Neonatal mortality, risk factors and causes: a prospective population-based cohort study in urban Pakistan
Imtiaz Jehan a, Hillary Harris b, Sohail Salat a, Amna Zeb a, Naushaba Mobeen a, Omrana Pasha a, Elizabeth M McClure b, Janet Moore b, Linda L Wright c & Robert L Goldenberg d
a. Aga Khan University, Karachi, Pakistan.
b. Research Triangle Institute, 3040 Cornwallis Road, Durham, NC, 27709, United States of America.
c. National Institute of Child Health and Human Development, Rockville, MD, USA.
d. Drexel University, Philadelphia, PA, USA.
Correspondence to Elizabeth M McClure (e-mail: firstname.lastname@example.org).
(Submitted: 09 January 2008 – Revised version received: 21 June 2008 – Accepted: 25 June 2008 – Published online: 06 January 2009.)
Bulletin of the World Health Organization 2009;87:130-138. doi: 10.2471/BLT.08.050963
Of the estimated 130 million infants born each year worldwide,1 4 million die in the first 28 days of life. Three-quarters of neonatal deaths occur in the first week, and more than one-quarter occur in the first 24 hours.1,2 Neonatal deaths account for 40% of deaths under the age of 5 years worldwide. Therefore, efforts to achieve the UN Millennium Development Goal 4 of reducing childhood mortality by two-thirds by 2015 are focused on reducing neonatal deaths in high-mortality countries.
Two-thirds of the world’s neonatal deaths occur in just 10 countries, mostly in Asia. Pakistan is number three among these countries. With an estimated 298 000 neonatal deaths annually and a reported neonatal mortality rate of 49 per 1000 live births, Pakistan accounts for 7% of global neonatal deaths.1–5 Infection (36%), preterm birth (28%) and birth asphyxia (23%) account for 87% of neonatal deaths worldwide.1,2,6 Since causes of neonatal death vary by country and with the availability and quality of health care, understanding neonatal mortality in relation to these factors is crucial.2,7–10 Data available on neonatal deaths in Pakistan come primarily from hospital studies, which have a selective referral bias, or from communities in which the cause of death is rarely recorded. Information on pregnancy complications and other events before delivery is limited.4,5,11
Given the paucity of reliable population-based information in Pakistan, this study was undertaken to examine the prevalence, sex distribution, timing and causes of neonatal death in a population-based pregnancy cohort in urban Pakistan. We hypothesized that the neonatal mortality rate in this urban population, with relatively good access to obstetric care and Caesarean section, would be substantially lower than that generally reported for Pakistan. This study therefore examines delivery outcomes in pregnant women with reasonably good access to professional health care who were enrolled at 20 to 26 weeks’ gestation and followed with their infants to 28 days postpartum.
This prospective population-based study was conducted from September 2003 to August 2005 in four of 12 administrative units in the town of Latifabad, Hyderabad, Pakistan. These four units covered an area with a low-to-middle income population of about 90 000 individuals, or about 9000 households. Permanent residents who planned to give birth in the catchment area were screened by lady health workers (LHWs) of the Pakistan Ministry of Health. These LHWs are female community residents who have had eight or more years of education and 15 months of government training. They provide basic maternal care, including child health services, and maintain logs of all pregnancies and birth outcomes among their assigned households. In the four study units, approximately 90 LHWs were trained in the research protocols, study recruitment, communication skills and confidentiality. Study nurses supervised the LHWs in the required fieldwork. We therefore believe that our study team was aware of nearly every pregnancy in the catchment area.
During their routine home visits, LHWs provided study information to pregnant women who were screened as eligible for the study. Women were eligible if they were aged 16 years or more, did not have a serious medical condition, planned to deliver in the catchment area and were at 20–26 weeks’ gestation at enrolment. Women who indicated interest were scheduled for an appointment at the research clinic closest to their home. At the research clinic visit, gestational age was determined by ultrasound to confirm eligibility, and a physical examination and anthropometric measurements were performed. Various demographic and health data and routine antenatal laboratory test results were collected on pretested study forms by trained female research staff, which included two doctors, one dentist, two health visitors and one midwife. Prior to enrolment, all eligible women provided informed consent.
