The Zimbabwe STI Etiology Study: Design, Methods, Study

The Zimbabwe STI Etiology Study: Design, Methods, Study Population
Response to CDC Review
We thank the reviewers of our manuscript for their thoughtful comments and
questions. Please see below a point-by-point response. Based on the reviews, we
revised the manuscript and have attached the new draft that we hope will meet your
approval.
Reviewer #1 (Meade Morgan)
As has become common since moving to the eClearance manuscript management
system, none of the required ancillary documents (concurrence letters from coauthors, publication clearance form, and CDC non-research determination or IRB
approval letters have been provided.
Response
We will make sure that these documents will be uploaded with the resubmission.
However, we have our own issues with this process; the reviews came in a barely
readable PDF (especially the table) that also proved to be difficult to share through
our online manuscript development process. In order to provide an orderly
response to the reviews, I (KR) had to retype all the comments. There should be a
way to provide the comments in an editable way.
A link to the study protocol was provided in reference #9 of the manuscript, though I
could not link to the site www.stdpreventiononline.org when I attempted to do so.
Response
Unfortunately the website was down during the review period. Access has now been
restored and the document can now be downloaded. We have attached the protocol
for your convenience.
Specific Comments
Abstract, page 3; Results pages 15-16 – there is a discrepancy between the abstract
and the text in the results section of the main text. The abstract states that “Specimen
collection and NAAT testing was complete for all patients”, but the results section
notes that there were few exceptions. Please fix the wording in the abstract.
Response
Thanks for pointing this out. In fact, specimen collection and m-PCR testing were
complete, and only two specimens were not analyzable on one of the NG/CT
platforms. We fixed the wording in both places.
Study design, pages 6-7 – the manuscript is a little unclear as to whether HIV
laboratory tests were performed. It appears that only self-reported HIV status was
used, though the statement that a secondary aim was to investigate “.. the associations
of HIV with STI syndromes and pathogens…” suggests that HIV testing might have
been done. Please clarify.
Reponse
HIV testing was done on study subjects who provided consent for phlebotomy (see
lines 222-224 and 242-244, including footnote on page 12. These results will be
included in the study outcomes papers; for this paper we only used self-reported
HIV status. We added additional clarification in the manuscript on pages 6 and 16.
Sample size, pages 8-9 – please indicate what statistical procedure was used to
calculate the confidence intervals that are being described (simple asymptotic,
asymptotic with continuity correction, Wilson Score, Clopperson-Pearson,”exact”, etc.).
Though I can’t tell in that I wasn’t able to reproduce the numbers described exactly, I
believe that either Wilson bounds or asymptomatic bounds with continuity correction
were use. Also I think that rather than discuss the “tightening” it would be appropriate
to describe the bounds that were calculated about the 15% prevalence figure for each
of the sample sizes considered (100, 200, 300, 400). The “tightening” discussion is a
little confusing and I think that the argument would be better made if the
investigators simply stated that in order to balance the needs for resources and
precision, they were willing to accept a confidence interval of 10.7% - 20.6% about a
point estimate of 15% prevalence as being acceptable.
Thanks very much for these very insightful comments. We rewrote parts of this
section of the paper using your suggestions. The confidence intervals were
calculated using the Vassar Stats online tool (http://vassarstats.net) that uses the
Wilson procedure with continuity correction.
Sample size, pages 8-9 – it is important that the sample size discussion reflect what is
in the protocol. However, the investigators should be aware that they made a very
strong assumption that there is no clustering of pathogens by site. This may have been
intentional in that the sites were purposively selected and to not represent a true
statistical sample of sites. Even then though, I think that a proper set of sample size
calculations would have allowed for an intra-site (intra-cluster) correlation. The
manuscript itself, at various places, discussed the heterogeneity across sites which
suggests that the authors feel that site clustering is an important consideration.
Whether or not the sample size calculations were done in the most appropriate way,
the manuscript does at least need to explicitly note that intra-sites correlations were
not considered when determining sample size. Also, if this was a conscious decision on
the part of the investigators, the explanation of doing so should also be provided. (I did
see in the discussion where the investigators note that the sites are probably not
representative of the country.)
Response:
We did consider that there would be site differences. In fact, that was the reason to
select a regionally (and ethnically) diverse set of clinics. As stipulated in the
manuscript (lines 182-183) we will include site-specific analyses in all outcome
papers.
