Recurrence-Aware Long-Term Cognitive Network for Explainable Pattern Classification

Machine-learning solutions for pattern classification problems are nowadays widely deployed in society and industry. However, the lack of transparency and accountability of most accurate models often hinders their safe use. Thus, there is a clear need for developing explainable artificial intelligence mechanisms. There exist model-agnostic methods that summarize feature contributions, but their interpretability is limited to predictions made by black-box models. An open challenge is to develop models that have intrinsic interpretability and produce their own explanations, even for classes of models that are traditionally considered black boxes like (recurrent) neural networks. In this article, we propose a long-term cognitive network (LTCN) for interpretable pattern classification of structured data. Our method brings its own mechanism for providing explanations by quantifying the relevance of each feature in the decision process. For supporting the interpretability without affecting the performance, the model incorporates more flexibility through a quasi-nonlinear reasoning rule that allows controlling nonlinearity. Besides, we propose a recurrence-aware decision model that evades the issues posed by the unique fixed point while introducing a deterministic learning algorithm to compute the tunable parameters. The simulations show that our interpretable model obtains competitive results when compared to state-of-the-art white and black-box models.

Patient-reported outcome measures on mental health and psychosocial factors in patients with Brugada syndrome

AIMS

Brugada syndrome (BrS) is a hereditary arrhythmic disease, associated with sudden cardiac death. To date, little is known about the psychosocial correlates and impacts associated with this disease. The aim of this study was to assess a set of patient-reported psychosocial outcomes, to better profile these patients, and to propose a tailored psychosocial care.

METHODS AND RESULTS

Patients were recruited at the European reference Centre for BrS at Universitair Ziekenhuis Brussel, Belgium. Recruitment was undertaken in two phases: phase 1 (retrospective), patients with confirmed BrS, and phase 2 (prospective), patients referred for ajmaline testing who had an either positive or negative diagnosis. BrS patients were compared to controls from the general population. Two hundred and nine questionnaires were analysed (144 retrospective and 65 prospective). Collected patient-reported outcomes were on mental health (12 item General Health Questionnaire; GHQ-12), social support (Oslo Social Support Scale), health-related quality of life, presence of Type-D personality (Type-D Scale; DS14), coping styles (Brief-COPE), and personality dimensions (Ten Item Personality Inventory). Results showed higher mental distress (GHQ-12) in BrS patients (2.53 ± 3.03) than in the general population (P < 0.001) and higher prevalence (32.7%) of Type D personality (P < 0.001) in patients with confirmed Brugada syndrome (BrS +). A strong correlation was found in the BrS + group (0.611, P < 0.001) between DS14 negative affectivity subscale and mental distress (GHQ-12).

CONCLUSION

Mental distress and type D personality are significantly more common in BrS patients compared to the general population. This clearly illustrates the necessity to include mental health screening and care as standard for BrS.

Challenges in describing the conformation and dynamics of proteins with ambiguous behavior

Traditionally, our understanding of how proteins operate and how evolution shapes them is based on two main data sources: the overall protein fold and the protein amino acid sequence. However, a significant part of the proteome shows highly dynamic and/or structurally ambiguous behavior, which cannot be correctly represented by the traditional fixed set of static coordinates. Representing such protein behaviors remains challenging and necessarily involves a complex interpretation of conformational states, including probabilistic descriptions. Relating protein dynamics and multiple conformations to their function as well as their physiological context (e.g., post-translational modifications and subcellular localization), therefore, remains elusive for much of the proteome, with studies to investigate the effect of protein dynamics relying heavily on computational models. We here investigate the possibility of delineating three classes of protein conformational behavior: order, disorder, and ambiguity. These definitions are explored based on three different datasets, using interpretable machine learning from a set of features, from AlphaFold2 to sequence-based predictions, to understand the overlap and differences between these datasets. This forms the basis for a discussion on the current limitations in describing the behavior of dynamic and ambiguous proteins.