The LHWs tracked all enrolled women until delivery. Research staff also developed liaisons with public and private delivery facilities and home birth attendants to ensure complete birth reporting for the study. Once a delivery had been reported, a study physician and nurse visited the woman at home or at the health facility within 48 hours to collect maternal data on the delivery and birth outcome. A postnatal visit for data collection was made on about day 28.
Outcomes for all fetuses and neonates delivered after enrolment were clearly defined. A stillbirth was defined as any fetus born without a heartbeat, respiratory effort or movement, or any other sign of life. The stillbirth rate is the number of stillbirths per 1000 births. Neonatal death rates included all of the deaths of live-born infants on or before 28 days postpartum and the early neonatal death rate included all deaths of live-born infants occurring on or before 7 days of age. Both are expressed per 1000 live births. The perinatal death rate is the sum of neonatal deaths and stillbirths per 1000 births. In addition, a perinatal mortality-1 rate was defined as the sum of all stillbirths and neonatal deaths on or prior to day 7 (henceforth 7-day neonatal deaths) per 1000 births, and a perinatal mortality-2 rate as the sum of all stillbirths and deaths on or prior to day 28 (henceforth 28-day neonatal deaths) per 1000 births.
For all neonatal deaths and stillbirths, the study physician and nurse interviewed the mother about the circumstances leading to the event. Because most neonatal deaths occurred in hospitals, maternal reports were supplemented by a review of hospital records by the study physician. Finally, the completed study forms on mothers who experienced a neonatal death were reviewed jointly by a neonatologist (SS) and the primary author (IJ) using the Pattinson et al.12 adaptation of the Aberdeen classification13 for developing countries. Details of stillbirths in our study community and the overall methodology have been reported previously.14
The Pattinson et al. adaptation of the Aberdeen classification was used because it identifies the potential for both preventing and reducing avoidable fetal and neonatal mortality.12 The primary obstetric cause of neonatal death was defined in the classification as the obstetric antecedent factor or event that initiated the process or sequence of events leading to the death of the neonate. The classification system is non-hierarchical and allows for the identification of the following obstetric causes of neonatal death: preterm labour (< 37 weeks) or premature rupture of membranes, antepartum haemorrhage, intrapartum asphyxia, infection, intrauterine growth retardation including postmaturity, hypertension, fetal abnormality, maternal disease, trauma and unexplained intrauterine death. For the classification of the primary obstetric cause of death, intrapartum asphyxia included prolonged labour, meconium aspiration and umbilical cord compression or accident. Fetal abnormalities included chromosomal and somatic abnormalities. A single obstetric cause was assigned to each neonatal death.
In addition, the final cause of a neonatal death was also assigned according to the event that caused the death, as follows: immaturity-related, birth asphyxia or hypoxia, infection, congenital abnormality, trauma, other or unknown. In this classification, immaturity-related deaths included those due to extreme multiorgan immaturity (only in infants born less than 28 weeks’ gestation) and hyaline membrane disease or clinical respiratory distress in the absence of any other detectable cause. Death due to birth asphyxia was recorded when a normally formed term baby was unable to initiate and sustain respiration at birth or had a low Apgar score or clinical signs of hypoxia or meconium aspiration. Congenital abnormalities included chromosomal and somatic malformations.
The study was undertaken under the auspices of the Global Network for Women’s and Children’s Health Research. The study was reviewed and approved by the Aga Khan University Ethical and Review Committee in Pakistan and institutional review boards at the University of Alabama at Birmingham and Research Triangle International in the United States of America.
For this study, we had sufficient resources to study approximately 1300 pregnant women and their neonates. Since we were evaluating many different risk factors and the sample size was fixed, the following calculation was performed to determine the likelihood that a specific risk factor would have a significant association with neonatal mortality: with an expected neonatal mortality rate of 50 per 1000 live births, the expected 95% confidence interval associated with a sample size of 1300 births is ±12 neonatal deaths per 1000 live births.
Data management and analysis
All data were entered centrally. Data audits, including inter- and intra-form consistency checks, were performed at data entry, and additional audits were performed by the data centre (i.e. Research Triangle International, NC, USA). Data were analysed using SAS version 9.1.3. (SAS Institute Inc., NC, USA). For the descriptive analysis, frequencies, percentages and rates were calculated, and 95% confidence intervals (CIs) were determined for mortality rates. Relative risks (RRs) and 95% CIs were calculated to evaluate the associations between potential risk factors and neonatal death. In addition, logistic regression models were used to compute adjusted odds ratios (AORs) and their associated 95% CIs.
