The data used in this study are based on live birth and stillbirth registrations and notifications of birth received at the British Columbia Vital Statistics Agency (BCVSA) for births from January 1, 1981 to December 31, 2000. Infant death registration records from the Agency's death registry were linked and added to the birth records (including any infant deaths in 2001 that occurred to infants born in 2000). Links were deterministically based on birth registration number, which appears on the death record for infant deaths. In the case of infant deaths to former residents of British Columbia, inter-provincial agreements assured that the death record was available for linkage. The procedure resulted in a 98.9% linkage rate based on BCVSA infant death tables for 1981–2000 . The confidentiality of BCVSA records was protected according to approved practices .
By British Columbia law, birth weight is recorded in hospital immediately after birth; <1% of deliveries occur out of hospital.
Since the early 1980s, ultrasound assessment is routinely performed in British Columbia early in the second trimester. The main source of BCVSA gestational age data prior to 1993 was the notice of birth completed by the attending physician (containing the gestational age as recorded by the physician, which is usually based on an early ultrasound estimate ), rather than the birth registration completed by the mother, and since 1993 the notice of birth has been the sole source. Furthermore, birth weights >4 SD at each week of gestation were identified during data analysis and corrected by accessing original documents, which were found for all but 4 cases; the latter were excluded from further analysis.
Analyses were restricted to singleton live births and stillbirths between 22 and 44 completed weeks of gestation with birth weights less than 7000 grams. Records where the weight, gestational age, or other study variables had missing or unknown values were excluded from the analysis.
For the purposes of this study, all births were allocated to one of four ethnic groups. Births were designated as Chinese if both the mother and father were born in the People's Republic of China, Hong Kong, Taiwan, Vietnam, or Singapore. Births were classified as South Asian if the mother and father were both born in Bangladesh, British India Ocean Territories, Sri Lanka, India, Nepal, or Pakistan. These locations were chosen in consultation with immigrant and cultural support organisations in British Columbia. Immigrants from these locations are considered to share Chinese or South Asian habits, culture, natality customs, and (most importantly) genetic heritage.
In recognition of current preferences, we use the term "First Nations" to refer to the third ethnic category. Status Indians are identified by means of a flag in the data set; their status is officially registered with the federal government and comprise the major part of the broad group of First Nations people in British Columbia, which also includes non-Status Indians, Inuit ("Eskimo"), and Metis. The major source for the data flag was the BCVSA statistical database of information extracted from the registration of births. Additional sources were the Indian Status Verification File provided by Health Canada's First Nations and Inuit Health Branch (which originates from the Department of Indian Affairs and Northern Development) and the Status Indian Entitlement files from the British Columbia Medical Services Plan. Using an extensive computer matching process, a birth was considered to be Status Indian if the mother was identified as a Status Indian in any of the three sources . The term First Nations can be considered synonymous with "North American Indian" for most intents and purposes.
Births not included in any of the 3 ethnic groups specified above were allocated to the "Other" category and thus comprise Caucasian (primarily), mixed (mother-father), non-immigrant Chinese and South Asian, and Black ethnicities. Blacks are present in very small numbers in British Columbia and are not identified on the birth record.
The data file contained 865,968 records of singleton births including 4,456 stillbirths and 4,808 infant deaths at 22–44 weeks of gestation to residents of British Columbia. Chinese births totalled 40,092, South Asian births 38,670, and First Nations births 56,097, with the remaining 731,109 births in the "Other" group.
Gestational age-specific perinatal mortality was calculated as the number of perinatal deaths [stillbirths plus early neonatal (<7 days) deaths] at each completed week of gestational age, divided by the number of fetuses at risk at each gestation . For example, perinatal mortality at 22 weeks gestation was calculated by dividing the number of perinatal deaths at 22 weeks by the number of live births plus stillbirths at 22 or more completed weeks of gestation, i.e., fetuses who delivered at 23, 24, 25, or more weeks of gestation were also at risk of live birth or stillbirth at 22 weeks.
Gestational age-specific patterns of fetal growth restriction were estimated using an indirect method based on the fetuses-at-risk approach. The number of "revealed" (see below) small-for-gestational-age (SGA) live births was determined for each group based on a birth weight < 10th percentile for gestational age according to two different standards: (1) the current British Columbia live birth standard  and (2) an ethnic-specific live birth standard produced for each of the four ethnic groups under study by using the birth weight-for-gestational-age distributions specific to each group. "Revealed" SGA rates were then calculated by dividing the number of gestational age-specific SGA live births by the number of fetuses at risk at that gestation.
Because of the low absolute number of events (perinatal deaths, revealed SGA births) at early gestational ages for the Chinese, South Asian, and First Nations groups, we analyzed rates for these events as 2-week prospective risks. In other words, the rates were calculated as the number of events occurring during a given 2-week gestational period divided by the number of fetuses alive (and thus at risk for these events) at the beginning of that period. Neonatal deaths and stillbirths were analysed using the same method as perinatal deaths, with similar results (available on request).
We have previously shown that risks based on the number of fetuses at risk, rather than the number of total births, provides greater coherence between birth rates (and thus risks of early preterm birth), fetal growth restriction, and perinatal mortality [17, 21–23]. One important consequence of using fetuses at risk rather than live births or total births as the denominator for calculating rates of gestational age-specific pregnancy outcomes is that perinatal mortality rates (and stillbirth and early neonatal mortality rates as well) rise with advancing gestational age. This may at first seem counter-intuitive, but conventional "rates" are actually ratiosof deaths to live births or total births at a given gestational age. They are not true proportions, because the denominator does not include all subjects (unborn fetuses) at risk for the events denoted by the numerator; all living fetuses are at risk for stillbirth, live birth, and early neonatal death in the succeeding week. Neonatalogists are (appropriately) concerned with mortality among live-born births at a given gestational age, but neither the pregnant woman carrying a live fetus at a given gestational age nor her obstetrician, family physician, or midwife has any way of knowing whether or not her fetus will be born in the next week. From the woman's and her unborn fetus's perspective, the risk of stillbirth or live birth and early neonatal mortality in the succeeding week does indeed increase with advancing gestation, because the likelihood of birth (either a live birth or a stillbirth) rises as gestation advances .
Because SGA cannot be determined among unborn fetuses (i.e., those remaining in utero), and because the weight of stillbirths may underestimate the fetal weight at the (earlier) time of fetal death, we have developed a proxy measure, "revealed SGA," that provides a tip-of-the-iceberg indication of fetal growth restriction. The revealed SGA rate is the number of live-born SGA infants at a given gestational age divided by the number of fetuses at risk [17, 21–23], where SGA is defined as a birth weight below the 10th percentile birth at the given gestational age for this data set (i.e., an internal standard). Since the revealed SGA rate depends on both the birth rate and the SGA rate among live births, it is far below 10%, except in the last gestational age category (42+ weeks) when all remaining fetuses are born. It thus relates the number of live-born SGA infants to the number of fetuses at a given gestational age who were at risk for bothSGA and birth during the subsequent week.
We used two different internal standards to define revealed SGA: (1) a single standard comprising all three study groups, and (2) a group-specific standard for each of the ethnic groups. We then graphically compared the patterns of gestational age-specific rates of live birth, revealed SGA, and perinatal death among the three study groups and compared the coherence of the patterns using the single vs group-specific SGA standards.
All statistical analyses were carried out using SAS-PC version 8.2. Specialized graphic output was produced using Microsoft Excel software Version 2002. Smoothing of the charts was accomplished using a 3rd order polynomial calculated as the least squares fit through data points according to the following equation: y = b + c1x + c2x2 + c3x3 where b and c are constants. Missing birth weight and gestational age values comprised <0.25% of total births in each group and were proportionally distributed across gestational age. Chinese and South Asians had the lowest percent missing, while First Nations was only slightly higher than Others. BCVSA makes a particular effort to include birth weight and gestational age values on all records. If either measure is not recorded on the notice of birth, the source is contacted before the record is processed.