Comparison of color discrimination in chronic heavy smokers and healthy subjects

Background: Cigarette smoke is probably the most significant source of exposure to toxic chemicals for humans, involving health-damaging components, such as nicotine, hydrogen cyanide and formaldehyde. The aim of the present study was to assess the influence of chronic heavy smoking on color discrimination (CD). Methods: All subjects were free of any neuropsychiatric disorder, identifiable ocular disease and had normal acuity. No abnormalities were detected in the fundoscopic examination and in the optical coherence tomography exam. We assessed color vision for healthy heavy smokers ( n = 15; age range, 20-45 years), deprived smokers ( n = 15, age range 20-45 years) and healthy non-smokers ( n = 15; age range, 20-45 years), using the psychophysical forced-choice method. All groups were matched for gender and education level. In this test, the volunteers had to choose the pseudoisochromatic stimulus containing a test frequency at four directions (e.g., up, down, right and left) in the subtest of Cambridge Colour Test (CCT): Trivector. Results: Performance on CCT differed between groups, and the observed pattern was that smokers had lower discrimination compared to non-smokers. In addition, deprived smokers presented lower discrimination to smokers and non-smokers. Contrary to expectation, the largest differences were observed for medium and long wavelengths. Conclusions: These results suggests that cigarette smoking, chronic exposure to its compounds, and withdrawal from nicotine affect color discrimination. This highlights the importance of understanding the diverse effects of nicotine on attentional bias.

Cigarette smoking is still a major source of exposure to chemicals that are toxic for humans. The compounds in cigarettes and cigarette smoke, such as nicotine, oxygen dioxide and formaldehyde, are highly harmful to health 1 . Data from the World Health Organization (WHO) hypothesize that by 2030, cigarettes could kill nearly 9 million people a year around the world 2,3 .
Cigarette nicotine deprivation in chronic users may impair cognitive and attentional abilities even after long time of cessation 4,5 . The neurotoxic effects of chronic use and smoking abstinence on the nervous system have not been extensively studied [6][7][8] . However, chronic cigarette smoking increases cardiovascular response 9 , which, in turn, affects retinal responses through altered blood flow 10 . In addition, tobacco compounds may increase free radical that would cause macular degeneration along with the action of ischemia 11 . Whereas smoking effects on color vision are understudied, the existing data are controversial and highlights the importance of a rigorous testing procedure that measures color discrimination 12,13 . Thus, to identify the mechanisms underlying neurotoxic smoking effects on multisensory integration, we need to understand how smoking may alter early visual processing.
A visual percept may consist of stimuli that vary over the space (spatial contrast), time (temporal contrast) or direction of motion, and vary in luminance (achromatic) and chromaticity (saturation and hue color) 12,14,15 . Thus, chromatic contrast involves chromaticity differences, which can be expressed by the distance in the CIE 1976 uniform chromaticity scale diagram and assessed by thresholds of vectors on the Cambridge Color Test (CCT), for example 16,17 . It has the advantage of being used to evaluate in detail whether these anomalies are due to congenital factors or acquired conditions 16,17 .
We base our rationale on the premise that chronic exposure to nicotine will led to receptor desensitization and not suffer influence of arousal and increase in attentional resources in smokers 18 . The purpose of the present study was to assess the influence of chronic heavy smoking on color discrimination (CD).

Participants
In this study, 15 non-smokers (mean age = 32.5 years; SD = 9.1; 7 male), 15 cigarette smokers (mean age = 32.1 years; SD = 5.7; 7 male) and 15 deprived smokers (mean age = 31.9 years; SD = 6.3; 7 male) who were staff or students at the Federal University of Paraiba were recruited through newspaper advertisements. The participants were 25-45 years old. The participants were excluded if they had any one of the following criteria: younger than 20 and older than 45 years (since effects of aging in the human visual system could superestimate the results 19,20 ); current history of neurological or cardiovascular disease; a history of head trauma, color blindness, current or previous drug abuse and current use of medications that may affect visual processing and cognition. Subjects were required to have good ocular health, with no abnormalities on fundoscopic examination or optical coherence tomography examination. All of the subjects were screened for color blindness using the test of Ishihara for color deficiency, and had normal or corrected-to-normal vision as determined by a visual acuity of at least 20/20.

Amendments from Version 2
The study was adapted according to other publications involving Trivector and English edition. The "cut-off" section of the Ms methods was removed, and some lines in introduction and methodology were added (and corrected).