Between September 2003 and August 2005, LHWs identified 2205 pregnant women from the study area, 25% of whom were not eligible for study enrolment. Of the 1659 eligible, 17% either refused to participate (6%) or failed to appear at the research clinic (11%). Thus, 83% of those eligible, or 1369 women, were enrolled at 20–26 weeks’ gestation. Birth outcome data were obtained for 1280 (94%) of the enrolled women and 28-day follow-up data, for 1121 women. The demographic characteristics of the women whose neonatal outcomes were known at 28 days were not significantly different from those who were lost to follow-up or who refused to participate (P > 0.05 for age, educational level, marital status, maternal height and maternal weight).
Rates were as follows: stillbirth, 33.6 per 1000 births (95% CI: 23.6–43.6); early neonatal mortality, 34.8 per 1000 live births (95% CI: 24.1–45.5); 28-day neonatal mortality, 47.3 per 1000 live births (95% CI: 34.9–59.7); perinatal mortality-1 (i.e. stillbirths plus all early neonatal deaths), 70.4 per 1000 births (95% CI: 55.7–85.1); and perinatal mortality-2 (i.e. stillbirths plus all neonatal deaths), 82.5 per 1000 births (95% CI: 66.7–98.3) (Table 1). Of the 53 neonatal deaths, 39 (75%) occurred in the first 7 days. Although the difference was not statistically significant, early neonatal and perinatal mortality-1 rates were slightly higher among males than females (35.0 versus 29.3 per 1000 live births, P = 0.60; and 72.6 versus 56.9 per 1000 births, P = 0.29, respectively). However, the 28-day neonatal mortality rate was slightly lower among males than females (40.5 versus 48.8 per 1000 live births; P = 0.51), as shown in Table 1. The reason for this difference was that the late neonatal mortality rate (i.e. for deaths from days 8 to 28) was significantly lower for males than females (5.5 versus 19.5 per 1000 live births, P = 0.05).
Table 1. Stillbirth and neonatal and perinatal mortality rates, by gender and in total, in a prospective study in an urban Pakistani population, 2003–2005
Table 2 compares the characteristics of mothers whose infants died within the 28 days following delivery with those of mothers whose infants survived. On univariate analysis, no maternal characteristic was found to be significantly associated with neonatal death. Regression analysis confirmed the absence of a significant association in this population.
Table 2. Demographic and clinical characteristics of the mothers of infants who died within 28 days postpartum or who survived, in a prospective study in an urban Pakistani population, 2003–2005
Table 3 shows a comparison of the delivery and clinical characteristics of infants who died in the 28 days following delivery and those who survived. Factors found to be significantly associated with neonatal death in the univariate analysis include gestational age < 37 weeks (RR, 5.8; 95% CI: 3.4–9.7), birth weight < 2000 g (RR, 5.6; 95% CI: 2.3–13.6), Caesarean section (RR, 2.3; 95% CI: 1.3–4.2), meconium-stained fluid (RR, 2.7; 95% CI: 1.4–5.3), foul-smelling amniotic fluid (RR, 3.1; 95% CI: 1.5–6.2) and excessive vaginal bleeding (more than blood-stained mucus) during labour (RR, 2.3; 95% CI: 1.0–5.1). Neither gender, premature rupture of membranes, prolonged labour nor maternal fever was significantly associated with neonatal mortality. When these factors (excluding birth weight because of its collinearity with gestational age) were evaluated together using regression analysis, gestational age < 37 weeks (AOR, 5.0; 95% CI: 2.5–9.9) and Caesarean section (AOR, 2.3; 95% CI: 1.1–4.7) were the only factors that remained significant predictors of neonatal mortality. These results suggest that the other factors found to be significant in the univariate analysis may have been acting through their relationship to gestational age or birth weight or to the mode of delivery. In a subsequent univariate analysis of delivery risk factors in which gestational age, birth weight and mode of delivery were excluded, significant relationships were found between neonatal mortality and foul-smelling amniotic fluid (AOR, 2.1; 95% CI: 1.0–5.4) and excessive vaginal bleeding (AOR, 1.6; 95% CI: 1.0–4.5).
