Obstetrics & Gynecology Email Alerts
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Obstetrics & Gynecology 2000;96:986-990
© 2000 by The American College of Obstetricians and Gynecologists
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by BLICKSTEIN, I.
Right arrow Articles by MAZKERETH, R.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by BLICKSTEIN, I.
Right arrow Articles by MAZKERETH, R.

ORIGINAL RESEARCH

Adaptive Growth Restriction as a Pattern of Birth Weight Discordance in Twin Gestations

ISAAC BLICKSTEIN, MD, RAN D. GOLDMAN, MD and RAM MAZKERETH, MD

From the Department of Obstetrics and Gynecology, Kaplan Medical Center, Rehovot, Israel; Department of Pediatrics, Sheba Medical Center, Tel Hashomer, Israel.

Address reprint requests to: Isaac Blickstein, MD, Kaplan Medical Center, Department of Obstetrics and Gynecology, Rehovot, 76100, Israel, E-mail: blick{at}netvision.net.il


   Abstract
Top
Abstract
Patients and Methods
 Results
 Discussion
 References
 
Objective: To challenge the hypothesis that discordant growth is a normal variation by relating birth weight discordance to total twin birth weight.

Method: Among 12,565 Israeli live-born twin pairs (1993–98), we compared total twin birth weight decile, the frequencies of three levels of discordance in the general population, over 25% discordance between like- versus unlike-sex pairs, and over 25% discordance between pairs delivered by primiparas versus multiparas.

Results: We found a marked change in the best-fit correlation function with increased discordance: level 15–24.9% was inversely linear whereas levels 25–34.9% and over 34.9% were inversely logarithmic (R2 = .47, .88, and .9, respectively). The best-fit correlation of frequencies of more than 25% discordance was inversely logarithmic and similar in like-and unlike-sexed twins across deciles functions (P = .7, odds ratio [OR] 1.0, 95% confidence interval [CI] 0.9, 1.2). The overall frequencies of discordance were also similar (9.3% versus 10.2%, P = .11, OR 0.9, 95% CI 0.8, 1.0). The frequencies of primiparas decreased linearly (R2 = .98) and the frequencies of more than 25% discordant pairs in multiparas and primiparas across the deciles had similar inversely logarithmic patterns (P = .55, OR 1.0, 95% CI 0.9, 1.2). Discordance over 25% was significantly more frequent among primiparas (P < .001, OR 1.45, 95% CI 1.3, 1.6).

Conclusion: The observed patterns of birth weight discordance did not substantiate normal variation but an adaptive growth restriction that might explain why the likelihood of discordant growth decreases as total twin birth weight increases.

Birth weight discordance is caused by inherent intertwin biologic diversity and extrinsic determinants of fetal growth, epitomized by the maxim "interplay between nature and nurture." Birth weight disparity is not necessarily synonymous with deviant growth, but is a specific measure of intrauterine development in multifetal gestations. However, it has not been established whether greater birth weight disparity represents transition from normal variation to growth deviation through a critical discordance level, or whether small and large discordance values represent distinct entities. Those uncertainties show the continued debate over whether intertwin birth weight discordance is a normal biologic variation or some form of growth restriction.

In a study of a different data set,1 we used an epidemiologic approach to examine the relationship between intertwin birth weight discordance and total twin birth weight to reject the null hypothesis that discordant twin growth is a normal variation. Assuming that total twin birth weight represents the uterine capacity for carrying twins, the frequency of over 25% discordant pairs was expected to follow a zero-slope linear function across the range of total twin birth weight deciles. However, we found an inverse logarithmic correlation, apparently valid for both like- and unlike-sexed twins. We concluded that the more favorable the uterine milieu for carrying twins (ie, allowing a larger total twin birth weight), the smaller the likelihood that the twins would have discordant growth.1

That database comprised nearly 9000 twin pairs from five different populations grouped into two cohorts. To confirm and further examine the hypothesis, a larger, single-population cohort also was necessary. After obtaining the required database, we examined the relationship between frequency of discordant pairs and total twin birth weight deciles. We first examined the hypothesis that smaller degrees of discordance should have flattened logarithmic curves whereas larger degrees should have accentuated inverse logarithmic curves, when describing frequencies of birth weight discordance as functions of total twin birth weight deciles. We then challenged the findings of previous studies1,2 and hypothesized that if discordance is a result of normal biologic variation, larger birth weight differences should be expected among unlike-sexed twins. The anticipated difference between uterine capacity of primiparas and multiparas led us to hypothesize that if a normal biologic variation explains discordance, each group of parturients would have different frequency patterns of divergent growth.