REVISED
The four-alternative forced-choice 16,22 (4-AFC) method was used, and the subjects' task was to identify, using a remote control response box, whether the Landolt 'C' stimulus was presented at the left, right, up or down side of the monitor screen. The participants were also instructed to respond whether they could not identify the stimulus gap 16 . After each correct answer, the chromaticity of the target proceeded closer to that of background, while each wrong answer or omission was followed by the presentation of the target at a greater chromatic distance from the background. After each correct answer, the chromaticity of the target proceeded closer to the background. Each incorrect answer or omission was followed by presentation of the target at a greater chromatic distance from the background. The experiment ended after 11 reversals for each axis and the threshold was estimated from the six final reversals 23 .
The trivector testing protocol estimates sensitivity for the short, medium and long wavelengths through the protanopic, deuteranopic, and tritanopic confusion axes, respectively 23,24 . Trivector protocol uses vectors as central measurement. In general, we used a default setting where the Landolt 'C' had an opening at 1° of visual angle, minimum luminance of 8 cd/m², maximum luminance of 18 cd/m², 6 s of response time for each trial and distance of 269 cm between participant and monitor screen.

Data analysis
The distributions for each group were compared with Shapiro-Wilk.Both groups showed non-normal distribution; therefore, nonparametric statistical methods were used to analyze the data. For group comparisons, the non-parametric univariate analysis was used, with pairwise comparisons by Mann-Whitney U test. Spearman's rank correlation coefficients (rho) were conducted to assess the relationship between outcomes of color discrimination data and biosociodemographic variables, such as age, gender and level of education. All the calculations were made using SPSS ® , version 21.0.
The effect size (r) estimation was used from the conversion of z-scores 25,26 . Values above .50 are considered as large effect size.
Results are presented as medians. Center lines show the medians; box limits indicate the 25th and 75th percentiles as determined by SPSS software; whiskers extend 1.5 times the interquartile range from the 25th and 75th percentiles (ends of the whiskers are the maximum and minimum values). When presented, errors bars represent standard deviations (SD) of the median based on 1000 bootstrap resamplings. Bonferroni correction was the method of adjusting the P-value that we used. P < 0.016 was accepted as statistically significant for multiple comparisons and P < 0.025 for pairwise comparisons.

Results
Color discrimination thresholds were obtained in u'v' units of the CIE 1976 color diagram, for protan, deutan, and tritan axes, respectively. Nonparametric analysis were carried out showing that there were significant differences in discrimination thresholds between groups along the protan (χ²(2) = 26.53, P < 0.001), deutan (χ²(2) = 22.40, P < 0.001) and tritan (χ²(2) = 14.93, P < 0.001) confusion axes. Thresholds for the smokers and deprived smokers were higher than the normative values observed in other studies. Therefore, there was a reduction in color discrimination in both groups. The results of the trivector measurements are shown in Figure 1.

