The final study sample consisted of 107 ex-smokers (age range = 46–95 years; 47% female) and 193 non-smokers (age range 42–97 years; 62% female), and was obtained through the Open Access Series of Imaging Studies-3 (OASIS-3) database (http://oasis-brains.org). To compile the dataset for the current study, first we inspected the participants’ accompanying clinical data excluding those with any documented cognitive impairment or acute disorder. Additional exclusion criteria were any diagnosis or history of neurological disorders, cerebrovascular disorder, psychiatric disorder, or head trauma (except for minor head trauma if it was neither acute nor recent). In order to avoid possible artefacts of temporary hypoperfusion and concurrent structural brain alterations, we also excluded participants with a history of cardiac arrest, congestive heart failure or cardiac bypass procedures. For the remaining 508 participants, we inspected the brain images and excluded those corrupted by artefacts or noise. The resulting sample of 485 healthy participants included 371 participants for whom smoking status had been established within one year before or after the brain scan. Of those, 14 active smokers were excluded (as the current study focused on ex-smokers and non-smokers), Of the remaining 357 participants, 300 participants (107 ex-smokers and 193 non-smokers) had complete and consistent accompanying information with respect to smoking/non-smoking. Sample characteristics are given in Table 1. All subjects gave written informed consent to participate and to publicly share their anonymized data [18, 19].

High resolution T1-weighted brain images were obtained on a 1.5T Siemens Vision scanner, a 1.5T Siemens Sonata scanner, a 3T Siemens Biograph scanner or a 3T Siemens Trio scanner. The acquisition parameters for each scanner are detailed elsewhere (http://oasis-brains.org). The voxel sizes differed slightly across the three scanners, measuring 1 $\times$ 1 $\times$ 1.25 mm${}^3$ (Vision), 1 $\times$ 1 $\times$ 1 mm${}^3$ (Sonata), 1.2 $\times$ 1.05 $\times$ 1.05 mm${}^3$ (Biograph) and 1 $\times$ 1 $\times$ 1 mm${}^3$ (Trio), which was accounted for by adding scanner as a nuisance variable to the statistical model (see Section 2.4). All brain images were processed using CAT12 (http://www.neuro.uni-jena.de/cat/), SPM12 (https://www.fil.ion.ucl.ac.uk/spm/software/spm12/) and MATLAB (https://au.mathworks.com/products/matlab.html) applying bias field corrections as well as rigid-body spatial normalizations to MNI space. In addition, images were tissue-classified into gray matter, white matter and cerebrospinal fluid to calculate the total intracranial volume (TIV) to be added as a nuisance variable to the statistical model (see Section 2.4).

Each corpus callosum was traced manually at the midsagittal brain section by a single rater (CD) blind to the participant’s smoking status. Before tracing this particular dataset, intra- and inter-rater reliability were established by tracing an independent set of twenty scans twice and comparing these traces to those of an experienced rater (FK). Both intra- and inter-rater reliability were high, with dice indices of 0.97 and 0.94 respectively. Seven callosal subareas were defined according to the Witelson parcellation scheme (see Fig. 1), followed by calculating the areas of the (a) rostrum, (b) genu, (c) rostral body, (d) anterior midbody, (e) posterior midbody, (f) isthmus and (g) splenium [20].

Fig. 1.

The Witelson parcellation scheme. Shown are callosal subareas/proportions (from anterior to posterior): The rostrum (a), genu (b), and rostral body (c) comprise the anterior third of the corpus callosum. The anterior midbody (d) and the posterior midbody (e) comprise the middle third of the corpus callosum. The splenium (g) constitutes the callosal posterior fifth and, together with the isthmus (f), makes up the posterior third of the corpus callosum.

These seven subareas were compared between ex-smokers and non-smokers using a one-way multivariate analysis of covariance (MANCOVA), with age, sex, TIV, and scanner treated as nuisance variables. A significant omnibus test was followed up by seven area-specific post hoc analyses using the same nuisance variables. To assess potential sex differences in the effects of smoking, a group-by-sex interaction was calculated, which failed to reach significance (p = 0.909), indicating that the effects of smoking were similar in males and females. The interaction term was therefore omitted in the main statistical model. In addition to these group comparisons (ex-smokers vs. non-smokers), correlation analyses were performed within the ex-smoking group, specifically relating the seven callosal area measures to (1) the years of abstinence; (2) the years of smoking; (3) the cigarette packs per day; and (4) the pack years (i.e., the cumulative smoking load, calculated as ‘years of smoking’ times ‘cigarette packs per day’). Again, age, sex, TIV, and scanner were treated as nuisance variables. Bonferroni corrections were applied to account for multiple comparisons, yielding an adjusted significance threshold of p $\le$ 0.007 (0.05/7).

The MANCOVA indicated a significant main effect of group (p = 0.032). The subsequent post hoc tests revealed that ex-smokers had a significantly smaller genu (p = 0.015) and posterior midbody (p = 0.002) than non-smokers, and also showed a trend for a significantly smaller isthmus (p = 0.059). However, only the effect within the posterior midbody survived statistical corrections for multiple comparisons (see Table 2).

No significant associations, not even at uncorrected significance levels, emerged between any of the callosal area measures and (1) years of abstinence; (2) years of smoking; (3) cigarette packs per day; and (4) pack years.

The smoking-related literature on the corpus callosum is sparse. Nevertheless, the nature of our findings seems to agree with previous studies reporting smaller callosal volumes [14], greater age-related callosal volume loss [15], as well as an altered callosal fiber integrity [13, 21, 22] in smokers. Another study outside the framework of smoking, points to negative links between callosal fiber integrity and positive drug tests (cocaine) over a period of eight weeks [23]. There is a lack of research on the corpus callosum within ex-smokers (rather than active smokers), but some structural alterations/normalizations have been reported [12, 17], as detailed in the introduction. One of those studies revealed that FA and MD (indicators of white matter fiber integrity) were aberrant in ex-smokers but normalized after 20 years [12] suggesting that negative effects of smoking may reverse eventually. However, this trend might not be true for all brain measures/regions. For example, the current study—where the average smoking cessation time was 31 years—revealed that callosal aberrations persist within the posterior midbody (and perhaps also within the genu and isthmus). In potential support of this assumption, there were no significant correlations between callosal area measures and the number of years abstained from smoking.

A smaller midsagittal callosal area may reflect a reduction in the number of interhemispheric axons and/or a reduction in the axonal fiber diameter, caused for example by decreasing myelin sheaths. As detailed in the following, smoking may affect the brain’s white matter (and as such also the corpus callosum) through different mechanisms, either separately or in conjunction.

Smoking has been shown to lead to progression in white matter hyperintensities proportionate to cumulative smoking load [24]. These smoking-induced hyperintensities are caused by potentially reversible changes in interstitial fluid mobility and water content that become permanent (due to axonal lesions and demyelination) when the smoking continues [25]. Interestingly, smoking has also been shown to affect axonal myelinization without directly damaging the axon by reducing the gene transcription factors of myelin-producing oligodendrocytes [26]. After smoking cessation, less than 25% of this demyelination seems to recover [26]. Finally, smoking drives the release of glutamate, which can be neurotoxic in high concentrations [27], with effects contributing to the gray matter loss in smokers in prefrontal, cingulate, temporal, thalamic, cerebellar and supplementary motor regions [6, 7, 8, 9, 28, 29]. These gray matter losses may result in the degeneration of the respective callosal fibers (i.e., those that connect some of the aforementioned areas) as a direct result of neuron loss.

The corpus callosum was smaller in ex-smokers within the posterior midbody as well as within the genu and isthmus (albeit the latter two only when using less conservative thresholds). Although the literature is sparse with respect to ex-smokers, these findings are in agreement with studies that observed an altered fiber integrity in the genu, posterior midbody, and isthmus of smokers compared to non-smokers [21, 22] as well as a negative association between fiber integrity and cumulative smoking load within the genu [13].

Fibers travelling through the posterior midbody connect the supplementary motor area (SMA) [30], a region associated with gray matter loss in smokers [28]. The SMA comprises part of the neural pathway governing response inhibition, with lower gray matter volumes linked to greater impulsivity [31]. Impulsivity is regarded as a manifestation of inhibitory control dysfunction underlying lapses into addictive behavior [32], such as smoking. Fibers travelling through the callosal genu connect the prefrontal cortex [30, 33], another region associated with gray matter loss in smokers [6, 8, 9, 10], potentially due to neurotoxic glutamate exposure [27] and associated with impaired self-control, executive function and emotion regulation [34]. Fibers travelling through the callosal isthmus connect the inferior temporal cortex, yet another region associated with gray matter loss in smokers [6, 28, 35]. The inferior temporal cortex has been linked to emotional regulation and semantic cognition [35] and any dysfunction may bias the emotional response to smoking-related cues in the environment, leading to the maintenance of addiction [35]. The interplay between these factors is likely complex: A loss of neurons in the aforementioned cortical regions cortex might cause (or be caused by) a loss of their respective transcallosal axons in the corpus callosum, which in turn may further enhance specific (smoking-related) impulses and habits.