Eugen Mattes, Timothy M E Davis, Danian Yang, Dorothy Ridley, Helen Lund and Paul E Norman
I think this article is a major contribution to our understanding of abdominal aortic aneurysms. There is considerable divergence of opinion as to whether occlusive atheroschlerotic disease and aneurysmal disease are manifestations of the same process or are different diseases. The lack of correlation with diabetes and aneurysms in this study would suggest the latter although the discussion does raise some interesting questions as to reasons why the incidence of aneurysms is in fact lower than in the general community.
However, the practical importance of this study is to indicate that it would not be cost effective to screen all diabetics with a view to determining they may have occult abdominal aortic aneurysms which would then require treatment. This is an important practical issue in the community.
This is a large group of patients studied by a group who have long had a particular interest in this field and who have effectively developed many of the original observations about the pathogenesis and epidemiology of abdominal aneurysms.
[Author's response: There are no concerns requiring a response from Referee 1.]
1. The study addresses the prevalence of abdominal aortic aneurysms (AAA) and makes inferences about the potential value of screening. Although it may be a purely academic point, such information should be based on the incidence of AAA and the modifiable adverse outcomes from them. For example, if the incidence of AAA in men with diabetes is higher than the general population (as might be expected) and the excess mortality associated with AAA is even higher in the diabetic compared to the general population (also seems likely) then the prevalence will be low in the diabetic population despite the higher incidence. This academic point would be practically relevant if the adverse outcomes associated with AAA in the diabetic population could be modified (e.g. by surgery preventing death from abdominal rupture). Case finding amongst the high risk population might then be justified (e.g., as suggested by the authors, in those with claudication). The authors discuss a possible selection bias (towards more severe cases of diabetes) because of the recruitment methods of the Fremantle Diabetes Study (FDS). This potential bias could magnify these effects.
Perhaps the authors could discuss the effects of the incidence of AAA, associated excess mortality and the potential for modifiable adverse outcomes in the diabetic and general populations.
[Author's response: Death from AAA under 60 years of age is very rare and there is no evidence to suggest that those who die from ruptured AAA at any age have an increase risk of diabetes (Reference: Collin J. Aortic aneurysm - screening and management. In. Taylor I, Johnson C, ed. Recent advances in Surgery; No 14. Edinburgh: Churchill Livingstone, 1991.). In addition, in this study we initially screened all men over the age of 50 years but found no AAA after screening 34 men aged 50 to 59 years (data not shown).]
2. As noted in the authors' discussion the limitation of the study is the comparison of one self selected group in one population (the FDS) and another selected group from a second population (reference 7). The authors may not be correct in their assumption that the differences in populations and in sample bias are unlikely to affect the outcome. However the broad similarity of the WA survey (reference 7) to other studies summarised in Table 3 is reassuring.
[Author's response: We feel that the referee's concern regarding sample bias has already been addressed in the text. Furthermore, referee 2 (last sentence) and referee 1 comments seem to support our use of comparative data and the qualified conclusions we draw from them.]
3. Given the potential influence of the sample selection in the study and comparative samples the conclusion seems too strong.
[Author's response: We have changed the wording in the conclusion of the abstract to read "Although a small number of diabetic men have undiagnosed AAA, the prevalence does not appear to be high enough to warrant targeted ultrasound screening ." The concluding statement in the discussion pertains to the results of "the present study" and in this context is not necessarily too strong a conclusion.]
4. It seems that the multi-variate analysis included variables which may not have been statistically significantly associated on univariate analysis. If so, could the authors provide some justification for the inclusion and exclusion of variables in the analysis.
Given the low prevalence of AAA, readers might be helped by the authors including the wide confidence intervals for the prevalence estimates and the level of significance of the apparent difference.
[Author's response: The reason for performing multivariate analysis is to examine for possible interactions between variables. Thus, if a variable is not significant on univariate analysis it may still become significant once included in an interaction or on adjustment for other variables, such as age, in a multivariate analysis.
As recommended we have included the following wording in the manuscript: "The variables included in the statistical analysis were risk factors for cardiovascular disease and diabetes-specific variables which are known to determine macrovascular status. No important known risk factors for AAA were excluded from the analysis."
Furthermore we have, as requested, incorporated the 95% confidence intervals for the prevalence estimates for our study and the significance of the apparent difference]
5. In Table 3 it might help readers to know the methods of assessment for AAA (unless all were by ultrasound).
[Author's response: We have amended the title of Table 3 to detail the method of assessment for
AAA as follows: "Table 3: Prevalence of AAA (
30 mm) as assessed by ultrasound in other
community-based studies".]
6. As I understand it "elderly" generally refers to those over the age of 65 years. The authors might use a different term for participants in their study of men over the age of 60 years.
[Author's response: We have changed the wording from "elderly" to "older" men throughout the manuscript.]
1. What is the proportion of elderly males?
[Author's response: As recommended, we have included a breakdown of the expected proportion of elderly men for FDS based on the available figures on men aged 55+ and 65+ years in the referenced study (Welborn TA, Knuiman MW, Bartholomew HC, Whittall DE. 1989-90 National Health Survey: prevalence of self-reported diabetes in Australia. Med J Aust 1995; 163: 129-132). ]
2. The Wilcoxon rank sum test DOES NOT test means.
[Author's response: We have amended the sentence to read: "... using Wilcoxon rank-sum test for continuous variables , ..."]
3. The authors mean backward and forward regression (not progression).
[Author's response: We have amended the sentence as recommended ]
4. I'm a bit puzzled about the comment no significant interactions were found! How many interactions were considered? What was the rationale for including squared terms? This section requires elaboration.
[Author's response: All possible interactions between risk factors for cardiovascular disease and diabetes-specific variables were tested. In addition, the squared terms were only created for all continuous variables such as age, duration of diabetes etc examining for a geometric rather than linear relationship between these independent variables and the outcome variable (ie the presence or absence of AAA). The text of the manuscript has been altered accordingly.]
5. Last paragraph is a bit ambitious. Surely after all the data analysis/testing P values are hardly strong evidence of any real effect (although they may be hypothesis generating!).
[Author's response: We agree with these comments and have amended the last sentence of that paragraph to read: "However, the association of AAA with hypertriglyceridaemia in our patients is consistent with elevated triglycerides being a possible risk factor for macrovascular disease in the diabetic population".]