Discussion
The data indicated that smokers groups, as a whole, had higher discrimination thresholds when compared to non-smokers (P < 0.05), indicating the existence of a diffuse impairment in visual processing. Results showed good agreement between the normative data of control groups, being the protan and deutan thresholds lower than tritan thresholds, a pattern repeatedly observed in adults tested with the CCT 17 . Moreover, the higher thresholds observed in the group of smokers and deprived smokers are in agreement with the differences observed in other studies using CCT. The effect sizes reached medium to high values.
Small differences in blue-yellow color processing suggest that sensor neurons responsive to the short wavelength may differently operate from those responding to medium and long wavelengths 27 . Indeed, the koniocelular pathway may not suffer from the influences of tobacco components.
Along the trivector protocol, smokers had more errors in protan and deutan confusion axes ( Figure 1). An effect size analysis confirmed that smokers had the largest discrimination errors for protan (r = -85) and deutan (r = -82) confusion axes when comparing against non-smokers. As stated, this result does not support the idea of channel selectivity. However, we base our rational on the existence of diffuse processing impairment, which may include magno-and parvocellular pathways 28 .
Nicotine enhances neurotransmission release through modulation of nicotinic acetylcholine receptors (nAChRs) located in the cortex 18,29 . There are also nAChRs and dopamine receptors on the retina; therefore, the chronic use of cigarette would enhance attentional resources 30-32 . However, there were no improvements in color discrimination. One may argue this could be due to desensitization effect, one of many brain changes caused by addiction 33 . Chronic nicotine exposure leads to nAChRs desensitization through brain upregulation 34,35 . The more exposure, the greater the need for it activate the receptors 36,37 . Whereas nicotine enhancing effects remain unchanged after chronic exposure, this may explain the lower discrimination, but the small similarity, between smokers and non-smokers in some of our data ( Figure 1).
Then, why did the deprived smokers group have less discrimination?
The withdrawal effect, which affect neurotransmission release 38,39 , reflecting both visual processing 40-42 and brain reward function 43 , may explain this. Visual attention plays a role for detection of environmental stimuli 44 .
As stated, impairments observed at color discrimination can occur due to cones saturation, amplification of the noise that reach visual cortex or by the action of nicotine in parvocelular pathway 45 .
In agreement with studies, color vision impairments may be related to ventral stream, which processes color 46 . However, our tests used pseudoisochromatic stimuli. Thus, color discrimination may have occurred through dorsal and ventral stream 28 . Too soon to conclude anything, but there may be nAChRs in both dorsal and ventral stream 47 and both streams may suffer from the action of neurotrasmission hypofunction, affecting directly visual processing 39-41,43 .
Knowing the existence of the expression of nAChRs in bipolar, amacrine and ganglionar cells 29,47,48 , we suggest that smoking affects visual processing, regardless of deprivation. Although the differences between smokers and nonsmokers were small, we could not ignore the existence of many harmful compounds to vision in cigarettes. As noted in others studies, exposure to cigarette smoking 49-55 and solvents 56,57 affects vision. Thus, smoking can be harmful even for passive smokers.
Our limitations need to be considered. We evaluated cigarette smoking as a whole, not the nicotine-only effects 51,52 . Which brings us to the idea of further studies, using nicotine gum and the same paradigm used here. Clearly, further work is needed, but this study highlights the relationship between smoking and color discrimination, involving short, medium and long wavelengths 27 . We conclude that cigarette compounds affect vision 54,55 more than nicotine separately 58-61 .

Data availability
Dataset Author contributions TM: design of the work, data collection and interpretation, and drafting the article. NL: data reanalysis and interpretation and revision of the updated manuscript. NA: design of the work, data analysis and interpretation, and critical revision of the article. All authors approved the final version to be published.

Competing interests
No competing interests were disclosed. From my point of view, the authors have made relevant changes to the paper and have met all the requests previously suggested. In a way that was reviewed by the authors, it was clarified the cut-off points used for the color discrimination, contributing to a better understanding of the results. In addition, the discussion has improved. Thus, I believe that now the paper should be accepted for publication. Goro Maehara Department of Human Sciences, Kanagawa University, Yokohama, Japan I found the manuscript was improved but still have some concerns.

Outcome Cutoffs
There are some difficulties to understand the Outcome Cutoffs section. The authors should state clearly how they determined and calculated the cutoff points. I also suggest explaining what the values in parentheses represent and what IQR means (interquartile range). Although the authors mentioned '(mean + median + IQR + upper and lower limits)', the plus signs in this phrase seem a bit odd.

The Results Section
Please clearly compare the thresholds with the cutoffs values. Not only the median thresholds for smokers, but those for normal observers also exceeds the cutoff values in the deutan and tritan tests, don't they? I think that the cutoff values might be too strict (low).
No competing interests were disclosed. Competing Interests:

I have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.
Author No matter how small the visual processing loss is, the differences that will be classified as impairments will be the statistically significant differences. If it was not clear, I will summarize. Conjecturing: If in one study, for the protan axis, the control group had 50 10 u' v' units and the experimental one had 80 10 u' v' units, having a value of p <.001, this is considered color impairment. Thus, the experimental group would present damage in color processing. This is what indicates most, if not 90% of all articles that used the Trivector as a measure. I hope it has helped in understanding how the Trivector test works. Even so, i will try to answer your questions below:

Outcome Cutoffs
There are some difficulties to understand the Outcome Cutoffs section. The authors should state clearly how they determined and calculated the cutoff points. This was explained. The mean or median (+ SD or + IQR) for the control group of all the studies that used the Trivector were summed and divided by the number of studies (a simple grand mean / median).

I also suggest explaining what the values in parentheses represent and what IQR means (interquartile range). "Based on previous studies, cutoff points (median + IQR/2) for the trivector test were established to
" As stated, median + IQR/2 was established by Paramei designate color vision impairment. et al. (2012). This was also explained. In my opinion, there is no need to explain why using IQR/2 or IQR, since they are simply representing the standard deviation in a non-parametric data

Although the authors mentioned '(mean + median + IQR + upper and lower limits)', the plus signs in this phrase seem a bit odd.
I understand your point of view. However, the reason for inserting this comment into the text was to explain to the reader that there are normative values for the Trivector. These values would be means, medians, tolerance values, etc. They can be observed in the reference article quoted in the cutoffs outcome section.

The Results Section Please clearly compare the thresholds with the cutoffs values. Not only the median thresholds for smokers, but those for normal observers also exceeds the cutoff values in the deutan and tritan tests, don't they? I think that the cutoff values might be too strict (low).
As explained, the use of cutoff values was only one way to meet the requests of the reviewers. However, in no study using trivector cutoff values are imposed. This is due to the fact that to consider impairment in color vision it is only necessary to have statistically significant differences and to have medium or high effect size.
Nevertheless, the medians of our results for the control group are "beating" with the cutoff values + IQR/2. So, the purpose of this comment was to try to show how the trivector works. Just read one or two papers that made use of the test and you will note that the differences need not exceed the set values of 100, 100 and 150 10 u' v 'units to prove impairments at color vision. Even differences observed in values smaller than 100 10 u'v 'units are significant differences, since this is a robust and reliable test.
According to Trivector's studies, there is no large or small differences, there is statistically significant differences, with large or small effect sizes. And this was what we observed in our study.
Thank you for the review.
Kind Regards, The authors have solved the major concern in my 1st response through the revision process.  set the cutoff values relatively high probably because the test was originally designed for detecting color blindness.
Here we need to tackle the issues about the present cutoff values.
I agree with another reviewer in that the standard of normality is useful for making a comparison between the present and previous studies. Any analysis in a paper must be logical even if authors think a specific analysis is not very important. In the present study, not only the thresholds for the smoker, but also those for normal controls exceeded the cutoff values for the tritan test (and deutan test, too?). The authors need to somehow solve this logical problem.
Thank you for the explanations on the cutoff values. That is, the cutoff values were defined as the average of normal observers' (median threshold + IQR/2) reported by the previous studies (or only by Paramei?). It will be helpful for readers to rewrite the Outcome Cutoff section as the authors commented above.
The cutoff values of (median threshold + IQR/2) mean that 25 % of the population is classified as abnormal. It seems reasonable for me to increase the cutoff values because 25 % of the population is too much. This might be a possible solution.
Another possible solution could be discussing why the thresholds for normal observers exceeded the cutoff values in the Discussion section. The authors could also emphasize the importance of statistical differences there.
If the authors have any idea for workable solutions, please feel free to let me know. The article investigates and compares color discrimination in chronic smokers and healthy individuals. The authors found a lower significant color discrimination in chronic smokers.
However, I believe that in order to clarify the cut-off point of color discrimination, authors should write in the standard of normality used to rate low or high color discrimination. In addition, in the Methods description of the , for better understanding, the authors should explain better the results (which are results below the expected) and then the significance of them.
The results suggest a possible, even small, important change to the color discrimination of smokers which deserves attention and should be better studied. Therefore, I believe that the paper should be accepted for publication with the aforementioned suggestions. We will try to answer your questions below: However, I believe that in order to clarify the cut-off point of color discrimination, authors 1.

5.
However, I believe that in order to clarify the cut-off point of color discrimination, authors should write in Methods the standard of normality used to rate low or high color discrimination. We appreciate the suggestion. These changes were made. We stand by calculating the average of the control group of all studies using Trivector in Brazil and using the normative data for age groups used by Paramei et al. In addition, in the description of the results, for better understanding, the authors should explain better the results (which are below the expected) and then the significance of them. We appreciate the suggestion. Although the description of Trivector's results is quite directive (for details, see ), we agreed that the way the results were presented was below expectations.These changes were made.
The results suggest a possible, even small, important change to the color discrimination of smokers which deserves attention and should be better studied. In agreement. Based on this suggestion, we added a few paragraphs on the importance of the present study, see Introduction and Discussion.
1. The thresholds for chronic smokers According to  , "normal limits for performance for first examination on the basic "Trivector" test are 100 (protan), 100 (deutan) and 150 (tritan)." The threshold medians for chronic smokers were lower than those values, except for the deutan threshold median for deprived smokers. Although there were statistically significant difference between chronic smokers and non-smokers, the thresholds for non-smokers were very low (about 40, 50, and 80 in 10 u' v' units for protan, deutan, and tritan, respectively). In addition, the differences in thresholds (about 60) make little change in color appearance. Taken together, it seems difficult to conclude that smoking impairs color discrimination abilities.
Minor concern 1. Equation 1 I am not sure why Weber contrast (equation 1) needs to be explained in the Methods section. The authors should state clearly how they used the equation in their experiment.

Results
The authors just listed the statistical results in the Results section. I suggest describing the results in the more detailed way (ex. thresholds were higher for smokers than non-smokers, = 132, = 0.002, = U P r -.61).

3rd paragraph in the Discussion section
We will try to answer your questions below: "Although their methods were scientifically sound, the thresholds for chronic smokers were not high enough to conclude that smoking impairs their color discrimination abilities." Although we did not observe large differences between the thresholds of the control group and the group of smokers, we agree with Lakens (2013) and Field (2013) noting that the statistically significant differences are consonant with the related effect sizes which fluctuated between mid-to-high values ( values between .50 and .61; chronic smokers x controls). Moreover, when r comparing the group of smokers with deprived smokers, we observed that these differences were not so large ( values reaching .50). We do not know if the smoking habit, cigarette compounds or r smoking , are responsible for the decrease in color discrimination. per se But there was a loss of color discrimination, suggesting the idea that visual color processing may be diffusely impaired in smokers (Besson et  Many thanks for the review. Based on our expertise in the use of Cambridge Colour Test: the minor the threshold, better . If a group (in this case, smokers group) has a higher threshold, this means that discrimination they needed more chromatic contrast to detect the stimuli. Thus, higher thresholds means (Hasrod & Rubin, 2015).

lower discriminiation along confusion axes
After the publication of Mollon and Reffin's about the CCT (2000), several studies using Trivector have been published. Including preliminary norms for the use of CCT (Ventura et al., 2003), which was considered by the creators of the CCT on the software website (http://www.crsltd.com/tools-for-vision-science/measuring-visual-functions/cambridge-colour-test/) We re-checked our Trivector data and compared to several studies and we observed that we're with similar values for control groups. If another group (such smokers or deprived smokers) had higher thresholds, it means that they differ from the standard values and are likely to have color vision impairments. Taken together, we can not ignore that, although not as large, the differences were significant in this sample. Based on previous studies, even though there are small differences, they need to be punctuated, since we agree that this is an important area that requires further research.

Minor concern
1. Equation 1; I am not sure why Weber contrast (equation 1) needs to be explained in the Methods section. The authors should state clearly how they used the equation in their experiment.
Ops! Many thanks! Since CCT already uses this default setting, we strongly agreed with your review. These changes will be in the second version of the manuscript (we will remove it).

Results;
The authors just listed the statistical results in the Results section. I suggest describing the results in the more detailed way (ex. thresholds were higher for smokers than non-smokers, U = 132, P = 0.002, r = -.61).
Many thanks again. We believe that the way you suggested will facilitate the reader's understanding and will enhance the scientific level of our writing. We fixed it. They will be more descriptive in the next version 3. 3rd paragraph in the Discussion section; The authors stated that "smokers were more sensitive to protanopic and deutanopic confusion axes." This sentence is confusing. Which does this sentence mean, "more sensitive than non-smokers" or "more sensitive than to tritanopic axes"? We appreciate the suggestion and agree that the use of two forms of explanation may actually confuse the reader. We will correct this.
However, when we mention that the smoking group was more sensitive to the protanopic or deutanopic axes, we simply mean that they made more errors than the control group, for example. That is, they needed more chromatic contrast (they were more sensitive) than the comparison group. The confusion axes refer to the red (protanopic), green (deutanopic) and blue (tritanopic) axes. Thus, if any group was more sensitive to the red confusion axis, forr example, it means that they possibly had impairments in the processing of this wavelength.
In this way, based on the apointments above, the status is honestly inconsistent 'not approved' with the content of the work. We ask you to reconsider your decision and we are grateful for the comment, reading, and review of the manuscript. Based on your suggestions, substantial changes were made.
We inserted a subsection in the methods where we explained the cutoff points (where the results would be normal and where the discrimination losses would be).
In addition, we better describe the results section, making it clear to the reader that the higher the threshold, the lower the color discrimination. Also, based on your last suggestion, we added a few paragraphs about the relationship between the harmful cigarette compounds and the damage they cause to the retina, and consequently, visual processing. I hope we have answered the suggestions. Again, we ask you to reconsider your decision about the status of our work.
Best regards, None. Competing Interests: