Author's resonse to reviews Title: Correlates of quality of life of overweight and obese atients: a harmacy-based cross-sectional survey Authors: Laurent Laforest (laurent.laforest@chu-lyon.fr) Eric Van Ganse (eric.van-ganse@chu-lyon.fr) Cecile Ritleng (cecile.ritleng@chu-lyon.fr) Gaelle Desamericq (gaelle.desamericq@chu-lyon.fr) Laurent Letrilliart (laurent.letrilliart@wanadoo.fr) Alain Moreau (almoreau@club-internet.fr) Sarah Rosen (sarah.rosen@hima.com) Hubert Mechin (hubert.mechin@mai-naxis.fr) Genevieve Chamba (genevievechamba@wanadoo.fr) Version: 2 Date: 26 March 2009 Author's resonse to reviews: see over
UNITE DE NEUROEPIDEMIOLOGIE ET DE PHARMACOEPIDEMIOLOGIE Hôital Pierre Wertheimer 59, boulevard Pinel 69677 BRON CEDEX Tel : 04 72 12 96 41 Fax : 04 72 12 97 14 March 26, 2009 Re : Revision of the manuscrit: Correlates of quality of life of overweight and obese atients: a harmacy-based cross-sectional survey Dear Sir, Dear Madam Please, find below the relies to the comments. We did our best to address the different oints raised by the reviewers. We have also added in the revised manuscrit the detailed contributions of the different co-authors. Please note that the last two authors may not meet the criteria to qualify as authors. Please feel free to move them to the acknowledgements section and consequently remove them from the list of authors. Should you have any additional questions, lease do not hesitate to contact us Best regards Dr Laurent Laforest on behalf of Dr Van Ganse Retrouvez toutes nos coordonnées sur notre site web htt://eia.univ-lyon1.fr
Reviewer's reort Title: Correlates of quality of life of overweight atients: a harmacy-based cross-sectional survey Version: 1 Date: 24 December 2008 Reviewer: Francesco Corica Reviewer's reort: Major Comulsory Revision 1. Background. This section start with a sentence which refer to chronic diseases in general. Successively, the Authors state about the revalence of obesity. This is somewhat confusing. I think it should be rehrased. We have extensively amended the Background section into 3 main aragrahs: 1- obesity, 2- QOL issues, 3- The reasons for investigating QOL correlates in obese atients (lease, see the revised version). 2. Methods. Only atients with BMI>=28 were enrolled. However, as the Authors state, the aim of the study was to investigate QOL in overweight atients. It seems to me that overweight atients are a minority in this study given the BMI cut off used to define the inclusion criteria. We acknowledge that the threshold chosen in the study ( 28 kg/m²) does not comrise all overweight atients. Our more restrictive definition of overweight ruled out borderline and less severe overweight atients. The Authors should exlain why the OSQOL was chosen for this study. This may result in its internal consistency, and should resent briefly revious exeriences with this instrument. The choice of OSQOL was motivated by the wish to have a secific questionnaire rather than a generic instrument. Unfortunately, we are not aware of any rior exerience with the OSQOL questionnaire. The identification of comorbid diagnoses is a very relevant roblem. The Authors acknowledged artially this roblem in the limitations section when they state that diagnoses for whom no drug was rescribed were not identifiable. However, can the Authors exlain how was determined the co-morbid diagnosis from drugs used for more than one ossible disease (e.g. ace-inhibitors: heart failure? hyertension? Diabetic nehroathy?)? Indeed, this is a ivotal oint. We acknowledge that our definitions of co-morbid conditions were somewhat aroximate. Indeed, some drugs were not secific to a studied disease, like ACE-inhibitors. In such cases, as only regular atients were included in the study, harmacists were oriented by atients medical history. We added the following sentence in the Method section ( Data collected, last sentence): In case of isolated unsecific theray (C03, C07, C08), harmacists were oriented by atient s medical history. Our aim was to assess a burden of associated co-morbid condition, as aroached by rescribed theray for the most common chronic diseases, rather than to define recise diagnoses. This is the reason why we used the number of associated diagnoses, rather than detailed secific conditions. 3. Results. BMI varied from 28 to 51 (median=32). Overweight atients (25<BMI<30) are only a minority. The Authors should consider to change the referral to overweight throughout the manuscrit. We thank the reviewer for noticing this oint. We have changed overweight into overweight or obese atients throughout the manuscrit. Univariate analysis: Please describe univariate analysis before introducing results from multivariable models (as on age 8). A secific aragrah is devoted to univariate results age 9 ( 3- Univariate correlates). 4. Discussion. 2
The Authors should consider to discuss the imact of obesity on QOL in the light of recent findings suggesting that sychological well being is one of the most imortant correlates of QOL, both in the hysical and in the mental domains (Corica et al, Int J Obes 2008). We warmly thank the reviewer for drawing our attention to this oint. This useful reference illustrates some of our results, such as the imact of BMI on the hysical comonent of QOL. The reference has been cited in the Discussion with comarison with our findings. The underlined statement has been added to the following sentence: Domains referring to relations with others and sychological distress were only artially studied as only a single item was dedicated to these dimensions in the OSQOL. Given the role of sychological welfare in QOL (Corica), further studies with more elaborated instruments are needed to investigate these toics more accurately. 3
Reviewer's reort Title: Correlates of quality of life of overweight atients: a harmacy-based cross-sectional survey Version: 1 Date: 13 January 2009 Reviewer: Katherine Alegate Reviewer's reort: Major Comulsory Revisions : NONE Minor Essential Revisions 1) Abstract: In the US, the grouing of self-emloyed workers and to executives would not occur because these labels are generally associated with different socio-economic categories. Please adjust or clarify the term self-emloyed for this context. Indeed, in our manuscrit Self-emloyed corresonded to rofessional ersons. We have changed this term throughout the manuscrit. 2) Background section: The first sentence of the second aragrah is difficult to read, lease rehrase. Also, same aragrah, change to between obesity and cardiovascular disease not diseases, no subject verb agreement. We acknowledge that these sentences were not clear. We have extensively rehrased this aragrah as follows: The consequences of obesity and more generally excess weight on mortality and morbidity (ref), notably cardiovascular diseases, are well established (ref). Likewise, the consequences on osteo-articular and endocrinal morbidity should not be overlooked (ref). 3) Background section, aragrah 3: The authors state that understanding factors that influence QoL could imact interventions for obesity. This is not necessarily accurate, but at least, should be more thoroughly exlained. We have elaborated this aragrah as follows: Increasing our existing awareness of factors influencing QOL in this oulation may be helful in terms of ublic health. Indeed the otential identification of sub-grous of atients with oor QOL may be a reliminary ste before imlementing reventive action for imroved management of overweight and obesity. 4) Background section, aragrah 3: Please exand on the age related reference as it relates to this study. This sentence was awkward as we had no reason to secifically focus on age. As a consequence, for higher clarity we have removed it. Moreover, the corresonding aragrah has been extensively amended. Nonetheless, this sentence could be re-inserted in the manuscrit if the reviewer or the editor judge it relevant. 5) Background section: Secific hyotheses to be tested by this study would be helful rior to beginning the Methods section. This would also better define the questions to be tested in the study. We thank the reviewer for drawing our attention to this oint. We aimed at knowing to which extent OSQOL scores varied according to atients and disease characteristics. Our aims are stated at the end of the Background section. 6) Background section: Please discuss in this section the rationale or benefit of using harmacies as the study site, as the recruitment location reresents a unique and relevant asect of this study. Indeed, this ivotal issue was missing. We thank the reviewer for drawing our attention to this oint. This is an interesting alternative to recruitment by hysicians. However, we judged this oint imortant enough to insert the additional aragrah in the Discussion section rather than in the Background. We have added the following to the last aragrah: An originality of the PRICARDO harmacy-based study was its design. Studies on chronic diseases have been successfully conducted in harmacies. Pharmacists with whom atients often have a relationshi of confidence are ideally ositioned to conduct such studies, notably in case of regular or chronic theray. Our results roved the feasibility of such a study in the context of overweight and obesity. On the other hand, only a secific sub-oulation of atients is more likely to be easily catured in harmacies: those needing a regular treatment. 4
7) Methods: How reresentative is this samle for France comared to the general oulation? This is an interesting question, although very difficult to answer. Our samle cannot claim to be reresentative in any way of the overall oulation of obese atients, which was acknowledged in the Discussion section. In turn, our study oulation may reflect that of obese atients receiving a regular theray. What ercentage of the oulation is on a rescrition medication? In our study all recruited atients resented at least one rescribed drug. 8) Methods: Please exand on why this assessment of QoL was used comared to other measures. Indeed, there were numerous alternative outcomes to investigate. In the resent manuscrit, we wanted to focus on quality of life with its different dimensions. Other outcomes may be studied in further manuscrits. 9) Methods: Data collection section, aragrah four has a selling error (our). Thank you for noticing this tyo that we have corrected. 10) Methods: Do many eole in this oulation articiate with more than one harmacy? This could affect the assigned medical diagnoses tally. In theory, this may have been ossible, the more so as self-questionnaires were anomymous. However, this was very unlikely. In our samle, no atient exhibited concomitantly the same age, gender, BMI, waist circumference and sociorofessional status. 11) Results: How may have the limited question items on dimensions 3 and 4 affected the results. Ideally, the assessment of these domains would have been more comrehensive. We acknowledge that this is a limitation of our aroach. This noteworthy oint has been mentioned in the Discussion section (limitations): Domains referring to relations with others and sychological distress were only artially studied as only a single item was dedicated to these dimensions in the OSQOL. Given the rominent role of sychological welfare in QOL [ref], further studies, with more elaborated instruments are needed to investigate these toics more accurately. 12) Discussion: Please clarify how the age difference (i.e. reference to younger atients) may have imacted relationshi domain. Data seems limited to conclude it is related to hysical aearance. Our conclusions have been downlayed as follows in the Discussion: Our data suggest that younger atients are more affected in their relationshis by their excess weight. A otential exlanation might be hysical aearance, as it may lay a more imortant role in the social life of younger atients. Indeed, excess weight may be a barrier to develoing social activities in younger atients whereas older overweight atients may have become accustomed to their aearance meaning that the imact on their social life is much less imortant. However, these hyotheses require confirmation and, more generally, a better understanding of the effects of age on relationshi domain is desirable. 13) Discussion: What criteria were the harmacists using to determine if a atient should be aroached to articiate in the study? Only hysical aearance was used by the harmacist to select atients. Then, selected atients were roosed to articiate and they comleted a selfquestionnaire. BMI was comuted based on height and weight reorted in the questionnaires. Patients were included if BMI 28 kg/m². 14) Discussion: Is there any data available on these atients mental health status or usage of sychotroic medication? This would have been of major interest to investigate a otential correlation between mental health and the amount of disensed sychotroic medication. Unfortunately, these data were not available in our study. Discretionary Revisions NONE 5
Level of interest: An article whose findings are imortant to those with closely related research interests Quality of written English: Accetable Statistical review: No, the manuscrit does not need to be seen by a statistician. Declaration of cometing interests:i declare that I have no cometing interests. 6
Reviewer's reort Title: Correlates of quality of life of overweight atients: a harmacy-based cross-sectional survey Version: 1 Date: 12 December 2008 Reviewer: Edwin Wouters Reviewer's reort: The article examines the effects of number of atient characteristics on different dimensions of QoL in samle of obese atients resenting themselves in a local harmacy. The authors conclude that obesity has a negative imact on all four dimensions of QoL and that future research on QoL in obese atients should take atient characteristics such as age, gender, SES, dietary habits and hysical activity into account. Although these conclusions are useful for future research, a number of issues arise when reading the manuscrit. These issues have to be addressed to make the manuscrit suitable for ublication. Major comulsory revisions: I will list them following the structure of the manuscrit. 1. The background section needs to be extended. For examle, the first aragrah contains two sentences, which are not well connected. Indeed, the first sentence was not fully aroriate in this context. We have extensively reorganised the Background section into 3 main aragrahs: 1- obesity, 2- QOL issues, 3- The reasons to investigate QOL correlates in obese atients. More background information would be useful. The background section has been extensively reorganized. More details have been rovided on the results of revious studies dealing with QOL of obese atients. Obesity may also have detrimental consequences on atients health-related QOL, articularly their hysical functioning [ref], In contrast, the imact of obesity on mental comonents of QOL yielded more controversial conclusions [ref]. Increasing our existing awareness of factors influencing QOL in this oulation may be helful in terms of ublic health. Indeed the otential identification of sub-grous of atients with oor QOL may be a reliminary ste before imlementing reventive action for imroved management of overweight. 2. The second aragrah exlains that excess weight is detrimental to atients QoL. One might wonder why the current study is needed when we already know that excess weight negatively influences different asects of QoL? The background aragrah has been extensively amended. The interest of our study was to determine the correlates that influenced QOL in a oulation of obese atients in harmacies. This oint has been stated in the second art of the Introduction. 3. The methods section: the way the resondents are selected is roblematic. First of all, the convenience samle is not reresentative for the obese oulation, which hamers generalizations and olicy recommendations. The authors state this in the limitations section. Indeed, our samle was not reresentative of the overall oulation of obese atients, which has been acknowledged in the Discussion section. In contrast, our study may be more reresentative of the sub-oulation of middle-aged/elderly overweight atients requiring frequent or chronic drug theray. We believe that this sub-oulation may not be uncommon and is of interest in terms of ublic health as chronic diseases are common in obese atients (Torres et al 2006), notably at a given age. In addition, our conclusions are in line with those of other authors. Another additional interest of the study was to investigate the ossibility of conducting in harmacies studies focused on obese/ overweight atients under treatment. Torres M, Azen S, Varma R; LALES Grou. Prevalence of obesity and associated co-morbid conditions in a oulation-based samle of rimarily urban Mexican Americans. Ethn Dis. 2006 Sring;16(2):362-9. But the second roblem is much more serious. The authors select their resondents in local harmacies, which results in the fact that only obese atients who visit their harmacy can be included 7
in the study (one can argue that all eole visit the harmacy at least from time to time, but this still means that atients who visit the harmacy more often have a greater chance to be included into the samle). This tye of samling results in a selection bias : atients who regularly visit the harmacy have some kind of medical roblem and are therefore exected to have oorer QoL. We have acknowledged that we focused on a secific sub-oulation of overweight/obese atients visiting regularly harmacies. Consequently, these atients may resent a higher burden of comorbid diagnoses than a more reresentative samle of obese atients. Nonetheless, we believe that such selection bias may not substantially affect our findings : - Firstly, it must be ket in mind that the OSQOL score is secific to obesity. - Although the burden of associated co-morbid diagnoses may contribute to imair the hysical and the vitality dimensions of OSQOL, the imact on the two others ( relation with others, sychological state ) may be more questionable. Interestingly, our results are suortive of this assumtion: the relationshi between the number of associated comorbid diagnoses and QOL scores reveals that hysical and vitality scores are significantly affected only beyond two co-morbid diagnoses (Table 3). Furthermore, no significant effect was noted in multivariate analysis (Table 4). Also, the influence of associated co-morbid diagnoses was more limited on the other OSQOL scores. These findings suggest that this selection bias may not substantially affect our conclusions. - Lastly, most of our findings are consistent with those of revious studies. We have added the following sentence in the Discussion section: Additionally, our study oulation recruited in community-harmacies may resent more associated diseases than a more reresentative samle of overweight and obese atients may. Nonetheless, we believe that such a selection bias may not substantially affect our findings: no significant influence of co-morbid diagnoses was noted for hysical and vitality scores in multivariate analysis (Table 5) and for the other dimensions in univariate analyses (Table 3). The authors select their resondents on the basis of a criterion that is heavily associated with the deendent variable (QoL). The samle will most robably underestimate the QoL of the obese oulation. Indeed, this risk cannot be formally discarded, as our study oulation was not reresentative of the general oulation of obese atients. Nevertheless, this otential underestimation of QOL did not revent us from identifying QOL correlates in accordance with those of revious ublished studies. This can also disturb the other associations (as for examle oor obese eole with a good QoL do not visit the harmacy and rich obese eole with good QoL do to buy exensive beauty roducts). This may not be the case in the French system. Virtually all French atients are covered by the National Health System for their rescribed medications, although a minority of homeless eole and atients facing serious financial difficulties may refrain from buying their theray. Even though beauty roducts are commonly sold in French harmacies, no luxury roduct can be found. More imortantly, the activity of harmacies remains mainly focused on disensing medications. Lastly, virtually all rescribed drugs can be urchased only in community harmacies. To conclude, even though under-reresentation of atients with low socio economic status and good QOL cannot be formally excluded its imact on our findings may be assumed to be limited. 4. Measuring atients relations with other eole and their sychological state with only one question seems too narrow. It is very difficult to measure these comlex interactions and emotions with only one item. The authors should exlain the limited validity of these measurements. We agree with the reviewer. We have exanded on this oint in the Discussion section (underlined art has been added): Domains referring to relations with others and sychological distress were only artially studied as only a single item was dedicated to these dimensions in the OSQOL. Given the role 8
of sychological welfare in QOL (Corica), further studies with more elaborated instruments are needed to investigate these toics more accurately. 5. The conclusions do not seem very innovative. The findings that older eole and eole of lower socio-economic status have oorer QoL are general socio-medical henomena that are not restricted to obese eole. This is true. However, the originality of our work was to study a sub-oulation of obese atients attending harmacies. Indeed, the correlates of QOL identified in this secific sub-oulation may not be systematically similar to those identified in the general oulation. Minor essential revision: 1. Tyo: second to last aragrah of the methods section: our should be oor This oint has been corrected. We thank the reviewer for drawing our attention to this oint. To conclude, the manuscrit rovides interesting information on the factors influencing different asects of QoL in an obese samle of atients. However, the conclusions do not seem really innovative (as conclusions have already been reorted in other studies) and there are serious roblems with the samling rocedure (selection bias robably causing an underestimation of the QoL) which curtail the reliability of the conclusions. All our conclusions are in line with those of revious surveys, as underlined by the reviewer. Nonetheless, an originality of our survey was that it was conducted in harmacies, which has been mentioned in the Discussion section. Furthermore, an additional analysis has been added to study the relationshi between BMI and the OSQOL scores according to the other factors (Table 4, revised version). This rovides an additional originality to our findings. Language: OK, although I am not a native seaker. Statistics: OK Level of interest: An article of limited interest Quality of written English: Accetable Statistical review: Yes, and I have assessed the statistics in my reort. Declaration of cometing interests:i declare that I have no cometing interests. 9
Reviewer's reort Title: Correlates of quality of life of overweight atients: a harmacy-based cross-sectional survey Version: 1 Date: 31 December 2008 Reviewer: Thomas von Lengerke Reviewer's reort: This study reorts data from a convenience samle with an unreorted and ossibly unknown resonse rate - of harmacy customers with a BMI >= 28 in the Rhone-Ales Region (France). Reorted data focus on obesity-secific quality of life. While being of otential interest to audiences with related research interests, to my assessment this manuscrit is undecided as to what it wants (see my remark 1.). Also, the outcome instrument (OSQOL) is very imerfect in terms of sychosocial QOL (remark 2). Finally, the limitations associated with studying harmacy customers should be discussed - not only the strengths of this design (remark 3). Also, there are a number of minor essential and discretionary revisions I suggest (remarks 4 + following). Major Comulsory Revisions 1. 1.1- To me, this manuscrit is undecided as to what it wants. Does it aim to identify correlates of OSQOL in a samle of adults with excess body weight, while adjusting for BMI? Or does it aim to exlain existing differences in OSQOL across BMI-grous by sociodemograhic/-economic factors, behavioural factors, and (co-)morbidities? If it is the former, then one comes to the descrition of the results the authors offer in the resective section of their abstract and, to be honest, in my view these variations of OSQOL by age, SES, and gender do not sufficiently add new insights to our field of inquiry. However, I suose that the authors real aim in fact is a different one - namely to exlain existing OSQOL-differences across BMI-grous by third variables, i.e. sociodemograhic, socioeconomic, and behavioral factors, and [co-]morbidities Our initial aim was to identify the correlates of OSQOL questionnaires, BMI and others. In our models, the severity of overweight was always taken into account in our multivariate analyses were systematically adjusted for BMI (Table 4). However, these hyotheses really deserve investigations. We have exanded on these oints in the Result section (see aragrah below). 1.2-. For instance, regarding Table 2 the authors state that the influence of excess weight on sychological and relational dimensions was also significant, with 22.3 % feeling ill-at-ease due to excess weight and 19.6 % feeling attacked when eole talked about their weight (. 8). Clearly, this refers to OSQOL-differences by BMI (even though surrisingly, such differences are not even reorted in the table), which should then be exlained by other variables. We acknowledge that this sentence was awkward as this aragrah only aimed at describing the different OSQOL scores (effect of BMI is investigated in the next aragrah (Table 3). Consequently, we rehrased the sentence as follows (RESULT section, 2- QOL scores): A noticeable roortion of atients reorted that they felt ill-at-ease due to excess weight (22.3 %), or attacked when eole talked about their weight (19.6 %). 1.3- Also, in the Discussion the authors state that the result that hysical OSQOL declined with BMI can be artly exlained by the osteo-articular and resiratory consequences of excess weight (. 10). This again oints to the aim of exlaining differences across BMI-grous, which however HAS not been done even though it COULD have been done with the resent data. This interesting analysis was not included in our initial objectives. We fully agree that it would rovide novelty and a substantial additional interest for our manuscrit. We warmly thank the reviewer for this helful suggestion. This analysis has been added in the manuscrit (Table 3, see end of the resent document). -The following aragrah has been added in the Method section (results aragrah): Comlementary analyses were also conducted to assess to which extent the relationshi between BMI and the different OSQOL scores varied according to other atients characteristics. 10
- The corresonding results are summarised in Table 3 (revised manuscrit) Comments have been added in the Result Section ( 3- Univariate correlates): The relationshi between BMI and the different OSQOL dimensions according to the other factors did not show significant changes (Table 3). and in Discussion (2 nd Interretation of results) As one final examle, later in the Discussion it s said that Regular exercising rovides hysical and sychological well-being to atients regardless of the severity of their excess weight... This was confirmed by our results where atients efforts to increase hysical activity significantly imroved hysical functioning (. 11-12). Well, no - the authors do not test (or at least not reort any test on) if/how hysical activity imroves OSQOL in each BMI-grou. We found that atients who reorted revious efforts for substantial changes in their hysical activities had significantly better scores for OSQOL hysical dimension (table univariate results). Besides, multivariate Model 1 adjusted for BMI yielded similar conclusions (Table 5, revised version). What the reviewer suggests (to which extent hysical activity interferes with the OSQOL-BMI relationshi) is interesting, but a distinct analysis from our initial goal. Given their originality and interest, these analyses have been added in the revised version (Table 4). Stratified analysis BMI x OSQOL according to hysical activity does not noticeably alter our revious conclusions, notably ercentages of atients resenting a low QOL across the different BMI categories (table below). 11
Statistical relationshi between OSQOL and BMI according to hysical activity Dimension 1 Dimension 2: Dimension 3 Dimension 4 Physical state Vitality desire to do things Relations with others (1) Psychological state (2) % Q25% score (3) % Q25% score (3) % Fairly true or absolutely true % Fairly true or absolutel y true n BMI (kg/m2) OVERALL <.0001 0.028 <.0001 <.0001 < 29.9 154 12.3 22.7 13.6 13.0 30 34.9 186 21.5 25.8 15.1 19.3 35 154 42.2 35.7 31.2 35.1 Previous efforts for substantial changes in hysical activities No substantial 0.0002 changes <.0001 0.2019 < 29.9 kg/m² 119 14.3 26.9 12.6 13.5 30 34.9 kg/m² 145 22.8 29.0 15.2 20.0 35 kg/m² 122 44.3 36.9 32.0 36.1 Major changes 0.0097 0.2248 0.6963 < 29.9 kg/m² 28 3.6 10.7 21.4 14.3 30 34.9 kg/m² 28 10.7 14.3 21.4 21.4 35 kg/m² 23 34.8 30.4 30.4 34.8 * Please, note that interretation of -values may be fallacious given the loss of statistical ower resulting from stratification. In addition, counts of atients who changed their habits were limited. <.0001 0.2165 When the relationshi between hysical activity and QOL was studied searately in each BMI class, no noticeable change was observed. This was corroborated by the absence of significant interaction between the corresonding variables: lease see aragrah (c) below. Of note, the difference in -values may only be attributed to the very unbalanced counts of atients who changed their habits and those who did not (n=79 vs. n=386) All in all, in order to make sufficiently more of the data, it should be reanalysed as follows: (a) Probably relacing Table 2 for reasons of sace, it should be reorted how the three BMIgrous (28-29.9, 30-34.9, >= 35) are comosed of in terms of age, gender, occuational status, current smoking, alcohol consumtion, number of associated co-morbid diagnoses, revious efforts to change diet and revious efforts to change hysical activity - and if BMI is associated with these variables. Results in Table 2 corresonded to our initial objective. However, as mentioned above, we are most hay to follow the reviewer s interesting suggestion as to comlementary analyses (Table 4, revised manuscrit): TABLE 4: Relationshi between OSQOL dimensions and BMI according to the other variables Dimension 1 Dimension 2: Dimension 3 Dimension 4 Physical state Vitality desire to do things Relations with others (1) Psychological state (2) 12
% Q25% score (3) % Q25% score (3) % Fairly true or absolutely true % Fairly true or absolutel y true n BMI (kg/m2) OVERALL <.0001 0.028 <.0001 <.0001 < 29.9 154 12.3 22.7 13.6 13.0 30 34.9 186 21.5 25.8 15.1 19.3 35 154 42.2 35.7 31.2 35.1 Age (years) < 60 <.0001 0.2268 0.0481 0.009 < 29.9 kg/m² 53 3.8 20.8 20.8 13.2 30 34.9 kg/m² 83 19.3 28.9 20.5 25.3 35 kg/m² 81 35.8 34.6 35.8 37.0 60 69.9 <.0001 0.0227 0.0286 0.005 < 29.9 kg/m² 38 7.9 21.1 13.2 5.3 30 34.9 kg/m² 47 27.7 17.0 10.6 17.0 35 kg/m² 38 57.9 42.1 31.6 34.2 70 0.072 0.8085 0.2447 0.075 < 29.9 kg/m² 63 22.2 25.4 7.9 17.5 30 34.9 kg/m² 56 19.6 28.6 10.7 12.5 35 kg/m² 35 40.0 31.4 20.0 31.4 Gender Male <.0001 0.298 0.0993 0.0176 < 29.9 kg/m² 84 8.3 22.6 9.5 7.1 30 34.9 kg/m² 111 15.3 23.4 13.5 13.5 35 kg/m² 64 43.8 32.8 21.9 23.4 Female 0.0049 0.1217 0.0031 0.0051 < 29.9 kg/m² 70 17.1 22.9 18.6 20.0 30 34.9 kg/m² 75 30.7 29.3 17.3 28.0 35 kg/m² 90 41.1 37.8 37.8 43.3 Socio-economic status Lower classes (4) 0.0005 0.4219 0.0029 0.0014 < 29.9 kg/m² 98 17.4 26.5 14.3 14.3 30 34.9 kg/m² 126 25.4 30.2 16.7 23.8 35 kg/m² 118 40.7 34.8 31.4 35.6 Uer classes (5) <.0001 0.2642 0.0207 0.0873 < 29.9 kg/m² 36 0.0 11.1 5.6 5.6 30 34.9 kg/m² 38 5.3 13.2 10.5 5.3 35 kg/m² 18 38.9 27.8 33.3 22.2 Current smoking 0.0006 No <.0001 0.1034 <.0001 < 29.9 kg/m² 134 13.4 22.4 9.7 14.2 30 34.9 kg/m² 156 21.2 25.0 15.4 19.9 35 kg/m² 125 40.8 33.6 29.6 33.6 Yes 0.0005 0.1067 0.0613 0.0115 < 29.9 kg/m² 16 0.0 18.8 31.3 6.3 30 34.9 kg/m² 19 21.1 42.1 10.5 21.1 35 kg/m² 23 52.2 52.2 43.5 47.8 Alcohol None 0.0002 0.2312 0.0057 0.0006 < 29.9 kg/m² 77 15.6 20.8 18.2 15.6 13
30 34.9 kg/m² 96 24.0 26.0 17.7 22.9 35 kg/m² 96 42.7 32.3 35.4 40.6 Yes <.0001 0.0841 0.0434 0.0727 < 29.9 kg/m² 74 9.5 24.3 8.1 10.8 30 34.9 kg/m² 87 19.5 26.4 12.6 13.8 35 kg/m² 56 41.1 41.1 23.2 25.0 Number of 0.0169 associated comorbid diagnoses 0.4444 0.0069 At most two 0.0056 17.4 10.1 13.0 < 29.9 kg/m² 69 10.1 21.3 13.8 20.0 30 34.9 kg/m² 80 16.3 26.8 30.4 33.9 35 kg/m² 56 32.1 Three or more <.0001 0.0932 0.0104 0.0006 < 29.9 kg/m² 85 14.1 27.1 16.5 12.9 30 34.9 kg/m² 106 25.5 29.3 16.0 18.9 35 kg/m² 98 48.0 40.8 31.6 35.7 Previous efforts for substantial changes in dietary habits No = as ou eu de changement 0.0001 0.2115 0.0313 <.0001 < 29.9 kg/m² 100 15.0 24.0 13.0 13.0 30 34.9 kg/m² 109 23.9 33.0 15.6 18.4 35 kg/m² 89 41.6 34.8 27.0 38.2 Yes = changement imortant ou très imortant <.0001 0.0109 0.0072 0.1299 < 29.9 kg/m² 50 8.0 22.0 16.0 14.0 30 34.9 kg/m² 70 15.7 14.3 15.7 21.4 35 kg/m² 60 45.0 36.7 36.7 30.0 Previous efforts for substantiay changes in hysical activities No = as ou eu de 0.0002 changement <.0001 0.2019 < 29.9 kg/m² 119 14.3 26.9 12.6 13.5 30 34.9 kg/m² 145 22.8 29.0 15.2 20.0 35 kg/m² 122 44.3 36.9 32.0 36.1 Yes = changement imortant ou très imortant 0.0097 0.2248 0.6963 < 29.9 kg/m² 28 3.6 10.7 21.4 14.3 30 34.9 kg/m² 28 10.7 14.3 21.4 21.4 35 kg/m² 23 34.8 30.4 30.4 34.8 <.0001 0.2165 (b) Logistic modelling should follow a hierarchical aroach, i.e. in such a way that variables that can be hyothesised to be able to exlain higher odds of oor OSQOL in severely and moderately obese comared to those in re-obesity range (BMI= 28-29.9) are entered consecutively after the BMI-groufactor to see if the ORs of the obese grous are attenuated. If this is the case, and at the same time there is an association between BMI and the variable under scrutiny, there is evidence for mediation (Baron RM, Kenny DA. J Pers Soc Psychol 1986;51: 1173-82). Our rimary aim was to determine 14
redicting factors of OSQOL scores. However, we added the interesting comlementary analyses suggested by the reviewer. (c) If really any moderation effects are to be tested, e.g. if hysical activity changes boosts OSQOL regardless of BMI, then interaction terms should be entered into modelling (again, see Baron & Kenny, 1986). We have checked this oint. The results are not suggestive of any interaction with hysical activity (Please, see Table age 3 and the corresonding comments aragrah 1.3 above). In addition, we formally tested this interaction for each score (Wald test). Results were non significant: =0.90, =0.58, =0.80, =0.70 for dimensions 1, 2, 3 and 4, resectively. 2. Even though the authors acknowledge the limitations of the OSQOL-instrument in the two sychosocial domains, they should be even more cautious in their interretations based on these dimensions: I dare to seriously doubt that the one item which indicates sychological state, i.e. I feel very ill-at-ease, has anything obesity-secific to it (making the interretation of the threefold higher odds of the severely vs. the non-obese even more difficult, secifically in terms of oor QOL - which is not the same as deressive mood!). Also, is the item I feel I am being attacked when eole talk about my weight really anywhere near sufficiency to assess overweight s imact on overweight eole s Relations with others? We acknowledge that this is a limitation of our aroach. This noteworthy oint has been mentioned in the Discussion section (limitations): Domains referring to relations with others and sychological distress were only artially studied as only a single item was dedicated to these dimensions in the OSQOL. Given the rominent role of sychological welfare in QOL [ref], further studies with more elaborated instruments are needed to investigate these toics more accurately. 3. Likewise, I strongly suggest a more balanced assessment of the ros and cons of harmacy customers as a oulation among which to test the BMI-QOL-association among overweight adults. Besides that as noted above no resonse rate is reorted (see also 10. below), a rather elderly, chronically (co)morbid, and ossibly also acutely ill samle has resulted (when do you go to a harmacy? in case of need, if nothing else, isn t it?). Unfortunately the rate of refusal was not documented. Our samle is not reresentative of the overall oulation of obese atients. Indeed, another aim of this study was to ascertain the cardiovascular risk level of overweight and obese atients. That is the reason why a samle of rather elderly atients with chronic conditions was targeted. It is noteworthy that the resence of chronic disease is common in this oulation. As a result, our results may be reresentative of a middle aged/ elderly oulation of obese atients who often resent chronic diseases. This is not to say harmacy customers are not an interesting oulation to study. However, a comarison of this samle with the French adult oulation or that of the Rhone-Ales Region should be added to at least roughly assess how reresentative of the overall oulation of overweight subjects (. 12) the samle is, esecially given the resent ublic health journal. As mentioned above, our samle cannot claim in any way to be reresentative of the general oulation of obese atients, even at a regional scale. An originality of this study was to show the feasibility of a harmacy-based survey to investigate health issues and quality of life of obese atients. The corresonding aragrah (last one in the Discussion section) has been amended and extended as follows: An originality of the PRICARDO harmacy-based study was its design. Studies on chronic diseases have been successfully conducted in harmacies. Pharmacists with whom atients with chronic treatment have often built a relationshi of confidence are ideally ositioned to conduct such studies, notably in case of regular or chronic theray. Our results roved the feasibility of such a study in the context of overweight and obesity. On the other hand, only a secific sub-oulation of obese atients is more likely to be easily catured in harmacies: those needing a regular treatment. Of course, Major Comulsory Revisions should have significant imact on all sections, including Discussion and Conclusions 15
