Frequency and Determinants of a Short-Interval Follow-up Recommendation After an Abnormal Screening Mammogram.

PURPOSE: After imaging assessment of an abnormal screening mammogram, a follow-up examination 6 months later is recommended to some women. Our aim was to identify which characteristics of lesions, women, and physicians are associated to such short-interval follow-up recommendation in the Quebec Breast Cancer Screening Program. METHODS: Between 1998 and 2008, 1,839,396 screening mammograms were performed and a total of 114,781 abnormal screens were assessed by imaging only. Multivariate analysis was done with multilevel Poisson regression models with robust variance and generalized linear mixed models. RESULTS: A short-interval follow-up was recommended in 26.7% of assessments with imaging only, representing 2.3% of all screens. Case-mix adjusted proportion of short-interval follow-up recommendations varied substantially across physicians (range: 4%-64%). Radiologists with high recall rates (≥15%) had a high proportion of short-interval follow-up recommendation (risk ratio: 1.82; 95% confidence interval: 1.35-2.45) compared to radiologists with low recall rates (<5%). The adjusted proportion of short-interval follow-up was high (22.8%) even when a previous mammogram was usually available. CONCLUSIONS: Short-interval follow-up recommendation at assessment is frequent in this Canadian screening program, even when a previous mammogram is available. Characteristics related to radiologists appear to be key determinants of short-interval follow-up recommendation, rather than characteristics of lesions or patient mix. Given that it can cause anxiety to women and adds pressure on the health system, it appears important to record and report short-interval follow-up and to identify ways to reduce its frequency. Short-interval follow-up recommendations should be considered when assessing the burden of mammography screening.

Clinical image quality in daily practice of breast cancer mammography screening.

OBJECTIVE: To assess the quality of screening mammograms performed in daily practice in the Quebec Breast Cancer Screening Program. SUBJECTS AND METHODS: Clinical image quality of a random subsample of 197 screening mammograms performed in 2004-2005 was independently evaluated by 2 radiologists based on the criteria by Canadian Association of Radiologists (CAR). When disagreement occurred for overall judgement or positioning score, the mammograms were reviewed by a third radiologist. Cohen's kappas for interrater agreement were computed. Multivariable robust Poisson regression models were used to study associations of overall quality and positioning with body mass index (BMI) and breast density. RESULTS: The CAR criteria were not satisfied for 49.7% of the mammograms. Positioning was the quality attribute most often deficient, with 37.2% of mammograms failing positioning. Interrater agreement ranged from slight (kappa = 0.02 for compression and sharpness) to fair (kappa = 0.30 for exposure). For overall quality, women with a BMI ≥ 30 kg/m(2) had a failure proportion of 67.5% compared with 34.9% for women with a BMI<25 kg/m(2) (risk ratio 2.1 [95% confidence interval, 1.5-3.0]). For positioning, women with a BMI ≥ 30 kg/m(2) had a failure proportion of 53.8% compared with 27.9% for women with a BMI < 25 kg/m(2) (risk ratio 1.9 [95% confidence interval, 1.2-3.1]). Effects of breast density on image quality differed among radiologists. CONCLUSION: Despite measures to ensure high-quality imaging, including CAR accreditation, approximately half of this random sample of screening mammograms failed the CAR quality standards. It would be important to define quality targets for screening mammograms carried out in daily practice to interpret such observations.

Protective efficacy of a polyvalent influenza A DNA vaccine against both homologous (H1N1pdm09) and heterologous (H5N1) challenge in the ferret model.

This study describes the protective efficacy of a novel influenza plasmid DNA vaccine in the ferret challenge model. The rationally designed polyvalent influenza DNA vaccine encodes haemagglutinin and neuraminidase proteins derived from less glycosylated pandemic H1N1 (2009) and H3N2 (1968) virus strains as well as the nucleoprotein (NP) and matrix proteins (M1 and M2) from a different pandemic H1N1 (1918) strain. Needle-free intradermal immunisation with the influenza DNA vaccine protected ferrets against homologous challenge with an H1N1pdm09 virus strain, demonstrated by restriction of viral replication to the upper respiratory tract and reduced duration of viral shedding post-challenge. Breadth of protection was demonstrated in two heterologous efficacy experiments in which animals immunised with the influenza DNA vaccine were protected against challenge with a highly pathogenic avian influenza H5N1 virus strain with reproducible survival and clinical outcomes.

Inactivated pandemic 2009 H1N1 influenza A virus human vaccines have different efficacy after homologous challenge in the ferret model.

BACKGROUND: The 2009 pandemic H1N1 (A(H1N1)pdm09) influenza A virus (IAV) has replaced the previous seasonal H1N1 strain in humans and continues to circulate worldwide. The comparative performance of inactivated A(H1N1)pdm09 influenza vaccines remains of considerable interest. The objective of this study was to evaluate the efficacy of two licensed A(H1N1)pdm09 inactivated vaccines (AS03B adjuvanted split virion Pandemrix from GlaxoSmithKline and referred here as (V1) and non-adjuvanted whole virion Celvapan from Baxter and referred here as (V2)) in ferrets as a pre-clinical model for human disease intervention. METHODS: Naïve ferrets were divided into two groups (V1 and V2) and immunised intramuscularly with two different A/California/07/2009-derived inactivated vaccines, V1 administered in a single dose and V2 administered in 2 doses separated by 21 days. Six weeks after the first immunisation, vaccinated animals and a non-vaccinated control (NVC) group were intra-nasally challenged with 106.5 TCID50 of the isolate A/England/195/2009 A(H1N1)pdm09 with 99.1% amino acid identity to the vaccine strain. Clinical signs, lung histopathology, viral quantification and antibody responses were evaluated. RESULTS AND CONCLUSIONS: Results revealed important qualitative differences in the performance of both inactivated vaccines in relation to protection against challenge with a comparable virus in a naive animal (ferret) model of human disease. Vaccine V1 limited and controlled viral shedding and reduced lower respiratory tract infection. In contrast, vaccine V2 did not control infection and animals showed sustained viral shedding and delayed lower respiratory infection, resulting in pulmonary lesions, suggesting lower efficacy of V2 vaccine.

Assessing the volume-outcome hypothesis and region-level quality improvement interventions: pancreas cancer surgery in two Canadian Provinces.

BACKGROUND: The volume-outcome hypothesis suggests that if increased provider procedure volume is associated with improved patient outcomes, then greater regionalization to high-volume providers should improve region-level outcomes. Quality improvement interventions for pancreas cancer surgery implemented in year 1999 in Ontario, Canada were designed to regionalize surgery to high-volume hospitals and decrease operative mortality. Similar interventions were not used in Quebec, Canada. We assessed the volume-outcome hypothesis and the impact of the Ontario quality improvement interventions. MATERIALS AND METHODS: Administrative databases helped identify pancreatic resections from years 1994 to 2004 and relevant patient and hospital characteristics. Hospitals were high-volume if they provided ≥10 procedures in a given calendar year. Outcomes were regionalization of surgery to high-volume providers and rates of operative mortality. RESULTS: From 1994 to 2004 the percentage of cases in high-volume hospitals increased from 33 to 71% in Ontario and from 36 to 76% in Quebec. Annual rates of operative mortality dropped in Ontario (10.4-2.2% or less) and changed little in Quebec (7.2-9.8%). Changes in measures over time in both provinces were similar before and after year 1999. CONCLUSIONS: Regionalization was associated with improved operative mortality in Ontario but not in Quebec, undermining the volume-outcome hypothesis. The Ontario quality improvement interventions likely were of little influence since patterns in regionalization and operative mortality were similar before and after year 1999.

Reproducibility of serologic assays for influenza virus A (H5N1).

Hemagglutination-inhibition (HI) and neutralization are used to evaluate vaccines against influenza virus A (H5N1); however, poor standardization leads to interlaboratory variation of results. A candidate antibody standard (07/150) was prepared from pooled plasma of persons given clade 1 A/Vietnam/1194/2004 vaccine. To test human and sheep antiserum, 15 laboratories used HI and neutralization and reassortant A/Vietnam/1194/2004, A/turkey/Turkey/1/2005 (clade 2.2), and A/Anhui/1/2005 (clade 2.3.4) viruses. Interlaboratory variation was observed for both assays, but when titers were expressed relative to 07/150, overall percentage geometric coefficient of variation for A/Vietnam/1194/2004 was reduced from 125% to 61% for HI and from 183% to 81% for neutralization. Lack of reduced variability to clade 2 antigens suggested the need for clade-specific standards. Sheep antiserum as a standard did not reliably reduce variability. The World Health Organization has established 07/150 as an international standard for antibody to clade 1 subtype H5 and has an assigned potency of 1,000 IU/ampoule.

