SEARCH

SEARCH BY CITATION

Keywords:

  • acute lymphoblastic leukemia;
  • Central America;
  • child;
  • developing countries;
  • treatment-related death

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. FUNDING SOURCES
  7. REFERENCES

BACKGROUND:

The objectives of this study were to describe the incidence, timing, and predictors of treatment-related mortality (TRM) among children with acute lymphoblastic leukemia (ALL) in El Salvador, Guatemala, and Honduras.

METHODS:

Patients aged <20 years who were diagnosed with ALL between January 2000 and March 2008, who received treatment in any of the 3 countries, and who started induction chemotherapy were included in the study. Almost all patients were treated on the El Salvador-Guatemala-Honduras II protocol, which was based on the St. Jude Total XIII and XV protocols. Biologic, socioeconomic, and nutritional variables were examined as predictors of TRM.

RESULTS:

Of 1670 patients, TRM occurred as a first event in 156 children (9.3%); TRM occurred during remission induction therapy in 92 of 156 children (59%), between remission induction and maintenance therapy in 27 of 156 children (17%), and during maintenance therapy in 37 of 156 children (24%). Although the TRM rate decreased in patients who were diagnosed after July 1, 2004 (11.2% vs 7.9%; P = .02), the rate of induction death did not change (5.2% vs 5.8%; P = .58). Independent predictors of induction death included higher risk ALL (odds ratio [OR], 1.84; 95% confidence interval [CI], 1.03-3.27; P = .04), lower initial platelet counts (OR per 10 × 109/L, 0.94; 95% CI, 0.89-0.98; P = .005), and longer travel time to the clinic (OR, 1.06 per hour; 95% CI, 1.01-1.14; P = .03).

CONCLUSIONS:

In Central America, TRM remains an important cause of treatment failure in children with ALL. A large proportion of TRM occurs in maintenance, although this proportion has decreased over time. Supportive care interventions should especially target children who present with low platelet counts. Further study on transfusion ability and the location of induction deaths is required. Cancer 2011;. © 2011 American Cancer Society.

The past several decades have seen significant improvements in the outcome of children with acute lymphoblastic leukemia (ALL) who are treated in high-income countries (HICs), in which >85% are cured.1, 2 However, the majority of children with ALL live in low-income countries (LICs), where the chance of a cure is far lower.3, 4 Several causes for this survival gap have been proposed: more advanced stage at diagnosis, differences in population biology, higher rates of relapse, abandonment of therapy, and higher rates of death from toxicity (treatment-related mortality [TRM]).5

TRM commonly causes treatment failure in LICs. Whereas ALL in HICs has been associated with TRM rates of 1% to 3%,6-11 the few studies that have investigated LICs have identified rates from 11% to 21%.4, 12-15 In El Salvador, 12.5% of children with ALL experienced TRM.16 Although we identified several predictors of overall TRM in the El Salvador cohort, the limited sample size precluded an analysis of the patterns and timing of TRM. A more detailed understanding is necessary for both the design of rational interventions and the identification of high-risk groups for which specific interventions should be developed. In the current study, we expanded our cohort to analyze children with ALL who were diagnosed over an 8-year period in 3 Central American countries—El Salvador, Guatemala, and Honduras—allowing a detailed examination of the incidence, causes, timing, and predictors of TRM in this population.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. FUNDING SOURCES
  7. REFERENCES

Study Population and Setting

In this retrospective cohort study, the patient sample consisted of children with ALL who started induction chemotherapy in any of the following 3 centers: Benjamin Bloom National Children's Hospital, San Salvador, El Salvador; the National Pediatric Oncology Unit, Guatemala City, Guatemala; or Hospital Escuela, Tegucigalpa, Honduras. These hospitals and their associated satellites are the sole pediatric oncology centers in their respective countries, which allowed us to conduct population-based analyses. We included those aged birth to 20 years at diagnosis with de novo ALL who were diagnosed between January 1, 2000 and March 28, 2008. Patients with mature B-cell leukemia were excluded.