Table 3. Clinical and delivery variables for infants who died within 28 days postpartum and for those who survived in prospective study of neonatal mortality in an urban Pakistani population, 2003–2005
Table 4 presents details of the medical care received by infants who died in the 28 days following delivery and their mothers both at birth and before neonatal death. Of the 53 neonatal deaths, 45% occurred in the first 48 hours and 73% within 7 days. Eighty per cent of infants who died were born in a hospital or maternity clinic, and 69% were delivered by a doctor (data not shown). Thirty-five per cent of all neonatal deaths and 55% of those that occurred within 48 hours followed delivery by Caesarean section. Seventy-five per cent of all neonatal deaths and 91% of deaths within 48 hours occurred in the hospital. Of the 45 infants who died within 28 days and who received medical treatment, 87% were treated in a hospital and 13% in a clinic.
Table 4. Medical care received by infants who died within 28 days postpartum and their mothers, both at delivery and before neonatal death, in prospective study in an urban Pakistani population, 2003–2005
Table 5 presents the primary obstetric and final causes of neonatal death as determined using the Pattinson et al. classification. The obstetric factors associated with neonatal death were: preterm labour (34%), intrapartum asphyxia (21%), antepartum haemorrhage (9%), infection (4%), congenital abnormality (4%) and intrauterine growth retardation (2%). No obstetric cause was found in 19% of cases.
Table 5. Primary obstetric and final causes of neonatal deaths that occurred within 28 days postpartum in prospective study in an urban Pakistani population, according to the time of death, 2003–2005
Almost 75% of neonatal deaths were attributed to three final causes: immaturity-related (26%), birth asphyxia or hypoxia (26%) and infection (23%). Congenital abnormality accounted for 8%. Almost all deaths classified as due to immaturity or asphyxia occurred during the first week of life. There were no deaths classified as infection-related (i.e. involving sepsis, pneumonia or meningitis) in the first 48 hours. Although low birth weight was not considered as an independent cause of death, in 54% of neonatal deaths the infant weighed < 2500 g at birth, with 87% of these low-birth-weight infants being preterm.
To address UN Millennium Development Goal 4 on reducing childhood mortality, there is a need for better population-based data on the rates and causes of neonatal death. Our prospective population-based study provided a rare opportunity to obtain reliable information on the rate, timing and direct cause of neonatal death. Because pregnant women from a defined population were enrolled at 20 to 26 weeks’ gestation and followed with their infants to 28 days postpartum, data on antepartum history, delivery and events before neonatal death, in addition to maternal interview data, were available, so we could determine the causes of death quite reliably. Our study therefore provides information on neonatal mortality in a population with relatively good access to professional maternity and neonatal care.
The high 28-day neonatal mortality rate of 47 per 1000 live births and the high 7-day perinatal mortality rate of 70 per 1000 births observed in our study are striking, since they represent outcomes in an urban cohort in which a high proportion of births took place in a health facility assisted by skilled attendants, and a high proportion of sick neonates were cared for in the formal health-care system. Since reported neonatal mortality rates for all of Pakistan are in the range of 45–50 per 1000 live births, our hypothesis that the neonatal mortality rate in this urban population would be substantially lower than that reported for the rest of Pakistan proved incorrect. We believe there are two reasons for this finding. First, we suspect that, because of underreporting, actual Pakistani neonatal mortality rates may be higher than reported and, second, that, despite the seemingly appropriate quantity of care provided for women in the study, the quality may have been suboptimal.
Recently there has been a growing demand for perinatal mortality data to be disaggregated by gender, geographic location and socioeconomic status, to enable programmes to improve resource allocation and monitoring.8 Our study reports gender-specific neonatal mortality rates in a defined urban population. The gender differential in early and late neonatal mortality is worth noting. Proportionately, there were more male deaths in the early neonatal period, a finding consistent with the well described biological survival advantage of girls in the neonatal period. In contrast, there were more female deaths in the late neonatal period. Reduced care-seeking for girls compared with boys has been reported in several settings, especially in south Asia.15,16
The most common primary obstetric causes of neonatal death were preterm delivery in 34%, intrapartum asphyxia in 21% and antepartum haemorrhage in 9%. The relative importance of these causes is reflected in the distribution of the final causes of neonatal death: immaturity-related in 26%, birth asphyxia or hypoxia in 26% and infection in 23%. These results are consistent with WHO reports on the causes of neonatal death in developing countries and also with other reports from Pakistan.2,4,5 Furthermore, our finding that infection, including sepsis, pneumonia and meningitis, is an important contributor to neonatal deaths that occur after 3 days postpartum among hospital-born neonates is consistent with recent studies from developing countries and emphasizes the importance of monitoring delivery and hospital-acquired infection.17
It could be argued that the high neonatal mortality seen in this population may be due to selective recruitment of high-risk women into the study. However, women were enrolled into the study by LHWs during routine home visits, and a substantial majority of all eligible women were enrolled. Study nurses cross-checked enrolment and LHW delivery logs to ensure that women enrolled prenatally and those for whom we had delivery results were the same women. In addition, the baseline characteristics of the study population are typical for the area. We therefore believe that the risk of biased reporting of neonatal mortality for this population is small.
