Ps and Qs: Quantization-Aware Pruning for Efficient Low Latency Neural Network Inference.

Efficient machine learning implementations optimized for inference in hardware have wide-ranging benefits, depending on the application, from lower inference latency to higher data throughput and reduced energy consumption. Two popular techniques for reducing computation in neural networks are pruning, removing insignificant synapses, and quantization, reducing the precision of the calculations. In this work, we explore the interplay between pruning and quantization during the training of neural networks for ultra low latency applications targeting high energy physics use cases. Techniques developed for this study have potential applications across many other domains. We study various configurations of pruning during quantization-aware training, which we term quantization-aware pruning, and the effect of techniques like regularization, batch normalization, and different pruning schemes on performance, computational complexity, and information content metrics. We find that quantization-aware pruning yields more computationally efficient models than either pruning or quantization alone for our task. Further, quantization-aware pruning typically performs similar to or better in terms of computational efficiency compared to other neural architecture search techniques like Bayesian optimization. Surprisingly, while networks with different training configurations can have similar performance for the benchmark application, the information content in the network can vary significantly, affecting its generalizability.

Effects of micronised microencapsulated ferric pyrophosphate supplementation in patients with advanced cancer and iron deficiency: a single-centre cohort pilot study.

BACKGROUND: Iron deficiency is the most common nutritional deficiency in advanced cancer patients and causes anaemia. Iron deficiency anaemia treatment (i.e. intravenous or oral iron administration) has been demonstrated to be effective but is often associated with adverse reactions. Micronised microencapsulated ferric pyrophosphate (MMFP) is a recently developed formulation characterised by a higher intestinal bioavailability due to the small particle size distribution at nanometer level. The aim of this study was to evaluate the efficacy of an oral administration of 30 mg of MMFP associated with 80 mg of ascorbic acid in advanced cancer patients with hyposideraemia. MATERIALS AND METHODS: This was an observational prospective cohort study (10 months) conducted on 42 adult patients with advanced cancer and serum iron levels lower than 60 µg/dL. All patients received one capsule/day for 30 days of a supplement containing 30 mg of MMFP and 80 mg of ascorbic acid. At enrolment (T0) and at 30 days (T1) patients were subjected to blood sampling for evaluation of serum iron, ferritinaemia and blood count. In addition, any undesirable effects reported by patients were evaluated. RESULTS: MMFP treatment increased sideraemia from 36.1±8.37 µg/dL to 73.22±28.60 µg/dL, haemoglobin from 10.43±1.09 g/dL to 11.52±1.90 g/dL, and ferritinaemia from 42.10±16.90 ng/mL to 123.33±55.79 ng/mL. No adverse effects were noted from the use of MMFP supplementation. DISCUSSION: The supplementation of 30 mg/d of MMFP in combination with 80 mg/d of ascorbic acid in advanced cancer patients with hyposideraemia led to a significant increase in sideraemia and ferritinaemia. Moreover, in some of the patients whose serum iron level did not increase, an increase in haemoglobin was observed.

Altered expression of uncoupling protein 2 in GLP-1-producing cells after chronic high glucose exposure: implications for the pathogenesis of diabetes mellitus.

Glucagon-like peptide-1 (GLP-1) is a gut L-cell hormone that enhances glucose-stimulated insulin secretion. Several approaches that prevent GLP-1 degradation or activate the GLP-1 receptor are being used to treat type 2 diabetes mellitus (T2DM) patients. In T2DM, GLP-1 secretion has been suggested to be impaired, and this defect appears to be a consequence rather than a cause of impaired glucose homeostasis. However, although defective GLP-1 secretion has been correlated with insulin resistance, little is known about the direct effects of chronic high glucose concentrations, which are typical in diabetes patients, on GLP-1-secreting cell function. In the present study, we demonstrate that glucotoxicity directly affects GLP-1 secretion in GLUTag cells chronically exposed to high glucose. Our results indicate that this abnormality is associated with a decrease in ATP production due to the elevated expression of mitochondrial uncoupling protein 2 (UCP2). Furthermore, UCP2 inhibition using small interfering RNA (siRNA) and the application of glibenclamide, an ATP-sensitive potassium (KATP(+)) channel blocker, reverse the GLP-1 secretion defect induced by chronic high-glucose treatment. These results show that glucotoxicity diminishes the secretory responsiveness of GLP-1-secreting cells to acute glucose stimulation. We conclude that the loss of the incretin effect, as observed in T2DM patients, could at least partially depend on hyperglycemia, which is typical in diabetes patients. Such an understanding may not only provide new insight into diabetes complications but also ultimately contribute to the identification of novel molecular targets within intestinal L-cells for controlling and improving endogenous GLP-1 secretion.

Randomized evaluation of intralesion versus intracoronary abciximab and aspiration thrombectomy in patients with ST-elevation myocardial infarction: The COCTAIL II trial.

BACKGROUND: Thrombus burden and distal embolization are predictive of no-reflow during primary percutaneous coronary intervention (PCI) in patients with acute ST-elevation myocardial infarction (STEMI). We sought to compare the efficacy of pharmacological and catheter-based strategies for thrombus in patients with STEMI and high atherothrombotic burden. METHODS: Between January 2012 and December 2013, 128 STEMI patients undergoing primary PCI at 5 centers were randomly assigned in a 2 × 2 factorial design to intracoronary (IC) abciximab bolus (via the guide catheter) versus intralesion (IL) abciximab bolus, each with versus without aspiration thrombectomy (AT). Study end points were residual intrastent atherothrombotic burden, defined as the number of cross-sections with residual tissue area >10% as assessed by optical coherence tomography, and indices of angiographic and myocardial reperfusion. RESULTS: Residual intrastent atherothrombotic burden did not significantly differ with IL versus IC abciximab (median [interquartile range] 6.0 [1-15] vs 6.0 [2-11], P = .806) and with AT versus no aspiration (6.0 [1-13] vs 6.0 [2-12], P = .775). Intralesion abciximab administration was associated with improved angiographic myocardial reperfusion in terms of thrombolysis in myocardial infarction (TIMI) flow (3 [3-3] vs 3 [2-3], P = .040), corrected TIMI frame count (12 ± 5 vs 17 ± 16, P = .021), and myocardial blush grade (3 [2-3] vs 3 [2-3], P = .035). In particular, IL abciximab was associated with higher occurrence of final TIMI 3 flow (90% vs 73.8%, P = .032) and myocardial blush grade 3 (71.6% vs 52.4%, P = .039). Conversely, AT had no significant effect on indices of angiographic or myocardial reperfusion. CONCLUSIONS: In patients with STEMI and high thrombotic burden, neither IL versus IC abciximab nor AT versus no aspiration reduced postprocedure intrastent atherothrombotic burden in patients with STEMI undergoing primary PCI. However, IL abciximab improved indices of angiographic and myocardial reperfusion compared to IC abciximab, benefits not apparent with AT.

Association between proximal stent edge positioning on atherosclerotic plaques containing lipid pools and postprocedural myocardial infarction (from the CLI-POOL Study).

Postprocedural myocardial infarction is an ominous complication of percutaneous coronary intervention (PCI). Despite several patient, lesion, and procedural factors that may affect its occurrence and severity, it is unclear if implanting a stent edge on a coronary lipid pool, as appraised by optical coherence tomography (OCT), adversely affects outcomes. The aim of this study was to assess the association between postprocedural myocardial infarction and the implantation of a stent edge on a lipid pool, as assessed by OCT. A database was screened for patients without ongoing myocardial infarctions; who underwent PCI with stenting for single, native, de novo lesions; without periprocedural side-branch occlusion or compromise; who underwent post-PCI OCT; and had postprocedural myocardial infarctions. These subjects were matched 1:1 with patients with similar features but without postprocedural myocardial infarctions. Plaque characterization with OCT was performed using established criteria. Specifically, lipid pools within stent edges were quantified by computing the number of involved quadrants and the degree of lipid arc on cross-sectional images. A total of 30 patients were included (15 with postprocedural myocardial infarctions and 15 controls without infarctions). Whereas no patient or control subject had lipid pools in correspondence to distal stent edges, landing of proximal stent edges on lipid pools was significantly more frequent in patients than in controls (10 [66%] vs 2 [13%], p = 0.009), Moreover, patients with postprocedural myocardial infarctions had more extensive lipid pools at proximal stent edges than those without postprocedural myocardial infarctions. Accordingly, lipid pool arc at proximal stent edge was significantly associated with peak post-PCI creatine kinase-MB/upper limit of normal ratio (Spearman's ρ = 0.49, p = 0.006). In conclusion, incomplete stent coverage of coronary lipid pools appears to be associated with an increased risk for postprocedural myocardial infarction in patients who undergo PCI.

Percutaneous coronary intervention of unprotected left main coronary artery disease as culprit lesion in patients with acute myocardial infarction.

OBJECTIVES: This study sought to evaluate short- and long-term outcomes of patients undergoing emergency percutaneous coronary intervention (PCI) for acute myocardial infarction due to a culprit lesion in an unprotected left main coronary artery. METHODS: In this retrospective, 2-center, international observational study, 5,261 patients were admitted between February 2005 and December 2008 with acute myocardial infarction and treated with PCI; of these, 1,277 were ST-segment elevation myocardial infarction and 3,984 non-ST-segment elevation myocardial infarction. We identified 48 patients among this cohort who underwent emergency PCI to an unprotected left main coronary artery culprit lesion. RESULTS: Mean age was 70 ± 12.5 years, and 45% of the patients presented with ST-segment elevation myocardial infarction or new left bundle branch block. Cardiogenic shock was present in 45%, and distal left main coronary artery disease was present in 71% of patients. Angiographic procedural success was achieved in 92% of patients. Overall in-hospital mortality was 21%, due in all cases to refractory, multiorgan failure. Twenty-five percent experienced major adverse cardiac events, defined as death, myocardial infarction, stent thrombosis, and target vessel revascularization. In patients presenting in cardiogenic shock, in-hospital mortality was 32%. At 1-year follow-up, in-hospital survivors had a mortality rate of 10.5%, whereas 18.4% experienced subsequent major adverse cardiac events. Long-term prognosis was excellent in hospital survivors with a 1-year survival rate of 89.5%. CONCLUSIONS: Patients with acute myocardial infarction and thrombosis of the unprotected left main coronary artery are a high-risk subgroup with a substantial mortality, particularly if they present in cardiogenic shock. We demonstrate that in these patients, PCI is a feasible treatment option associated with reasonably good outcomes. Long-term prognosis is excellent in hospital survivors with an 89.5% survival rate at 1 year.

Safety and feasibility of frequency domain optical coherence tomography to guide decision making in percutaneous coronary intervention.

AIMS: The purpose of this single centre registry is to assess safety and feasibility of the frequency domain optical coherence tomography (FD-OCT) system during coronary interventions. METHODS AND RESULTS: Ninety patients with unstable or stable coronary artery disease were included in this study. OCT imaging was performed in a first group of 40 patients (group 1), to evaluate ambiguous/intermediate lesions (24 patients in group 1 had OCT also done post-PCI, for assessment of stent deployment); and in a second group of 50 patients (group 2), to address the adequacy of stent deployment. Therefore, 74 patients underwent FD-OCT after stent implantation. A complex-lesion population was studied (B2 type lesion=72.2% and C type lesion=20.3%). The mean time of a FD-OCT pull-back (from the set up to the completion of the pull back) was 2.1 min and in all but one (99.1%) the procedure was successful. No patients experienced major complications in terms of death, myocardial infarction, emergency revascularisation, embolisation, life-threatening arrhythmia, coronary dissection, prolonged and severe vessel spasm and contrast induced nephropathy. In the ambiguous lesion group, 60% of patients were treated with PCI, whilst in the others, PCI were deferred. In total, 113 deployed stents (33,6% chromium cobalt stent, 66,4% drug eluting stent) were imaged with OCT. OCT findings led to additional interventions in 24 out of 74 patients (32%): 15 had further balloon inflations, nine had additional stent deployment whilst two had both treatments. At clinical follow-up, (4.6 ± 3.,2 months), there were no death, acute myocardial infarctions and cases of stent thrombosis, whilst two patients underwent revascularisation for recurrence of angina. CONCLUSIONS: The present registry shows that FD-OCT is a feasible and safe technique for guidance of coronary interventions. Randomised studies will confirm whether the use of FD-OCT will improve the clinical outcome.

Randomized trial of intravenous iron supplementation in patients with chemotherapy-related anemia without iron deficiency treated with darbepoetin alpha.