Results, pages 14-15; table 1 – There are several discrepancies in between the text and
the table a noted below. These should be corrected as appropriate.
(1) the text states that the mean age if the study population was 28.7 years but
table 1 indicates that it was 28.6 years.
(2) The text states that the lifetime history of an STI was 36% overall, but the table
shows the figure as 35%
(3) The text states that condom use at last sex with a casual partner overall was
24% but the table indicates that this was 36% (which I think needs to be
corrected in the table to 37%).
Response: the table was revised as per the reviewer’s suggestion and discrepancies
with the text have been resolved.
References – please check the URL reference to the protocol. I could not access either
the protocol or the site when I tried to so as a part of this review
Please see above. The site was down during the review period, but is now available.
Table 1 – please see the scanned image below for possible errors in the table. It needs a
good bit of cleaning and possible reformatting.
Note that the handling of missing and unknown (and not applicable data) does not
appear t be consistent. For example, some of the counts for age category do not sum
properly, but there is no missing row. However, for condom use at last casual sex, HIV
stats and STD history, there are missing and unknown categories. I think it would be
helpful if counts for all categories are provides such that the counts always sum to the
proper total. This would mean adding “not applicable” rows for some of the variables
such as condom use. If the percentages are to be calculated without considering the
not applicable rows, then those rows can be asterisked and a footnote added.
Response: We thank the reviewer for his careful review of the table. We have made
the requested changes.
Also in the last column, some of the percentages are shown with a .5% rather than
rounded to whole numbers. I understand why this was done in that the decimal
percent is exactly .5 with a denominator of 600. However, for consistency with the rest
of the table the counts need to be rounded up. A footnote stating that the counts may
not add to 100% due to rounding may be added.
Response: We recalculated all percentages to 1 decimal.
Finally, there are a number of variable labels that need to be clarified. For the sexual
risk factors, write out “in the last 3 months” rather than the somewhat cryptic “<3
months”. In addition, it would be appropriate to note that “HIV status” is really” selfreported HIV status” instead of any sort of confirmed status based on medical records
or new HIV test results. While the text does make this clear, it should be better
reflected in the table. The same is true for STI history, though perhaps the word
“history” implies self-report.
We made the changes as requested.
Reviewer 2 (Dennis)
Line 1: Table lacks a legend.
Response: Table 1 has a title.
Line 65: How do HIV self-reporting compare to testing data apparently done?
Response: HIV testing data will be reported elsewhere. We added a clarification to
this effect (line 311).
Line 73: This belongs in method section.
Response: We respectfully disagree; this is, in essence, a methods paper and
completion of specimen collection and testing is a result.
Line 76: Self reported. Was this confirmed?
Reponse: See comment and response above. We changed to “reported to be HIV
infected”.
Line 216: Self reporting was indicated in the abstract – which is it?
Response: Clarified above.
Line 328: Study limitations were not provided
Major study limitations are related to generalizability of study results from a limited
number of purposively selected clinics, as detailed in lines 333-342.
Reviewer 3 (Carol Ciesielski)
Laboratory procedures: Did NICD send the laboratory sample to the laboratories at
the U of Zimbabwe and the Flowcytrometry laboratory or were they sent by Wilkins
Hospital?
Response: All specimens were initially sent to the receiving lab at Wilkins and from
there to the other labs. We provided clarification in line 253.
Was the urine aliquoted at Wilkins Hospital?
Yes – see line 252.
Sample Size: Did you have targets per clinic for number of participants or by
syndromes. Did the study team have a set number of days to be at a clinic or did they
stay until a target number was obtained? A am asking primarily because of the low
number of participants at Gutu Road clinic. Was this the last stop and hence there
were only few patients remaining to reach the targets?
Response: Please see clarification in lines 281-287
Line 300 – add “%” after 24.5
Response: This has been fixed
Line 303 – according to Table 1, 35% reported STD (not 36%). Since DZ clinic reported
the clinic average, I don’t understand why you singled it out in line 305.
Response: This has been fixed, thanks.
Line 308 – according to Table 1, the number should be 36%, not 41%
Response – This has also been fixed. See also the comments regarding the Table in
response to Reviewer #1.