We were surprised by the high Caesarean section rate of 19% in this community. However, in comparison with rural areas of Pakistan, where home birth and a very low Caesarean section rate are the norm, there is now a growing trend in urban areas towards hospital birth and associated Caesarean delivery.18,19 We therefore believe that the high Caesarean section rate seen in this urban population is closer to the norm than is commonly realized. The UN recommends a Caesarean section rate of 5–15% to optimally minimize maternal and neonatal mortality rates.4,20 This recommendation presumes that these Caesarean sections are performed in a timely manner on appropriate women. Evidence from other areas of Pakistan suggests that this may not be the case.21
Recent reports from developing countries have shown improvements in perinatal and neonatal outcomes with increased coverage by health services and skilled birth attendants.2,8,22 However, our findings suggest that women experienced avoidable antenatal and obstetric complications despite good availability of antenatal care, a high rate of births in a health facility with skilled birth attendants, and a “high” Caesarean section rate. Intrapartum events such as heavy vaginal bleeding and the presence of foul-smelling amniotic fluid were significant risk factors for early neonatal death. Deaths associated with these factors may be modifiable through effective antenatal and intrapartum care. Furthermore, while the majority of sick neonates received some formal health care before death, we suspect that higher quality neonatal care would also result in reduced mortality in these cases.
Without improved quality, increased health-care coverage is unlikely to substantially improve perinatal and neonatal outcomes.2,8,22,23 Recent reports from Pakistan and other low-resource settings indicate that substandard care, inadequate training, low staff competence and a lack of resources, including equipment and medication, are all factors that contribute to neonatal deaths.23,24 Since the quantity of care in this study’s setting was relatively high for a developing country, it is likely that the quality of care will need to be improved if the neonatal mortality rate is to be substantially reduced. We therefore speculate that improvements in health-system performance and ongoing clinical audits25 could reduce the high rates of adverse perinatal outcomes in a population such as ours with a high level of health-care coverage. ■
Funding: The study was supported through a cooperative agreement with the Bill and Melinda Gates Foundation and the US National Institute of Child Health and Human Development (NICHD) Global Network for Women’s and Children’s Health Research (U01HD040607 and U0104063606).
Competing interests: None declared.
- World health report 2005: Make every mother and child count. Geneva: WHO; 2005.
- Lawn JE, Cousens S, Zupan J. 4 million neonatal deaths: When? Where? Why? Lancet 2005; 365: 891-900 doi: 10.1016/S0140-6736(05)71048-5 pmid: 15752534.
- Neonatal and perinatal mortality: country, regional and global estimates. Geneva: WHO; 2006.
- Bhutta ZA. Maternal and child health in Pakistan: challenges and opportunities. Oxford University Press; 2004.
- Jalil F. Perinatal health in Pakistan: a review of the current situation. Acta Paediatr 2004; 93: 1273-9 doi: 10.1080/08035250410017022 pmid: 15499944.
- Lawn JE, Cousens SN, Wilczynska K. Estimating the causes of four million neonatal deaths in the year 2000: statistical annex. In: The world health report 2005. Geneva: WHO; 2005.
- Darmstadt GL, Bhutta ZA, Cousens S, Adam T, Walker N, de Bernis L, et al., et al. Evidence-based, cost-effective interventions: how many newborn babies can we save? Lancet 2005; 365: 977-88 doi: 10.1016/S0140-6736(05)71088-6 pmid: 15767001.
