagnostic pathway is reliable so that appropriate treatment may be provided. Although the performance of the aldosterone-renin ratio (ARR) is well known and the preanalytical factors that affect its reliability are widely reported (6-11), the reproducibility of AVS results remains unknown.

AVS requires at least 3 expert parties: the radiologist who performs the technical procedure, the laboratorian who performs the biochemical assays, and the clinician who interprets the results, so theoretically, there is the potential for error at each level. Much attention has been paid to the problem of clinical adrenal vein interpretation criteria and the extent to which the final subtype diagnosis may vary, depending on which biochemical interpretation protocol is applied (12, 13). Additionally, the effects of laboratory assay variation on clinical interpretation (within one interpretive regimen) has recently been reported (14). However, the extent of diagnostic variation based on technical aspects of AVS is still unknown, probably because of the ethical impracticality of having a PA patient submit to multiple, invasive AVS procedures purely for research purposes.

At times, however, repeat AVS in a PA patient is clinically indicated, if a first attempt is completely or partially technically unsuccessful or if the observed results suggest a final subtype diagnosis that is at odds with the clinical and imaging data. So-called “discordant” AVS results are common (15) and always raise the possibility of process error leading to sample mislabeling and subsequent confused interpretation. We performed a retrospective analysis of our single-operator AVS database to analyze AVS data in patients who underwent the procedure twice. We sought to determine the overall correlation between AVS biochemical data from 2 separate AVS catheterizations; based on the agreement (or lack thereof) of repeat AVS procedures, we aimed to describe the overall reproducibility of AVS diagnoses and the frequency with which a repeat AVS procedure could add diagnostic clarity to uncertain initial AVS results.

Materials and Methods

We performed a retrospective analysis of results from The University of Calgary AVS Program Database. The Calgary AVS Program was established in 2000, and all AVS procedures since 2005 have been performed by a single operator (B.S.) using a consistent, standardized protocol that has been published previously (16). The database consists of all patients who underwent AVS from 2005 onward and includes clinical and biochemical data pertaining to PA diagnosis, adrenal imaging, and AVS results. Clinical outcomes following treatment are routinely collected at the Endocrine Hypertension Clinic for patients living in the region. For patients referred from out of region (eg, neighboring provinces) to Calgary solely for AVS, clinical outcomes were not consistently available.

Briefly, patients referred for AVS at our center all had a clinical and biochemical diagnosis of PA made by the referring endocrinologist consistent with guidelines or local practice (17). Patients refrained from using mineralocorticoid receptor antagonist medications for at least 6 weeks before AVS; the use of angiotensin-converting enzyme inhibitors, angiotensin-receptor blockers, and diuretics was discouraged and stopped at least 2 weeks prior if possible. α-Blockers, β-blockers, calcium-channel blockers, and hydralazine were used as needed for blood pressure control instead. Hypokalemia, if present, was corrected with oral supplements before AVS. In the angiography suite, after intravenous sedation, the radiologist collected simultaneous baseline blood samples from the adrenal veins and peripheral sheath (iliac vein) for cortisol and aldosterone measurement. After a 250-µg cosyntropin bolus with subsequent infusion of 6.25-µg cosyntropin over 15 minutes, a second set of samples was obtained.

We used a selectivity index (SI) (adrenal_cortisol/inferior vena cava [IVC]_cortisol) of greater than 2:1 at baseline and greater than 3:1 post cosyntropin to define technical catheterization success. Unilateral aldosterone excess was diagnosed biochemically when the lateralization index (LI) ([aldosterone]_dominant/[cortisol]_dominant/[aldosterone]_nondominant/[cortisol]_nondominant) exceeded 3:1 at any time (18). The aldosterone and cortisol assays have been previously described (19); the Roche Cortisol II assay was adopted in 2014 to replace the previous Roche Cortisol (20), but for this analysis no AVS pairs involved different assays.