The vast majority of patients in all 3 sites were treated according to the El Salvador-Guatemala-Honduras II protocol, which was based on the St Jude Total XIII and XV protocols.1, 17 Important modifications from the St Jude protocols included the use of only 2 risk groups, standard and high. Standard-risk patients were defined as those ages 1 to 10 years at diagnosis who presented with white blood cell (WBC) counts <50 × 109/L, a DNA index between 1.16 and 1.6, and the absence of any high-risk features (central nervous system [CNS] or testicular involvement, T-cell immunophenotype, M3 bone marrow on Day 15, or M2/M3 bone marrow on Day 36). Etoposide was omitted from the St. Jude protocols, and the dosing of high-dose methotrexate was modified to a 3-hour infusion of 2 g/m2 for standard-risk patients and 3 g/m2 for high-risk patients.

Programs targeting hand washing were in place during the entire study period. Patients received cotrimoxazole for Pneumocystis jiroveci prophylaxis; no other prophylactic antibiotics or antifungals were used. Central lines were not used routinely in the care of children with ALL. All children with febrile neutropenia were admitted to the hospital and treated with broad-spectrum antibiotics. All 3 centers broadened antibiotics in case of hemodynamic instability and initiated antifungal coverage in patients with prolonged fever; the specific choice of antibiotics and antifungals varied by center and period. Intensive care and mechanical ventilation were available throughout the study period in all 3 centers. Colony-stimulating factors were not used for prophylaxis or for the treatment of febrile neutropenia.

Blood banking practices and availability of blood products varied among centers. Both El Salvador and Honduras had blood banks available on site, whereas the center in Guatemala was supported by an off-site, private blood bank. Blood products were obtained from both family members and random donors in all 3 sites. Hemoglobin transfusion thresholds for stable patients were 7.0 g/L in Guatemala and Honduras and 8.0 g/L in El Salvador. Platelet transfusion levels for asymptomatic patients were 10,000 × 109/L in Guatemala and Honduras and 40,000 × 109/L in El Salvador. No systematic data on the length of time between blood product request and administration were available.

Allopurinol was given to all children with ALL in El Salvador in an effort to prevent tumor lysis syndrome (TLS); its use in the other 2 centers was restricted to those patients who had high WBC counts, high uric acid levels, or signs of TLS. Rasburicase and leukophoresis were not available at any site, although hemodialysis was available at all 3 centers.

Abandonment of therapy is an important treatment challenge in many LICs and can represent the most common cause of treatment failure.18 In all 3 centers, social workers attempted to contact families who missed appointments to encourage the resumption of treatment. This often involved visits to the patient's community and home. Treatment was provided at no financial cost to families; accommodation and child care also were offered at no cost to families living significant distances from the treatment center. This was funded primarily through the support of local nongovernmental organizations and international partnerships. These efforts have reduced abandonment rates to 13% for children in El Salvador.19

Procedure

In all 3 centers, data managers abstracted information from patient charts in real time. Data were confirmed by the treating oncologists and then entered in the Pediatric Oncology Networked Database (POND) (http//:www. pond4kids.org accessed March 15, 2011). POND is an online database for pediatric cancer patient information that was designed to permit users at multiple locations to store and analyze data that include patient demographics, diagnoses, treatments, and outcomes in a secure environment with stringent control of access and privacy.20 Audits of POND data quality in Honduras revealed that accuracy for basic data fields was 99%.20

Outcome Measures

The primary outcome was TRM, which was defined as any induction death or any death in complete remission. Induction death was defined as any death that occurred within 42 days of starting induction treatment. Because date of remission was not captured in the database, death in complete remission was defined as any death that occurred ≥42 days after diagnosis in patients who did not relapse or have a second malignancy. These definitions are consistent with those used by others.7, 10, 11

TRM was further subdivided into specific causes: infection, bleeding, and other causes (including metabolic derangements and organ dysfunction). An additional category of disease-related death was included for patients who died during remission induction therapy. Cause of death was determined by the local site. Timing of TRM was categorized as follows: 1) induction (from Day 0 to Day <42), 2) postinduction (from Day 42 to the beginning of maintenance therapy), and 3) after the initiation of maintenance therapy. The study period was divided into an early period and a late period, which we defined as before and after July 1, 2004, respectively (the approximate midpoint of the study period).

Potential Predictors

We examined predictors of TRM during 2 phases of therapy, during remission induction and after initiating maintenance therapy, as the majority of deaths occurred during these periods. Potential predictor variables were chosen based on previous literature and data availability across the 3 sites and were categorized as biologic, socioeconomic, or nutritional. Different sites began to collect data on covariates at different times during the study period, resulting in various data availability between variables. Biologic variables included demographic features, such as age and sex, and disease-related features, such as risk status (standard vs high), immunophenotype (B-cell vs T-cell), DNA index, CNS status, and initial blood counts.

Socioeconomic variables included monthly income, maximal parental education (secondary level or greater vs primary level or lower), number of household family members, and time traveling to the clinic. Monthly incomes were measured across the 3 countries in US dollars (USD), which, in turn, were corrected for purchasing power parity (PPP). Simple exchange rates between countries do not reflect the individual purchasing power of a currency within a country. PPP is calculated between countries by comparing the price of a standard “basket” of goods and services. 2005 PPP conversion factors provided by the World Bank were used: El Salvador, 0.55; Guatemala, 4.54; and Honduras, 9.66.21 Thus, 4.54 units of the local Guatemalan currency have the same purchasing power as 9.66 units of the local Honduran currency. Dividing these figures by the average USD-local currency exchange rate for 2005 (1:1, 1:7.73, and 1:19.6 for El Salvador, Guatemala, and Honduras, respectively), we calculated that 0.55 USD in El Salvador had the same purchasing power as 0.587 USD in Guatemala and 0.49 USD in Honduras. Therefore, the monthly income of each patient in USD was multiplied by the appropriate correction to give the monthly PPP.

Nutritional variables included body mass index percentile, triceps skin-fold thickness percentile, middle-upper arm circumference percentile, and initial albumin. Body mass index percentile was calculated relative to growth charts published by the Centers for Disease Control and Prevention in 2000.22 Triceps skin-fold thickness provides a measure of fat mass, whereas middle-upper arm circumference is a measure of lean mass; previously, these have been suggested as “gold-standard” measures of nutritional status in both general pediatric and oncologic populations.23 Both were calculated with reference to previously collected population norms.24

Statistical Methods

The distribution of TRM causes by phase of therapy and the proportion of patients experiencing TRM by early and late periods were compared using the chi-square test or the Fisher exact test, as appropriate. Predictors of induction death were examined using logistic regression. Cox proportional hazards models were used to explore predictors of maintenance TRM in which relapsed and second malignancies were considered competing events. Variables that were significant at the P < .1 level in univariate analyses were included in multivariate models. Statistical analyses were performed using SAS-PC software (version 9.2; SAS Institute, Cary, NC). Statistical significance was defined as P < .05. Because of the nature of this study, consent from patients was not required. This study was approved by the institutional research ethics boards at The Hospital for Sick Children in Toronto, Canada and at each of the 3 local sites.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. FUNDING SOURCES
  7. REFERENCES

The study sample included 1792 patients who were diagnosed within the specified period. Of these patients, 122 (6.8%) did not start any chemotherapy; thus, 1670 patients with ALL were included in the analysis. Demographic characteristics of the patients are provided in Table 1. Of the children who started any chemotherapy, 156 (9.3%) experienced TRM. Of these 156 deaths, 92 (59%) occurred in induction (before Day 42), 27 (17%) occurred after induction (from Day 42 to before the start of maintenance therapy), and 37 (24%) occurred during maintenance therapy or later.

Table 1. Demographic Characteristics of the Study Sample
CharacteristicNo. With Data AvailableaNo. (%)Median [IQR]
  • IQR indicates interquartile range; CNS, central nervous system; WBC, white blood cell

  • a

    Different sites began to collect covariates at different times during the study period, resulting in different data availability between variables.

Biologic   
 Boys1670942 (56.4) 
 Age, y1670 6.3 [0.0-19.9]
 High risk1658859 (51.8) 
 CNS positive1648172 (10.4) 
 B-lineage immunophenotype16391504 (91.8) 
 DNA index1347 1.00 [1.00-1.16]
 Initial WBC, ×109/L1663 9600 [3600-37,000]
 Initial hemoglobin, g/L1529 7.4 [5.5-9.6]
 Initial platelets, ×109/L1527 41,000 [17,000-106,000]
Socioeconomic   
 Monthly purchasing power parity units538 82.5 [55-275]
 Parental education at least secondary756238 (31.5) 
 No. of family members679 5 [4-6]
 Travel to clinic, h1223 3 [1-3]
Nutritional   
 Body mass index percentile1000 49 [16-79]
 Triceps skin-fold thickness percentile565 17.5 [7.5-62.5]
 Middle/upper arm circumference percentile564 17.5 [2.5-37.5]

Table 2 indicates that the distribution of subcauses of TRM differed significantly by phase of therapy. Infection accounted for a greater proportion of TRM in maintenance therapy, whereas bleeding and other causes, primarily metabolic, were responsible for more deaths in the induction and postinduction periods (P = .01).When examining the proportion of patients who experienced TRM by treatment period, the TRM rate decreased in those who were diagnosed in the early period compared with the late period (11.2% vs 7.9%; P = .02). However, the rate of induction death did not change between the 2 periods (5.2% vs 5.8%; P = .58). Indeed, induction death represented an increasing proportion of total TRM between the 2 periods (46% vs 73%; P = .002), whereas the proportion of TRM that occurred in maintenance decreased (30% vs 17%; P = .002).

Table 2. Number of Treatment-Related Deaths by Phase and Cause of Deatha
 Cause of Death
Treatment PhaseDiseaseInfectionBleedingOtherb
  • a

    P = .01.

  • b

    Other causes of death included organ failure (hepatic, 3 patients; respiratory, 2 patients; renal, 1 patient; and multiorgan, 3 patients), tumor lysis syndrome and electrolyte disturbances (5 patients), pancreatitis (1 patient), and other/unknown (17 patients).

Induction (D 0 to D <42)7471721
Postinduction (D 42 to maintenance)01772
Maintenance or later02719

The median time of induction death was 17.5 days from the start of induction therapy, with an interquartile range (IQR) of 12 to 26 days. Thus, 75% of induction deaths occurred on or after Day 12 of induction therapy. The median time of maintenance TRM was 171 days (IQR, 78-532 days) from the start of maintenance therapy.

The univariate analysis of predictors of induction death is provided in Table 3. Predictors of induction death included high-risk status (odds ratio [OR], 1.77; 95% confidence interval [CI] 1.14-2.79; P = .01), higher initial WBC count (OR per 10 × 109/L, 1.02; 95% CI, 1.00-1.04; P = .02) lower initial platelets (OR per 10 × 109/L, 0.94; 95% CI, 0.90-0.97; P < .0001), and longer travel time to the clinic (OR per hour, 1.05; 95% CI, 1.01-1.11; P = .02). Nutritional variables did not predict induction death. When placed in a multivariate model that contained these 4 variables, the initial WBC count no longer was associated with induction death (OR per 10 × 109/L, 1.01; 95% CI, 0.98-1.03; P = .56). However, high-risk status (OR, 1.84; 95% CI, 1.03-3.27; P = .04), lower initial platelets (OR per 10 × 109/L, 0.94; 95% CI, 0.89-0.98; P = .005), and longer travel time to the clinic (OR per hour, 1.06; 95% CI, 1.01-1.14; P = .03) remained independently predictive of induction death. TRM during maintenance therapy was not predicted significantly by any of the examined variables.

Table 3. Univariate Predictors of Induction Death
VariableOR (95% CI)P
  • OR indicates odds ratio; CI, confidence interval; CNS, central nervous system; WBC, white blood cell.

  • a

    Significant difference.

Biologic  
 Boys1.16 (0.75-1.78).502
 Age, per y0.99 (0.93-1.04).576
 High risk1.77 (1.14-2.75).011a
 CNS negative1.04 (0.51-2.11).919
 B-lineage immunophenotype0.82 (0.40-1.67).579
 DNA index0.46 (0.06-3.55).453
 Initial WBC count, 10×109/L1.02 (1.00-1.04).023a
 Initial hemoglobin, g/L1.00 (0.92-1.08)1.000
 Initial platelet count, 10×109/L0.94 (0.94-0.97).001a
Socioeconomic  
 Monthly purchasing power parity, per unit1.00 (1.00-1.00).994
 Parent education at least secondary0.65 (0.28-1.54).330
 No. of family members0.90 (0.74-1.10).314
 Travel to clinic, h1.06 (1.01-1.11).021a
Nutritional  
 Body mass index percentile1.00 (0.99-1.01).426
 Triceps skin-fold thickness percentile0.99 (0.98-1.01).370
 Middle/upper arm circumference percentile1.00 (0.98-1.01).528

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. FUNDING SOURCES
  7. REFERENCES

The current study of 1670 children with ALL in Central America represents, to the best of our knowledge, the largest cohort of pediatric ALL patients studied in the LIC setting to date. We observed that TRM represented an important cause of treatment failure in this population and that 24% of these events occurred during maintenance therapy. Patients with high-risk status, lower initial platelet counts, and longer travel times to the clinic were at the greatest risk of induction death; however, we were unable to identify a group at greater risk of TRM during maintenance therapy. Although TRM rates decreased over the study period, the decrease mainly reflected a decrease in maintenance TRM, and induction death rates remained constant.

The TRM rate of 9.3% in this study was higher than equivalent rates in HIC settings that reportedly ranged from 1% to 3%.6-11 However, our incidence was somewhat lower than ALL TRM rates reported in other LIC populations; studies in India, Pakistan, Brazil, and Thailand have described TRM rates between 11.5% and 16%.12-15 It is noteworthy that, in a study conducted in Honduras of 168 children with ALL between January 1999 and January 2002, the TRM rate was 20.8%,4 which was higher than our more recent and much larger cohort. Indeed, our results verify that TRM is decreasing over time in Central America; 11.2% of children who were diagnosed before July 1, 2004 experienced TRM versus 7.9% of children who were diagnosed after that date. Improvements in supportive care and experience with the protocol were likely responsible for the decline. Over the study period, for example, weekly meetings conducted online at www.cure4kids.org were initiated, during which the supportive care of challenging individual patients was discussed. In the HIC setting, a “learning curve” with intensive treatment protocols has been associated with a reduction in TRM over time once centers gain experience with a specific protocol. For example, in pediatric patients with acute myeloid leukemia (AML), a reduction in the TRM rate from 19% to 12% over time was demonstrated in the Children's Oncology Group CCG-2961 trial,25 and a reduction from 18% to 10% was reported in the Medical Research Council AML-10 trial.26 Such a learning curve also may explain the reduction in TRM observed in Central America over time in the current study.

Induction death, however, did not decline, remaining similar across the study period (5.2% vs 5.8%). This is in contrast to recent reports from HICs.11 Over half of these deaths were secondary to infections, with another 18.5% caused by bleeding. It is noteworthy that the median time to induction death was at Day 17.5, and the majority of deaths occurred after Day 12 of therapy. These data suggest that there is opportunity and time for interventions to address induction deaths.

Few studies have analyzed predictors of TRM in LICs. In HICs, age, sex, high-risk status, higher initial WBC counts, T-cell lineage immunophenotype, and trisomy 21 have been associated with an increased risk of TRM.7-9, 11 In an Indian study, malnutrition, lymphadenopathy, and lower hemoglobin levels were associated with TRM.13 In our previous study in El Salvador, we observed that socioeconomic variables predicted TRM, whereas biologic parameters did not.16

This study is the first to examine predictors of TRM in an LIC subdivided by phase of therapy. The results may lead to the more accurate identification of high-risk groups, because predictors of TRM occurring during induction therapy are likely to be different from TRM occurring during maintenance therapy. Indeed, we observed that children with high-risk status, lower platelet counts, and a longer travel time to the treating center were at greater risk of induction death. These results have several implications. First, given the risk associated with a low initial platelet count and the finding that almost 20% of induction deaths were secondary to bleeding, children in this setting may benefit from more aggressive platelet transfusion support. These data suggest that careful review of platelet availability and assessment of outcomes using various platelet thresholds during the induction period may also be useful. Second, given the late median time of induction death and the risk associated with longer travel times, evaluation of inpatient versus outpatient location at the time of death is warranted. If deaths are occurring as outpatients, then consideration should be given to housing patients at a location close to the hospital for children who live far from the hospital, similar to the system used for children at risk of abandonment.

