Exploring the diagnostic and prognostic significance of circulating tumor cells in stage II-IV colorectal cancer using a nano-based detection method

Abstract

Background:

Colorectal cancer (CRC) is a leading cause of cancer mortality globally, underscoring the urgency for a noninvasive and effective biomarker to enhance patient prognosis. Circulating tumor cells (CTCs), a potential marker for real-time tumor monitoring, are limited in clinical utility due to the low sensitivity of existing detection methods. Previously, we introduced a novel nano-based CTCs detection method that relies on the electrical properties of cell surfaces, thus eliminating the need for specific molecular biomarkers. In this study, we used this technique to evaluate the diagnostic and prognostic value of CTCs in stage II-IV CRC.

Methods:

A total of 194 participants were included, consisting of 136 CRC patients and 58 healthy individuals. The peripheral blood of the participants was collected, and CTC enumeration was performed utilizing the nano-based detection method that we newly developed. The receiver operating characteristic (ROC) curve and multivariate Cox proportional-hazards analysis were used to assess the effectiveness of CTCs for diagnosing CRC and predicting patient prognosis.

Results:

The nano-based method demonstrated an ability to differentiate CRC patients from healthy individuals with a sensitivity of 84.6% and a specificity of 94.8%. Furthermore, baseline CTC levels were predictive of progression-free survival (PFS) in CRC patients, with lower levels associated with longer PFS compared to higher levels (4.5 vs 8.0 months at 15 CTCs/mL, p = 0.016; 4.4 vs 8.0 months at 20 CTCs/mL, p = 0.028). We also explored the dynamic changes in the number of CTCs after 1 to 5 cycles of chemotherapy. Patients with increasing CTC levels typically experienced disease progression (PD), while those with decreasing levels often achieved a partial response (PR) or maintained stable disease (SD). These findings suggest that the dynamic fluctuations in CTC counts are closely tied to the clinical course of the disease.

Conclusion:

Our study indicates the potential of nano-based CTCs detection in diagnosing and predicting outcomes for patients with stage II-IV CRC.

1. INTRODUCTION

Colorectal cancer (CRC) is the third most prevalent cancer worldwide and the second leading cause of cancer-related deaths.¹ Mortality associated with CRC is primarily attributed to distant metastases. While the 5-year survival rate for patients with localized CRC can be as high as 90%, this figure plummets to 14% for those diagnosed at the metastatic stage.² Consequently, early detection and diagnosis are critical to improving patient outcomes.

To date, tissue biopsy remains the gold standard for determining CRC status, but its invasiveness and discomforting properties make it unsuitable for longitudinal surveillance. Moreover, due to the spatial and temporal heterogeneity of tumors, a single biopsy fails to provide comprehensive information about malignancy and dynamic changes within the tumor.³ Therefore, additional tumor biomarkers are required to accurately represent the tumor information. Carcinoembryonic antigen (CEA) is a recommended biomarker for diagnosing tumors and monitoring treatment response in CRC according to current treatment guidelines. However, CEA and other serological markers such as CA199 and CA72-4 have limited sensitivity and specificity,⁴ underscoring the need for a more effective, noninvasive liquid biomarker.

Circulating tumor cells (CTCs), disseminated from the primary or metastatic tumors into the bloodstream, offer a promising noninvasive biomarker for real-time assessment of cancer burden and progression.⁵ Recognized as a critical element in the metastatic cascade of CRC and other solid tumors,⁶ CTCs have been extensively studied. These investigations consistently indicate that high levels of CTCs correlate with a poor prognosis and increased risk of recurrence.^7–9 Additionally, dynamic changes in CTC counts have proven to be predictive of chemotherapy efficacy in CRC patients,¹⁰ highlighting the significant diagnostic and prognostic value of CTCs detection and characterization.

