Population-based birth weight reference percentiles for Chinese twins

Abstract

Background: Birth weight percentiles by gestational age are important for assessing prenatal growth and predicting postnatal outcomes of newborns. Several countries have developed nation-specific birth weight references for twins, but China still lacks such references.

Methods and results: Birth weight data for twins born between October 2006 and September 2015 were abstracted from the China National Population-based Birth Defects Surveillance System. A total of 54,786 live twin births aged ≥28 weeks of gestation without birth defects were included in the analysis. The LMS method was adopted to generate gestational age-specific birth weight percentiles and curves for male and female twins separately. Significant differences were observed between the current reference and other references developed for Chinese or non-Chinese twins. The neonatal mortality rate in this cohort was 12.3‰, and much higher rates at very early gestation weeks were identified in small-for-gestational-age twins grouped by the newly developed reference cutoffs.

Conclusions: The established birth weight centiles represent the first birth weight norm for contemporary Chinese twins and can be a useful tool to assess growth of twins in clinical and research settings.

Key Messages

• There have been no population-based birth weight percentiles for Chinese twins prior to this study. The established birth weight centiles for female and male twins are markedly lower than those for Chinese singletons. Twin-specific curves should be used for determining inappropriate for gestational age in twins rather than using existing singleton reference.

• The birth weight percentiles for twins differed significantly from those for non-Chinese twins. In addition to ethnic influences, the observed differences could be ascribed to variations in prenatal care, fetal or maternal nutrition status or other environmental factors.

• Neonatal mortality rates varied considerably among twins grouped by the newly developed reference percentiles. Small-for-gestational-age twins had much higher mortality than did appropriate-for-gestational-age twins, highlighting the need to reduce postnatal mortality by improving perinatal health care for twins.

Keywords:

• Birth weight
• reference percentile
• Chinese
• twins

Introduction

Birth weight remains the most commonly used indicator of fetal growth (1,2). Newborns can be classified as small, appropriate, and large for gestational age according to their birth weights, birth lengths, and/or head circumference. A weight below the 10th percentile for the gestational age is usually defined as small, while weight above the 90th percentile is referred as large (1). Both small and large for gestational ages are associated with an increased risk of perinatal and infant mortality and morbidity, and with some long-term problems (3–5). Therefore, correct classification of birth weights by using right references is very important in both clinic and research works for a more accurate risk prediction and more effective postnatal health care (1,2).

Recently, a number of nation-specific birth weight percentiles for gestational age have been developed. Most of them are created for singletons, a few for twins. Twins’ intrauterine growth pattern diverges from singletons’ in the third trimester (6–8), and varies across racial groups (6,9). Twin pregnancies are more prone to perinatal and infant mortality than singletons (10–13), which can be mainly attributed to the high proportion of preterm delivery, low birth weight, small for gestational age, and intrauterine growth restriction (6,14). Increasing evidence has indicated the need of establishing the plurality-specific birth weight norm in the growth evaluation of newborns (3,8).

Twins account for 2–4% of all births worldwide and the number continues to increase in developed countries (15) as well as in some developing countries including China (16). Due to their contribution to postnatal adverse outcomes, twin pregnancies and related issues have attracted more attention recently. Several countries, including Australia (17), Argentina (18), Canada (19), Finland (8), German (20), Japan (21), Norway (22), and United States of America (7), have developed population-based twin-specific birth weight references in order to accurately evaluate twin growth. The population in China accounts for one-fifth of the world population (23), but there were no birth weight references for Chinese twins. Using data from the National Population-based Birth Defects Surveillance System (23,24), we constructed a national reference of gestational age-specific birth weight percentiles for Chinese twins born between 2006 and 2015.

