The 10-Year Study of the Impact of Particulate Matters on Mortality in Two Transit Cities in North-Eastern Poland (PL-PARTICLES)

Round 1

Reviewer 1 Report

This study examined associations between short-term exposure to PM2.5 & PM10 and mortality in two cities of Poland using a time-stratified case-crossover design. The research and the manuscript need to be improved for publication.

A. Title & Abstract: 

please clarify the study population. “Transit cities in Eastern Europe” is unclear

B. Introduction: 

1. 1. The authors appear to put emphasis on investigations of the health effect of air pollution in the setting of low concentrations. However, there already exist many studies about the health effect of low concentrations of PM, such as cities in the United States, Canada, and Europe. If they wanted to emphasize “rural cities”, please clarify why. (It seems to me that PM2.5 concentration in Lomza and Suwalki is not low). 

1. 1. What is the main purpose of this study? The authors mentioned that “Taking this into consideration we decided to analyze clinical data with both air quality and meteorological data from two small cities over a period of 10 years.” However, given the methods and results, I don’t understand what they actually considered. 

C. Methods: 

1. 1. What is the ICD code for mortality? What is the main outcome of the analysis? Is this all-cause mortality, including external causes? Please clarify
2. 2. I would suggest to include a geographical map with monitoring sites to show the location of study populations and whether monitoring stations well represent exposure to background ambient air pollution.
3. 3. Given Table 5, the authors adjusted PM-mortality association for linear-term of temperature. This is not a standard method in air pollution epidemiology. The authors need to adjust for temperature using a non-linear term such as splines. Also, since the authors explore lags, different lags of temperature should also be considered as a potential confounder.

D. Results:

1. 1. Please clarify abbreviations when they appear first in the text, not the table. CDR/SDR
2. 2. What is the percentage in Table 3? Are there missing values of PM or temperature? How did they treat them? 

E. Discussion: In general, the impression is like over-interpretations. 

1. 1. The authors did not analyze PAH, but their discussion seems focused on PAH. As the authors mentioned, PM is a complex mixture of chemical compositions and different sized particles. Unless the authors (quantitatively) show that PAH is a significant contributor to PM in these study populations, focusing on PAH is inappropriate.
2. 2. The authors tried to explain their results in the context of cause-specific mortality. However, they did not show associations between PM and cause-specific mortality (they only presented the contribution of cause-specific mortality to all-cause mortality). Air pollution studies usually categorize all-cause mortality (except for non-external causes), cardiovascular mortality, and respiratory mortality. The authors should have analyzed associations between PM and those outcomes. I strongly suggest that such categorizations unless they investigate associations between PM and very-specific mortality.
3. 3. The authors need to discuss how exposure misclassification in this type of studies may have affected their findings.

F. Limitations:

1. 1. The authors conducted a time-stratified case-crossover design. Why is the lack of information about tobacco smoking a limitation?
2. 2. What does “it made its modifying effect on particulate matter impossible to count?” Does this mean that gaseous pollutant decreases/increases PM? 

G. English: the authors need to check English throughout the manuscript.

Comments for author File: Comments.pdf

Author Response

Dear Editorial Board and reviewers of JCM,

Thank you very much for considering our manuscript for publication in JCM. We are grateful for the time and effort invested in reviewing our manuscript and for the many thoughtful comments and suggestions from the reviewers. We have considered these comments carefully and feel that we have fully addressed the reviewer’s concerns.  We have replied to each comment below and edited the manuscript accordingly. We believe that this input has improved this manuscript greatly. Complying with all the suggestions, we hope that you will find the manuscript suitable for publication in JCM.

On behalf of the authors
Łukasz Kuźma MD

Review 1:

This study examined associations between short-term exposure to PM2.5 & PM10 and mortality in two cities of Poland using a time-stratified case-crossover design. The research and the manuscript need to be improved for publication.

 We are very grateful for the time and effort invested in reviewing our manuscript.

1. Title & Abstract: 

please clarify the study population. “Transit cities in Eastern Europe” is unclear

As suggested, we have decided to make alterations that will narrow the population in this study by adding a more precise location: north-eastern Poland, Europe. The studied population and city characteristics are described under the “2.2 Region's characteristics“ section.

1. Introduction: 

1. 1. The authors appear to put emphasis on investigations of the health effect of air pollution in the setting of low concentrations. However, there already exist many studies about the health effect of low concentrations of PM, such as cities in the United States, Canada, and Europe. If they wanted to emphasize “rural cities”, please clarify why. (It seems to me that PM2.5 concentration in Lomza and Suwalki is not low). 

Firstly, both of the cities lie in the region commonly known as The Green Lungs of Poland. Theoretically, they should be relatively free of pollution. We wanted to confront this common belief in clean air in relatively small towns, often near rural regions, with air pollution data. As we can see, the results do not support this thesis. Secondly, there is a lack of air pollution studies from Poland. We wanted to fill this gap by conducting a case-crossover study from our region, Podlaskie Voivodeship. We believe that it is important not to concentrate only on big cities, as, according to Polish National Statistical Office, only about 28% of Polish citizens live in cities with population over 100,000 inhabitants.

1. 1. What is the main purpose of this study? The authors mentioned that “Taking this into consideration we decided to analyze clinical data with both air quality and meteorological data from two small cities over a period of 10 years.” However, given the methods and results, I don’t understand what they actually considered.

Our aim was to analyze the short-term impact of PM on all-cause, cardiovascular, and respiratory mortality in district transit cities in north-eastern Poland.

1. Methods: 

1. 1. What is the ICD code for mortality? What is the main outcome of the analysis? Is this all-cause mortality, including external causes? Please clarify

We have added the following information regarding ICD codes in the Study Design section: “ According to codes in the International Classification of Diseases—10th Revision, we extracted the data for cardiovascular-related mortality (ICD-10 from I.00 to I.99), pulmonary-related mortality (ICD-10 from J.00 to J.99).” The main outcome of the analysis is that in the whole studied region despite differences in air quality, the influence of PMs on all-cause mortality was observed. This effect was prolonged up to one and two days after the exposure. Cardiovascular mortality in both of the cities was influenced by PM_2.5 on lag0, whereas PM₁₀ was associated only with higher mortality rate in Suwalki. Pulmonary mortality rate was associated with increase in PM_2.5concentrations only in Lomza on lag1. This is all-cause mortality with the inclusion of external causes.

1. 2. I would suggest to include a geographical map with monitoring sites to show the location of study populations and whether monitoring stations well represent exposure to background ambient air pollution.

We thank the reviewer for this great suggestion, according to it we added such a map, which can be found under "Figure 1".

1. 3. Given Table 5, the authors adjusted PM-mortality association for linear-term of temperature. This is not a standard method in air pollution epidemiology. The authors need to adjust for temperature using a non-linear term such as splines. Also, since the authors explore lags, different lags of temperature should also be considered as a potential confounder.

As suggested by the reviewers we have adjusted our analysis. The association between PMs and the occurrence of deaths was estimated by odds ratios (OR) with 95% confidence intervals (CI) using conditional logistic regression (CLR). Meteorological data including temperature during the same lag period were used as covariate in the CLR model. We used a natural cubic spline with 5 degrees of freedom for the temperature–mortality function. We made separated models for total mortality, cardiovascular mortality and pulmonary mortality.

1. Results:

1. 1. Please clarify abbreviations when they appear first in the text, not the table. CDR/SDR

It has now been corrected.

1. 2. What is the percentage in Table 3? Are there missing values of PM or temperature? How did they treat them? 

The whole study covered 10 years. During this period we recorded 15,568 deaths in the examined cities. There were missing values of PM. We excluded from the study all of the days with missing data of PM. We presented absolute value of days with observations and its percentage of 3653 days (total number of days in the analyzed 10 years). Missing data can also be observed in Figure 2.

1. Discussion: In general, the impression is like over-interpretations. 

1. 1. The authors did not analyze PAH, but their discussion seems focused on PAH. As the authors mentioned, PM is a complex mixture of chemical compositions and different sized particles. Unless the authors (quantitatively) show that PAH is a significant contributor to PM in these study populations, focusing on PAH is inappropriate.

Thank you for your suggestion. We deleted all of the fragments that mentioned PAH. The initial idea was to bring closer the differences between PM origin and its components and PAH was used as an example.

1. 2. The authors tried to explain their results in the context of cause-specific mortality. However, they did not show associations between PM and cause-specific mortality (they only presented the contribution of cause-specific mortality to all-cause mortality). Air pollution studies usually categorize all-cause mortality (except for non-external causes), cardiovascular mortality, and respiratory mortality. The authors should have analyzed associations between PM and those outcomes. I strongly suggest that such categorizations unless they investigate associations between PM and very-specific mortality.

We are grateful for this suggestion. As you suggested, we analyzed associations between PM all-cause mortality, respiratory mortality and cardiovascular mortality. New results lead us to the new conclusions.

1. 3. The authors need to discuss how exposure misclassification in this type of studies may have affected their findings.

We thank the reviewer for this suggestion, we added the last paragraph that discusses exposure misclassification and its influence on our study’s results.

1. Limitations:

1. 1. The authors conducted a time-stratified case-crossover design. Why is the lack of information about tobacco smoking a limitation?

This is a very good point as indeed, each case serves as its own control. However, it would be interesting to separate the group of smokers and non-smokers in order to be able to compare the impact of smoking - the lack of such data in this context was considered a limitation of our study. As this may be misunderstood by potential readers, we decided to remove this from the limitations.

1. 2. What does “it made its modifying effect on particulate matter impossible to count?” Does this mean that gaseous pollutant decreases/increases PM?

This sentence lacked precision. The main thought was that gaseous pollutants might have had a synergistic detrimental influence on mortality. We clarified that in more accessible way.  

1. English: the authors need to check English throughout the manuscript.

As suggested, we have checked the language in the manuscript and made appropriate corrections.

Reviewer 2 Report

As a reviewer I have the following remarks.

1. Abstract: “The analysis with a time-stratified bidirectional case-crossover 18 design was performed.” – I think you don’t need “bidirectional” rather “a time-stratified case-crossover”. PM2.5 – is not defined, also other terms OR, IQR, CI. We know it’s known terms, buy…. You may say fine particulate matters (PM2.5). It can stay as is but usually/professionally the abbreviation is explained. OR are reported for IQR – we don’t know what IQR=? Remove P-value, say IQR value. Spell names in Polish.
2. Line 40: “We can also distinguish ultrafine particles with a diameter below 0.1 μg.” – should be measurements units.
3. Line 60: “The data of air pollutions and gases were obtained from…” but in Line 39: “Air pollution is a complex mixture of particles and gases”. Thus gases are different?
4. Line 62: “In the analysis, we used the concentration of particulate matter…” – why did you introduced gases, probably better to say that only PM are studied. It can stay, but….
5. Line 64: “part of the city - PL0151A” – just I am guessing, is it ID of the monitor station?
6. Line68: “the temperature form”?
7. Line 73: “The study material lacked about 14.07%” – I think better 14.1%.
8. Line 76: “Podlaskie Voivodeship” - In general I suggest to use Polish correct names and in () English translation/meanings.
9. Line 90: “time-stratified bidirectional case-crossover” – already time-stratified defines the used schema. See remarks on Abstract.
10. Line 90. The case-crossover method invented Maclure (ref. 13, but another paper from 1991). I don’t know the paper ref 12, but in my opinion time-stratified approach was proposed in the following work: Janes H, Sheppard L, Lumley T. Case-crossover analyses of air pollution exposure data: referent selection strategies and their implications for bias. Epidemiology. 2005 Nov;16(6):717-26. doi: 10.1097/01.ede.0000181315.18836.9d. PMID: 16222160. I suggest Ref12: Maclure M. Am J Epidemiol. 1991 Jan 15;133(2):144-53. doi: 10.1093/oxfordjournals.aje.a115853. PMID: 1985444/ and Ref13: Janes H, Sheppard L, Lumley…. Maclure invented this method, Janes et al. proposed time-stratified. I read you Ref 12 – nothing on the method you used here.
11. Line 91: “control periods included all days that were from the same day of the week in the same month” – nothing on bidirectional, It’s OK, as I asked to remove this term.
12. Line 96: “temperature as covariates in the CLR model” – linear, spline?
13. Line 107: “The male gender” – I think should be “sex”, we don’t know the gender in this study.
14. Line 109: “Suwalki was 71.68 years (SD=16.56),: - de we need such accuracy? 71.7 – it’s more informative and simpler to read.
15. Table 2. “All, % (N) 100 (7486) 100 (8082)” I think better N, % (one digit after the dot).
16. Simplify the table 3. Reduce the number after the dot. Think about a potential reader.
17. Line 140: “vs.” – ““vs.” (American English) or “vs” (British English) can be used as an abbreviation.”
18. Line 149: “(OR 1.075, 95% CI 1.015–1.138, P=0.018, lag 1) - (OR = 1.075, 95% CI: 1.015–1.138, P=0.018, lag 1). I am not sure that we need P-value as we have 95%CI.
19. In my opinion the paper/notation should be simplified. For example (from the above). “(OR 1.075, 95% CI 1.015–1.138, P=0.018, lag 1) could be written (1.075; 1.015 – 1.138), if in the first use we say OR, 95%CI. Why do you complicate life? 
20. Thank you.
21.  

Author Response

Dear Editorial Board and reviewers of JCM,

Thank you very much for considering our manuscript for publication in JCM. We are grateful for the time and effort invested in reviewing our manuscript and for the many thoughtful comments and suggestions from the reviewers. We have considered these comments carefully and feel that we have fully addressed the reviewer’s concerns.  We have replied to each comment below and edited the manuscript accordingly. We believe that this input has improved this manuscript greatly. Complying with all the suggestions, we hope that you will find the manuscript suitable for publication in JCM.

On behalf of the authors
Łukasz Kuźma MD

As a reviewer I have the following remarks.

1. Abstract: “The analysis with a time-stratified bidirectional case-crossover 18 design was performed.” – I think you don’t need “bidirectional” rather “a time-stratified case-crossover”. PM2.5 – is not defined, also other terms OR, IQR, CI. We know it’s known terms, buy…You may say fine particulate matters (PM2.5). It can stay as is but usually/professionally the abbreviation is explained. OR are reported for IQR – we don’t know what IQR=? Remove P-value, say IQR value. Spell names in Polish.

We thank the reviewer for this suggestions. Bidirectional refers to the selection of control days. Unlike the unidirectional retrospective control sampling (selecting control times only prior to death) - in our model, which we wanted to highlight, control days are days from the same month. However we decided to remove this term from the manuscript. Additionally, we have inserted the abbreviation extensions in the Abstract. Lastly, cities names are now in correct Polish spelling.

1. Line 40: “We can also distinguish ultrafine particles with a diameter below 0.1 μg.” – should be measurements units.

We apologize for missing this, indeed, it should be measurements units - μm. This has been addressed in the revised version.

1. Line 60: “The data of air pollutions and gases were obtained from…” but in Line 39: “Air pollution is a complex mixture of particles and gases”. Thus gases are different?

It was a simple error in the use of words. We removed the information about the gases in line 60 because, as it was rightly noted, it was unnecessary to mention them there. The word we wanted to use there, and at the same time this will answer question number 6, was “temperature”. We apologize for the this mistake and thank the reviewer for noticing this, it has now been corrected

1. Line 62: “In the analysis, we used the concentration of particulate matter…” – why did you introduced gases, probably better to say that only PM are studied. It can stay, but….

We agree with this suggestion, we decided to remove the mention of gases from the Study Design section and focus only on the thing we studied - PMs.

1. Line 64: “part of the city - PL0151A” – just I am guessing, is it ID of the monitor station?

Yes, that is in fact the ID of the monitor station. Now it is clearly shown on the map attached to the manuscript, which demonstrates the names of the stations and their locations in parts of cities.

1. Line68: “the temperature form”?

We thank the reviewer for noticing this typo, it has already been corrected. The word "from" should be there instead od “form”.

1. Line 73: “The study material lacked about 14.07%” – I think better 14.1%.

It has now been corrected.

1. Line 76: “Podlaskie Voivodeship” - In general I suggest to use Polish correct names and in () English translation/meanings.

According to the suggestion of using correct Polish names, we changed the names "Lomza" to Lomza and "Suwalki" to Suwalki throughout the manuscript. However, as far as the name of the voivodeship goes, “Podlaskie” Voivodeship is written in the same way in Polish and English.

1. Line 90: “time-stratified bidirectional case-crossover” – already time-stratified defines the used schema. See remarks on Abstract.

It has now been corrected.

1. Line 90. The case-crossover method invented Maclure (ref. 13, but another paper from 1991). I don’t know the paper ref 12, but in my opinion time-stratified approach was proposed in the following work: Janes H, Sheppard L, Lumley T. Case-crossover analyses of air pollution exposure data: referent selection strategies and their implications for bias. Epidemiology. 2005 Nov;16(6):717-26. doi: 10.1097/01.ede.0000181315.18836.9d. PMID: 16222160. I suggest Ref12: Maclure M. Am J Epidemiol. 1991 Jan 15;133(2):144-53. doi: 10.1093/oxfordjournals.aje.a115853. PMID: 1985444/ and Ref13: Janes H, Sheppard L, Lumley…. Maclure invented this method, Janes et al. proposed time-stratified. I read you Ref 12 – nothing on the method you used here.

We thank the reviewer for this comment. We have decided to change these references in the bibliography to the suggested.

1. Line 91: “control periods included all days that were from the same day of the week in the same month” – nothing on bidirectional, It’s OK, as I asked to remove this term.

It has now been adjusted according to previous question.

1. Line 96: “temperature as covariates in the CLR model” – linear, spline?

As suggested by the reviewers we have adjusted our analysis. The association between PMs and the occurrence of deaths was estimated by odds ratios (OR) with 95% confidence intervals (CI) using conditional logistic regression (CLR). Meteorological data including temperature during the same lag period were used as covariate in the CLR model. We used a natural cubic spline with 5 degrees of freedom for the temperature–mortality function. We made separated models for total mortality, cardiovascular mortality and pulmonary mortality.

1. Line 107: “The male gender” – I think should be “sex”, we don’t know the gender in this study.

Thank you for this important suggestion. We fully agree with this and have made the appropriate changes to the manuscript.

1. Line 109: “Suwalki was 71.68 years (SD=16.56),: - de we need such accuracy? 71.7 – it’s more informative and simpler to read.

As we have changed the values of numbers to 1 decimal in table number 1, there has also been a change in the text - as suggested - the age number is now 71.7

1. Table 2. “All, % (N) 100 (7486) 100 (8082)” I think better N, % (one digit after the dot).

In this line the order of % and N was reversed, this error has already been fixed in the manuscript and the numbers in the table are in the correct order - % (N).

1. Simplify the table 3. Reduce the number after the dot. Think about a potential reader.

As suggested, we simplified tables 1-3 to 1 decimal place.

1. Line 140: “vs.” – ““vs.” (American English) or “vs” (British English) can be used as an abbreviation.”

Thank you for that note. We have unified the entire manuscript to the American English form of vs.

1. Line 149: “(OR 1.075, 95% CI 1.015–1.138, P=0.018, lag 1) - (OR = 1.075, 95% CI: 1.015–1.138, P=0.018, lag 1). I am not sure that we need P-value as we have 95%CI.

Thank you for this comment, we are aware that the confidence interval indirectly indicates relevance, but we have decided to include P for the clarity of the message also for those less familiar with statistics. Additionally, to show that some results are statistically highly significant with P<0.001

1. In my opinion the paper/notation should be simplified. For example (from the above). “(OR 1.075, 95% CI 1.015–1.138, P=0.018, lag 1) could be written (1.075; 1.015 – 1.138), if in the first use we say OR, 95%CI. Why do you complicate life? 

We thank the reviewer for these suggestions. We simplified the manuscript by rounding the numbers to 1 decimal.

1. Thank you.

We are very grateful for the time and effort invested in reviewing our manuscript.