Due to the extreme rarity of CTCs in the peripheral blood of cancer patients, efficient enrichment of CTCs is essential for their analysis and clinical application. The CellSearch system, which targets the single cell surface marker epithelial cell adhesion molecule (EpCAM), is the only FDA-approved CTCs detection method.¹¹ The feasibility of CellSearch has been widely validated in multiple studies, establishing it as the gold standard for CTCs detection in various carcinomas.¹² However, its sensitivity and specificity are compromised due to the epithelial-to-mesenchymal transition (EMT) observed in CTCs, rendering it incapable of recognizing mesenchymal CTCs and those lacking the EpCAM phenotype.¹³ To address this issue, we have recently developed a new nano-based technique for the detection of CTCs. This method exploits the unique negative surface charge of cancer cells, which arises from increased lactate secretion due to their enhanced glycolysis activity, compared to other peripheral blood cells. By leveraging this characteristic, our nano-based method can detect and capture cancer cells without relying on specific surface markers.^14,15 Despite the capabilities of this innovative technique, the potential of using CTCs as a diagnostic tool for CRC with this method remains underexplored. Therefore, in this study, we used our newly developed nano-based CTCs detection method to assess the diagnostic and prognostic potential of CTCs as a noninvasive biomarker for patients with stage II-IV CRC.

2. METHODS

2.1. Patient collection

Between December 2020 and December 2022, a cohort of 136 patients diagnosed with CRC was enrolled at the Department of Throatic Surgery, East Hospital affiliated with Tongji University. Diagnosis was confirmed for all patients through histopathology. Peripheral blood samples were systematically collected before treatment initiation. Additionally, for 34 patients, blood samples were collected both before and after 1 to 5 circles of chemotherapy. All blood samples were carefully preserved at room temperature before undergoing CTCs enrichment and identification. The study was thoroughly reviewed and approved by the Ethics Committee of East Hospital, Tongji University, and conducted in accordance with the Declaration of Helsinki. To ensure ethical standards, all patients provided signed informed consents before participating in the study.

2.2. Cell culture

MDA-MB-231, A549, H460, and HeLa cells purchased from ATCC were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin at 37°C under 5% CO₂. Cells in the exponential growth phase were detached using trypsin-EDTA and collected for subsequent spiking experiments.

2.3. Cell spiking experiments

To validate the efficacy of the nano-based detection method, an equivalent number of tumor cells were introduced into PBS or whole blood obtained from the healthy individuals. Subsequently, the isolated tumor cells were enumerated in accordance with the “Isolation and Characterization of CTCs” protocol outlined below. All experiments were conducted in triplicate to ensure robustness and reliability.

2.4. Isolation and characterization of CTCs

For each sample, peripheral blood was processed using the nano-based CTCs detection method within 12 hours. Briefly, hematopoietic cells and platelets were removed by density centrifugation, and the mononuclear cell layers containing CTCs were collected and rinsed in PBS. Subsequently, 30 μL of activated nanoparticles were added into the isolated cells, followed by incubation at 4°C for 10 minutes to facilitate CTCs binding to the nanoparticles. The nanoparticles-captured CTCs were isolated under a magnetic field for 10 minutes and then rinsed with PBS. Aliquots of enriched CTCs in PBS were placed on slides and fixed with 4% paraformaldehyde (PFA) for 10 minutes at room temperature. The slides were then stained with Diff-Quick Stain, air-dried, and cover-slipped. CTCs were reviewed and identified by cytopathologists under a bright-field microscope based on the following criteria: (i) cell long-end diameter >15 μm; (ii) visible presence of cytoplasm, and nuclear-cytoplasmic ratio >0.8; (iii) nuclear atypia or irregular nuclear morphology; (iv) thickened, wrinkled, and jagged nuclear membrane; (v) hyperchromatic nuclei and nonhomogeneous staining. Typically, each slide was independently examined and reviewed by two cytopathologists to ensure the accuracy of the results.

2.5. Statistical analysis

All statistical analyses were performed using R version 4.1.3. The t test was used for comparing continuous variables, while the chi-square test was utilized for analyzing categorical data. A receiver operator characteristic (ROC) curve was plotted to determine the CTCs cutoff value based on the Youden index. Progression-free survival (PFS) was defined as the time from the start of chemotherapy to the date when progressive disease was confirmed at a follow-up visit. Kaplan-Meier survival plots were used to estimate the median PFS along with 95% CI. These plots were generated based on the CTCs count, and comparisons were conducted using the log-rank test. The multivariate Cox proportional-hazards analysis was applied to estimate the hazard ratio (HR). For all analyses, a significance level of p < 0.05 was considered statistically significant.

3. RESULTS

3.1. Patient characteristics

As shown in Fig. 1, a total of 136 CRC patients and 58 healthy individuals were enrolled for diagnosis analysis. Initial blood samples were taken from all CRC patients and healthy individuals. Subsequently, second blood samples were collected from 36 CRC patients for treatment effectiveness evaluation, excluding two patients due to the absence of radiological imaging.

Fig. 1.

Flowchart of patient recruitment and sample collection.

The median age of the CRC patients, comprising 53 females and 83 males, was 61 years, ranging from 36 to 86 (Table 1). At baseline, 46.3% (63/136) patients exhibited CEA levels above 5 ng/mL and 33.8% (46/136) presented with carbohydrate antigen 19-9 (CA199) levels of 37 U/mL or higher. Of all patients, 80.1% (109/136) were diagnosed with adenocarcinoma, with 77 having colon cancer and 59 having rectal cancer. According to tumor node metastasis (TNM) staging, 23 patients (16.9%) were at stage II, 40 patients (29.4%) at stage III, and 73 patients (53.7%) at stage IV. Among the 81 patients with a specified differentiation grade, 71 were diagnosed with a median grade of differentiation. Most metastatic patients were diagnosed with liver or lung metastases.

Table 1.

Clinical characteristics of CRC patients

Characteristics	Subgroups	Values
Age	Median	61
	Range	36-86
Sex	Male	83 (61.0%)
	Female	53 (39.0%)
CEA	≤5 ng/mL	73 (53.7%)
	>5 ng/mL	63 (46.3%)
CA199	≤37 U/mL	90 (66.2%)
	>37 U/mL	46 (33.8%)
Histology subtypes	Adenocarcinoma	109 (80.1%)
	Unknown	27 (19.9%)
TNM stage	II	23 (16.9%)
	III	40 (29.4%)
	IV	73 (53.7%)
Differentiation grade	Poor	9 (6.6%)
	Medium	71 (52.2%)
	High	1 (0.7%)
	Unknown	55 (40.4%)
Metastasis	No	63 (46.3%)
	Liver	25 (18.4%)
	Lung	8 (5.9%)
	Bone	4 (2.9%)
	Peritoneum	1 (0.7%)
	Abdominal cavity	4 (2.9%)
	Liver/lung/bone/peritoneum/brain	15 (11.0%)
	Unknown	16 (11.8%)
Type of cancer	Colon	77 (56.6%)
	Rectum	59 (43.4%)
CTCs	Positive	116 (85.3%)
	Negative	20 (14.7%)

CRC = colorectal cancer; CTC = circulating tumor cell; TNM = tumor node metastasis.

3.2. Performance of nano-based method using tumor cell lines as models

To assess the performance of nano-based CTCs detection method, positive control experiments were performed in parallel using tumor cells from A549, HeLa, H460, and MDA-MB-231 cell lines spiked into peripheral blood from healthy volunteers and PBS, respectively. Blood samples were processed through the nano-based method, and the enriched cells were subsequently analyzed and enumerated using bright-field images. The recovery rates for the cell lines spiked in PBS at a concentration of 1000 cells/mL were 92.6% for A549, 92.1% for HeLa, 90.3% for H460, and 78.2% for MDA-MB-231 (Fig. 2A). Recovery rates in healthy blood were 56.6% for A549, 55.2% for HeLa, 53.1% for H460, and 51.8% for MDA-MB-231. A significant advantage of the nano-based method is its ability to preserve cell morphology, as demonstrated by Diff-Quick staining. As shown in Fig. 2B, enriched CTCs (CTCs-1 and CTCs-2) were larger, had higher nuclear-cytoplasmic ratios, and stained more intensely compared to normal leukocytes. Overall, these results demonstrate the efficiency and accuracy of the nano-based method in isolating and identifying CTCs.

Fig. 2.

Validation of the nano-based CTCs detection method in tumor cells model. A, Capture efficiency of spiked A549, Hela, H460, and MDA-MB-231 cells in whole blood obtained from healthy donors with equal number of cells. B, Representative images of CTCs (CTCs-1 and CTCs-2) captured by the nano-based CTCs detection method and normal white blood cells (granulocytes and monocytes). CTCs-1 and CTCs-2 were isolated from the peripheral blood of a colon cancer and a rectal cancer patient, respectively. Cell morphology was visualized through Diff-Quick staining. The red arrow indicates cells, and the black arrow indicates nanoparticles. All images were captured under an oil microscope at ×100 magnification. CTC = circulating tumor cell.

3.3. Diagnostic performance of nano-based method in CRC patients

Subsequently, we analyzed the baseline CTC counts in both healthy individuals and CRC patients. Specifically, the presence of CTCs was observed in 116 (85.3%) of all 136 CRC patients, with the number of CTCs per mL of blood ranging from 0 to 102. In contrast, only 4 of 58 healthy individuals (6.9%) had detectable CTCs, with counts ranging from 0 to 20. As shown in Fig. 3A, the mean CTCs count in CRC patients was significantly higher compared to healthy individuals (p < 0.001), with mean counts of 0 and 23, respectively. Furthermore, the ability of CTCs detected by the nano-CTCs assay to discriminate between CRC patients and healthy individuals was evaluated using ROC curves (Fig. 3B). For patients with early-stage disease (stage II-III), the ROC curve analysis yielded an AUC of 0.888, a sensitivity of 81%, and an optimal threshold of 1.5 CTCs/mL. For advanced-stage (stage IV) CRC, the AUC was 0.933, with a sensitivity of 87.7% and the same optimal threshold. When combining CRC patients across stages II-IV, the optimal cutoff value was established at 1.5 CTCs/mL of blood, resulting in an ROC curve with an AUC of 0.912, and a sensitivity and specificity of 84.6% and 94.8%, respectively. Taken together, these data indicate that the baseline CTCs captured by the nano-based CTCs detection method effectively differentiate stage II-IV CRC patients from healthy individuals.

Fig. 3.

Capability of the nano-based CTCs detection method to distinguish CRC patients from healthy individuals. A, Comparative analysis of CTCs count between CRC patients and healthy individuals. B, ROC curve evaluating CTCs as a biomarker for distinguishing healthy individuals from CRC patients at stage II-III, stage IV, and stage II-IV, respectively. CTC = circulating tumor cell; ROC = receiver operating characteristic.

3.4. Relevance between baseline CTCs and clinical characteristics

Additionally, t test analyses were conducted to explore the correlation between the baseline CTCs count and various clinical characteristics of CRC patients. As illustrated in Fig. 4A, the CTCs level in CRC patients exhibited a positive correlation with pathological stage, showing a significant difference between stage II and IV patients (15 vs 26 CTCs/mL, p = 0.017). Data in Table 2 indicate that stage IV patients had significantly higher CTC levels compared to those at stages II and III combined (18 vs 26 CTCs/mL, p = 0.027). Additionally, as shown in Fig. 4C, patients with higher CEA levels exhibited a slight tendency to have higher CTC counts (21 ± 19 vs 25 ± 22 CTCs/mL). However, no significant differences were observed in CTC numbers based on age, sex, cancer type, or CA199 levels (Fig. 4B, D and Supplementary Figure S1, http://links.lww.com/JCMA/A274). These findings suggest that the enumeration of CTCs captured by the nano-based CTCs detection method possesses the capability to discriminate between patients with early or advanced CRC.

Fig. 4.

Comparative analysis of CTCs count using nano-based method across various clinical subgroups of CRC. A, Boxplot illustrating CTCs count in CRC patients categorized by clinical pathological staging. B, Boxplot displaying CTCs count in patients with colon cancer and rectal cancer. C, Boxplot displaying CTCs count in CRC patients within different CEA subgroups. D, Boxplot displaying CTCs count in CRC patients across CA199 subgroups. CEA = carcinoembryonic antigen; CRC = colorectal cancer; CTC = circulating tumor cell.

Table 2.

The association between CTCs level and clinical characteristics

Features	CTCs detection rate	χ 2	p	CTCs count	t	p
Gender
Female	84.9% (45/53)	0.000	1.000	21 ± 18	0.655	0.514
Male	85.5% (71/83)			23 ± 22
Age
≤60	85.9% (55/64)	0.033	0.856	24 ± 22	0.782	0.436
>60	84.7% (61/72)			21 ± 18
CEA
≤5 ng/mL	83.6% (61/73)	0.342	0.559	21 ± 19	1.064	0.289
>5 ng/mL	87.3% (55/63)			25 ± 22
CA199
≤37 U/mL	85.6% (77/90)	0.000	1.000	23 ± 21	0.009	0.993
>37 U/mL	84.8% (39/46)			23 ± 20
Cancer
Colon	85.7% (66/77)	0.035	0.852	24 ± 22	1.002	0.318
Rectum	84.8% (50/59)			21 ± 18
Stage
II-III	81.0% (51/63)	0.711	0.399	18 ± 17	2.238	0.027*
IV	89.0% (65/73)			26 ± 22

CEA = carcinoembryonic antigen; CTC = circulating tumor cell.

^*p <0.05.

3.5. Prognostic value of baseline CTCs on PFS

A total of 38 CRC patients were successfully followed up to monitor survival according to RECIST criteria. A range of CTC cutoff values from 5 to 30 per mL of blood was analyzed to identify the most predictive threshold for rapid disease progression. The log-rank test revealed significant differences in PFS at CTC thresholds of 15 (p = 0.016) and 20 (p = 0.028) (Fig. 5). The median PFS for patients with CTCs <15 was 8 months (95% CI, 7.8-NA), while for those with CTCs ≥15, the median PFS decreased to 4.5 months (95% CI, 3.5-7.0). At a threshold of 20 CTCs, patients with CTCs ≥20 had a median PFS of 4.4 months (95% CI, 3.2-8.2), compared to 8 months for those with <20 CTCs (95% CI, 7.0-16.2). Other thresholds analyzed (5, 10, 25, 30 CTCs) showed reduced PFS for patients exceeding these levels, although no statistically significant differences were found (Supplementary Figure S2, http://links.lww.com/JCMA/A274). In addition, CTCs emerged as a reliable predictor of PFS in multivariate Cox proportional analysis, with an HR of 1.022 (95% CI, 1.001-1.04; p = 0.038) (Supplementary Figure S3, http://links.lww.com/JCMA/A274). Collectively, these results highlight the clinical significance of CTC enumeration in the prognosis and therapeutic management of CRC patients.

Fig. 5.

Kaplan-Meier curves showing PFS of CRC patients, stratified by distinct thresholds of baseline CTCs. A, CTCs ≥15 and CTCs <15. B, CTCs ≥20 and CTCs <20. CTC = circulating tumor cell; PFS = progression-free survival.

3.6. Interplay between short-term CTC dynamics and therapeutic response

To evaluate the short-term dynamics of CTCs, blood samples were collected from 34 patients undergoing chemotherapy at baseline and after 1 to 5 circles of treatment. Posttreatment evaluation categorized patients into stable disease (SD), partial response (PR), and progressive disease (PD) subgroups according to RECIST criteria. As shown in Fig. 6A, CTC levels significantly decreased in all three PR patients post-chemotherapy. Among the four PD patients, three exhibited increased CTC levels, while one showed a decrease after treatment. In the 27 SD patients, CTC levels decreased or remained stable in 22 patients, with an increase observed in 5 patients after short-term chemotherapy. A statistically significant difference was observed in the change in CTC count before and after treatment within the combined PR/SD group (p < 0.001) (Fig. 6B). Conversely, no significant change in CTC levels was noted in PD patients (p = 0.106). In summary, the reduction in the number of CTCs can serve as a predictor of therapeutic efficacy.

Fig. 6.

Association between CTC levels and chemotherapy response in CRC patients. A, The changes in CTCs count after 1 to 5 circles of chemotherapy among patients with SD, PR, and PD at baseline. B, The changes in CTCs count in PR/SD and PD patients before or after 1 to 5 circles chemotherapy. CRC = colorectal cancer; CTC = circulating tumor cell; PD = progressive disease; PR = partial response; SD = stable disease.

4. DISCUSSION

CTCs are considered highly promising as a noninvasive biomarker for cancer diagnosis, therapeutic monitoring, and prognosis prediction.¹⁶ However, current CTC detection technologies, characterized by relatively low sensitivity, are inadequate for capturing rare CTCs in peripheral blood.¹⁷ CTCs undergoing EMT exhibit dynamic phenotypic changes, including the downregulation of epithelial markers like EpCAM and the upregulation of mesenchymal markers such as Vimentin. Consequently, a subset of CTCs becomes negative for epithelial markers.¹⁸ The FDA-approved CellSearch system, which relies on EpCAM for CTCs capture, fails to detect CTCs with minimal to no EpCAM expression.¹⁹ In contrast, the nano-based CTCs detection method, utilizing cell surface electrical properties for CTCs isolation, has been shown to be effective in capturing CTCs in peripheral blood. This method is also noted for its technical simplicity and affordability.²⁰ In this study, we have demonstrated the clinical feasibility of using CTCs as a liquid biomarker for the diagnosis and prognosis of CRC patients through the application of this novel nano-based CTCs detection method.