   Patients and Methods
Top
Abstract
 Patients and Methods
Results
 Discussion
 References
 
We used data from the Israeli Birth Registry through the Central Bureau of Statistics, which includes all deliveries in Israel, as required by law. The database of 12,785 twin pairs delivered from 1993–1998 was used to select a cohort of 12,777 pairs (99.9%) with known birth weights and genders. Maternal parity was known in 12,565 (98.3%). The range of maternal age was 18–44 years, the range of parity 0–7. All women were Israeli citizens, roughly comprising 80% Jewish, 15% Moslem, and 5% other faiths. All but 1.5% were registered as married. Most (92%) resided in towns. Most (75%) of the 8% of rural residents were Moslems. We used the Microsoft Excel program (Microsoft Corporation, Redmond, WA) to sort total twin birth weights (twin A + twin B) into deciles. Level of discordance was calculated for each pair [as 100 x (birth weight difference/birth weight of heavier twin)].

The first hypothesis was that frequencies of discordant pairs across the range of total twin birth weight deciles have a similar descriptive function irrespective of the level of discordance. To test that hypothesis, we calculated in each decile the frequency of discordant pairs in the following three categories: 15–24.9%, 25–34.9%, and 35% or more, which roughly corresponded to moderate, severe, and extreme discordance, respectively. Then we calculated the best-fit correlation of frequencies across the range of deciles and the respective R2 value for each discordance category. We compared the slopes and intercepts using regression analysis.

The second hypothesis was that greater intertwin birth weight variation (ie, discordance) is expected in unlike-sexed pairs across the range of total twin birth weight deciles. To test that hypothesis, we calculated the frequency of over 25% discordant pairs in like-versus unlike-sexed twins in each decile. Then the best-fit correlation for each type and its R2 value were evaluated.

The third hypothesis was that primiparas and multiparas have different birth weight discordance frequencies across the range of total twin birth weight deciles. To test that hypothesis, we calculated the frequency of primiparas versus multiparas and the frequency of over 25% discordant pairs delivered to primiparas versus multiparas in each decile. The best-fit correlation and its R2 value were then examined.

We used True EPISTAT Software (Math Archives, Round Rock, TX) for statistical analysis. Owing to the large database, the 95% power of our sample ({alpha} = .05) was sensitive enough to detect differences of less than 0.5% when comparing unpaired samples. The Mantel-Haenszel {chi}2 test (with Woolf test for heterogeneity) was used to compare pooled categorical results, allowing us to test the hypotheses of no overall relationship derived from a case-control (with and without discordance) frequency matching for the ten categories (deciles) of total birth weight. The best-fit correlation between the decile order and frequencies of discordance was computer generated. P < .05 was considered statistically significant.


   Results
Top
Abstract
 Patients and Methods
 Results
Discussion
 References
 
Frequencies of the variables studied are shown in Table 1Go. The numbers of discordant pairs in the categories 15–24.9%, 25–34.9%, and 35% or more were plotted as functions of the total twin birth weight deciles. The best-fit correlation for discordance level 15–24.9% was inversely linear (percentage of discordant twin pairs = -0.44 [decile order] + 21.8) but with a relatively low R2 value of 0.47 and the slope was not different than zero (P = .02; 95% confidence interval [CI] for the slope -.8, -.1). The best-fit correlation for discordance level 25–34.9% was inversely logarithmic (percentage of discordant twin pairs = -3.0 ln [decile order] + 11.2) but with a high R2 value of 0.88. The best-fit correlation for discordance level over 35% was also inversely logarithmic, with a high R2 value of 0.9, but with an accentuated decrease (percentage of discordant twin pairs = -4.4 ln [decile order] + 9.7). Comparison of the slopes and intercepts of the functions were statistically different (P < .001; 95% CI for the slope -1.4, -.4). Figure 1Go shows the scalar and semilogarithmic trend lines of those functions. Those findings led us to conclude that frequencies of discordant twin pairs across the range of total twin birth weight deciles have different, discordance level–dependent functions.


