Reviewer's report
Title: Correlation of respiratory symptoms and spirometric lung patterns in a rural
community setting, Sindh, Pakistan: a cross sectional survey
Version: 1 Date: 26 August 2012
Reviewer: Nigel Bruce
Reviewer's report:
MAJOR ESSENTIAL REVISIONS
1. Abstract/methods: please provide information on number (%) male and females, and
some information on age distribution. Other revisions following on comments below on
assessing the validity of the questionnaire would need to be reflected in the abstract.
Response:page2
Description regarding gender, age, linear regression, validity and other revision mentioned
in the comments ahead has been added in the abstract.
2. Background/methods: the authors make a good case for having valid screening
questionnaires for the settings they describe in Pakistan, and in other low-income
countries. It is important however to formulate the research question more clearly, in
respect of testing the validity of the questionnaire as means of identifying subjects with
potentially important lung function deficit and developing or developed COPD. To do
this, they need to assess the questionnaire using standard measures of validity, that is
sensitivity and specificity (and predictive value), and examine not just the correlation,
but the extent to which the questionnaire detects people with significant lung function
deficit, and also the degree of detection of false positives. This is possible with the data
they have, albeit not with great statistical power – but it is recognised this is small, pilot
study. The authors should consider revising the approach to assessing validity of the
questionnaire.
Response: Validity assessment done and results are given. Research question (mentioned
in the last paragraph of introduction) revised to include the validity assessment of
questionnaire as follows: page 4, 5
The objective of this study was to determine the correlation of respiratory symptoms and
spirometric lung patterns as well as to determine the validity of the ATS respiratory
questionnaire in a rural community setting of Sindh, Pakistan.
3. Methods/respiratory questionnaire: generally definitions of symptoms are clear, but
for wheeze – which is a term for which the meaning is notoriously difficult in most
languages however it is defined – needs some further description. The authors should
explain what measures were taken to check the comprehension of the term (as
whistling sound heard on expiration) in the local Sindhi language.
Response: Following description added to the methods section under the heading of
respiratory questionnaire: page 6
During the data collection the study team was particularly trained to make sure that the
respondents were able to truly comprehend the meaning of all questions, especially those
pertaining to presence of respiratory symptoms, such as wheeze.
4. Methods/lung function. Can the authors explain what predicted values were based
on, and how appropriate they were to the population studied. Also what criteria were
used for quality assessment of the spirometry manoeuvre – for example ATS
repeatability criteria?
Response: page 6-7
As mentioned in methods section under the heading of lung function measurements
Predicted values were based on age, height and sex
Quality of spirometry manoeuvre was assessed using ATS repeatability criteria
description for which has been added in methods section under the heading of lung
function measurement last paragraph as follows:
Participants were explained the procedure of spirometry in detail and were allowed to
practice until they felt comfortable. Participants were asked to refrain from smoking for
at least 1 hour prior to the procedure. Spirometry was conducted in standing position
without nose clips and ATS repeatability criteria were used for quality assessment of the
spirometry maneuver. Results of three acceptable readings were recorded and best of the
three readings was used for analysis in this study. Spirometry was conducted by a trained
technician who was also well versed with local language.
Furthermore, spirometry reports were verified by a pulmonologist description for which
has been added in the last lines of the above mentioned paragraph as follows:
Spirometry reports were further assessed by a pulmonologist for quality assurance.
5. Methods/statistical analysis/paragraph 3: The use of logistic regression is not
inappropriate, but a more efficient alternative may be to use linear regression, with the
lung function value as the dependent variable, which would show the average
decrement in lung function associated with the presence of the symptom. On the other
hand, in order to assess validity as recommended above, it would be helpful to have a
dichotomous outcome. The authors should however explain the rationale for the cutoff values used in defining the dichotomous lung function outcome variables. One way
to address this question of which outcome to use, is to use both continuous (for
average effect) and dichotomous (for validity assessment) outcomes.
Response: page 7-8 and 14
Linear regression (table 4) and validity assessment (table 5) done as suggested by
reviewers.
6. Results/paragraph 1: the results for FVC and FEV1 are stated to be in millilitres, yet
the values of less than 100 are far too small (one would expect average values for both
of these lung function measures to be in the range 2,000 to 3,000 ml for a broad age
group of adults). Are these % predicted values? This needs clarification.
Response: page 8
Revised as suggested. These are % predicted values.
7. Discussion/Paragraph1: In the Discussion, the authors need to pay careful attention
to reporting which findings were significant, and which were not. In respect of this
(see comment on Table 3), p-values should be given to 3 significant places (not 2).
Also, it is not clear that findings for asbestos workers are relevant to this study
population.
Response:
p values reported to 3 decimal places. (Page 3)
Asbestos workers’ reference has been removed
8. Discussion/paragraph 4: the authors discuss the higher prevalence of asthma in their
study, in comparison with the WHO (WHS) data. However, the value of
5.5% has a 95% CI of 2.34 – 8.66, which includes the range quoted for the WHS. This
should be recognised, and less made of the difference, for which random sampling error
is one (likely) explanation. (Same paragraph) Also, what is your reasoning for stating
that you believe ‘these estimates may be generalisable to similar settings in rural
Pakistan? You may well be right, but this needs some more justification.
Response: page 17
The description in 3rd paragraph regarding WHO comment has been rephrased as follows.
Furthermore, comment regarding random samling error has also been incorporated:
Prevalence reported in this study is somewhat similar as compared to that reported using
similar definition by WHO for Pakistan and neighboring countries like India, Bangladesh
and Srilanka where it ranged between 2.6 to 3.16.
Regarding comment on generalizability, justification is provided as follows:
Although our study reports the prevalence of asthma from a localized community setting
of a rural area, we believe that these estimates may be generalizable to similar rural
settings in Pakistan along the river Indus.