The AVS database was searched to select all patients with 2 separate AVS procedures or a record denoting that the AVS was a repeat procedure. Records were excluded from the analysis cohort if one of the AVS predated the database creation, if an AVS was performed outside Calgary, or if any of the AVS procedures had been terminated for patient issues before completion. After establishing the analysis cohort, each patient record was manually searched to abstract the recorded reason for repeat AVS being performed in addition to basic patient and case demographics, along with final diagnosis and outcome, if available.

To study the reproducibility of technically successful AVS, we compared the following 4 sets of results: aldosterone/cortisol (A/C) ratios collected from adrenal veins pre or post cosyntropin, and A/C ratios collected from the iliac vein, pre or post cosyntropin. An A/C ratio was considered valid for analysis if the SI proved a technical success for that particular sampling and if there was a matching A/C ratio from an equally successful second procedure. The correlation between a precosyntropin or postcosyntropin LI was described only for cases in which each element of both LI calculations was achieved through appropriate SI criteria for success.

Manual review of individual case results was performed to calculate the percentage of technically successful AVS when the second procedure was performed because of a failed SI on the first attempt. Given that many repeat AVS were performed because of an apparent diagnostic discordance with imaging data, as an exploratory analysis in a post hoc fashion we considered the possibility that a consistently discordant AVS result might reflect cortisol cosecretion from an adrenal mass (21, 22). This possible diagnosis was considered when there was contralateral lateralization in the setting of a known adrenal mass, in addition to a low SI (< 2) opposite the mass (due to adrenal vein cortisol suppression) despite an aldosterone_adrenal/aldosterone_IVC greater than 2.0.

Simple descriptive statistics were used to describe the study population and case details. A Shapiro-Wilk test for normality was performed on each variable prior to Spearman correlation for nonparametrically distributed data. A 2-tailed P less than .05 was considered to be statistically significant. GraphPad Prism 6.0.2 was used for statistical analysis, and the project was approved by the Conjoined Health Regional Ethics Board. Owing to patient privacy restrictions, data sharing is not available.

Results

The construction of the analytical cohort was as follows (Fig. 1): Of a total of 550 AVS procedures, 59 were recorded as repeat AVS procedures; 2 were excluded for missing laboratory data (preceding 2005), 9 were excluded because the first AVS was performed elsewhere, and 2 were excluded for incomplete procedures (AVS aborted because of patient-related factors). After applying exclusionary criteria, there were 46 sets of AVS from 23 patients who were included in the final study cohort. The basic clinical and biochemical data from each individual are recorded in Table 1. Of 23 patients, there were 12 men, and 16 of the 23 had a known adrenal lesion. The median (interquartile range) ARR was 2.2 (4.0)-fold greater than the locally validated renin-specific upper reference limit for ARR (reported as such because of differing renin assays used across the study follow-up). Table 2 shows the detailed information for each patient in terms of imaging results, indication for repeat AVS, and main AVS findings along with outcome data where available. In 8 of 23 cases, the indication for repeat AVS was failure to catheterize at least one adrenal vein according to the calculation of the SI. In the remaining 15 cases, the repeat AVS was performed to exclude sample mislabeling on account of having found an LI that was discordant to the clinical and imaging data. The median time between repeat AVS was 3 months (range, 1-47 months).

Figure 1.

Study cohort construction.

Table 1.

Baseline characteristics of the study cohort

Characteristics	All (n = 23)
Age, median (IQR), y	54 (20)
Age bands, No., (%)
< 40 y	3 (13.0)
40-49 y	6 (26.1)
50-59 y	7 (30.4)
≥ 60 y	7 (30.4)
Male, No. (%)	12 (52.2)
BMI, median (IQR), kg/m2	28.8 (10.1)
SBP, median (IQR), mm Hg	139 (21)
DBP, median (IQR), mm Hg	91 (17)
PAC, median (IQR), pmol/La	474.0 (334.0)
ARR, median (IQR), % ULNb	2.2 (4.0)
Serum potassium, median (IQR), mmol/L	3.3 (0.8)
Hypokalemia (< 3.6 mmol/L), No. (%)	17 (73.9)
Serum creatinine, median (IQR), µmol/L	76.0 (27.0)
eGFR, median (IQR), mL/min/1.73 m2	80.9 (29.1)

Abbreviations: ARR, aldosterone-to-renin ratio; AVS, adrenal vein sampling; BMI, body mass index; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate; IQR, interquartile range; PAC, plasma aldosterone concentration; SBP, systolic blood pressure; ULN, upper limit of normal.

^a Two patients had missing data for baseline plasma aldosterone concentration.

^b Different renin assays were used at different time points; to facilitate standardization, the degree of elevation in ARR specific to the renin assay in use is reported instead.

Table 2.

Imaging and adrenal vein sampling details according to indication for repeat sampling

Patient	Imaging results	Reason for repeat AVS	AVS 1 result	Time interval, mo	AVS 2 result	Final outcome, management details
1	No lesion	Failed R cath	bRSI 1.0 sRSI 1.5	4	bRSI 1.8 sRSI 35.7	Success; lateralized R; normal BP on MRA
2	14 mm L	Failed R cath	bRSI 1.0 sRSI 1.2	3	bRSI 0.9 sRSI 1.0	Failed AVS; L ADx based on CT
3	15 mm L	Contralateral LI	bLI 6.0 (R)	7	bLI 5.1 (R)	Same results
4	16 mm L	Contralateral LI	bLI 15.3 (R)	2	bLI 4.2 (R)	Same resultsa NP-59 scan shows hyperfunctioning L adrenal mass; surgery to remove tumor was curative
5	9 mm R	Failed L cath	bLSI 1.1 sLSI 2.0	7	bLSI 13.2 sLSI 86.2	Success; lateralized R
6	9 mm L	Contralateral LI	bLI 6.0 (R) sLI 3.7 (R)	5	bLI 1.6 (R) sLI 9.1 (R)	Same results
7	28 mm R	Contralateral LI	bLI 3.1 (L) sLI 4.1 (L)	2	bLI 2.6 (L) sLI 3.7 (L)	Same resultsa Left ADx but persistent high BP and high ARR
8	11 mm L	Contralateral LI	bLI 4.6 (R) sLI 14.2 (R)	2	bLI 14.4 (R) sLI 9.2 (R)	Same results—R ADx shows 15-mm R mass not seen on CT; normal ARR and clinical cure
9	29 mm L	Failed R cath	bRSI 1.1 sRSI 1.1	2	bRSI 4.7 sRSI 52	Success but contralateral LI; MRA therapya
10	No lesion	Lateralization with no adrenal mass	bLI 34.7 (R) sLI 27.6 (R)	43	bLI 114 (R) sLI 60.8 (R)	Same results—Lateralized R
11	No lesion	Failed L cath	bLSI 1.0 sLSI 1.8	4	bLSI 1.6 sLSI 27.7	Success but no lateralization: MRA therapy
12	8 mm L	Contralateral LI	sLI 4.7 (R)	7	bLI 18.6 (R) sLI 13.1 (R)	Same results
13	17 mm R	Contralateral LI	bLI 7.2 (L)	4	bLI 89.3 (R) sLI 4.1 (R)	Same resultsa, left ADx but persistent high BP and ARR
14	No lesion	Failed R cath	bRSI 1.1 sRSI 1.1	2	bRSI 1.7 sRSI 105	Success but no lateralization: MRA therapy
15	No lesion	Failed R cath	bRSI 1.2 sRSI 0.9	2	bRSI 2.4 sRSI 31.8	Success but no lateralization: MRA therapy
16	16 mm R	Contralateral LI	bLI 2.7 (L) sLI 6.6 (L)	3	bLI 2.0 (L) sLI 1.5 (L)	Same resultsa,b NP-59–positive mass, DHEAS suppressed. Surgery refused because of severe obesity
17	No lesion	Lateralization with no mass	bLI 3.9 (R) sLI 1.35 (R)	20	bLI 2.1 (R) sLI 1.94 (L)	Success but no lateralization: MRA therapy
18	10 mm L	Contralateral LI	bLI 7.4 (R) sLI 6.6 (R)	3	bLI 11.0 (R) sLI 7.0 (R)	Same results; R ADx. Pathology 0.9-cm adrenal mass, no hyperplasia. Normal ARR, BP improved
19	24 mm L	Contralateral LI	bLI 11.8 (R) sLI 4.7 (R)	3	bLI 3.1 (R) sLI 2.4 (R)	Same resultsa,b NP59-positive, ACTH suppressed. L ADx; postop ACTH now high but ARR still high with hypertension
20	18 mm L	Discordant— nonlateralizing	bLI 1.8 (L) sLI 3.2 (R)	1	bLI 12.0 (R) sLI 1.6 (R)	Same results
21	29 mm R	Contralateral LI	bLI 2.8 (L) sLI 3.4 (L)	4	bLI 1.4 (L) sLI 1.9 (L)	Same resultsa
22	No lesion	Failed R cath	bRSI 0.9 sRSI 1.0	3	bRSI 0.7 sRSI 1.6	Failed AVS
23	28 mm L	Contralateral LI	bLI 4.7 (R) sLI 1.1 (R)	47	bLI 2.0 (R) sLI 2.4 (L)	Same results (basal)a,b NP-59–positive mass, surgery L mass, normal ARR and BP improved

Abbreviations: ACTH, adrenocorticotropin; ADx, adrenalectomy; ARR, aldosterone-renin ratio; AVS, adrenal vein sampling; bLI, basal lateralization index; bLSI, basal left selectivity index; BP, blood pressure; bRSI, basal right selectivity index; cath, catheterization; CT, computed tomography; DHEAS, dehydroepiandrosterone sulfate; L, left; MRA, mineralocorticoid receptor antagonist; NP-59, ¹³¹I-iodocholesterol; R, right; sLI, stimulated lateralization index; sLSI, stimulated left selectivity index; sRSI, stimulated right selectivity index.

^a Query cortisol cosecretion.

^b See Supplementary material for more case details.

After confirmation of technical success by SI criteria, there were 27 and 39 pairs of adrenal vein A/C ratios pre and post cosyntropin, respectively, along with 23 paired measurements of IVC A/C ratios both for precosyntropin and postcosyntropin samples. Spearman correlations between these repeat paired measures are shown in Fig. 2 which shows there is a moderate correlation for adrenal and IVC A/C ratios at baseline, (r = 0.42 and 0.59, respectively, each P < .05). Following cosyntropin infusion, the correlations were closer, a trend seen both in adrenal and IVC measures (r = 0.67 and 0.76, respectively, each P < .05). For paired comparison of precosyntropin LI, there were only 5 from technically successful AVS procedures and these were not analyzed further because of small numbers. There were 16 pairs of postcosyntropin LI from technically successful AVS, and there was moderate correlation in the calculated LI between both AVS procedures (r = 0.53, P < .05) Fig. 3.

Figure 2.

Correlation of aldosterone/cortisol (A/C) ratios measured pre or post cosyntropin infusion, sampled from successfully catheterized adrenal veins (selectivity index > 2) or femoral sheath/inferior vena cava (IVC). ACTH, (1-24)adrenocorticotropin/cosyntropin.

Figure 3.

Correlation of adrenal vein sampling lateralization indices ([aldosterone]_dominant/[cortisol]_dominant/[aldosterone]_nondominant/[cortisol]_nondominant) in repeat, technically successful sets of adrenal vein sampling. ACTH, (1-24)adrenocorticotropin/cosyntropin, LI, lateralization index

In Table 2, the clinical impact of repeat AVS is demonstrated in the records of whether the second AVS agreed with the initial AVS in diagnosis. Among 15 patients in whom AVS was repeated because of imaging discordance, the final AVS-determined diagnosis was exactly the same as in the first set of results, effectively ruling out sample mislabeling/mishandling as a cause of diagnostic confusion in all instances. For the 8 patients in whom AVS was repeated for initial catheterization failure, 6 repeat procedures (75%) were technically successful on the second attempt. The radiology notes suggested that the initial failures were thought to be due to inadvertent catheterization of the hepatic vein, duplicated, tiny, or atypical right adrenal veins, catheter positional instability, inadvertent sampling of the left phrenic vein, or simple absence of a visible right adrenal vein for access.

Given a surprising number of repeatedly discordant lateralization results when comparing AVS with cross-sectional imaging, we performed an exploratory analysis around a hypothesis of possible cortisol cosecretion as a reason for reverse AVS lateralization (due to suppression of adrenal cortisol in the uninvolved gland, inflating the A/C ratio and reversing the LI). Among the 15 cases with discordant results, 12 were in patients with an adrenal mass of 10 mm or greater and 7 patients met the proposed biochemical criteria for possible cortisol cosecretion. In 3 of these 7, a ¹³¹I-iodocholesterol (NP-59) scan suggested unilateral steroidogenesis in the mass; another 2 of the 7 had surgery to remove the contralateral (normal) adrenal based on the AVS results and in both cases, the surgery failed to achieve biochemical or surgical cure, again suggesting probable cosecretory contralateral lesions as the correct diagnosis (see Supplementary material [23]) for full case details). After removing suspected cosecretors from the data set, a repeat of the primary analysis showed improved correlations in all measures (r = 0.58-0.79, P < .05, Supplementary Figs. 1 and 2 [23]).

Discussion

Repeat AVS is very uncommon, accounting for approximately 7% of all AVS performed at our center over the past 15 years. However, our collected experience provides a first report of AVS result reproducibility, at least in the hands of a single experienced operator using a consistent protocol for all procedures. Our data suggest that with standardized approaches to AVS and in expert hands, AVS generates largely reproducible results with moderate-to-good correlations between all similar-paired measures. The correlations are particularly good in the postcosyntropin samples, likely reflecting the ability of cosyntropin to eliminate fluctuations and variability in cortisol levels (24). Most important for clinical practice, the clinical subtype interpretations were largely consistent with repeated AVS procedures, even when lateralization appeared discordant from anatomical imaging. Moreover, repeat AVS was successful in most cases when cannulation was not achieved on the first attempt.

Because confirmation of unilateral PA subtype is a prerequisite prior to surgery for most cases (25), the importance of accurate AVS cannot be overstated. Novel nuclear medicine tracers may some day provide an alternative to AVS (26, 27) but outside a research setting, AVS is an absolute requirement in the surgical PA workup, particularly for patients with normal or indistinct adrenal imaging results (28, 29). If AVS is unsuccessful, it is probable that most patients lose the opportunity for surgical cure, which is unfortunate given the superior outcomes seen with curative surgical therapy compared to medical treatment (4, 30, 31), particularly when unilateral aldosterone excess is biochemically confirmed (32). Our data suggest that with an experienced operator, there is actually a high rate of AVS success on a second AVS attempt when the first one fails. In most of our cases, initial AVS failure arose in the setting of aberrant adrenal venous anatomy; repeating AVS another day allows the radiologist to reanalyze prior imaging and venograms to tailor a search strategy for the repeat procedure. Improved hydration before repeat AVS (such as with intravenous saline bolus prior to testing) may also make repeat catheterization easier.