View this table:
[in this window]
[in a new window]
  		Table 1. Number and Percentage of Twins by Total Twin Birth Weight Deciles
 


View larger version (20K):
[in this window]
[in a new window]
  		Figure 1. Frequency of birth weight discordant pairs by total twin birth weight deciles. (A) Trend lines on scalar scales (right y-axis for discordance >= 35%). (B) Trend lines on semi-logarithmic scale.

 
The frequencies of more than 25% discordant pairs in like- versus unlike-sexed twins across the total twin birth weight deciles are shown in Figure 2Go. The two best-fit correlations were inversely logarithmic and very similar (percentage of discordant twin pairs = -7.4 ln [decile order] + 20.5, and percentage of discordant twin pairs = -7.2 ln [decile order] + 21.3, respectively). The R2 values were markedly high (0.89 for both). The statistics showed no significant difference between the frequency curves (P = .7; odds ratio [OR] 1.0, 95% CI 0.9, 1.2; Woolf test for heterogeneity, P = .29). That similarity suggests that unlike-sexed pairs did not have greater discordance levels than like-sexed pairs, which also was evident from the overall similar frequencies of discordant pairs (9.3% versus 10.2%; P = .11; OR 0.9, 95% CI 0.8, 1.0).



View larger version (11K):
[in this window]
[in a new window]
  		Figure 2. Frequency of birth weight discordance over 25% by total twin birth weight deciles in unlike-sexed (thin line) and like-sexed (heavy line) pairs.

 
The frequencies of primiparas decreased as an almost absolute linear function of the total twin birth weight deciles (percentage of primiparas = -3.6 [decile order] + 58.1, R2 = 0.98). The ratio of multiparas to primiparas was 0.8 and 0.9 for the lower two deciles, changed to 1.2 for the third decile and further increased to 3.2 for the tenth decile (Figure 3Go). However, the frequencies of more than 25% discordant pairs in multiparas and primiparas across the total twin birth weight deciles (Figure 4Go) showed similar logarithmic patterns (percentage of discordant twin pairs = -6.5 ln [decile order] + 18.8, and percentage of discordant twin pairs = -7.2 ln [decile order] + 21.2 respectively). The R2 values were also very high (0.88 for both). Statistics showed no significant difference between the two frequency curves (P = .55; OR 1.0, 95% CI 0.9, 1.2, Woolf test for heterogeneity P = .91). That assessment suggested that frequencies of discordant twin pairs across the range of total twin birth weight deciles among primiparas were not different than those for multiparas. The likelihood of discordance decreased with increased total twin birth weight, and because multiparas were more likely to deliver larger twins (Figures 3Go and 4Go), the frequency of birth weight discordance (over 25%) was also significantly higher among primiparas (P < .001; OR 1.45, 95% CI 1.3, 1.6).



View larger version (10K):
[in this window]
[in a new window]
  		Figure 3. Frequencies of primiparas (thin line) and multiparas (heavy line) in each total twin birth weight decile.

 


View larger version (10K):
[in this window]
[in a new window]
  		Figure 4. Frequencies of birth weight discordance over 25% by total twin birth weight deciles in primiparas (thin line) and multiparas (heavy line).