Long short-term cognitive networks

In this paper, we present a recurrent neural system named long short-term cognitive networks (LSTCNs) as a generalization of the short-term cognitive network (STCN) model. Such a generalization is motivated by the difficulty of forecasting very long time series efficiently. The LSTCN model can be defined as a collection of STCN blocks, each processing a specific time patch of the (multivariate) time series being modeled. In this neural ensemble, each block passes information to the subsequent one in the form of weight matrices representing the prior knowledge. As a second contribution, we propose a deterministic learning algorithm to compute the learnable weights while preserving the prior knowledge resulting from previous learning processes. As a third contribution, we introduce a feature influence score as a proxy to explain the forecasting process in multivariate time series. The simulations using three case studies show that our neural system reports small forecasting errors while being significantly faster than state-of-the-art recurrent models.

Fuzzy-Rough Cognitive Networks: Theoretical Analysis and Simpler Models

Fuzzy-rough cognitive networks (FRCNs) are recurrent neural networks (RNNs) intended for structured classification purposes in which the problem is described by an explicit set of features. The advantage of this granular neural system relies on its transparency and simplicity while being competitive to state-of-the-art classifiers. Despite their relative empirical success in terms of prediction rates, there are limited studies on FRCNs' dynamic properties and how their building blocks contribute to the algorithm's performance. In this article, we theoretically study these issues and conclude that boundary and negative neurons always converge to a unique fixed-point attractor. Moreover, we demonstrate that negative neurons have no impact on the algorithm's performance and that the ranking of positive neurons is invariant. Moved by our theoretical findings, we propose two simpler fuzzy-rough classifiers that overcome the detected issues and maintain the competitive prediction rates of this classifier. Toward the end, we present a case study concerned with image classification, in which a convolutional neural network is coupled with one of the simpler models derived from the theoretical analysis of the FRCN model. The numerical simulations suggest that once the features have been extracted, our granular neural system performs as well as other RNNs.

Interpreting a black box predictor to gain insights into early folding mechanisms

Protein folding and function are closely connected, but the exact mechanisms by which proteins fold remain elusive. Early folding residues (EFRs) are amino acids within a particular protein that induce the very first stages of the folding process. High-resolution EFR data are only available for few proteins, which has previously enabled the training of a protein sequence-based machine learning 'black box' predictor (EFoldMine). Such a black box approach does not allow a direct extraction of the 'early folding rules' embedded in the protein sequence, whilst such interpretation is essential to improve our understanding of how the folding process works. We here apply and investigate a novel 'grey box' approach to the prediction of EFRs from protein sequence to gain mechanistic residue-level insights into the sequence determinants of EFRs in proteins. We interpret the rule set for three datasets, a default set comprised of natural proteins, a scrambled set comprised of the scrambled default set sequences, and a set of de novo designed proteins. Finally, we relate these data to the secondary structure adopted in the folded protein and provide all information online via http://xefoldmine.bio2byte.be/, as a resource to help understand and steer early protein folding.

An evidence-based bundle improves the quality of care and outcomes of patients with candidaemia

Background

Candidaemia is a leading cause of bloodstream infections in hospitalized patients all over the world. It remains associated with high mortality.

Objectives

To assess the impact of implementing an evidence-based package of measures (bundle) on the quality of care and outcomes of candidaemia.

Methods

A systematic review of the literature was performed to identify measures related to better outcomes in candidaemia. Eight quality-of-care indicators (QCIs) were identified and a set of written recommendations (early treatment, echinocandins in septic shock, source control, follow-up blood culture, ophthalmoscopy, echocardiography, de-escalation, length of treatment) was prospectively implemented. The study was performed in 11 tertiary hospitals in Spain. A quasi-experimental design before and during bundle implementation (September 2016 to February 2018) was used. For the pre-intervention period, data from the prospective national surveillance were used (May 2010 to April 2011).