9. Discussion/paragraph 5: the discussion at the end of this paragraph relates to the
earlier comments on testing the validity of the questionnaire against LFT results, and
should include assessment of false positives (that is, the number and % of people with
symptoms who have lung function in the ‘good’ range).
Response: the paragraph has been revised as suggested, as follows: page 18
Spirometry has an added advantage of identifying impaired lung function among those
who do not have manifest symptoms and it may not be possible to capture such
impairment by use of questionnaire alone, as indicated by the results of our study, that
shows low sensitivity and negative predictive values while high specificity and positive
predictive values for all the frequent or less severe as well as chronic or more severe
symptoms in the ATS questionnaire.
10. Table 1: it would be very helpful to have a breakdown of some of these key
variables by sex, even though it is recognised that the numbers are relatively small.
However, given that probably few women smoke (and men do), but women are
exposed to wood/dung smoke in the kitchen, this sex disaggregation would be
important to see. Also, note earlier comment on the values for FEV1 and FVC which
cannot be in millilitres. Finally, the n=188 implies some lsot data probably due to
inadequate spirometry – this needs to be described clearly as to how many were
inadequate, and how this was determined.
Response:
Breakdown by sex shown in table 1 (page 10)
Description regarding fuel and kitchen type has been added in the results section 1st
paragraph. (page 8)
Reason for lost data has been added in the results section 2nd paragraph, which was poor
quality of spirometry as determined by pulmonologist. (page 8)
11. Table 2: Denominators should be provided in this table
Response: page 11
Revised as suggested
12. Table 3: P-values should be given to 3 significant figures, e.g. p=0.002, or p<0.001,
or p=0.045. The issue of numbers mentioned under Table 1 is also important for the
analysis of restrictive (n=118) and obstructive (n=125) patterns. First, given that these
use the same two criterion measures, it is unclear why the numbers available differ.
Second, that the numbers are far lower than the 188 quoted in Table 1 indicates that
many subjects failed to produce adequate values for both FEV1 and FVC. This also
needs thorough description and discussion.
Response: page 13
P values changed to 3 decimal places
Analysis for restrictive and obstructive lung patterns was re-visited. After correction of
missing values, a composite variable containing both the obstructive and restrictive
disease was created and analysis was re-run with variables as the dependent variable
(table 3).
MINOR ESSENTIAL REVISIONS:
13. Generally the English is clear and well written, but the authors need to address the
frequent lack of use of the definite article (the) and in some cases the indefinite article
(a/an) where this really is needed to improve the flow of the text, and seek advice on this
if needed.
Response: Revised.
DISCRETIONARY REVISIONS
14. One of the issues not addressed by the paper is the reasons for the symptoms and
lung function values seen in some subjects. This was not the objective of the paper, and
it is noted that the role of arsenic is included in analysis. It would however be of interest
to examine this, and one factor that is not mentioned at all is the use of biomass fuels
and traditional stoves. Were any data collected on cooking fuel and stove? If so, can this
be included at least in descriptive analysis (e.g. in Table 1), to provide background
information on this important cause of COPD in women, especially non-smoking
women.
Response: description regarding biomass use has been provided. (page 8)
Level of interest: An article whose findings are important to those with closely related
research interests
Quality of written English: Needs some language corrections before being
published
Statistical review: No, the manuscript does not need to be seen by a
statistician.
Declaration of competing interests:
I declare that I have no competing interests
Reviewer's report
Title: Correlation of respiratory symptoms and spirometric lung patterns in a rural
community setting, Sindh, Pakistan: a cross sectional survey
Version: 1 Date: 4 July 2012
Reviewer: Mark Weatherall
Reviewer's report:
The manuscript presents a cross-sectional study that relates responses to an adapted
international questionnaire for respiratory symptoms to simple spirometry, and in
particular whether the strength of association may be sufficient to indicate that the
questionnaire may be used as surrogate markers of air-flow obstruction or restrictive
pattern spirometry in a poor rural community.
The authors found that there is an association between symptoms and
spirometric abnormalities.
I think the manuscript would be improved by, in table 3, listing the actual N/N for the
numbers of participants with or without the symptoms in relation to the
cut-offs for abnormal spirometry and the study definitions of obstructive lung disease
and restrictive lung disease. This would enable readers to directly calculate more
useful indicators of whether the questionnaire may be a useful surrogate for
spirometry such as sensitivity, specificity, and likelihood ratios. I wondered if this an
approach the authors might consider as they are interested perhaps more in the ability
of the questionnaire to identify participants with abnormal spirometry in particular
patterns. It is not particularly surprising that a respiratory questionnaire is associated
with abnormal spirometry.
Response:
Table 3 has been revised as suggested (page 13)
Denominators for the symptoms have been provided in table 2. (page 11)
Analysis has been done to assess the validity of the questionnaire and results are reported
(table 5). (page 15)
Other issues the authors might like to consider:
Page 4: The definition of COPD proposed by the authors does not correspond to the
GOLD definition and the references are to questionnaires rather than definition papers
Response: definition has been revised and reference has also been changed (page 4)
Page 7: It looks like the summary figures quoted for FVC and FEV1 are percent
predicted rather than ml (similar comment for Table 1, looks like percent predicted); I
guess the FEV1/FVC ratio is just the actual percentage and not percent predicted?
Response: correction has been made as suggested and these are % predicted values.
(page 8)
Table 2: Suggest putting the denominator for each numerator as well particularly as it
looks like there are missing data involved.
Response: Revised as suggested (page 11)
Level of interest: An article whose findings are important to those with closely
related research interests
Quality of written English: Acceptable
Statistical review: Yes, and I have assessed the statistics in my report.
Declaration of competing interests:
None