 

   Discussion
Top
Abstract
 Patients and Methods
 Results
 Discussion
References
 
We used an epidemiologic approach to differentiate between patterns of discordant growth in twins. There is evidence that frequencies of divergent birth weight have a pattern based on total twin birth weight, suggesting that the greater the uterine capacity to nurture twins is, the smaller the likelihood of birth weight discordance.1

The first aspect we examined was the transition from low- to high-level birth weight discordance. Figure 1Go shows that with more than 25% discordance, there is an inverse logarithmic relation to the total birth weight. From a purely mathematical viewpoint, there was little difference at the first two deciles, followed by a decrease in frequencies until the median total birth weight (ie, the fifth decile), and a final part, in which changes occurred almost as a linear function. The comparison of the trend lines is best seen on the left panel of Figure 1Go where the semilogarithmic depiction suggests that the principal difference between the frequencies of levels 25–34.9% and 35% or more is the markedly decreased rate in the latter between the second and fifth deciles. An explanation for that unique pattern is that the uterine environment efficiently limits discordant growth up to the median total birth weight. The total birth weight is smaller in growth-discordant pairs than concordant pairs of the same gestational age, so the pattern might suggest that growth discordance is an adaptive mechanism imposed by the limited uterine milieu. That adaptive measure might increase gestational age at a given uterine size and environment by inducing growth restriction of one twin, ie, discordant growth.

It follows that outcomes might be adverse if adaptation fails. Clinical correlates of that implication were published recently. Hollier et al3 examined outcomes of 1370 consecutive twin births, stratified according to intrapair birth weight differences. Greater birth weight discordance was associated with signs of uterine maladaptation, ie, significantly more preterm deliveries as a result of intervention, low birth weight, intensive care admissions, respiratory distress, fetal abnormalities, and fetal deaths. Rydhstrom4 also showed that like-sexed twin pairs had higher pregnancy loss rates with increased absolute discordance levels, with relative risks (RRs) ranging from 2.1 for a 500–749-g difference to 10.9 for over 1000-g birth weight difference.

When twins overcame the risks of preterm birth (ie, were delivered after 37 weeks) and low birth weight (ie, the lighter twin weighed more than 2500 g), there was no excess risk of adverse outcome for discordant twins.5 Cheung and coworkers6 defined the increased risk of discordant growth among preterm twins, and Yalcin et al7 observed that the same level of discordance had different clinical connotations in smaller twins. Our data support those views and suggest that discordance might be more important at lower birth weights.

Unlike-sexed twins are dominant contributors to increased intertwin birth weight differences because male fetuses are more often heavier than female fetuses.8,9 It is debatable whether large birth weight differences in dizygotic pairs of opposite sex can be explained by fetal sex and birth order only,10 or whether a specific effect on fetal growth is present only in unlike-sexed pairs.11 The debate has been revived recently by Glinianaia et al,9 who found a tendency for birth weight in females to be influenced by a male cotwin, with decreasing intertwin birth weight disparity in unlike-sexed twins. The clinical correlate of that finding comes from Rydh-strom4 who found that pregnancy loss was twice as high in like-sexed twins than unlike-sexed twins, and only in like-sexed pairs was pregnancy loss significantly correlated with birth weight discordance. In accordance with the model of uterine maladaptation, Rydhstrom’s finding suggests that discordance is more important to fetal well being in like-sexed than in unlike-sexed pairs. Thus, like- and unlike-sexed twins might have identical discordant growth patterns, but significant restricted fetal growth imposed by the uterine milieu might be more critical to like-sexed pairs who otherwise have concordant growth potential.

Our third hypothesis was based on the notion that the uteri of multiparas are more efficient promoters of twin growth than those of primiparas.12–14 Indeed, parous women are much more likely to deliver larger twins (Figure 3Go), and therefore less likely to deliver discordant pairs, which agrees with the model of uterine maladaptation. If birth weight discordance results from normal variation, one would expect to find different functions of discordance frequencies across the range of deciles. However, Figure 4Go excludes that assumption and confirms that maladapted uteri in multiparas are as likely to influence growth as maladapted uteri in primiparas.