We established an optimal cutoff of 1.5 CTCs per mL of peripheral blood, effectively distinguishing CRC patients from healthy individuals. The nano-based approach demonstrated a sensitivity of 84.6% and a specificity of 94.8% for CRC detection. In contrast, previous studies using the CellSearch system reported CTCs detection rates of only 39% to 42% in metastatic CRC (mCRC) patients.^21,22 Notably, the nano-based approach identified CTCs in 89.09% of late-stage IV CRC patients (65/73), highlighting its superior detection capability in mCRC patients. Additionally, among stage II-III CRC patients (n = 63), a relatively high CTCs detection rate of 80.95% was observed, indicating a high sensitivity in early-stage CRC. This contrasts with a previous study reporting positive rates of only 20.7% and 24.1% for stage II and III patients, respectively.²³ Notably, the PePiTA clinical trial, investigating the prognostic role of CTCs in stage II and III colon patients, was discontinued due to a low CTCs detection rate of 6%.²⁴ Therefore, CTCs detected through the nano-based method serve as a valuable biomarker for CRC patients.

Additionally, multiple studies have consistently demonstrated an association between CTCs and prognosis.²⁵ A previous study involving 430 mCRC patients revealed that CTC counts ≥3 per 7.5 mL of blood served as a poor prognostic indicator for PFS.²⁶ Similarly, another study of 467 advanced CRC patients also reported a prognostic role for CTCs.²⁷ In our study, we validated the prognostic value of CTCs for PFS in CRC patients. Our data demonstrated that patients with lower CTC levels exhibited significantly longer PFS compared to those with higher CTC levels, with cutoff value of 15 CTCs (4.5 vs 8.0 months, p = 0.016) and 20 CTCs (4.4 vs 8.0 months, p = 0.028) per mL of peripheral blood. Moreover, we enrolled 34 patients exhibiting the highest objective response rate (ORR) of 8.82% (PR) and a disease control rate of 88.24% (PR/SD) to assess the potential of CTCs detection for monitoring short-term chemotherapy efficacy. Among the PR patients, 100% (3/3) of PR patients showed a decline in CTCs count. Among PD patients, three of four (75%) exhibited an elevated CTCs count, while one showed a decrease. Conversely, among SD patients, 22 of 27 (81.5%) demonstrated a decline in CTC levels, and 5 showed an increase. Generally, patients identified with an increased CTCs level tended to be PD patients, while those with a decreased level tended to be PR/SD patients. This suggests that monitoring dynamic changes in CTC levels may assist oncologists in predicting clinical responses to chemotherapy. However, because our study was limited to evaluating the short-term predictive value of CTCs in CRC patients, further studies with longer follow-ups and larger cohorts are necessary to explore the potential of nano-based CTCs detection for real-time therapeutic monitoring. Additionally, further investigation is required to elucidate the prognostic value of CTCs for overall survival.

In conclusion, our study marks the first demonstration of the potential diagnostic, prognostic, and predictive value of CTCs in patients with stage II-IV CRC using nano-based CTCs detection technology. The results showed that CTCs are effective in differentiating between CRC patients and healthy individuals with a sensitivity of 84.6% and a specificity of 94.8%. Furthermore, baseline CTCs emerged as a promising independent prognostic biomarker for PFS, while short-term changes in CTCs showed potential in predicting drug efficacy in CRC patients undergoing chemotherapy. Larger and more comprehensive studies are required to further validate the clinical significance of CTCs with our nano-based detection method.

ACKNOWLEDGMENTS

This research was funded by the Shanghai Health and Family Planning Commission Project (202240009), Natural Science Foundation of Fujian Province (2019J01596), Shanghai Pujiang Program (2020PJD052), and Jinjiang High-Level Talent Project (2017014).

APPENDIX A. SUPPLEMENTARY DATA

Supplementary data related to this article can be found at http://links.lww.com/JCMA/A274.