Innate Handedness and Disease-Specific Mortality in Women

Abstract

Background: 

Left-handedness has been reported to be associated with reduced life expectancy, but the evidence is far from conclusive.

Methods: 

We studied the association between innate handedness and total mortality, as well as cause-specific mortality, in a cohort of 12,178 middle-aged Dutch women who were followed for almost 13 years. The relation between handedness and mortality was analyzed using Cox regression in a case-cohort approach, in which a random sample of 1500 women was used to represent person-years under observation for the entire cohort.

Results: 

During a median follow-up of 12.6 years, 252 women died. Hazard ratios comparing left-handed women with other women were 1.4 for all-cause mortality (95% confidence interval = 0.9–2.0), 1.7 for total cancer mortality (1.0–2.7), 2.0 for breast cancer mortality (0.8–4.6), 4.6 for colorectal cancer mortality (1.5–14.3), 1.3 mortality from diseases of the circulatory system (0.5–3.3), and 3.7 for cerebrovascular mortality (1.1–12.1), after adjusting for potential confounders (socioeconomic status, age, body mass index, and cigarette smoking status at study recruitment).

Conclusions: 

Left-handedness is associated with higher mortality in women.

Approximately 10% of the population are left-handed.¹ There is an increasing body of evidence in support of a genetic basis for hand preference,^1,2 but many observations suggest environmental influences in early life as well.^3,4 One of the prevailing explanations is that left-handedness stems from an adverse prenatal environment, such as excessive exposure to testosterone, which influences brain lateralization.³

In recent decades, there is a growing interest in early life origins of adult chronic diseases. Furthermore, childhood growth may modify the effect of birth weight on later life diseases risk^5,6 and mortality.⁷ Hypotheses regarding fetal and infant origins of adult diseases⁸ may include mechanisms leading to both left-handedness and increased morbidity and mortality in later life.⁹ Just as an adverse environment in utero may lead to greater mortality from certain adult chronic diseases,¹⁰ left-handedness may similarly be associated with reduced survival. Indeed, left-handers are reported to be underrepresented in the older age groups, although such findings are still much debated.^11,12

Several studies^11,13–20 have attempted to explain the decreased number of left-handed people among elderly populations through a “modification” hypothesis, meaning cultural or social pressure against left-handedness, or an “elimination” hypothesis, meaning greater rates of mortality among the left-handed group.¹¹ The modification hypothesis could not explain the reduced proportion of left-handedness in the elderly, as the estimated proportion of learned cases of left-handedness was much smaller than the estimated life-span shift toward dextrality.¹¹ The elimination hypothesis has been studied to some extent, but the results are largely inconsistent.¹² Two studies reported a relation between left-handedness and increased mortality,^11,17 whereas 2 others found left-handedness to be associated with a survival benefit.¹⁸ However, many studies did not find any relation.^{13–16,19–21} Most of the studies above used cross-sectional designs, cohorts with short follow-up, or very selective groups of participants.

In this work, we explored the relationship between handedness and cause-specific mortality in a large prospective, population-based, cohort study of middle-aged women.

METHODS

In 1974, a population-based project was started in Utrecht, The Netherlands, and its surrounding municipalities, to study the early detection of breast cancer by mammographic screening.²² Within the project, there were 4 sub-birth cohorts. Women who participated in the third cohort (birth date 1932–1941) were asked by questionnaire about their handedness. Participation in this third cohort was approximately 40% from all invited women, with a total of 12,178 women recruited from 1982 to 1985.

These women filled out extensive questionnaires about reproductive history, demography, and lifestyle habits. They were asked about their innate hand preference: “Are you left or non left-handed by birth?” Of all women, 269 (2%) did not answer the question about handedness, and they were excluded from the analyses. Trained assistants took anthropometric measures before mammography screening. Body mass index (BMI) was calculated as body weight in kilograms divided by the square of body height in meters. From the questionnaires, we derived some potential confounders such as cigarette smoking status (never, past, and current smoker), age at recruitment, and socioeconomic status (SES), which was based on type of health insurance: private (higher status), civil servant (intermediate status), and sick funds (lower status).

Information on date and cause of death was provided to the project registration office by the regional municipalities and general practitioners until January 1996. However, date and destiny of emigration of study participants were not monitored, hampering precise measurement of person-years lived under observation in the entire cohort. The active assessment of complete follow-up information, which was necessary for obtaining adequate information regarding loss to follow-up for the whole cohort, is costly and time consuming. We therefore randomly selected a sample of 1500 women (12%) from the third cohort. For this random sample, we ascertained emigration data and vital status through regional municipality registries until January 2000; for the current analysis follow-up was truncated at January 1996. Because these women were randomly selected, their accrued person-years of follow-up were used to represent person-years lived in the entire cohort (case-cohort design).²³

Within the cohort, 252 women died during the follow-up period until January 1996. Causes of death were classified by the International Classification of Diseases, ninth revision (ICD-9) criteria. We investigated effects on total mortality, but also on disease-specific mortality: total cancer (ICD-9: 140–209), breast cancer (ICD-9: 174), colorectal cancer (ICD-9: 153–154), diseases of the circulatory system (ICD-9: 390–459), cerebrovascular diseases (ICD-9 430–438), and other causes. More detailed analyses on cause-specific mortality, such as other types of cancer or ischemic heart disease, were hampered by the increasingly smaller number of cases for other causes of death.

Women were excluded from the analysis if there were no data available on innate handedness or other covariates. From the women who died (cases), 11/252 (4.4%) had missing data on innate handedness, compared with 30/1500 (2.0%) for the women in the random sample, and 269/12,178 (2.2%) for the entire cohort. Similarly, 1/252 (0.4%) of the case group had missing data on covariates, compared with 4/1500 (0.3%) for the random sample and 28/12,178 (0.2%) for the entire cohort. An additional 11 women were excluded from the random sample as they were lost to follow-up immediately following examination, leaving 1455 women from the random sample available for the analyses, together with 240 deaths.

The case–cohort design was first introduced by Miettinen²⁴ and later extended to a failure time analysis design by Prentice.²⁵ Age-adjusted absolute incidence rates were computed using direct standardization, with the use of the random sample population as the standard. For these analyses, the person-years lived in the random sample were extrapolated to the total cohort.

To assess the relation between handedness and mortality (overall and specific causes), we used weighted Cox regression analysis. The methods for such analysis are similar to standard Cox regression and previously have been described by Barlow et al.²³ In the case–cohort design, the standard errors of risk estimates need to be corrected by a weighting scheme.²³ We chose to use the weighting scheme proposed by Prentice²⁵ because it was found to provide estimates that best resemble those from a full-cohort analysis.²³ Follow-up time started from study inclusion (between 1982 and 1985) and ended at the date of death. Women who remained alive (or died of other causes when a specific cause was under investigation) during the observation period were censored at date of movement, date of death or at 1 January 1996, whatever occurred first.

Analyses were performed with SAS (version 8.2, SAS Institute Cary, NC) by use of a macro (available at http://lib.stat.cmu.edu/general/robphreg) that computes the weighted estimates together with a robust standard error, from which we calculated 95% confidence intervals (CIs). The proportionality of the hazards over time was evaluated with log-minus-log plots. Because cohort members were relatively healthy at recruitment, log-minus-log plots showed that mortality hazards were virtually identical between left- and non-left-handed women for the first 5 years after recruitment and started to deviate from each other afterward to reach parallelism. Therefore, to fully fulfill the proportional hazards assumption, we also analyzed the data excluding the first 5 years of follow-up. Because of limited numbers of events, this analysis was restricted only to overall mortality.

Univariate and multivariate models were run to consider potential confounders. Continuous variables were introduced into the models, and for categorical variables dummies were created. Hazard ratios (HRs) are reported with corresponding CIs.

RESULTS

At the end of follow-up, 90% of the random sample of 1455 women were still alive, 2.6% had died, 6.5% had migrated from the region, and 1.4% were lost to follow-up. A total of 17,567 person-years were accrued in the random sample, with a median time of follow-up of 151 months (12.6 years). Taking into account the sampling fraction, we extrapolated to 143,521 person years accrued in the total cohort, during which 240 women died (overall mortality rate: 1.7 per 1000 person-years).

Of the study population, 11.5% reported that they were left-handed “by birth.” Baseline characteristics according to handedness are presented in Table 1. Left- and non-left-handed women did not materially differ in age, BMI, SES, or smoking habits.

TABLE 1: Characteristics at Study Recruitment According to Innate Handedness

The left-handed group had a crude mortality rate for all causes of 2.3 (adjusted for age by direct standardization = 2.1) compared with 1.5 (adjusted for age = 1.6) per 1000 person-years in the non-left-handed group.

Table 2 shows that, after adjustment for age, SES, BMI, and cigarette smoking status, left-handed women had a 1.36 times higher risk of dying from all causes than non-left-handed women. The adjusted HR for total mortality, after excluding the first 5 years of follow-up time, was 1.58 (95% CI = 1.03–2.42).

TABLE 2: Innate Handedness and Mortality

With regard to cancer mortality, left-handed women had a 1.7 times greater risk of dying from any type of cancer (CI = 1.0–2.7), a 4.6 times higher risk of dying from colorectal cancer (1.5–14), and a 2.0-fold higher risk of dying from breast cancer (0.83–4.6). Handedness was weakly associated with overall mortality from diseases of the circulatory system (1.3, 0.54–3.3), although left-handed women had a 3.7 times greater risk of dying from cerebrovascular diseases than non-left-handed women. Left-handedness was not associated with mortality attributable to causes other than the above-mentioned.

DISCUSSION

In this cohort of middle-aged women, we have shown that left-handedness is associated with increased mortality, particularly from cancer and cerebrovascular diseases.

Certain features of our study need to be addressed. The participation rate in our cohort was 40%, and women were selected from participants in a breast cancer-screening project. Our population may be healthier compared with the general population, because women who voluntarily join screening programs are more likely to have healthier lifestyles and to be higher educated.²⁶ However, there is no reason to assume that the relationship between handedness and mortality would differ by participation.

A limitation of the present study is the small number of cases, which prohibited more detailed analyses on left-handedness and cause-specific mortality. Another limitation is that we could not study the effects of handedness in men, a group with a greater percentage of left-handedness.¹

Measuring handedness using one question about writing hand or self-assessment may introduce misclassification.³ Because our focus originally was not on handedness, we asked only for innate hand preference (“at birth”). However, this method would presumably underestimate associations because this misclassification is likely to be at random (non differential). The proportion of innate left-handed people in our population is similar to previous studies.¹

The data on mortality were obtained from active follow-up through regional municipalities and general practitioners, and we therefore believe it to be largely complete. The observed mortality rate in our cohort (2.3 per 1000 person-years) is lower than what would be expected from data from the general Dutch population (3.4 per 1000 person years).²⁷ This rate probably reflects better average health among our participants.

We could think of only a limited number of possible confounders, in the sense of external factors biasing associations between handedness and mortality. Previously, it has been shown that age, BMI, SES, and cigarette smoking status are independent risk factors for mortality (overall and also for some specific causes of death)^28,29 and some may also be related to hand preference.³ We adjusted for these factors (obtained at study recruitment), even though the left- and non-left-handed women showed only very minor differences with regard to these variables. Consequently, there were no material changes in the HRs after adjustment, leading us to conclude that in our data these factors seem to be neither confounders nor intermediate components of the causal pathway linking handedness to mortality. Finally, it is possible that our findings are due to chance.

There have been conflicting results in studies on the relationship between handedness and mortality. A study by Halpern and Coren¹⁷ found that left-handed people have a 9 years’ shorter life expectancy than their right-handed counterparts. Earlier studies¹¹ also found a shorter life span in the left-handed group. These studies^11,17 were widely criticized regarding their validity. More recent studies found no difference in mortality between the left- and non-left-handed subjects^{13,15,16,19,20} except for injury-related death.^13,19 One found longer survival for the left-handed.¹⁸ All the aforementioned studies used current handedness, with various methods to measure handedness, in specific age and sex groups or in cohorts with shorter follow-up duration. The use of current handedness, especially in older generations could, as a result of societal pressure towards right-handedness, introduce misclassification of many innate left-handers as right-handers.

Left-handedness has been related to a higher occurrence of breast cancer in one case–control study³⁰ and recently in our own cohort study³¹ and a lower occurrence of one type of brain tumor.³² To our knowledge, no previous study has shown a relationship between left-handedness and mortality because of cancer. Left-handedness was not related to cardiovascular disease mortality in previous studies.^14,15,21 We could not examine the effect of left-handedness on injury-related mortality, because of a very small number of such events in our population. However, our study supports the hypothesis that fewer left-handed people among the elderly might be caused by elimination as the result of higher susceptibility to certain diseases or higher case-fatality among the left-handed, rather than an incapability to adapt to a right-handed world, leading to elimination by accident-related death.

If our observations about handedness and mortality are true, there could be several explanations. So far, the most plausible theory of handedness is the genetics theory from Annett et al² and McManus et al,¹ which still allows for a contribution from nongenetic factors. Nongenetic factors may cause some fraction of individuals to divert from their genetically designated handedness. The nongenetic factors may include exposure to an adverse environment during fetal life or birth that leads to “atypical” laterality or pathologic handedness,^3,4 disease susceptibility (inadequate immune system),³ and development of unhealthy lifestyles, such as smoking or alcoholism.³³ However, we did not find clues that lifestyle differences between the handedness groups explained our findings. There is no evidence so far suggesting that genes involved in hand preference also act as an underlying factor for susceptibility for certain adult diseases.

In our study, left-handedness was related to a variety of major diseases with largely different causal mechanisms. Because breast cancer constitutes a majority of cancer deaths in our cohort, it dominates the relationship between left-handedness and cancer mortality. It has been suggested that breast cancer may be initiated in utero by high exposure to steroid hormones³⁴ that may also cause left-handedness.³ For cardiovascular diseases, the link may be less straightforward. Nevertheless, as left-handedness is also more common in extremely low birth weight babies³⁵ and low birth weight is related to cardiovascular mortality,¹⁰ the link might lay in the common intrauterine environment.⁹

In summary, the results of our study among 12,178 women followed for 13 years support the view that left-handedness is associated with higher mortality, especially as the result of cancer and cerebrovascular disease.