Handedness in twins: meta-analyses

Handedness in twins: meta-analyses. Lena Sophie Pfeifer, Judith Schmitz, Marietta Papadatou-Pastou, Jutta Peterburs, Silvia Paracchini & Sebastian Ocklenburg. BMC Psychology volume 10, Article number: 11. Jan 15 2022. https://bmcpsychology.biomedcentral.com/articles/10.1186/s40359-021-00695-3

Abstract

Background: In the general population, 10.6% of people favor their left hand over the right for motor tasks. Previous research suggests higher prevalence of atypical (left-, mixed-, or non-right-) handedness in (i) twins compared to singletons, and in (ii) monozygotic compared to dizygotic twins. Moreover, (iii) studies have shown a higher rate of handedness concordance in monozygotic compared to dizygotic twins, in line with genetic factors playing a role for handedness.

Methods: By means of a systematic review, we identified 59 studies from previous literature and performed three sets of random effects meta-analyses on (i) twin-to-singleton Odds Ratios (21 studies, n = 189,422 individuals) and (ii) monozygotic-to-dizygotic twin Odds Ratios (48 studies, n = 63,295 individuals), both times for prevalence of left-, mixed-, and non-right-handedness. For monozygotic and dizygotic twin pairs we compared (iii) handedness concordance Odds Ratios (44 studies, n = 36,217 twin pairs). We also tested for potential effects of moderating variables, such as sex, age, the method used to assess handedness, and the twins’ zygosity.

Results: We found (i) evidence for higher prevalence of left- (Odds Ratio = 1.40, 95% Confidence Interval = [1.26, 1.57]) and non-right- (Odds Ratio = 1.36, 95% Confidence Interval = [1.22, 1.52]), but not mixed-handedness (Odds Ratio = 1.08, 95% Confidence Interval = [0.52, 2.27]) among twins compared to singletons. We further showed a decrease in Odds Ratios in more recent studies (post-1975: Odds Ratio = 1.30, 95% Confidence Interval = [1.17, 1.45]) compared to earlier studies (pre-1975: Odds Ratio = 1.90, 95% Confidence Interval = [1.59–2.27]). While there was (ii) no difference between monozygotic and dizygotic twins regarding prevalence of left- (Odds Ratio = 0.98, 95% Confidence Interval = [0.89, 1.07]), mixed- (Odds Ratio = 0.96, 95% Confidence Interval = [0.46, 1.99]), or non-right-handedness (Odds Ratio = 1.01, 95% Confidence Interval = [0.91, 1.12]), we found that (iii) handedness concordance was elevated among monozygotic compared to dizygotic twin pairs (Odds Ratio = 1.11, 95% Confidence Interval = [1.06, 1.18]). By means of moderator analyses, we did not find evidence for effects of potentially confounding variables.

Conclusion: We provide the largest and most comprehensive meta-analysis on handedness in twins. Although a raw, unadjusted analysis found a higher prevalence of left- and non-right-, but not mixed-handedness among twins compared to singletons, left-handedness was substantially more prevalent in earlier than in more recent studies. The single large, recent study which included birth weight, Apgar score and gestational age as covariates found no twin-singleton difference in handedness rate, but these covariates could not be included in the present meta-analysis. Together, the secular shift and the influence of covariates probably make it unsafe to conclude that twinning has a genuine relationship to handedness.

Discussion

In three sets of meta-analyses, we examined the influence of twin status and twin zygosity on handedness prevalence and handedness concordance. Our first set of meta-analyses confirmed that in line with Sicotte et al. [10], left-handedness (OR = 1.40, Fig. 3) and non-right-handedness (OR = 1.36, Fig. 5) occur more often among twins than among singletons. Moderator analyses found elevated levels of non-right-handedness among twins to be independent of all variables tested with respect to a potential moderating effect. However, we found that more recent studies reported smaller differences in prevalence of left-handedness between twins and singletons (Fig. 6). To test whether there is a higher left-handedness prevalence in twins compared to singletons in more recent studies at all, we estimated twin-to-singleton ORs for left-handedness for studies published pre and post 1975 separately. With a pre-1975 OR of 1.90 (95% CI = [1.59, 2.27]) and a post-1975 OR of 1.30 (95% CI = [1.17, 1.45]), ORs for more recent studies were smaller, but still indicated a significant twin effect on left-handedness.

Overall, the decrease in twin-to-singleton ORs might either be explained by a decrease in left-handedness in twins or an increase of left-handedness in singletons, or both. As already mentioned, complications occur more often in the course of multiple births [11,12,13], which might contribute to the development of atypical handedness [10]. However, most individual studies included in our meta-analysis did not provide information on pre- or perinatal conditions, so we could not test for a moderating effect of these conditions on the twin-to-singleton OR. Along these lines, future research might have a closer look on the relation between birth complications and handedness.

Assuming that higher proportions of left-handedness among twins might be the by-product of birth complications, a decrease in atypical handedness in twins must be assigned to a decrease in the occurrence of these complications. In fact, it is well conceivable that medical progress over the last decades, that is clearly detectable, e.g. in the United States [92, 93], may have helped to equalize the risks associated with multiple and single births. Such assumptions are supported by a study by Heikkilä et al. [77] who showed differences in left-handedness in twins and singletons to disappear when controlling for birth weight, Apgar score, and gestational age. We therefore tested whether there is evidence for a decrease in left-handedness prevalence in twins (Fig. 6b) by running meta-analyses on left-handedness prevalence in twins and singletons separately while including publication year as a moderator variable. However, while there was no evidence for an effect of publication year on left-handedness prevalence in twins, there seemed to be a trend towards an increase of left-handedness prevalence in singletons (Fig. 6c).

The overall prevalence of atypical handedness in our study was lower than expected. We found 9.13% of twins and 6.97% of singletons to be left-handed (Table 1), while Papadatou-Pastou et al. [3] reported a figure of 10.6% (95% CI 9.71%, 11.50%) for the general population. The low values in our study might be the result of a general effect of publication year in singletons, given that the prevalence of left-handedness has been shown to be higher in younger than in older cohorts [25, 94, 95]. The social stigma associated with left-handedness in the last century [96] may have driven left-handers to conceal their preference in self-reports [97] and to retrain to use their right hand [25, 98]. Most of the studies included in our meta-analysis were published in the previous century and their participants could have been subjected to environmental pressures against left-handedness, leading to underestimation of the true population prevalence of left-handedness. Similarly, we found low overall prevalence of mixed-handedness (3.39% in twins and 2.67% in singletons, Table 1), whereas Papadatou-Pastou et al. [3] gave a point estimate of 9.3% for the general population. This might also be due to an effect of publication year. Moreover, three of five studies that provided data for mixed-handedness classified handedness as writing hand so that data extracted from these studies most likely reflect not mixed-handedness, but ambidexterity, which is much rarer [99].

Our second set of meta-analyses found no difference in the prevalence of atypical handedness between MZ and DZ twins (left-handedness OR = 0.98, mixed-handedness OR = 0.96, non-right-handedness OR = 1.01, Table 3). This result is consistent with the meta-analysis by Sicotte et al. [10] who interpreted this null-effect as indication against mirror imaging theories designed to explain heightened frequencies of left-handers and frequent handedness discordance among MZ twins [100,101,102]. Indeed, it weakens the hypothesis suggesting that the monozygotic twinning process is responsible for atypical handedness [10]. Moreover, it indicates that the overall heightened frequencies of left- and non-right-handers among twins are independent of the twins’ zygosity. A moderator analysis showed that this effect was not influenced by the method used to determine the twins’ zygosity, thus refuting the idea that the result was affected by the accuracy with which twins were classified as monozygotic or dizygotic. All in all, revealing comparable prevalence of atypical handedness for MZ and DZ twins cannot enrichen knowledge about genetic contribution to handedness per se. As already recognized by Sicotte et al. [10], to do so, it is crucial to look at pairwise handedness concordance or discordance of MZ and DZ twin pairs.

Our third set of meta-analyses found a small yet significant effect (OR = 1.11, Fig. 7) for higher handedness concordance among MZ (80.49%) as compared to DZ (79.27%) twins, consistent with the meta-analysis by Sicotte et al. [10]. Even though other publications have demonstrated the occurrence of handedness discordance among MZ twin pairs [100, 101, 103, 104], it was estimated to concern a minority of 20–25% of cases [2]. Stronger phenotypic variation among DZ compared to MZ pairs indicates a certain genetic foundation of that phenotype [2, 26]. Therefore, our results confirm handedness to rely on genetic factors to some extent [10] and are consistent with heritability estimates of 0.24–0.26 [31,32,33]. A moderator analysis suggested that the frequencies of handedness concordance did not differ between studies included in the meta-analysis by Sicotte et al. [10], studies explicitly excluded from Sicotte et al. [10], and more recent studies.

To allow future meta-analyses to perform comparisons on handedness prevalence in twins more specifically (e.g., handedness in male vs. female twins, or handedness in same sex pairs vs. opposite sex pairs), it is desirable that researchers report results broken down for parameters like zygosity, sex, and consider data on birth complications. As this might be beyond the scope of individual papers, we encourage authors to provide open raw data in publicly accessible repositories such as the osf.io.

The present study is not without limitations. We did not investigate relative hand skill but were restricted to hand preference. Measuring hand preference is far more established as compared to assessing relative hand skill, as it is easier and more convenient [105]. Most of the studies included in our meta-analysis only provided information on hand preference, not allowing for an additional analysis for hand skill. Moreover, hand preference and hand skill correlate to some extent [106,107,108], and the distribution of handedness categories overlaps for preference- and skill-related criterions in 90% of the cases [109].

Similarly, our study only dealt with handedness direction in terms of categorial handedness classification which does not take into account the fact that individual handedness can further be defined regarding its strength or its degree. Along these lines, other approaches consider handedness as a continuum, extending the question to how strong or how consistently one hand is preferred, used, or skilled over the other. Indeed, several findings obtained within laterality research on associations between handedness and structural brain lateralization [110] or cognitive performance [111, 112] as well as concerning the genetic foundation of handedness [113, 114] are linked to strength but not direction of handedness. However, since most studies included in the present meta-analyses did not assess handedness in a continuous manner, we were unable to account for handedness strength. Therefore, it falls to future studies to extend their assessment repertoire by measures of handedness strength.

From a methodological point of view, it is further crucial to mention that overall, our moderator analyses are low in power due to the investigated study sample sizes. Of note, in some cases, moderator levels included only three data points calling for an interpretation of these findings with caution.