Comparison of single radial immunodiffusion, SDS-PAGE and HPLC potency assays for inactivated influenza vaccines shows differences in ability to predict immunogenicity of haemagglutinin antigen.

The current gold-standard potency test for inactivated influenza vaccines is the single radial immunodiffusion (SRD) assay. A number of alternative potency tests for inactivated influenza vaccines have been proposed in recent years. Evaluation of these new potency tests commonly involves comparison with SRD, in order to ascertain that the new method obtains values that correlate with those measured by the standard potency test. Here, we extended comparison of two methods, reverse-phase HPLC and SDS-PAGE, with SRD by assessing the methods' capacity to detect loss of potency induced by various deliberate treatments of vaccine samples. We demonstrate that neither of these methods detected the loss of potency observed by SRD; importantly, neither SDS-PAGE nor reverse-phase HPLC reflected results from mouse experiments that showed decreased immunogenicity and protection in vivo. These results emphasise the importance of assessing the stability-indicating nature, ie the ability to measure loss of vaccine potency, of any potential new potency assay.

Volume of screening mammography and performance in the Quebec population-based Breast Cancer Screening Program.

BACKGROUND: In the Quebec Breast Cancer Screening Program (Programme quebecois de depistage du cancer du sein [PQDCS]), radiologists' and facilities' volumes of screening mammography vary considerably. We examined the relation of screening-mammography volume to rates of breast cancer detection and false-positive readings in the PQDCS. METHODS: The study population included 307,314 asymptomatic women aged 50-69 years screened during 1998-2000. Breast cancer detection rates were analyzed by comparing all women with screening-detected breast cancer (n = 1709) and a 10% random sample of those without (n = 30,560). False-positive rates were analyzed by comparing the 3159 women with false-positive readings and the 27,401 others in the 10% random sample. Characteristics of participants, radiologists and facilities were obtained from the PQDCS information system. Data were analyzed by means of logistic regression. RESULTS: The rate of breast cancer detection appeared to be unrelated to the radiologist's screening-mammography volume but increased with the facility's screening-mammography volume. The breast cancer detection rate ratio for facilities performing 4000 or more screenings per year, compared with those performing fewer than 2000, was 1.28 (95% confidence interval [CI] 1.07-1.52). In contrast, the frequency of false-positive readings was unrelated to the facility's screening volume but was inversely related to the radiologist's screening volume: the rate ratio for readers of 1500 or more screenings per year compared with those reading fewer than 250 was 0.53 (95% CI 0.35-0.79). INTERPRETATION: Radiologists' and facilities' caseloads showed independent and complementary associations with performance of screening mammography in the PQDCS. Radiologists who worked in larger facilities and read more screening mammograms had higher breast cancer detection rates while maintaining lower false-positive rates.

Matrix M H5N1 Vaccine Induces Cross-H5 Clade Humoral Immune Responses in a Randomized Clinical Trial and Provides Protection from Highly Pathogenic Influenza Challenge in Ferrets.

BACKGROUND AND METHODS: Highly pathogenic avian influenza (HPAI) viruses constitute a pandemic threat and the development of effective vaccines is a global priority. Sixty adults were recruited into a randomized clinical trial and were intramuscularly immunized with two virosomal vaccine H5N1 (NIBRG-14) doses (21 days apart) of 30 µg HA alone or 1.5, 7.5 or 30 µg HA adjuvanted with Matrix M. The kinetics and longevity of the serological responses against NIBRG-14 were determined by haemagglutination inhibition (HI), single radial haemolysis (SRH), microneutralization (MN) and ELISA assays. The cross-H5 clade responses in sera were determined by HI and the antibody-secreting (ASC) cell ELISPOT assays. The protective efficacy of the vaccine against homologous HPAI challenge was evaluated in ferrets. RESULTS: The serological responses against the homologous and cross-reactive strains generally peaked one week after the second dose, and formulation with Matrix M augmented the responses. The NIBRG-14-specific seroprotection rates fell significantly by six months and were low against cross-reactive strains although the adjuvant appeared to prolong the longevity of the protective responses in some subjects. By 12 months post-vaccination, nearly all vaccinees had NIBRG-14-specific antibody titres below the protective thresholds. The Matrix M adjuvant was shown to greatly improve ASC and serum IgG responses following vaccination. In a HPAI ferret challenge model, the vaccine protected the animals from febrile responses, severe weight loss and local and systemic spread of the virus. CONCLUSION: Our findings show that the Matrix M-adjuvanted virosomal H5N1 vaccine is a promising pre-pandemic vaccine candidate. TRIAL REGISTRATION: ClinicalTrials.gov NCT00868218.