Methods and materials

The data on twin births between October 2006 and September 2015 were obtained from the nation-wide birth defects surveillance system that covers 64 counties and districts in 30 provinces or municipalities in mainland China (23,24). In each province, one county in rural area and one district in a city were chosen as surveillance sites except for in Sichuan province (two rural sites) and in Tianjin city (six urban districts). Since the surveillance system did not collect data on fetuses or infants less than 28 weeks of gestation, live births and stillbirths of at least 28 weeks gestation who were born to women living in the defined surveillance areas for one year or more were recruited and followed till 42 d after birth. Data collected by surveillance staff include demographic features, birth date, sex, gravidity, parity, gestational age, birth weight, outcomes, and the identified birth defects (major malformations and chromosomal aberrations diagnosed within 42 d after birth and coded according to the International Classification of Diseases 10th edition). This surveillance database represented a wide array of geographical locations and socioeconomic status. More details on data collection and quality control in the surveillance system were described previously (23,24).

The gestational age was calculated as the number of completed weeks from the first day of the last menstrual period to the day of birth. In the surveillance areas, women with suspected pregnancy had an ultrasound examination for confirmation in the first trimester according to obstetric clinical guidelines. For women with irregular menses and those who could not remember the last menstrual date, gestational ages were estimated based on sonography examination. The birth weight of each neonate was measured and recorded to the nearest 5 g by a trained midwife within 1 h after birth. All the information of maternal and infant characteristics was included in hospital delivery records and abstracted by surveillance staff using uniformed datasheet (23,24). From October 2006 to September 2015, a total of 28,670 pairs of twins (57,340 individuals) were identified among 2,981,630 newborns by the surveillance system, of which male–male, male–female, and female–female pairs accounted for 38.2% (10,949), 34.9% (10,018), and 26.9% (7703), respectively. Infants of foreign origin (n = 14), stillbirths (n = 773, 1.35‰), and infants affected by congenital anomalies (n = 1630, 2.84%, including100 stillbirths), were first excluded from the analysis. Among the rest of 55,023 records, 126 records (0.23%) were subsequently excluded due to missing gestational age or birth weight. Finally, Tukey’s methodology (25) was adopted to screen implausible combinations of gestational age and birth weight in infants of ≤34 weeks gestation. Birth weights for each gestational age and sex combinations that were beyond the range of median ±2 × (interquartile range) were considered as outliers and were excluded. Meanwhile, the obviously erroneous records with gestational age above 34 weeks were eliminated by using pediatrician's professional judgment. Following this procedure, 111 records (0.20%, including 108 outliers and three errors) were removed from downstream analysis, yielding a final sample size of 54,786 for this study.

According to the geographical location of the surveillance sites, all the subjects were divided into three groups (costal, inner land, and remote areas). Infants born to women who lived in urban districts were classified into urban group, while newborns whose mothers lived in rural counties fell into rural group (24). Univariate analysis was used to examine the birth weight distributions and to calculate the interquartile range for each gestational age and sex. Standard Chi-square test was adopted to compare differences in categorical variables between different groups, while linear Chi-square test was used to examine time trends in preterm births, low birth weight, and adverse outcomes. The LMS method, which adopts a Box–Cox transformation based semiparametric technique and solves penalized likelihood equation, was utilized to generate birth weight centiles (26). The analysis was carried out by using R package VGAM (27).

Results

Table 1 shows the maternal and infant characteristics of the study subjects. The final sample for analysis included 51.5% male and 48.5% female births. Urban and rural births accounted for 54.5% and 45.5% of the cohort, respectively; while newborns whose mothers lived in coastal regions, inland, and remote areas accounted for 50.6%, 26.4%, and 23.1% of all births, respectively. The vast majority (92.8%) of the mothers were Han Chinese, and the rest (7.2%) were minorities. Most (87.1%) mothers aged 20–34 years at the time of delivery, with few (1.1%) aged ≤20 years and aged ≥35 years (11.8%). More than 70% of infants were born to primiparous women. Both maternal (age, ethnicity, parity, birth area, and geographic region) and infant (gestational age and sex) characteristics were associated with birth weight (Supplementary Table 1).

Table 1. Maternal and infant characteristics of live twin births in this study.

	Boys		Girls		Total
Groups	Number	%	Number	%	Number	%
Birth area
Urban	15,237	54.0	14,594	54.9	29,831	54.5
Rural	12,965	46.0	11,990	45.1	24,955	45.5
Geographic region
Coastal	14,204	50.4	13,506	50.8	27,710	50.6
Inner land	7554	26.8	6891	25.9	14,445	26.4
Remote	6444	22.8	6187	23.3	12,631	23.1
Maternal age (years)^a
<20	307	1.1	302	1.1	609	1.1
20–24	5448	19.3	5169	19.5	10,617	19.4
25–29	10,917	38.7	10,358	39.0	21,275	38.9
30–34	8197	29.1	7583	28.5	15,780	28.8
35–39	2872	10.2	2709	10.2	5581	10.2
≥40	443	1.6	442	1.7	885	1.6
Maternal ethnicity
Han	26,149	92.7	24,673	92.8	50,822	92.8
Minorities	2053	7.3	1911	7.2	3964	7.2
Parity
Nulliparous	20,082	71.2	19,160	72.1	39,242	71.6
Parous	8120	28.8	7424	27.9	15,544	28.4
Gestational age (weeks)
28–32	1235	4.4	1140	4.3	2375	4.3
33–36	8850	31.4	8223	30.9	17,073	31.2
≥37	18,117	64.2	17,221	64.8	35,338	64.5
Birth weight (g)^b
<1000	58	0.2	66	0.2	124	0.2
1000–1499	603	2.1	711	2.7	1314	2.4
1500–2499	10,241	36.3	11,700	44.0	21,941	40.0
≥2500	17,300	61.3	14,107	53.1	31,407	57.3
Postnatal outcomes^c
Neonatal death	337	1.2	339	1.3	676	1.2
Mortality within 42 d	356	1.3	353	1.3	709	1.3
Survived ≥42 d	27,846	98.7	26,231	98.7	54,077	98.7

^a Thirty infants with unknown maternal age.

^b Birth weight differed significantly between sex groups (χ²=383.9, p < 0.001).

^c Survival status within the first 42 days of life.

The overall percentage of low birth weight was 42.7%, and females tended to have a higher rate than males (46.9% versus 38.7%). Preterm births accounted for 35.5% of total twins. The annual preterm delivery rates varied between 28.8% and 39.7%, and presented an upward trend. About 1.3% of twins died within 42 days after birth, whereas the postnatal mortality rates declined from 21.7‰ in 2008 to 8.4‰ in 2015 (Table 1 and Figure 1).

Figure 1. Time trends in the rates of preterm births, low birth weight, and infant deaths within 42 d after birth during the study period (statistical year was defined as the period from the last October to this September).

By using LMS method, the smooth-estimated percentile values were generated for male and female twins, respectively (Table 2 and LMS parameters were presented in Supplementary Table 2) and smoothed growth curves for the 1st, 3rd, 5th, 10th, 25th, 50th, 75th, 90th, 95th, 97th, and 99th percentiles were constructed (Figure 2). The corrected median birth weights for boys at 28–42 weeks were 1.1–6.0% heavier than for girls. The sex differences in the 10th and 90th percentiles varied between 1.2–6.7% and 0.9–6.1%, respectively. Greater sex differences were observed in the birth weight percentiles for preterm births. For example, the differences could be as great as 8.3%, 7.5%, 7.2%, and 6.7% in the 1st, 3rd, 5th, and 10th percentiles at very early gestations (less than 33 gestation weeks), respectively.

Figure 2. Percentile charts for Chinese newborn twins. P1, P3 to P99 denote the 1st, 3rd to 99th percentile curves, respectively.

Table 2. Smoothed birth weight percentiles for Chinese twin boys and girls.

			Boys													Girls
Gestation (weeks)	Number	Mean ± SD	P1	P3	P5	P10	P25	P50	P90	P75	P95	P97	P99	Number	Mean ± SD	P1	P3	P5	P10	P25	P50	P75	P90	P95	P97	P99
28	154	1206 ± 279	577	697	759	853	1008	1176	1341	1487	1574	1630	1735	131	1146 ± 220	555	665	722	809	953	1110	1264	1402	1484	1537	1636
29	138	1310 ± 189	698	823	888	987	1149	1326	1500	1654	1745	1804	1915	118	1229 ± 245	654	773	835	929	1085	1255	1423	1573	1661	1719	1827
30	221	1468 ± 279	823	954	1022	1125	1296	1482	1664	1827	1923	1985	2102	201	1420 ± 267	762	888	954	1055	1222	1404	1584	1744	1839	1901	2016
31	236	1618 ± 287	953	1090	1161	1270	1449	1645	1838	2009	2110	2176	2299	263	1514 ± 302	880	1014	1083	1190	1366	1559	1749	1919	2019	2085	2207
32	486	1796 ± 313	1089	1232	1307	1422	1611	1817	2020	2201	2308	2378	2508	427	1717 ± 314	1013	1152	1225	1336	1520	1722	1921	2099	2204	2272	2401
33	704	2005 ± 303	1229	1380	1459	1580	1778	1996	2211	2401	2515	2588	2725	716	1875 ± 290	1157	1301	1377	1492	1683	1893	2101	2286	2396	2467	2601
34	1332	2149 ± 340	1370	1528	1612	1739	1948	2177	2404	2605	2724	2801	2946	1165	2054 ± 313	1301	1452	1531	1652	1852	2072	2289	2483	2599	2673	2813
35	2174	2348 ± 361	1514	1679	1766	1898	2116	2355	2591	2801	2926	3006	3158	2083	2240 ± 350	1444	1601	1684	1810	2020	2250	2478	2681	2802	2880	3027
36	4640	2507 ± 367	1656	1825	1913	2048	2271	2516	2757	2972	3100	3182	3337	4259	2408 ± 359	1586	1747	1831	1961	2176	2412	2645	2854	2978	3058	3209
37	7651	2655 ± 357	1770	1940	2030	2167	2393	2641	2886	3104	3233	3317	3475	7147	2552 ± 342	1703	1866	1952	2083	2300	2539	2776	2987	3113	3194	3348
38	5412	2718 ± 381	1816	1992	2084	2226	2460	2716	2969	3194	3328	3414	3577	5187	2625 ± 370	1753	1922	2011	2147	2373	2621	2867	3086	3217	3301	3460
39	3027	2732 ± 395	1809	1994	2091	2239	2483	2751	3016	3251	3391	3481	3651	2860	2643 ± 389	1755	1933	2027	2170	2408	2669	2927	3158	3295	3384	3551
40	1865	2739 ± 415	1773	1969	2072	2228	2486	2769	3048	3297	3444	3540	3719	1832	2683 ± 405	1733	1922	2022	2174	2426	2703	2977	3222	3367	3462	3639
41	125	2747 ± 537	1723	1932	2041	2208	2483	2784	3080	3344	3501	3602	3792	157	2730 ± 476	1697	1900	2006	2170	2439	2736	3029	3290	3445	3546	3735
42	37	2688 ± 396	1668	1891	2007	2185	2478	2797	3112	3392	3559	3666	3867	38	2649 ± 514	1651	1870	1985	2160	2450	2767	3081	3361	3528	3635	3838

When comparing with the latest charts for Chinese singletons (24), the current 10th and 50th percentiles for twins were significantly smaller at almost all gestational weeks except at very early gestations (Table 3). The differences in the 10th percentiles could be reached 36.9% and 32.6% for term males and females, respectively, while the differences in median birth weights for term infants varied between 14.3% and 26.1% (Table 3).

Table 3. Relative differences at 10th and 50th percentile between the current charts and previously published references.

	China singleton		Hu 2015, Taiwan^a		Li 2015, Australia		Voigt 2014, German		Ooki 2003, Japan^b				Min 2000, USA
Gestation (weeks)	P10	P50	P10	P50	P10	P50	P10	P50	P10^c	P10^d	P50^c	P50^d	P10	P50
Boys
28	−0.2	−2.5	2.2	−2.7	−15.6	−1.4	−0.4	−2.2	–	–	–	–	8.4	−1.4
29	1.9	−1.5	1.2	−2.0	−23.3	0.1	−0.7	−1.2	–	–	–	–	–	–
30	3.3	−0.9	0.7	−1.5	−17.4	−0.3	−0.5	−0.8	–	–	–	–	3.5	−3.3
31	4.3	−0.6	0.4	−1.3	−11.8	−0.8	0.0	−0.3	–	–	–	–	–	–
32	5.1	−0.2	0.0	−1.7	−17.0	−1.0	0.6	0.2	−5.6	4.7	−4.6	−2.0	−0.2	−4.4
33	6.0	0.7	−0.7	−2.4	−10.7	0.7	1.3	0.2	−7.8	1.7	−8.5	−2.8	–	–
34	7.2	2.5	−1.3	−3.2	−12.0	0.1	2.4	0.6	−8.8	−0.4	−10.3	−3.2	−3.4	−5.7
35	8.9	5.3	−2.0	−4.0	−7.5	1.0	2.7	0.6	−9.3	−1.9	−10.8	−3.4	–	–
36	12.1	9.5	−2.5	−4.2	−6.7	3.2	3.0	1.0	−9.1	−2.9	−10.2	−3.2	−3.3	−4.4
37	16.9	14.3	−1.9	−3.3	−3.1	4.7	4.5	1.9	−8.0	−2.5	−8.8	−2.2	–	–
38	23.1	18.4	0.6	−0.9	0.6	6.6	6.9	4.2	−5.1	−0.1	−6.5	−0.2	−0.1	−0.6
39	27.4	20.7	4.5	2.7	0.9	8.3	9.9	7.2	−1.2	3.6	−4.2	2.1	–	–
40	30.6	22.2	8.5	6.7	2.1	9.8	13.1	10.9	2.5	7.6	−3.1	3.9	4.7	2.9
41	33.5	24.0	13.3	15.5	–	8.3	–	–	5.1	10.9	−4.1	4.4	–	–
42	36.9	26.1	–	–	–	6.5	–	–	–	–	–	–	–	–
Girls
28	−10.6	−3.3	1.9	−2.0	7.6	0.2	0.1	−1.8	–	–	–	–	8.7	1.2
29	−3.7	−1.3	1.2	−2.1	0.1	−1.6	−1.0	−1.2	–	–	–	–	–	–
30	1.2	0.1	1.0	−2.0	−4.3	−2.1	−0.5	−1.0	–	–	–	–	3.6	−1.2
31	4.4	1.3	0.8	−2.1	3.1	0.6	0.0	−0.6	–	–	–	–	–	–
32	6.2	2.2	0.2	−2.5	2.2	0.5	1.0	−0.1	4.9	23.7	7.6	9.8	−0.1	−2.4
33	7.3	3.4	−0.7	−3.3	2.5	0.4	1.9	0.4	3.0	11.4	0.4	2.9	–	–
34	8.8	5.0	−1.6	−4.2	2.0	0.9	2.3	0.4	0.4	3.7	−4.1	−1.4	−3.6	−3.4
35	11.1	7.4	−2.3	−5.1	2.2	0.9	2.7	0.5	−2.3	−0.9	−6.8	−3.9	–	–
36	14.0	11.1	−2.7	−5.2	3.8	2.0	3.0	0.8	−4.7	−3.3	−7.9	−4.8	−4.0	−4.0
37	17.6	15.4	−2.0	−4.1	6.8	4.4	4.3	2.0	−5.9	−3.6	−7.4	−4.2	–	–
38	23.5	18.8	0.9	−1.4	9.4	5.5	6.6	3.8	−5.0	−1.2	−5.6	−2.3	−3.2	−0.2
39	27.4	20.6	5.0	2.4	11.5	7.7	9.2	6.8	−2.8	2.5	−3.3	0.1	–	–
40	29.9	21.4	9.2	6.5	8.9	11.0	11.3	9.1	−0.1	6.2	−1.6	1.7	7.9	2.8
41	31.8	22.3	15.6	16.3	–	7.2	–	−	2.7	8.8	−1.4	1.8	–	–
42	32.6	23.4	–	–	–	5.2	–	–	–	–	–	–	–	–

The relative differences were calculated according to the formula, difference = (other − current)/current ×100.

The symbol “−” indicates that the difference could not be calculated.

^a All infants ≥41 gestational weeks were combined into the group of week 41.

^b Infants of gestational age between 27 and 32 weeks were combined into the group of week 32.

^c Percentiles of infants born to primiparous women.

^d Percentiles of infants born to multiparous women.

There were significant differences between the current curves and previously published charts for twins. Moderate differences were found between ours and those based on samples from Taiwan (28) and the United States of America (7), whereas greater differences were observed as compared with Australian (17), German (20), and Japanese references (21). Interestingly, the 10th percentile values for Australian preterm male twins were lower (17), but such values for female twins were higher than current centile values at almost all gestation weeks except at 30 weeks of gestation. Taking our percentiles as reference, the degree of differences could be as great as 23.3% for 10th percentiles and 11.0% for medium birth weight (Table 3).

Finally, we analyzed the postnatal outcomes in twin groups categorized by current reference centiles. As shown in Table 4, the overall neonatal and postnatal mortality rates in the surveillance period (from birth to 42 d) were 12.3‰ and 12.9‰, nearly six-fold greater than the counterparts in non-malformed singleton cohorts (2.1‰ and 2.2‰) in the same areas. The neonatal and postnatal mortality rates of preterm twins (30.7‰ and 32.1‰) were considerably higher than those of term twins (2.2‰ and 2.4‰), but slightly lower than the rates of preterm singletons (33.9‰ and 34.8‰). Notably, the gestational-age specific mortality rates of very early preterm twins (<32 weeks gestation) were far greater than 100‰. Almost at all gestational weeks the mortality rates in small-for-gestational-age groups were higher than rates in large- and appropriate-for-gestational-age groups except for at 28 weeks’ gestation. The overall neonatal and postnatal mortality rates of small-for-gestational-age twins can be as high as 33.1‰ and 33.7‰, respectively (Table 4).

Table 4. Neonatal mortality and infant mortality within first 42 d of life by birth weight percentiles for gestational age.

	Small for gestational age			Large for gestational age			Appropriate for gestational age			Total
Gestation (weeks)	Number	Neonatal mortality [n (‰)]	Postnatal Mortality [n (‰)]	Number	Neonatal mortality [n (‰)]	Postnatal Mortality [n (‰)]	Number	Neonatal mortality [n (‰)]	Postnatal Mortality [n (‰)]	Neonatal mortality [n (‰)]	Postnatal Mortality [n (‰)]
28	13	6 (461.5)	6 (461.5)	40	19 (475.0)	19 (475.0)	232	112 (482.8)	116 (500.0)	137 (480.7)	141 (494.7)
29	14	11 (785.7)	11 (785.7)	14	4 (285.7)	4 (285.7)	228	76 (333.3)	79 (346.5)	91 (355.5)	94 (367.2)
30	47	22 (468.1)	22 (468.1)	43	4 (93.0)	5 (116.3)	332	45 (135.5)	50 (150.6)	71 (168.2)	77 (182.5)
31	59	16 (271.2)	17 (288.1)	37	5 (135.1)	5 (135.1)	403	47 (116.6)	50 (124.1)	68 (136.3)	72 (144.3)
32	92	21 (228.3)	21 (228.3)	86	3 (34.9)	3 (34.9)	735	38 (51.7)	39 (53.1)	62 (67.9)	63 (69.0)
33	106	8 (75.5)	8 (75.5)	112	2 (17.9)	2 (17.9)	1202	18 (15.0)	19 (15.8)	28 (19.7)	29 (20.4)
34	256	21 (82.0)	23 (89.8)	208	7 (33.7)	7 (33.7)	2033	36 (17.7)	37 (18.2)	64 (25.6)	67 (26.8)
35	401	11 (27.4)	11 (27.4)	381	2 (5.2)	2 (5.2)	3475	22 (6.3)	22 (6.3)	35 (8.2)	35 (8.2)
36	873	19 (21.8)	19 (21.8)	841	–	–	7185	23 (3.2)	27 (3.8)	42 (4.7)	46 (5.2)
37	1117	11 (9.8)	11 (9.8)	1412	1 (0.7)	1 (0.7)	12,269	23 (1.9)	24 (2.0)	35 (2.4)	36 (2.4)
38	937	8 (8.5)	8 (8.5)	1136	1 (0.9)	1 (0.9)	8526	13 (1.5)	13 (1.5)	22 (2.1)	22 (2.1)
39	547	4 (7.3)	4 (7.3)	495	1 (2.0)	2 (4.0)	4845	7 (1.4)	9 (1.9)	12 (2.0)	15 (2.5)
40	364	1 (2.7)	1 (2.7)	322	–	1 (3.1)	3011	5 (1.7)	7 (2.3)	6 (1.6)	9 (2.4)
41	31	1 (32.3)	1 (32.3)	31	–	–	220	1 (4.5)	1 (4.5)	2 (7.1)	2 (7.1)
42	9	1 (111.1)	1 (111.1)	5	–	–	61	–	–	1 (13.3)	1 (13.3)
Total	4866	161 (33.1)	164 (33.7)	5163	49 (9.5)	52 (10.1)	44,757	466 (10.4)	493 (11.0)	676 (12.3)	709 (12.9)

Note: Mortality rate was defined as the number of deaths per 1000 live births. During the study period, neonatal and postnatal mortality rates in live non-malformed singletons (2,863,340 individuals) were 33.9 and 34.8 per 1000 for preterm births, 1.2 and 1.3 per 1000 for term infants, and 2.1 and 2.2 per 1000 for all singletons, respectively.

"–" No infant death was reported during the study period, therefore the corresponding mortality rate was not calculated.

Discussion

Population-based birth weight references by sex and gestational age are essential for prenatal growth evaluation, postnatal care, and the risk prediction of long-term outcomes. However, there were no published percentiles for twins in China. For the first time, we developed a population-based, gestational age-specific birth weight reference for Chinese twins based on a large and nationally representative database, which could be of great value in evaluating contemporary Chinese twins.

Due to the lack of appropriate reference tools, birth weight percentiles for singletons are commonly used in clinical practice in China (29), which might cause a large portion of misclassification in twin birth weights. The current medium birth weight and 10th percentiles are smaller for term and moderate preterm births as compared to the newly updated Chinese singleton reference which is also based on data from the surveillance system (24) (Table 3). From around 30 gestational weeks onwards, birth weight of both twin boys and girls diverged progressively from singletons in similar pattern as observed in previous studies (6–8,22). For instance, the relative differences in median weights at 30 weeks’ gestation between singletons and twins were 0.9% and 0.1% for male and female newborns, respectively. Notably, the corresponding differences in the 50th percentiles at 37 weeks’ gestation increased to 14.3% and 15.4% for males and females, respectively. The large divergences clearly indicate that twin-specific curves should be used for determining inappropriate for gestational age of twins rather than using existing singleton reference.

Consistent with previous studies, this research found that birth weight patterns of twins differed significantly among populations. The smoothed median birth weights for gestational age of Chinese term twins were smaller, but the median weights of preterm twins were larger than those counterparts in Argentina (18), Japan (21), Taiwan (28), and the United States (7). For twins born between 33 and 42 gestational weeks, the current median weights are markedly lighter than those of non-Asian twins such as Australian (17), Canadian (19), Finnish (8), German (20), and Norwegian (22). Remarkably, the 10th percentiles at 33 gestation weeks onwards were consistently smaller than such references for Finnish and German twins (8,20), whereas the 10th centiles at gestation <37 weeks were larger than those for Argentinians (18) and Japanese twins (21), and for Australian twin boys (17). These findings suggest that racial disparities are apparent in twins as in singletons although the underlying mechanisms are still unknown. In addition to ethnic influences, the observed differences in nation-specific percentiles could be associated with variations in perinatal care, maternal nutrition status, or other environmental factors. Smaller percentile values at very early gestations in developed countries might be correlated with a higher live birth rate for fetuses prone to be premature, which is usually attributed to better maternal nutrition status and improved prenatal health care. Similarly, improvements in perinatal care for preterm births might contribute to the downward trend in twin mortality in our study (Figure 1).

It is well known that both environmental factors and genetic backgrounds contribute to birth weight patterns (30,31). In the current study, twin birth weight varied by socio-demographic, maternal, and infant characteristics. Twins in urban area or in coastal region who generally had better socioeconomic conditions were heavier than those in rural area, or in inner land and remote area. Older women tended to give birth to heavier babies than younger women, and twins born to primiparous women were lighter than those born to multiparous women. In our study, the difference of mean birth weight between maternal–parity groups was relatively smaller, 1.4% (38 g) and 1.1% (29 g) in boys and girls at 40 gestation weeks. There is still some controversy about whether maternal parity should be included in evaluating twin growth (8). Only a few of studies constructed parity-specific birth weight percentiles for twins (8,20). Considering the small difference in observed mean birth weight and small sample size at some gestation weeks, we did not create percentile curves by maternal parity for each sex. Other factors like maternal cigarette smoking, obesity, and mode of delivery may affect birth weight distributions, but the effects on centiles cannot be assessed due to limited data.

Birth weight for gestational age not only reflects intrauterine growth but also correlates postnatal outcomes. The neonatal mortality rate in this study (12.3‰) was greater than the latest rates in general Chinese population (i.e. 5.9‰ in 2014) (32) and in singleton cohort in the same areas (2.1‰). We identified an overall postnatal mortality rate of 33.7‰ for small-for-gestational-age twins (Table 4). Given the short-time period for surveillance (from 28 weeks of gestation to 42 d after birth), the actual infant mortality of small-for-gestational-age twins as well as of the whole twin population might be even higher. Although there was a substantial decline (61.1%) in the postnatal mortality during the study period (Figure 1), the high mortality rates indicated an urgent need to improve the perinatal health care for twins, particularly for small and premature twins (Table 4). It is obvious that the newly established birth weight reference can provide more accurate risk prediction for early intervention for twins.

As ethnicity and other socio-demographic factors correlate with birth weight distributions, researchers have suggested that the birth weight references should be updated every 5–10 years (28). In the past three decades in China, marked changes have occurred in socioeconomic, maternal, and infant characteristics along with the fast economic development, such as increased family income, improved education level of the parents, increments in infant weight and length, as well as improvements in prenatal nutritional status (33,34). In this study, the use of a large, nationally representative population-based sample of twins ensures a more representative and accurate estimate of percentiles. The ethnic and urban–rural distributions are in proportion to those from the National Census 2010 (http://www.stats.gov.cn/tjgb/rkpcgb/). The minorities and urban births accounted for 7.2% and 54.5%, respectively, highly comparable with 8.5% and 49.7% in the Census 2010. Although the male–male, male–female, and female–female twin pairs distribute unequally in our cohort, the sample size is large enough to generate sex-specific percentiles. Based on the experience in previous studies (22,35), we used Tukey’s methodology to screen outliers at 28–34 weeks of gestation, and removed erroneous measurements at 35–42 weeks by clinical judge. The careful exclusion of subjects with missing key variables or outlier values also reduces errors in the estimated percentiles.

Our study has several limitations. First, discrepancies in birth weight measurement could influence the accuracy of percentiles, whereas the effects were likely minimal due to systemically trained professionals in surveillance sites. Second, we could not develop percentiles for newborns below 28 weeks of gestational age, nor adjust reference centiles for maternal cigarette smoking, obesity, diabetes, in vitro fertilization treatment, and caesarean delivery, because these data were not collected. Third, birth weight alone is insufficient for a complete evaluation of fetal and neonatal growth, we were not able to create birth length and head circumference reference charts for twins due to limits of data. Future studies are warranted to develop references for other anthropometric measurements for Chinese twins as well as for singletons.

In summary, our newly developed gestational age-specific birth weight percentiles for Chinese twins are based on up-to-date data from the largest national registry. They can provide clinicians and researchers with a valuable reference of identifying high-risk twins in China.

Acknowledgements

The authors thank the obstetricians, pediatricians, and other staff members involved in data collection. The opinions expressed in this manuscript are the views of the authors. They do not reflect the official position of the National Center for Birth Defects Monitoring, P.R. China.

Disclosure statement

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of this article.

Additional information

Funding

This work was supported by Grant 81373069 from the National Natural Science Foundation of China and Grant 2013SZ0179 from the Science and Technology Department of Sichuan Province.