Having excluded the probability of normal biologic variation as the underlying cause, we proposed that birth weight discordance is a result of uterine inability to nurture twins equally. Therefore, if the purpose of nature is primarily to promote appropriate gestation, compensation in the form of adaptive growth restriction of one twin, causing mild discordance, begins early. Twins who exhibit marked differences early3,6,7 or defy inborn similarities4,9 are predisposed to adverse outcomes. Conversely, uteri that can efficiently nurture twins are associated with increased total birth weight and decreased discordance frequencies. Effective adaptation results in discordant pairs that are delivered at advanced gestational ages and that fare as well as growth concordant pairs.5

As in similar mechanisms, the proposed adaptation should be a continuum, interrupted by delivery or fetal death. Because perinatal outcome depends on the degree of the adaptive growth restriction, successful adaptation might result in pairs with divergent birth weights that nonetheless have normal perinatal courses, whereas failure to adapt might cause single fetal death and adverse outcomes among discordant pairs.


   Footnotes
 
PII S0029-7844(00)01079-6

Received March 7, 2000. Received in revised form July 14, 2000. Accepted September 7, 2000.


   References
Top
Abstract
 Patients and Methods
 Results
 Discussion
 References
 
1. Blickstein I, Goldman RD, Smith-Levitin M, Greenberg M, Sherman D, Rydhstroem H. The relation between inter-twin birth weight discordance and total twin birth weight. Obstet Gynecol 1999;93:113–6.[Abstract/Free Full Text]

2. Naeye RL, Benirschke K, Hagstrom JWC, Marcus CC. Intrauterine growth of twins as estimated from liveborn birth-weight data. Pediatrics 1966;37:409–16.[Abstract/Free Full Text]

3. Hollier LM, McIntire DD, Leveno KJ. Outcome of twin pregnancies according to intrapair birth weight discordance. Obstet Gynecol 1999;94:1006–10.[Abstract/Free Full Text]

4. Rydhstrom H. Discordant birthweight and late fetal death in like-sexed and unlike-sexed twin pairs: A population-based study. Br J Obstet Gynaecol 1994;101:765–9.[Medline]

5. Blickstein I, Shoham (Schwartz) Z, Lancet M. Growth discordancy in appropriate-for-gestational-age term twins. Obstet Gynecol 1988;72:582–4.[Abstract/Free Full Text]

6. Cheung VY, Bocking AD, Dasilva OP. Preterm discordant twins: What birth weight difference is significant? Am J Obstet Gynecol 1995;172:955–9.[Medline]

7. Yalcin HR, Zorlu CG, Lambet A, Ozden S, Gokmen O. The significance of birth weight difference in discordant twins: A level to standardize? Acta Obstet Gynecol Scand 1998;77:28–31.[Medline]

8. Essel JK, Opai-Tetteh ET. Twin birth-weight discordancy in Transkei. S Afr Med J 1994;84:69–71.[Medline]

9. Glinianaia SV, Mangus P, Harris JR, Tambs K. Is there a consequence for fetal growth of having an unlike-sexed cohabitant in utero? Int J Epidemiol 1998;27:657–9.[Abstract/Free Full Text]

10. Blickstein I, Weissman A. Birth weight discordancy in male-first and female-first pairs of unlike-sexed twins. Am J Obstet Gynecol 1990;162:661–3.[Medline]

11. Orlebeke JF, van Baal GC, Boomsma DI, Neeleman D. Birth weight in opposite sex twins as compared to same sex dizygotic twins. Eur J Obstet Gynecol Reprod Biol 1993;50:95–8.[Medline]

12. Blickstein I, Zalel Y, Weissman A. Pregnancy order: A factor influencing birth weight in twin gestations. J Reprod Med 1995;40: 443–6.[Medline]

13. Rydhstrom H. A birthweight-for-gestation standard based on 4737 twins born in Sweden between 1983 and 1985. Acta Obstet Gynecol Scand 1992;71:506–11.[Medline]

14. Fakeye O. Twin birthweight in Nigeria and the effect of sex-pair and parity. Trop Geogr Med 1986;38:265–70.[Medline]




This article has been cited by other articles:


Home page
 Int J EpidemiolHome page
D. Phillips
Commentary: Twins, low birthweight and type 2 diabetes
Int. J. Epidemiol., October 1, 2004; 33(5): 953 - 954.
[Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by BLICKSTEIN, I.
Right arrow Articles by MAZKERETH, R.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by BLICKSTEIN, I.
Right arrow Articles by MAZKERETH, R.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS