GENERAL COMMENTS

This is an interesting and relevant study investigating the potential influence of multi-drug use on optimal treatment of heart failure, by use of rather old data (2001) from a GP database.

I have some substantial questions and requests for changes to be made, before I can suggest that this manuscript can be published:

Introduction

• The opening statement that "the prevalence of HF is increasing in older populations" needs a reference. The use of multiple drugs is a potential problem not only among patients with HF, but also in the general elderly population. I miss a statement of this, for example with the reference to Nyborg G et al: Inappropriate prescribing for the elderly—a modern epidemic? Eur J Clin Pharmacol 2012;68:1085-94. You correctly say that "…a patient‘s multi-drug therapy may result in sub-optimal drug prescribing due to the clinical problem of drug interactions …." This leads to an important question that should be addressed: What do you mean by "sub-optimal" in such a context? Is it always optimal to adhere to several guidelines at the same time? Or is "optimal prescribing" to omit one drug to be able to prescribe another one, even in a patient with HF? Maybe the ―gold standard" is not always to adhere to guidelines, but to evaluate the patient‘s total need for prescriptions? The aims of the study are clearly described.

Methods Design:

• OK

Study population:

• You linked the diagnosis made by the GPs to data on drug prescriptions. A major concern is that GPs‘ use of a diagnosis is not reliable enough to allow for valid conclusions. It is valuable to establish clinical GP databases for the purpose of research, but has the use of diagnostic codes been quality assured in relation to these databases? In the Discussion section you just state that "accuracy of recording of morbidity and prescription data in GP is also well recognised" (p.10) and you give a reference – ref 38. I think you are rather over-confident in your conclusion based on this systematic review of Thiru et al. They say in the Abstract: "…. Prescribing data were generally of better quality than diagnostic or lifestyle data." And they conclude that ― The lack of standardised methods for assessment of quality of data in electronic patient records makes it difficult to compare results between studies.― A German study investigating inter-rating reliability of the ICPC-2 in a GP setting concluded that "The fair to moderate reliability on the single code level should be considered when designing studies and interpreting data…"( Swiss Med Wkly. 2012 Aug 22;142:w13621. doi: 10.4414/smw.2012.13621) So I am not convinced that the quality of the GPs‘ use of diagnoses is sufficiently high in your database to enable reliable results. What do you know about this?

Study groups:

• You identified 4 898 middle aged/elderly patients who had received at least one CVD drug during the study period. You selected those who had a diagnosis of HF into one group (n = 170) and those with other CVD diagnoses into another (n = 2 985). Then you present a third group (n = 1 739) which I find highly problematic: This group used CVD drugs, but had no CVD diagnosis during the study period. You call this group "non-CVD reference group". Reference for what, or compared to what? Which patients are included in this "group"? Do they use CVD drugs they do not need? Or are their GPs prone to inaccurate use of diagnoses? I do not see why inclusion of this "group" should be interesting, so I suggest that you exclude these patients from your analyses. In my opinion these are patients who can not be included into the study because of uncertainty regarding their reason for taking CVD drugs.

Definition of multi-drug therapy:

• OK.

Definition of CVD drug measures:

• OK

Statistical analysis:

• OK, but it should relate to two study groups, not three. You say in the last paragraph that "the adjusted associations between the HF group, the non-HF CVD group and the CVD drug measures compared to the reference group are presented". As stated above, I cannot see why this is of interest – and do not agree that this uncertain "group" of patients can be used as a reference for anything. So any comparing of the study groups of interest (the HF-group and the non-HF CVD group) to this so-called reference group should be omitted,and instead these two groups should be compared to each other.

Results

• Study population: If you omit the "uncertainty group" (n = 1 739) you are left with a total of 3155 patients. Out of these, 170 have a diagnosis of HF. How does this number fit in with the expected prevalence of HF in this age group? This may give some information regarding the accuracy of diagnoses in this database.

• Multi-drug prescribing: The word "only" should be omitted , as the finding that 43% in the age group 50-59 y use 4 or more non-CVD drug groups in addition to their CVD-drugs, is a really high prevalence!

• Associations between study groups and CVD drugs: OK, apart from use of the "reference group". Compare the HF-group to the non-HF CVD-group instead.

Discussion

• The first statement should be reformulated into: In a large cardiovasculat GP population, our study showed that multi-drug prescribing was common, but it did not influence optimal prescribing in the HF-group. The second statement is inaccurate: The older HF group had the highest non-CVD multi-drug therapy, yet relatively (compared to what?) they were stillmuch more likely (compared to whom?) to receive…..

• Your main finding was that despite the high level of multi-drug use, as many as ca 90% of HF-patients received "optimal" therapy, and that this did not decrease with increasing non-CVD prescriptions. As you say, other studies have found that multi-drug use did influence "optimal" HF-therapy. You should include a discussion of the possibility that the high adherence to guidelines found in your study in spite of multi-drug use, could lead to problems: drug-drug interactions, side-effects, etc. (Ref. Nyborg G et al: Inappropriate prescribing for the elderly—a modern epidemic? Eur J Clin Pharmacol 2012;68:1085-94. , and Brekke M et al: Pharmacologically inappropriate prescriptions for elderly patients in general practice: How common? Baseline data from The Prescription Peer Academic Detailing (Rx-PAD) study. Scand J Prim Health Care. 2008;26(2):80-5. doi: 10.1080/02813430802002875.)

Strengths and limitations

• As stated above, I am not convinced that ref. 38 supports your statement of accuracy of morbidity data in GP. I would rather prefer a humble attitude and a discussion of the limitations of diagnostic accuracy in such settings. I think that the high number of patients in your study (n =1 739) taking a CVD drug, but without a CVD diagnosis in the study period, reflects this fact.

Implications

• You say that …."there is clearly scope for improvements…" Based upon your findings, which improvements do you mean? You further say that "further work is still needed to investigate the impact of multi-drug therapy on optimal therapy in terms of dose and duration…" Should there be such an impact? What should be the gold standard? To adhere to all relevant guidelines in one single patient, or to carry out adjustments based on an overview?

Tables

• OK, but will be changed when omitting the "reference" goup.

Some improvements of language are needed.

GENERAL COMMENTS

I am generally pleased that the authors have followed my suggestions (which were partly consistent with those of Reviewer 1) in their revised manuscript.

I am especially pleased that they have excluded the study group without the CDV label in the 2-year time window and now carry out thel analyses in the HF-group and the CVD-group exclusively. I strongly suggest, however, that the Supplementary Table 3 , as well as the corresponding text, is omitted, as I think it is more likely to confuse readers than to enlighten them. I also do not think that it brings new and valuable information, compared to what is already given by the HF-group and the CVD-group (see my original comments).

Your Ref. 9 is identical to Ref. 42.

GENERAL COMMENTS

I have no further objections to the present version of this paper - the authors have followes my suggestions.