Intranasal vaccination with a plant-derived H5 HA vaccine protects mice and ferrets against highly pathogenic avian influenza virus challenge.

Highly pathogenic avian influenza H5N1 infection remains a public health threat and vaccination is the best measure of limiting the impact of a potential pandemic. Mucosal vaccines have the advantage of eliciting immune responses at the site of viral entry, thereby preventing infection as well as further viral transmission. In this study, we assessed the protective efficacy of hemagglutinin (HA) from the A/Indonesia/05/05 (H5N1) strain of influenza virus that was produced by transient expression in plants. The plant-derived vaccine, in combination with the mucosal adjuvant (3',5')-cyclic dimeric guanylic acid (c-di-GMP) was used for intranasal immunization of mice and ferrets, before challenge with a lethal dose of the A/Indonesia/05/05 (H5N1) virus. Mice vaccinated with 15 µg or 5 µg of adjuvanted HA survived the viral challenge, while all control mice died within 10 d of challenge. Vaccinated animals elicited serum hemagglutination inhibition, IgG and IgA antibody titers. In the ferret challenge study, all animals vaccinated with the adjuvanted plant vaccine survived the lethal viral challenge, while 50% of the control animals died. In both the mouse and ferret models, the vaccinated animals were better protected from weight loss and body temperature changes associated with H5N1 infection compared with the non-vaccinated controls. Furthermore, the systemic spread of the virus was lower in the vaccinated animals compared with the controls. Results presented here suggest that the plant-produced HA-based influenza vaccine adjuvanted with c-di-GMP is a promising vaccine/adjuvant combination for the development of new mucosal influenza vaccines.

The breadth of cross sub-type neutralisation activity of a single domain antibody to influenza hemagglutinin can be increased by antibody valency.

The response to the 2009 A(H1N1) influenza pandemic has highlighted the need for additional strategies for intervention which preclude the prior availability of the influenza strain. Here, 18 single domain VHH antibodies against the 2009 A(H1N1) hemagglutinin (HA) have been isolated from a immune alpaca phage displayed library. These antibodies have been grouped as having either (i) non-neutralising, (ii) H1N1 restricted neutralising or (iii) broad cross-subtype neutralising activity. The ability to neutralise different viral subtypes, including highly pathogenic avian influenza (H5N1), correlated with the absence of hemagglutination inhibition activity, loss of binding to HA at acid pH and the absence of binding to the head domain containing the receptor binding site. This data supports their binding to epitopes in the HA stem region and a mechanism of action other than blocking viral attachment to cell surface receptors. After conversion of cross-neutralising antibodies R1a-B6 and R1a-A5 into a bivalent format, no significant enhancement in neutralisation activity was seen against A(H1N1) and A(H5N1) viruses. However, bivalent R1a-B6 showed an 18 fold enhancement in potency against A(H9N2) virus and, surprisingly, gained the ability to neutralise an A(H2N2) virus. This demonstrates that cross-neutralising antibodies, which make lower affinity interactions with the membrane proximal stem region of more divergent HA sub-types, can be optimised by bivalency so increasing their breadth of anti-viral activity. The broad neutralising activity and favourable characteristics, such as high stability, simple engineering into bivalent molecules and low cost production make these single domain antibodies attractive candidates for diagnostics and immunotherapy of pandemic influenza.

Geographic access to mammography screening centre and participation of women in the Quebec Breast Cancer Screening Programme.

BACKGROUND: This study evaluated the impact of distance between women's residences and designated screening centres (DSC) on participation in the Quebec Breast Cancer Screening Programme, whether this impact varied according to the rural-urban classification and the proportion of participants who used the DSC nearest to their home. METHODS: Travel distance between the residence of 833 856 women and the nearest DSC (n=85) was estimated. Data were obtained from administrative and screening programme databases. The analysis made use of a log-binomial regression model adjusting for age and material and social deprivation. The proportions of participants who used the DSC nearest to their residence were measured. RESULTS: Compared to women living <2.5 km from a DSC, absolute decreases of 6.3% and 9.8% in participation rate were observed for distances of 50.0 to <75.0 km (rate ratios (RR)=0.88, 95% CI 0.86 to 0.89) and ≥75.0 km (RR=0.81, 95% CI 0.79 to 0.83), respectively. The lowest participation (42%) was observed in Montreal Island. The distance at which participation started to decrease materially varied according to rural-urban classification. Participation rates decreased at distances of ≥25.0 km in the Montreal suburbs and midsize cities, at ≥12.5 km in small cities and at ≥50.0 km in rural areas (interaction p<0.0001). The proportion of participants who had their mammography at the nearest DSC decreased with increasing distance. CONCLUSIONS: Distance affects participation and this effect varies according to rural-urban classification. The lower participation in Montreal Island, where all women lived <12.5 km from a DSC, argues for a major impact of other characteristics or other dimensions of accessibility.

Access to breast cancer screening programs for women with disabilities.

BACKGROUND: The goal of this study was to identify measures to facilitate access to the Quebec Breast Cancer Screening Program for women with activity limitations, considering the barriers to screening uptake in that population. METHODS: The study was carried out in three stages. First, 124 semi-structured interviews were conducted in five regions of Quebec with five groups of key informants. The content analysis lead to the identification of 64 proposals, which were submitted to 31 experts through a two-round Delphi survey process. Finally, consultations were held with 11 resource people to determine which decision-making levels (local, regional, provincial) could play a key role in implementing the proposals. RESULTS: A strong consensus (≥80%) was achieved for 25 proposals seen as highly relevant and feasible. DISCUSSION: The implementation of such proposals could substantially improve access to screening, given the prevalence of activity limitations in the age group targeted by the program.

Reproducibility of serology assays for pandemic influenza H1N1: collaborative study to evaluate a candidate WHO International Standard.

Haemagglutination-inhibition (HI) and virus neutralisation (VN) assays are used to evaluate immunogenicity of pandemic H1N1 vaccines; however these bioassays are poorly standardised leading to inter-laboratory variation. A candidate International Standard (IS) for antibody to H1N1 pdm virus (09/194) was prepared from pooled sera of subjects who had either recovered from H1N1 pdm infection or who had been immunised with an adjuvanted subunit vaccine prepared from reassortant virus NYMC X-179A (derived from A/California/7/2009 virus). Ten laboratories from seven countries tested the candidate IS, 09/194 and a panel of human sera by HI and VN using the A/California/7/2009 virus (six laboratories) and/or the reassortant virus NYMC X-179A (ten laboratories). As expected, the inter-laboratory variability for HI and VN assay results was high. For results of antibody tests to NYMC X-179A, the % geometric coefficient of variation (%GCV) for 09/194 between laboratories was 83% for HI and 192% for VN. For tests of all sera, the median %GCV ranged from 95 to 345% for HI (80-fold variation) and 204 to 383% for VN (109-fold variation), but for the titres relative to 09/194 the median %GCV was much reduced (HI 34-231%; VN 44-214%). For tests of antibody to the A/California/7/2009 wild type virus there were similar reductions in %GCV when 09/194 was used. These results suggest that 09/194 will be of use to standardise assays of antibody to A/California/7/2009 vaccine and 09/194 has now been established by WHO as an IS for antibody to A/California/7/2009 with an assigned potency of 1300 IU per ml.

Matrix-M adjuvanted virosomal H5N1 vaccine confers protection against lethal viral challenge in a murine model.

BACKGROUND: A candidate pandemic influenza H5N1 vaccine should provide rapid and long-lasting immunity against antigenically drifted viruses. As H5N1 viruses are poorly immunogenic, this may require a combination of immune potentiating strategies. An attractive approach is combining the intrinsic immunogenicity of virosomes with another promising adjuvant to further boost the immune response. As regulatory authorities have not yet approved a surrogate correlate of protection for H5N1 vaccines, it is important to test the protective efficacy of candidate H5N1 vaccines in a viral challenge study. OBJECTIVES: This study investigated in a murine model the protective efficacy of Matrix-M adjuvanted virosomal influenza H5N1 vaccine against highly pathogenic lethal viral challenge. METHODS: Mice were vaccinated intranasally (IN) or intramuscularly (IM) with 7·5 µg and 30 µg HA of inactivated A/Vietnam/1194/2004 (H5N1) (NIBRG-14) virosomal adjuvanted vaccine formulated with or without 10 µg of Matrix-M adjuvant and challenged IN with the highly pathogenic A/Vietnam/1194/2004 (H5N1) virus. RESULTS AND CONCLUSIONS: IM vaccination provided protection irrespective of dose and the presence of Matrix-M adjuvant, whilst the IN vaccine required adjuvant to protect against the challenge. The Matrix-M adjuvanted vaccine induced a strong and cross-reactive serum antibody response indicative of seroprotection after both IM and IN administration. In addition, the IM vaccine induced the highest frequencies of influenza specific CD4+ and CD8+ T-cells. The results confirm a high potential of Matrix-M adjuvanted virosomal vaccines and support the progress of this vaccine into a phase 1 clinical trial.