PURPOSE: Unresponsiveness to erythropoiesis-stimulating agents, occurring in 30% to 50% of patients, is a major limitation to the treatment of chemotherapy-related anemia. We have prospectively evaluated whether intravenous iron can increase the proportion of patients with chemotherapy-related anemia who respond to darbepoetin. PATIENTS AND METHODS: Between December 2004 and February 2006, 149 patients with lung, gynecologic, breast, and colorectal cancers and >or= 12 weeks of planned chemotherapy were enrolled from 33 institutions. Patients were required to have hemoglobin or= 12 g/dL or >or= 2 g/dL increase). RESULTS: Hematopoietic response by intention-to-treat analysis was 76.7% (95%CI, 65.4% to 85.8%) in the darbepoetin/iron group and 61.8% (95%CI, 50.0% to 72.7%) in the darbepoetin group (P = .0495). Among patients fulfilling eligibility criteria and having received at least four darbepoetin administrations, hematopoietic responses in the darbepoetin/iron group (n = 53) and in the darbepoetin-only group (n = 50) were 92.5% (95% CI, 81.8% to 97.9%) and 70% (95% CI, 55.4% to 82.1%), respectively (P = .0033). Increase of hemoglobin during treatment period showed a time profile favoring darbepoetin/iron with statistically significant effect from week 5 on. The safety profile was comparable in the two arms. CONCLUSION: In patients with chemotherapy-related anemia and no iron deficiency, IV iron supplementation significantly reduces treatment failures to darbepoetin without additional toxicity.

A phase II study of induction chemotherapy followed by concurrent chemoradiotherapy in elderly patients with locally advanced non-small-cell lung cancer.

The optimal management of unresectable locally advanced non-small-cell lung cancer in older patients has not been defined to date. The present phase II study was planned to evaluate the activity and safety of platinum-based induction chemotherapy followed by concurrent chemoradiotherapy in elderly patients with locally advanced non-small-cell lung cancer. Patients received two cycles of paclitaxel (175 mg/m) and carboplatin (area under the curve: 5) day 1, every 3 weeks. Chemoradiotherapy (thoracic radiation therapy) was initiated on day 42 and consisted of 1.8 Gy daily, five times per week over 5 weeks (45.0 Gy target dose) followed by 10 2.0 Gy daily fractions. Concomitant chemotherapy was weekly paclitaxel 50 mg/mq followed by weekly carboplatin at an area under the curve of 2. The eligibility for patients: age 70 or older and histologically documented untreated non-small-cell lung cancer, locally advanced, unresectable, stage III A N2 bulky or III B. Thirty consecutive patients were enrolled onto the study. The median age was 73 (range 70-76). According to the intention-to-treat analysis, 1 month after the end of combined chemoradiotherapy, we observed complete and partial responses in one and 19 of the 30 patients, respectively, for an overall response rate of 66% (95% confidence interval, 45-76%). Median progression-free survival was 8.7 months (95% confidence interval, 3.4-37.8) and median survival was 15 months (95% confidence interval, 4.2-52.1). During the treatment, 12 patients (40.0%) experienced grade 3-4 neutropenia, two patients neutropenic fever, and three patients grade 3 anaemia and grade 3 thrombocytopenia, respectively. Grade 3 oesophagitis, during concomitant radiotherapy, was observed in six patients (20.0%). No treatment-related mortality was reported. The investigated sequential approach including induction chemotherapy followed by concurrent chemoradiotherapy appears safe and seems a reasonable chance for the treatment of locally advanced non-small-cell lung cancer in the elderly population.

The RESTART project: high-dose atorvastatin helps control RESTenosis AfteR sTenting procedures at high risk for restenosis. Results from a prospective study and a matched control group comparison.

AIMS: The long-term results of bare stenting in complex procedures are troubled by a high incidence of restenosis. The use of drug-eluting stents lead to excellent long-term angiographic results, but require, long-term antithrombotic treatment to prevent stent thrombosis. We evaluated the efficacy of high-dose atorvastatin as an adjuvant in limiting clinical and angiographic restenosis after bare stenting procedures which carry an high risk of restenosis. METHODS AND RESULTS: Between November 2003 and February 2004, 91 consecutive patients (with 152 target lesions) underwent successful interventional procedures with bare stents (158 stents; 1.7 stent/patient). The patients were included in the study in the presence of at least one of the following criteria: diabetes mellitus (14.6%); acute myocardial infarction (8.8%); lesion length >15 mm (27.5%); lesion in vessels with a reference diameter of <2.5 mm (24.2%); >2 treated lesions (58.2%); side-branch lesions (3.3%). All patients were started on atorvastatin (80 mg/day) on the day of the procedure, and continued receiving it for at least nine months.Six of the 91 patients (6.5%) experienced a clinical restenosis and underwent target lesion revascularisation; a further two underwent re-PTCA due to progressive atherosclerosis in untreated sites. Seventy-one patients (78.0%; 125 stents) were followed up with angiography after a mean 8.5 months (7-10 months): binary angiographic restenosis was observed in 18/71 patients (25.3%) (in-stent restenosis 21.1%; in-segment restenosis 4.2%), and 17/125 stents (13.6%).We matched 122 lesions (125 stents) from the present population with a reference group of 600 consecutive lesions treated at the Ospedale Maggiore della Carità in Novara during the same period of time, but without high-dose statin treatment: 88 lesions (94 stents) were successfully matched on the basis of the clinical and angiographic variables of diabetes, acute myocardial infarction, treated vessel, reference diameter and lesion length. The incidence of late loss was significantly lower in the atorvastatin group (p=0.027). CONCLUSIONS: The RESTART prospective study showed that the adjunct of atorvastatin 80 mg leads to excellent clinical and angiographic outcomes. These results were confirmed in a matched lesion comparison.

A phase II study of carboplatin and paclitaxel as first line chemotherapy in elderly patients with advanced non-small cell lung cancer (NSCLC).

INTRODUCTION: Lung cancer is the leading cause of tumour-related deaths in the elderly population but the optimal management of advanced NSCLC in older patients has not been defined to date. The present phase II study was planned to evaluate the efficacy and toxicity of the combination of carboplatin and paclitaxel in elderly patients with advanced NSCLC. PATIENTS AND METHODS: Patients (>70 years old) who had pathologically been proven to have a NSCLC and measurable lesions were treated with paclitaxel (175 mg/m2 for 3h) and carboplatin [area under the concentration-time curve (AUC=5)] on day 1 every 3 weeks. RESULTS: Forty patients were enrolled into the study. The median age was 74 years (range, 70-78 years). Approximately 85% of the patients had stage IV and 80% had a performance status (PS) of 0-1. Nine of the 40 (22.5%; 90% CI 17.6-28.1) included patients had a partial response; one patient (2.5%; 90% CI 1.7-3.2) achieved a complete response. The overall response rate was 25% (90% CI 15.3-38.6). In addition stable disease was observed in 13 patients (32.5%; 90% CI 24.3-40.7). The median survival was 7.8 months (95% confidence interval, 5.1-11.8 months). The actual 1-year survival was 18% (95% confidence interval, 12-29%). The median time to disease progression was 4.1 months (95% CI 2.8-8.5). Overall, 37.5% of patients experienced grade 3-4 neutropenia of any duration with only two patients (5%) developing neutropenic fever. Grade 3 or 4 non-haematological toxicity was uncommon apart alopecia. CONCLUSIONS: In the present phase II study the combination of paclitaxel and carboplatin has demonstrated to be active and safe in an age-selected population.

Paclitaxel, carboplatin and gemcitabine combination as induction chemotherapy for stage IIIA N2 bulky non-small cell lung cancer.

BACKGROUND: Induction chemotherapy followed by surgical resection or definitive radiotherapy for patients affected by stage IIIA N2 bulky non-small cell lung cancer (NSCLC) has been investigated in several trials. PATIENTS AND METHODS: In this present study, 52 patients with stage IIIA N2 bulky NSCLC with cytologically or histologically confirmed mediastinal lymph node involvement received paclitaxel 175 mg/mq on day 1, carboplatin AUC 5 on day 1 and gemcitabine 1,000 mg/mq on day 1 and 8 every 3 weeks for three cycles as induction chemotherapy. RESULTS: Objective response (4 complete remission and 36 partial remission) was achieved in 40/52 patients. No early or toxic deaths were observed. Twenty-two patients were surgically explored. Fifteen were excluded for resection for biopsy-proven residual tumour in mediastinal nodes. Complete surgical resection was performed in 15 patients with confirmed pathological downstaging. Pathological complete response was achieved in 4 patients. No surgery-related mortality or significant morbidity was reported. Adjuvant radiotherapy was delivered in 15 patients, and 30 patients received definitive radiotherapy. CONCLUSION: In the present study, the combination of paclitaxel, carboplatin and gemcitabine has been a safe and active regimen in poor-prognosis stage IIIA N2 bulky NSCLC.

Oral temozolomide in heavily pre-treated brain metastases from non-small cell lung cancer: phase II study.

INTRODUCTION: The primary tumour type most likely to metastasize to the brain is lung cancer. In heavily pre-treated patients, limited therapeutic option is available and the results of availability therapies reported in literature are disappointing. The present phase II study was designed to assess the efficacy and safety of temozolomide (TMZ) as palliative treatment for brain metastases (BrM) in NSCLC patients pre-treated with WBRT and at least one line of chemotherapy for metastatic brain disease. MATERIAL AND METHODS: Temozolomide was administered orally at 150 mg/mq/day for five consecutive days for the first cycle, doses were increased to 200 mg/mq/day for 5 days every 28 days for subsequent cycles if no grade 3/4 haematological toxicity was observed. Eligibility criteria included cytological or histological confirmed NSCLC; BrM, recurrent or progressing after WBRT and at least one line of chemotherapy. A total of 30 consecutive patients entered the study and received the allocated treatment. RESULTS: Three patients (10%) achieved an objective response (OR) of BrM with two complete remission. Stable disease and progressive disease were achieved in 3 (10%) and 24 patients (80%), respectively. A correlation between response to TMZ and sensitivity to the previous first line chemotherapy was reported. Time to progression and overall survival were examined both for responder patients and for all included patients. For long-term survivors, we considered the patients who survived >12 months after the start of TMZ. According to this definition, three patients resulted long-term survivors: 2 with OR and 1 with stable brain disease. No grades 3 or 4 toxicity occurred. The total of treatment-related adverse events were mild or moderate (G1-2) in intensity. No patients discontinued TMZ as a result of treatment-related toxicity. DISCUSSION: The results of the present trial clearly demonstrates that TMZ is active and safe in BrM NSCLC patients previously treated with WBRT and at least one line of chemotherapy.

Stenting of culprit lesions in unstable angina leads to a marked reduction in plaque burden: a major role of plaque embolization? A serial intravascular ultrasound study.

BACKGROUND: Intravascular ultrasound (IVUS) studies have shown that a mechanism of plaque compression/embolization contributes toward the poststenting increase in lumen area. The aim of this IVUS study was to compare the mechanisms of lumen enlargement after coronary stenting in 54 consecutive patients with unstable angina (UA) (group 1) and 56 with stable angina (group 2) to verify whether plaque embolization plays a major role in the former. METHODS AND RESULTS: Both groups underwent the IVUS assessment (speed, 0.5 mm/sec) before the intervention and after stent implantation. The lumen area, the external elastic membrane area, and the plaque+media area (PA) were measured at 0.5-mm intervals. PA reduction in the lesion site was significantly greater in group 1 (-2.50+/-1.97 versus -0.53+/-1.43 mm2, P<0.001). After stenting, 47% of the lumen area increase in group 1 was obtained by means of PA reduction, and 53% was attributable to external elastic membrane area increase; the corresponding figures in group 2 were 13% and 87% (P<0.05). Decrease in PA after stenting was the only significant predictor of the MB fraction of creatinine kinase (CK-MB) release in a multiple regression model (P=0.047). CONCLUSIONS: Serial volumetric IVUS assessment revealed in UA lesions a marked poststenting reduction in plaque volume, which is significantly greater than in stable angina and is associated with postprocedural CK-MB release. The decrease in PA during the procedure predicts CK-MB release in a multiple regression model. These findings suggest that stent deployment is often associated with plaque embolization in patients with UA.

Normal distribution of an intravascular ultrasound index of vessel remodeling.

BACKGROUND: As a consequence of plaque accumulation, coronary arteries may undergo both compensatory enlargement and paradoxical constriction. The aim of this study was to address the distribution of the different remodeling patterns in patients with obstructive coronary atherosclerosis. METHODS: Eighty-seven non-branching segments of native coronary arteries with de novo, focal, non-ostial lesions were imaged at intravascular ultrasound (IVUS). Images were acquired with a motorized pull-back at a speed of 0.5 mm/s. The cross-sectional area (CSA) circumscribed by the external elastic membrane (EEM), the plaque + media complex and the lumen area were measured at its narrowest site (CSA with the minimal lumen area) and in the reference segment (average of proximal and distal reference cross-sections, defined as the most normal looking sites). The IVUS index of vessel remodeling (VRI) was calculated using the following formula: (narrowest site EEM CSA--reference EEM CSA)/reference EEM CSA*100. The index was tested for normality using the Kolmogorov-Smirnov goodness-of-fit test. RESULTS: The frequency distribution of VRI was found to have a normal unimodal distribution (p = 0.60). VRI ranged from -60 to +164, with a mean of 9.3 +/- 28.0 and a median of 3.6. Frequency distribution of VRI slightly skewed towards right (skewness index 1.69). None of the analyzed clinical and morphological variables predicted the presence of compensatory enlargement as opposed to paradoxical constriction. CONCLUSIONS: The frequency distribution of the vascular remodeling of de novo coronary lesions is unimodal. Therefore, compensatory enlargement and paradoxical constriction represent the extremes of a continuous spectrum.