- Martines J, Paul V, Bhutta ZA, Koblinsky M, Saucat A, Walker N, et al., et al. Neonatal survival: a call for action. Lancet 2005; 365: 1189-97 doi: 10.1016/S0140-6736(05)71882-1 pmid: 15794974.
- Kulmala T, Vaahtera M, Ndekha M, Koivisto AM, Cullinan T, Salin ML, et al., et al. The importance of preterm birth for peri and neonatal mortality in rural Malawi. Paediatr Perinat Epidemiol 2000; 14: 219-26 doi: 10.1046/j.1365-3016.2000.00270.x pmid: 10949213.
- Zupan J. Perinatal mortality in developing countries. N Engl J Med 2005; 352: 2047-8 doi: 10.1056/NEJMp058032 pmid: 15901857.
- Fikree FF, Azam SI, Berendes HW. Time to focus child survival programmes on the newborn: assessment of levels and causes of infant mortality in rural Pakistan. Bull World Health Organ 2002; 80: 271-6 pmid: 12075362.
- Pattinson RC, De Jong G, Theron GB. Primary causes of total perinatally related wastage at Tygerberg Hospital. S Afr Med J 1989; 75: 50-4 pmid: 2643836.
- Baird D, Walker JR, Thomsan AM. The causes and prevention of stillbirths and first week deaths. J Obstet Gynaecol Br Emp 1954; 61: 433-48 pmid: 13192514.
- Jehan I, McClure EM, Salat S, Rizvi S, Pasha O, Harris H, et al. Stillbirths in an urban community in Pakistan. Am J Obstet Gynecol 2007;197:257.e1-257.e8.
- Bonde JP, Wilcox A. Ratio of boys to girls at birth. BMJ 2007; 334: 486-7 doi: 10.1136/bmj.39141.622917.80 pmid: 17347191.
- Ulizzi L, Zonta LA. Sex differential patterns in perinatal deaths in Italy. Hum Biol 2002; 74: 879-88 doi: 10.1353/hub.2003.0012 pmid: 12617496.
- Zaidi AKM, Huskins WC, Thaver D, Bhutta ZA, Abbas Z, Goldman DA. Hospital-acquired neonatal infections in developing countries. Lancet 2005; 365: 1175-88 doi: 10.1016/S0140-6736(05)71881-X pmid: 15794973.
- Qazi GR, Akhtar S. Obstetrical correlates of the first time cesarean section, compared with the repeated cesarean section. J Coll Physicians Surg Pak 2007; 17: 611-4 pmid: 17999852.
- CORONIS Trial Collaborative Group. The CORONIS Trial. International study of caesarean section surgical techniques: a randomised fractional, factorial trial. BMC Pregnancy Childbirth 2007; 7: 24- doi: 10.1186/1471-2393-7-24 pmid: 18336721.
- Betrán AP, Merialdi M, Lauer JA, Bing-Shun W, Thomas J, Van Look P, et al., et al. Rates of caesarean section: analysis of global, regional and national estimates. Paediatr Perinat Epidemiol 2007; 21: 98-113 doi: 10.1111/j.1365-3016.2007.00786.x pmid: 17302638.
- Fikree FF, Mir A, Haq IU. She may reach a facility but will still die! An analysis of quality of public sector maternal health services, District Multan, Pakistan. J Pak Med Assoc 2006; 56: 156-63 pmid: 16711335.
- Zaidi S. The role of obstetrician in reducing perinatal mortality. In: Zaidi S, ed. Maternal and perinatal health. Karachi: Twel Publisher; 1992. pp. 115-20.
- Hasan IJ, Khanum A. Health care utilization during terminal child illness in squatter settlements of Karachi. J Pak Med Assoc 2000; 50: 405-9 pmid: 11191439.
- Korejo R, Bhutta S, Noorani KJ, Bhutta ZA. An audit and trends of perinatal mortality at the Jinnah Postgraduate Medical Centre, Karachi. J Pak Med Assoc 2007; 57: 168-72 pmid: 17489521.
- Baker R, Sullivan E, Camosso-Stefinovic J, Rashid A, Farooqi A, Blackledge H, et al., et al. Making use of mortality data to improve quality and safety in general practice: a review of current approaches. Qual Saf Health Care 2007; 16: 84-9 doi: 10.1136/qshc.2006.019885 pmid: 17403750.