One of the most interesting applications of repeat AVS is in the patient with successful catheterization (as judged by selectivity indices) but contralateral lateralization to the side opposite a known adrenal mass. This is not a rare situation and has been reported by our group and several others (17, 33, 34). Faced with this finding, the clinician has traditionally considered 3 possible explanations: 1) AVS blood sample mislabeling, 2) the presence of a nonfunctional adrenal mass in a PA patient, or 3) AVS results that simply fail to accurately lateralize (because AVS is not necessarily infallible). The first explanation is always a possibility and is usually the primary reason for repeat AVS. This emphasizes the importance of ongoing collaboration with the biochemistry laboratory (5) on continuous quality-control exercises that encompass diagnosis-to-surgery frames for the PA patient. Indeed, our data show that we did not have a single case of repeat AVS for which the original discordance was “corrected” and thus explained by probable mishandling of specimens. The possibility of a nonfunctional incidental adrenal mass must always be considered (35), and false lateralization is a theoretical possibility although rarely proved because many patients with persistent, unexplained computed tomography–AVS discordance are not offered definitive surgery.

We would like to add a fourth possible explanation for consideration—that of reverse lateralization in AVS due to aldosterone-cortisol cosecretion from an obvious adrenal mass. This situation has been previously reported but is usually diagnosed on the basis of a dexamethasone-suppression test (20, 36) and suppressed dehydroepiandrosterone sulfate (DHEAS) levels (37), although postoperative low cortisol has also been suggested (38), as have presurgical steroid metabolomic signatures in research laboratories (38). We acknowledge that the physicians ordering AVS in our center did not systematically screen for unexpected, coexisting subclinical cortisol excess using dexamethasone-suppression tests but the present results emphasize the importance of considering this test in all patients presenting with an adrenal mass, regardless of PA diagnosis. In addition, in light of the very preliminary experience reported here, we suggest that criteria for AVS-detected cortisol cosecretion be developed and validated prospectively. Measurement of adrenal vein metanephrine may be considered as an alternative “control” hormone in cases where cortisol is expected to be unreliable or asymmetric (39). Recognition of this phenomenon could be very important to ensure that persistently discordant imaging–AVS results are not ignored; if anything, cortisol cosecretion might be an even greater reason to pursue surgery for an affected patient (40).

Our study’s strengths include the use of a consistent AVS protocol and a single, experienced AVS operator throughout the entirety of the study time frame, which typically enabled us to achieve an AVS success rate of at least 97% (17). There are some limitations, however; our results may not be applicable to centers with less AVS experience. The number of repeat AVS tests was still small, despite being a high-volume AVS center. However, a larger, multicenter study of AVS reproducibility could be more difficult to interpret given the known variations in patients, AVS performance, and interpretation reported by different investigators (41). We acknowledge that we lack detailed data regarding medication use in every AVS case and, although we followed the principles of patient preparation outlined earlier, it is possible that some variation could be explained by slight differences in antihypertensive drug use between repeat AVS sets. Last, the persistent discordance/cortisol cosecretion hypothesis was a post hoc hypothesis that was not clinically suspected or confirmed at the time, and therefore still requires clinical validation in our patients and those from other centers before any routine adoption into AVS interpretation regimens.

Summary

This analysis of repeat AVS provides reassuring evidence that where AVS is performed by an expert center with ongoing clinical laboratory and endocrinologist collaboration, the AVS-generated results are largely reproducible both in unique A/C ratios as well as LIs. Repeat AVS was feasible and usually successful when performed for failed initial catheterization; blood sample mislabeling was not discovered and apparently discordant results were fully reproducible, suggesting an alternative explanation for the finding, possibly cortisol cosecretion. Repeat AVS should be considered for patients in whom initial AVS is unsuccessful or yields unexpected results.

Glossary

Abbreviations

A/C

aldosterone/cortisol

ACTH

adrenocorticotropin

ARR

aldosterone-renin ratio

AVS

adrenal vein sampling

IVC

inferior vena cava

LI

lateralization index

NP-59

¹³¹I-iodocholesterol

PA

primary aldosteronism

SI

selectivity index

Additional Information

Disclosures: The authors have nothing to disclose.

Data Availability

Restrictions apply to the availability of some or all data generated or analyzed during this study to preserve patient confidentiality. The corresponding author will on request detail the restrictions and any conditions under which access to some data may be provided.