Results

A total of 385 and 263 episodes were included in the pre-intervention and intervention groups, respectively. Adherence to all QCIs improved in the intervention group. The intervention group had a decrease in early (OR 0.46; 95% CI 0.23–0.89; P = 0.022) and overall (OR 0.61; 95% CI 0.4–0.94; P = 0.023) mortality after controlling for potential confounders.

Conclusions

Implementing a structured, evidence-based intervention bundle significantly improved patient care and early and overall mortality in patients with candidaemia. Institutions should embrace this objective strategy and use the bundle as a means to measure high-quality medical care of patients.

FCM Expert: Software Tool for Scenario Analysis and Pattern Classification Based on Fuzzy Cognitive Maps

Fuzzy Cognitive Maps (FCMs) have become a suitable and proven knowledge-based methodology for systems modeling and simulation. This technique is especially attractive when modeling systems characterized by ambiguity, and/or non-trivial causalities among its variables. The rich literature that is found related to FCMs reports very clearly many successful studies solved through the use of FCMs; however, when it comes to software implementations, where domain experts can design FCM-based systems, run simulations or perform more advanced experiments, not much is found or documented. The few existing implementations are not proficient in providing options for experimentation. Therefore, we believe that a gap exists, specifically between the theoretical advances and the development of accurate, transparent and sound FCM-based systems; and we advocate for the creation of more complete and exible software products. The goal of this paper is to introduce “FCM Expert”, a software tool for fuzzy cognitive modeling, where we focus on scenario analysis and pattern classification. The main features of FCM Expert rely on Machine Learning algorithms to compute the parameters that might define a model, optimize its network topology and improve the system convergence without losing information. Also, FCM Expert allows performing WHAT-IF simulations and studying the system behavior through a friendly, intuitive and easy-to-use graphical user interface.

Interventions to decrease short-term peripheral venous catheter-related bloodstream infections: impact on incidence and mortality

BACKGROUND

Short-term peripheral venous catheters are a significant source of healthcare-acquired bloodstream infections and a preventable cause of death.

AIM

To assess the effectiveness of interventions applied to reduce the incidence and mortality associated with short-term peripheral venous catheter-related bloodstream infections (PVCR-BSIs).

METHODS

The intervention included continuous PVCR-BSI surveillance, implementation of preventive measures related to catheter insertion and maintenance in accordance with evidence-based recommendations and the hospital's own data, front-line staff educational campaigns, and assessment of adherence to hospital guidelines by ward rounds. A Poisson regression model was used to estimate the trend of rate per year.

FINDINGS

From January 2003 to December 2016, 227 episodes of PVCR-BSI were identified among hospitalized patients at a university hospital. The mean age of patients was 67 years (standard deviation 14 years), 69% were male and the median Charlson score was 3 (interquartile range 2-5). Staphylococcus aureus caused 115 (50.7%) episodes. Thirty-day mortality was 13.2%. After implementation of the intervention, the incidence of PVCR-BSIs decreased significantly from 30 episodes in 2003 (1.17 episodes/10,000 patient-days) to eight episodes in 2016 (0.36/10,000 patient-days). The number of episodes caused by S. aureus decreased from 18 episodes in 2003 (0.70/10,000 patient-days) to three episodes in 2016 (0.14/10,000 patient-day), and mortality decreased from seven cases in 2003 (0.27/10,000 patient-days) to zero cases in 2016 (0.00/10,000 patient-days).

CONCLUSIONS

Surveillance, implementation of a multi-modal strategy and periodical assessment of healthcare workers' adherence to hospital guidelines led to a sustained reduction in PVCR-BSIs. This reduction had a major impact on S. aureus BSI rates and associated mortality.

Mortality risk factors among non-ICU patients with nosocomial vascular catheter-related bloodstream infections: a prospective cohort study

BACKGROUND

Vascular catheter-related bloodstream infections (CRBSIs) are highly preventable hospital-acquired infections and a major threat to patient safety. While there is significant information regarding CRBSI outcome among intensive care unit (ICU) patients, data regarding non-ICU patients are scarce.

AIM

To determine the risk factors associated with 30-day mortality among non-ICU patients with nosocomial CRBSIs.

METHODS

Prospective cohort study of non-ICU patients with nosocomial CRBSIs in a tertiary care centre between January 2004 and December 2014. The primary outcome was 30-day mortality, defined as death from any cause within 30 days of CRBSI. Follow-up was performed 30 days after CRBSI onset. Time until death was the dependent variable in Cox regression analysis.

FINDINGS

In total, 546 cases of CRBSI were identified. The mean age of patients was 64.5 years [interquartile range (IQR) 55-75 years], 66% were male, and the mean Charlson score was 3.59 (IQR 2-5). Of the 546 cases, 58.4% resulted from central venous catheters and 41.6% from peripheral venous catheters. The causative agents were Gram-positive cocci (70.1% of cases), Gram-negative bacilli (31.1%) and Candida spp. (1%). Mortality within 30 days was 13.9%, with no significant changes over the study period. Independent risk factors for 30-day mortality were Charlson score ≥4 [hazard ratio (HR) 1.80, 95% confidence interval (CI) 1.19-2.73], Staphylococcus aureus infection (HR 2.67, 95% CI 1.61-4.43) and Candida spp. infection (HR 6.1, 95% CI 2.08-18.04). Age; area of admission; type, use and site of vascular catheter; and administration of appropriate empirical antibiotic treatment were not independent risk factors for 30-day mortality.

CONCLUSION

Nosocomial CRBSIs outside ICUs are associated with high risk of mortality, particularly among patients with a higher Charlson score and bloodstream infections caused by Staphylococcus aureus and Candida spp.

Hypervirulent Klebsiella pneumoniae clones causing bacteraemia in adults in a teaching hospital in Barcelona, Spain (2007-2013)

Virulent hypermucoviscous Klebsiella pneumoniae strains associated with the magA and rmpA genes have mainly emerged in Asia. We analysed the frequency and the clinical and molecular epidemiology of K. pneumoniae bacteraemia isolates obtained over a 7-year period (2007-2013). Fifty-three of 878 K. pneumoniae invasive isolates (5.4%) showed a hypermucoviscous phenotype (by the string test). Of these, 16 (30.2%) were magA(+)/rmpA(+), 12 (22.6%) were magA(-)/rmpA(+), and the remaining 25 (47.2%) were magA(-)/rmpA(-). After multilocus sequence typing and wzi sequencing, all magA(+)/rmpA(+) isolates were serotype K1 and sequence type (ST)23. Of the 12 magA(-)/rmpA(+) isolates, nine were K2 (ST380, ST86, ST65, ST25 and ST493), and three magA(-)/rmpA(+) isolates had the new wzi allele 122, an unknown serotype, and the new ST1013. The remaining isolates, which were magA(-)/rmpA(-), showed different serotypes and STs. Patients with magA(+)/rmpA(+) or magA(-)/rmpA(+)K. pneumoniae bacteraemia more frequently had pyogenic liver abscesses (PLAs) and pneumonia than patients with magA(-)/rmpA(-)K. pneumoniae bacteraemia (respectively: 21.4% vs. 8%, p 0.26; and 17.9% vs. 0%, p 0.05). In fact, magA(-)/rmpA(-) isolates were similar to the those termed 'classic' K. pneumoniae isolates causing bacteraemia, the urinary and biliary tracts being the main foci of infection. In conclusion, hypervirulent clones (CC23K1, CC86K2, CC65K2, and CC380K2) were infrequent among K. pneumoniae isolates causing bacteraemia in our geographical area. A hypermucoviscous phenotype as determined with the string test is not enough to recognize these clones; additional molecular studies are needed. Patients with magA(+) and/or rmpA(+)K. pneumoniae bacteraemia more frequently had PLAs and pneumonia than patients without hypermucoviscosity genes.

Evolution and genetic diversity of the Spain23F-ST81 clone causing adult invasive pneumococcal disease in Barcelona (1990-2012)

Objectives

We aimed to analyse the clinical epidemiology and genetic diversity of invasive pneumococcal disease (IPD) episodes attributed to the Spain^23F-ST81 (PMEN1) clone.

Methods

Fifty-eight (2.7%) of 2117 invasive pneumococci isolated from adult patients during the 1990–2012 period shared a PFGE pattern related to the PMEN1 clone. The genotype was confirmed by multilocus sequence typing. The pbp2x, pbp1a, pbp2b and pspA genes were PCR-amplified and sequenced. Polymorphisms in the pspC gene were identified by PCR restriction fragment length polymorphism. The presence of transposons with erythromycin and tetracycline resistance determinants was detected by PCR.

Results

The prevalence of the PMEN1 clone increased from 0.8% in 1991 to 6.2% in 2001, and decreased to 0% in 2010–12, concomitant with the introduction of the seven-valent pneumococcal conjugate vaccine for children. A total of 93.1% of patients had pneumonia, meningitis or peritonitis; 87.9% of patients had associated underlying diseases, mainly cancer, chronic obstructive pulmonary disease and diabetes. Two closely related sequence types (STs) (ST81, n = 52; ST85, n = 6) were detected, with different serotypes: 23F (n = 42), 19A (n = 9) and 19F (n = 6). All the isolates were resistant to penicillin, co-trimoxazole and chloramphenicol. All the isolates also shared the same pbp1a allele, whereas multiple alleles of pbp2b, pbp2x, pspA and pspC were detected. Of the isolates, 89.7% were tetracycline resistant and 60.3% (n = 35) were macrolide resistant, and resistance was associated with different Tn916-like transposons.

Conclusions

Adult IPD caused by this clone was mainly detected in patients with underlying conditions, and genetic variability was observed among PMEN1 isolates collected in our area over the past 20 years.

Clinical Characteristics and Response to Newer Quinolones inLegionellaPneumonia: A Report of 28 Cases

Twenty-eight (11.6%) out of 241 Spanish patients enrolled in an international phase III clinical trial of mild to moderate community-acquired pneumonia (CAP) comparing gemifloxacin vs. trovafloxacin were diagnosed of Legionnaires' disease. A definite diagnosis was established by seroconversion in 13 patients of whom only 2 had a positive Legionella urinary antigen. The remaining 15 patients were possible Legionella infections based on a single elevated IgG titer (> or = 1:512). All patients had a radiologically confirmed diagnosis of pneumonia, 5 (19%) patients were older than 65, comorbidity was present in 9 (33%), and 10 (36%) had to be hospitalized. Fifteen patients were treated with oral gemifloxacin (320 mg/day) and 13 with oral trovafloxacin (200 mg/day). Overall, clinical success occurred in 25 (89.3%) patients after 7 days of treatment and only 1 patient needed a 14-day treatment. There were only one adverse event withdrawal and one clinical failure, and no patients died. In light of the favorable clinical outcome, the use of newer fluoroquinolones seems adequate for the treatment of suspected or proven Legionella pneumonia.

Trends in mortality and antibiotic resistance among HIV-infected patients with invasive pneumococcal disease

OBJECTIVES

The aim of the study was to describe trends and risk factors for mortality and changes in antibiotic resistance, serotypes and clones among HIV-infected patients with invasive pneumococcal disease (IPD).

METHODS

A prospective study of 199 episodes of IPD occurring in a cohort of 4011 HIV-infected patients was carried out. Predictors of mortality included clinical and microbiological data. The 7-valent pneumococcal conjugate vaccine (PCV7) for children was introduced in late 2001. Time periods were classified for mortality studies as pre- (1986-1996), early (1997-2001) and late (2002-2007) highly active antiretroviral therapy (HAART) era, and for serotype studies as pre-PCV7 (1986-2001) and PCV7 (2002-2007) era.

RESULTS

Of 199 IPD episodes, 71 (36%) occurred in HIV-infected patients with associated comorbidities (mainly liver cirrhosis; 52 of 71), which increased in recent years. The incidence of IPD decreased from the pre-HAART era to the early HAART era and then remained stable in the late HAART era (24.1, 8.4 and 7.4 episodes per 1000 patient-years, respectively). Rates of 30-day mortality have risen over the three periods (8, 19 and 25%, respectively; P = 0.017). In multiple logistic regression analysis, predictors of mortality were shock at presentation [odds ratio (OR) 7.01; 95% confidence interval (CI) 2.05-23.87] and associated comorbidities (OR 4.27; 95% CI 1.53-11.92). In the PCV7 era, IPD caused by non-PCV7 serotypes increased, and resistance to betalactams decreased. The most frequent genotypes were Spain(9V)-ST156, Spain(23F)-ST81, ST88(19F), Sweden(1)-ST304 and Spain(6B)-ST90.

CONCLUSIONS

In the late HAART era, the incidence of IPD has not significantly decreased. Mortality from IPD has risen in association with an increase in comorbidities such as liver cirrhosis. New vaccination strategies are needed to diminish the burden of IPD in the HIV-infected population.

Bacteremic pneumococcal cellulitis compared with bacteremic cellulitis caused by Staphylococcus aureus and Streptococcus pyogenes

In order to better characterize bacteremic cellulitis caused by Streptococcus pneumoniae, a review was conducted of 10 cases of bacteremic pneumococcal cellulitis, which represented 0.9% of all cases of pneumococcal bacteremia (n=1,076) and 3.2% of all cases of community-acquired bacteremic cellulitis (n=312) that occurred in the Hospital de Bellvitge, Barcelona, from 1984 to 2001. In addition to these 10 cases, 28 cases of bacteremic pneumococcal cellulitis from the literature (Medline 1975-2001) were reviewed. Pneumococcal cellulitis of the face, neck, and trunk was observed more frequently in patients with systemic lupus erythematosus and hematologic disorders, while pneumococcal cellulitis of the limbs was more common in patients with diabetes, alcoholism, and parenteral drug use. In the Hospital de Bellvitge group, bacteremic cellulitis due to Streptococcus pneumoniae was more frequently associated with severe underlying diseases than that due to Staphylococcus aureus or Streptococcus pyogenes (100%, 57%, and 72%, respectively;P=0.01). A concomitant extracutaneous focus of infection (e.g., respiratory tract infection) suggesting hematogenous spread with metastatic cellulitis was more frequent in patients with pneumococcal cellulitis, while a local cutaneous entry of microorganisms was feasible in most patients with Staphylococcus aureus or Streptococcus pyogenes cellulitis. The 30-day mortality was 10% in patients with pneumococcal cellulitis, 13% in patients with Staphylococcus aureus cellulitis, and 23% in patients with Streptococcus pyogenes cellulitis (P=0.3). Thus, bacteremic pneumococcal cellulitis is an unusual manifestation of pneumococcal disease and occurs mainly in patients with severe underlying diseases. In most cases, pneumococcal cellulitis has a different pathophysiologic mechanism than cellulitis caused by Staphylococcus aureus or Streptococcus pyogenes.

Pneumococcal peritonitis in adult patients: report of 64 cases with special reference to emergence of antibiotic resistance

BACKGROUND

Few data are available regarding pneumococcal peritonitis. We studied the clinical characteristics of intra-abdominal infections caused by Streptococcus pneumoniae and its prognosis in relation to antibiotic resistance.

METHODS

We reviewed all cases of culture-proved pneumococcal peritonitis. Patients with liver cirrhosis and primary pneumococcal peritonitis were compared with patients with Escherichia coli peritonitis.

RESULTS

Between January 1, 1979, and December 31, 1998, we identified 45 cases of primary pneumococcal peritonitis in patients with cirrhosis and 19 cases of secondary (or tertiary) pneumococcal peritonitis. Patients with cirrhosis and primary pneumococcal peritonitis vs those with primary E coli peritonitis had more frequent community-acquired infection, 73% vs 47%; pneumonia, 36% vs 2%; and bacteremia, 76% vs 33%; and higher attributable mortality (early mortality), 27% vs 9% (P<.05 for all). Secondary (or tertiary) pneumococcal peritonitis was associated with upper or lower gastrointestinal tract diseases; in most cases, the infection appeared after surgery. A hematogenous spread of S pneumoniae from a respiratory tract infection might be the most important origin of peritonitis; also, S pneumoniae might directly reach the gastrointestinal tract favored by endoscopic procedures or hypochlorhydria. There was an increased prevalence of penicillin and cephalosporin resistance up to 30.7% and 17.0%, respectively, although it was not associated with increased mortality rates.

CONCLUSIONS

Primary pneumococcal peritonitis in patients with cirrhosis more often spread hematogenously from the respiratory tract and was associated with early mortality. In secondary (and tertiary) pneumococcal peritonitis, a transient gastrointestinal tract colonization and inoculation during surgery might be the most important mechanisms. Current levels of resistance were not associated with increased mortality rates.

A randomized study comparing triple versus double antiretroviral therapy or no treatment in HIV-1-infected patients in very early stage disease: the Spanish Earth-1 study

BACKGROUND

Most current guidelines state that antiretroviral therapy should be considered for HIV-infected patients with plasma HIV RNA > 5000-10000 copies/ml and CD4 cells > 500 x 10(6) cells/l. However, there is increasing concern about whether this is the optimal point to begin treatment or whether it is better to delay the initiation to more advanced stages.

OBJECTIVE

To study the immunological and virological benefits of starting antiretroviral therapy at these early stages.

METHODS

A total of 161 HIV-infected asymptomatic patients with CD4 cell count > 500 x 10(6) cells/l and viral load > 10000 copies/ml were randomly assigned to one of five treatment groups: no treatment, twice daily zidovudine and thrice daily zalcitabine (ZDV-ddC), twice daily zidovudine and didanosine (ZDV-ddI), twice daily stavudine and didanosine (D4T-ddI), or a twice daily three-drug regimen with stavudine and lamivudine and ritonavir. The endpoints were progression to < 350 x 10(6) cells/l CD4 cells, to < 500 x 10(6) cells/l with either two Centers for Disease Control class B symptoms or an increase of viral load > 0.5 log10 copies/ml above baseline, or to AIDS or death. In various substudies, the lymphoid tissue and cerebrospinal fluid viral load, development of genotypic resistance, proliferative responses to mitogens and cytomegalovirus, and HIV-1 specific antigens and other immunophenotypic markers were also analysed.

RESULTS

Progression rates to study endpoints within 1 year were greater in the control group (31%) than in all groups receiving antiretroviral therapy pooled together (5%; estimated hazard ratio 7.41; 95% confidence interval 5.72-74.55; P < 0.001). The peak mean viral load decrease was greater in the three-drug group when compared with any of the three groups with a two-drug regimen (2.32, 1.65, 1.72 and 1.84, respectively; P < or = 0.001). At 1 year, viral load remained below 20 copies/ml in 30 out of 33 patients in the three-drug group (91%) and in only eight out of 94 patients (9%) in two-drug groups (P = 0.001). The peak mean increase in CD4 cells was also greater in the three-drug group than in the double treatment arms (259 versus 85, 144 and 145 x 10(6) cells/l, respectively; P = 0.001). By comparison, 36% of patients in the three-drug group regimen had to change the therapy as a result of adverse events. Substudies were performed in 60 patients recruited at two sites. Tonsillar tissue HIV RNA was measured in seven patients (two in the two-drug groups and five in the three-drug group) in whom plasma HIV RNA was < 20 copies/ml at 1 year. It was 15151 and 133333 copies/mg tissue in the two patients from the two-drug group, < 40 copies/mg tissue in four patients in the three-drug group, and 485 copies/mg in one patient in the three-drug group. At 1 year there was a mean increase of 4.21+/-2.94% in CD8+CD38+ cells in the control group and a decrease of 9.48+/-3.36% in the two-drug groups (P = 0.01), and 19.87+/-3.64 in the three-drug group (P = 0.001 and P = 0.05, for comparisons with control group and two-drug groups, respectively). Although proliferative responses to cytomegalovirus antigens were significantly greater in those receiving antiretroviral therapy, response to HIV-1 p24 antigen was not detected in any patient in either treatment group.

CONCLUSIONS

This study supports the recommendation to start antiretroviral therapy with a three-drug combination during very early stages of HIV-1 disease, at least if viral load is above a cut-off point (10000 copies/ml in our study). The risk of progression was sevenfold higher in non-treated patients at 8 months of follow-up. Some immune system parameters improved toward normal values after 1 year of antiretroviral therapy, but the proliferative response of CD4 T lymphocytes against the p24 HIV-1 antigen was not recovered. Therapeutic approaches with more potent, better-tolerated and more convenient regimens will increasingly favour early intervention with antiretroviral t

Analysis of the discontinuation of protease inhibitor therapy in routine clinical practice

We evaluated the frequency of and reasons for discontinuation of protease inhibitor therapy in a cohort of HIV-infected patients in a prospective observational study. We included 230 HIV-infected patients who had started protease inhibitor therapy between November 1996 and July 1997. Mean baseline CD4 count was 138 cells/microl and HIV-RNA 4.5 log10. Forty-five percent of patients had prior AIDS and 77% had been treated with nucleoside analogues. Saquinavir-treated patients were at a less advanced stage of HIV disease. Overall, 41.3% of patients discontinued therapy, and their last HIV-RNA measured higher than that of patients who continued therapy: 4.07 vs. 2.70 log10 (p < 0.0001). Reasons for discontinuation of therapy were poor adherence (including abandonment) (18.6%), drug intolerance (12.1%), virological failure (7%) and physician decision (3.5%). In a multivariate model, factors associated with drug discontinuation were not taking indinavir (OR 0.26, 95% CI 0.12-0.59) and being pretreated with nucleoside analogues (OR 3.42, 95% CI 1.58-7.42). We concluded that in routine clinical practice a high proportion of patients discontinued protease inhibitors during the first 6 months of therapy, the main reason being the patient's own decision (abandonment or poor adherence). Psychological support and counselling are warranted in patients when initiating protease inhibitor therapy.

Indium-111 pentetreotide in the diagnostic work-up of patients with bronchogenic carcinoma

In a prospective study we examined 38 patients with primary bronchogenic carcinoma to validate the use of indium-111 pentetreotide (IPT) as a diagnostic tool. Of these 38 patients, 25 had small cell lung cancer (SCLC) and 13, non-small cell lung cancer (NSCLC). The aim of the study was to investigate whether (a) the disease can be reliably detected, (b) IPT allows differentiation between SCLC and NSCLC and (c) IPT provides further information on metastatic disease. After giving their informed consent the patients were injected and imaged 4 and 24 h later using a planar whole-body technique. In addition single-photon emission tomography of the thorax and, if necessary, other areas of the body was performed at 24 h. In the 25 patients with SCLC 22 sites of primary tumour were correctly identified (true-positive, TP); one was false-negative (FN) and two were true-negative (TN), the patients being in full remission. Metastases were correctly identified in ten instances (lung, bone and brain), while the findings were FN in five cases. An additional six FN findings resulted in the area of the upper abdomen due to the physiological uptake in the liver, spleen and kidneys. In the 13 patients with NSCLC, ten findings were TP and 3 FN with respect to the primary tumour. Two FNs were squamous cell carcinoma, and one, adenocarcinoma. Metastases were TP in nine cases and FN in one.(ABSTRACT TRUNCATED AT 250 WORDS)