It is striking that approximately 25% of the TRM in our cohort occurred during maintenance therapy, and that most were infectious in nature. The median time of maintenance TRM was 171 days from the start of maintenance therapy, although the IQR was between 78 days and 532 days, indicating that these events were spread out throughout maintenance. This contrasts sharply with the patterns observed in HICs, where TRM in maintenance is considered a rare event. The etiology of the different pattern observed in our study population is unclear. Differences in risk and severity of infectious events, nutrition or immune status, or access to care during febrile episodes may account for the difference, although our study was unable to demonstrate a relation between nutritional or socioeconomic variables and a greater risk of maintenance TRM. Although maintenance TRM did decline over the study period, further study is needed to understand the mechanisms behind this event to facilitate further decreases. One way to gain more insight into the causes of TRM would be to carefully examine episodes of febrile neutropenia and determine whether there are barriers and delays to the administration of empiric antibiotic therapy.

The role of socioeconomic status (SES) in LIC pediatric oncology has yet to be fully elucidated. Measures of SES have been linked to outcome in ALL in several LICs, both for settings in which the cost of treatment was borne by families and for settings in which treatment was free.3, 27, 28 Work limited to El Salvador indicated that income predicted the risk of TRM, abandonment overall, and event-free survival in children with standard-risk ALL, although not in high-risk children.16, 19, 29 In the current study, however, which included the center in El Salvador and also the centers in Guatemala and Honduras, income did not predict induction death or maintenance TRM. The only socioeconomic variable that was predictive was travel time to the clinic. The reason for the different effect of SES in the 2 studies is unknown. The lack of data on parental education and income for a large number of patients may have weakened our ability to examine the effect of SES. However, it is also possible that the impact of SES differs in these 3 countries, thus explaining the negative association between SES and TRM in the combined analysis.

Strengths of our study include its population-based approach and the size of the cohort, which allowed for a more detailed look at TRM in this setting. However, several limitations deserve mention. First, our study did not include children who died in hospital before chemotherapy was initiated or those who died before reaching the hospital. Second, our definition of TRM was similar to that used by some, but not all, cooperative groups.7, 10, 11 The Nordic and Dutch groups, for example, in their definition of induction death, do not include children who die after achieving remission.8, 9 Thus, comparing our results with the results reported by these groups is problematic. Indeed, standard international definitions of TRM in acute leukemia are needed.

Another limitation of our study is that POND did not contain detailed information on the cause of infectious TRM. In part, the lack of this information is related to limited availability of diagnostic capabilities in individual centers, including the absence of blood culture media during some periods in some centers. Although the availability of diagnostic testing for bacteria, viruses, and fungi varied across centers and through time within each center, an important future goal is to examine the etiology of infectious mortality such that interventional trials may be planned. This may be especially helpful in understanding deaths that occur during maintenance therapy. For example, a clinical diagnosis of pneumonia as the cause of death is not sufficient to develop prevention strategies, because the optimal strategy would differ if pneumonias were predominantly bacterial versus secondary to Pneumocystis jiroveci. Finally, just as our study indicates that TRM in Central America follows a markedly different pattern than in HICs, caution should be exercised when generalizing our results to other low-income settings, in which different conditions may apply. Our current data demonstrates that the collection of local data is critical to designing the most appropriate interventions.

Future work should focus on identifying interventions that can improve outcomes in these settings. Interventions to consider include a focus on platelet transfusion support, housing during induction therapy, and support around febrile neutropenia. However, further research carefully delineating platelet product availability, location of induction death, time to empiric antibiotics in febrile neutropenia, and the microbiology of infectious TRM are prerequisites to designing rational interventional trials.

FUNDING SOURCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. FUNDING SOURCES
  7. REFERENCES

L.S. is supported by the Canadian Childhood Clinician Scientist Training Program from the Canadian Institutes of Health Research. This work was supported in part by the Pediatric Oncology Group of Ontario and the American Lebanese-Syrian Associated Charities. Funding sources had no role in the writing of the article or the decision to submit it for publication.

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. FUNDING SOURCES
  7. REFERENCES