The development of vaccine viruses against pandemic A(H1N1) influenza.

Wild type human influenza viruses do not usually grow well in embryonated hens' eggs, the substrate of choice for the production of inactivated influenza vaccine, and vaccine viruses need to be developed specifically for this purpose. In the event of a pandemic of influenza, vaccine viruses need to be created with utmost speed. At the onset of the current A(H1N1) pandemic in April 2009, a network of laboratories began a race against time to develop suitable candidate vaccine viruses. Two approaches were followed, the classical reassortment approach and the more recent reverse genetics approach. This report describes the development and the characteristics of current pandemic H1N1 candidate vaccine viruses.

Improved haemagglutinin antigen content in H5N1 candidate vaccine viruses with chimeric haemagglutinin molecules.

The candidate vaccine virus NIBRG-14 was derived by reverse genetics and comprises the haemagglutinin (HA) and neuraminidase (NA) genes derived from the clade 1 virus A/Viet Nam/1194/2004 on an A/Puerto Rico/8/34 (PR8) backbone. The HA gene was modified to remove the multibasic cleavage site motif associated with high pathogenicity. Reports from manufacturers, confirmed by data generated in this laboratory, have shown that this virus yields a low amount of HA antigen. We have generated a panel of new viruses using reverse genetics in which each virus consists of the PR8 backbone, the NA gene from A/Viet Nam/1194/2004 and a chimeric HA gene with sequences from both PR8 and A/Viet Nam/1194/2004. Here we show that a number of these viruses have improved HA antigen content and yield and are therefore better candidate vaccine viruses for use in production of H5N1 vaccine.

An adjuvanted pandemic influenza H1N1 vaccine provides early and long term protection in health care workers.

Mass vaccination was the most effective prophylaxis for protecting the population during the influenza H1N1 pandemic. We have evaluated the tolerability, immunogenicity and kinetics of the antibody response to a monovalent oil-in-water (AS03) adjuvanted human pandemic split influenza A/California/7/2009 H1N1 (3.75 µg haemagglutinin) vaccine in health care workers. Vaccination elicited a rapid and early protective level of haemagglutination inhibition antibody from 6 to 7 days post vaccination, and by 14 to 21 days post vaccination, up to 98% of vaccinees had protective antibody titres which persisted for at least 3 months in 84-92% of subjects. A rapid induction of protective antibody is important in reducing community spread of pandemic influenza and in helping maintain the integrity of the health care system during the pandemic.

Factors associated with death in the emergency department among children dying of complex chronic conditions: population-based study.

OBJECTIVES: To determine the percentage of deaths occurring or confirmed in an emergency department (ED) among children dying of complex chronic conditions and identify factors associated with that percentage. METHODS: The population and variables of this population-based study were derived from three administrative databases. The study focuses on all children aged 1-19 years who died of complex chronic conditions in Quebec in 1997-2001. Children not hospitalized on seventh day before death were considered at risk of ED death at that time. The percentage of ED deaths was measured in association with year of death, sociodemographic characteristics, outpatient visits, and hospitalizations in the last 6 months of life. RESULTS: Among all 506 deaths, 13.8% died in an ED. Among the 300 children not hospitalized on the seventh day before death, 21.7% had an ED death. Compared to children dying from malignancies, the adjusted odds of ED deaths were higher for those with cardiovascular conditions (odds ratio [OR] = 6.3; 95% confidence interval [CI] = 2.3-17.5), metabolic and other congenital or genetic defect (OR = 4.5; 95% CI = 1.5-13.5) and neuromuscular conditions (OR = 3.7; 95% CI = 1.5-9.4). The adjusted odds of ED deaths increased over time and were lower for children with hospitalizations in tertiary pediatric centers (OR = 0.3; 95% CI = 0.1-0.8), compared to those with no hospitalization. CONCLUSIONS: EDs play an important role in end-of-life care of children with complex chronic conditions. Multidisciplinary teams of tertiary pediatric centers may be better able to assess prognosis and provide appropriate advanced care planning.