Minor Essential Revisions 4. Abstract, Results, first sentence: This sentence is unclear: Affected by what? For higher clarity, we rehrased the sentence as follows: QOL was inadequate for all dimensions in the 494 atients included in the study 5. Abstract, Conclusions, first sentence: This statement is not justified by the results. It should read e.g. Severe obesity imairs three of four dimensions of QOL. This sentence has been removed from the abstract. 6. Background -section, first aragrah: This aragrah is too scanty. 2-3 sentences on QoL in chronic diseases, and 2-3 sentences on obesity should more thoroughly introduce readers to the aer. We have extensively reorganised the Background section into 3 main aragrahs: 1- obesity, 2- QOL issues, 3- The reasons to investigate QOL correlates in obese atients. The following sentences have been added with resect to QOL issues (notably with resect to chronic diseases): Quality of life (QOL) gave rise to an ongoing interest these ast years. QOL is a major tool to estimate atients erceived burden of diseases, for research uroses as well as for medical ractice [ref]. It has become a common end-oint in clinical trials, along with clinical outcomes. In the absence of ersective of recovery, QOL remains a useful criterion in the management of chronic diseases [ref]. 7. Background -section, third aragrah: Particularly because it introduces the central theme of the aer, this is both over-stated and under-cited. 7.1- It s over-stated because articularly (severe) obesity (rather than excess weight in general ) may be (not is ) detrimental for adults hysical health-related QoL. 7.2- It s under-cited because the one study cited for mental health-related QoL (Dinc et al. 2006) on the one hand is rather secific (women in a Turkish city with a high obesity revalence) and unreresentative for Western Euroe, and on the other hand virtually all of a huge number of studies have shown no noteworthy association between body mass and mental QoL. Not to refer to this is inadequate. We thank the reviewer for drawing our attention to these critical oints. The sentence has been rehrased as follows: Obesity may also have detrimental consequences on atients healthrelated QOL, articularly their hysical functioning [Katz Corica Jia], In contrast, the imact of obesity on mental comonents of QOL yielded more controversial conclusions [Katz Corica Jia marchseini]. We fully agree that the initial cited reference was awkward. It has been relaced by the following references: - Katz DA, McHorney CA, Atkinson RL. Imact of obesity on health-related quality of life in atients with chronic illness.gen Intern Med. 2000,15:789-96. - Corica F, Corsonello A, Aolone G, Mannucci E, Lucchetti M, Bonfiglio C, Melchionda N, Marchesini G. Metabolic syndrome, sychological status and quality of life in obesity: the QUOVADIS Study Int J Obes (Lond). 2008,32:185-91. - Jia H, Lubetkin EI. The imact of obesity on health-related quality-of-life in the general adult US oulation. J Public Health (Oxf). 2005,27:156-64. - Marchesini G, Natale S, Tiraferri F, Tartaglia A, Moscatiello S, Marchesini Reggiani L, Villanova N, Forlani G, Melchionda N. The burden of obesity on everyday life: a role for osteoarticular and resiratory diseases. Diabetes Nutr Metab. 2003,16:284-90. 16
8. Background -section, fourth and fifth aragrah: Have to be amended when changes according to Major Comulsory Revision 1. have been imlemented. As mentioned above, Background section has been extensively amended. The following sentence has been added: We also investigated whether the relationshi between OSQOL scores and BMI varied according to atients other characteristics. 9. Study design and oulation -section: How was the actual BMI of articiants determined? As stated in the Limitations-section of the Discussion, Only atients resenting a robable excess weight according to the harmacist s judgement were asked to articiate (which is critical, as the authors themselves state), but how was the actual BMI of articiants assessed? BMI was collected in self-questionnaires from reorted weight and height. Analyses excluded 25 atients with BMI below 28 kg/m². In actual conditions of ractice it would have been too time-consuming for the harmacist to comute the BMI himself/herself. In addition, some obese atients would not have felt at ease to be overtly asked their weight. By measurement through the harmacist or by self-reort of the articiants (given that it s certainly not by estimation by the harmacist). Also, the fact only atients resenting a robable excess weight according to the harmacist s judgement were asked to articiate should already be mentioned in Study design.... Indeed, atients were selected according to harmacists judgment. This oint is mentioned in the corresonding aragrah: A survey was conducted in 2005 in 76 French community-based harmacies of the Rhone- Ales Region. A convenience samle of atients with robable excess weight visiting the study harmacies were consecutively recruited and were asked to articiate in the study. 10. Study design and oulation -section: More information is needed: (a) How were the instructions for the harmacists on how to select those to be asked to articiate among those with an estimated BMI >= 28? Pharmacists were asked to roose the study to any atient with visible overweight and regular customers of the harmacy (at least 12 months of drug disensing recorded in the harmacy). Patients who acceted to articiate in the study comleted self-questionnaires in which they mentioned their weight and height. Only atients whose BMI 28 kg/m² (based on the reorted values) were considered for the analyses. The corresonding aragrah has been re-organised. (b) How many harmacy customers were asked to articiate? I.e., how was the resonse rate? Unfortunately, refusals were not documented. However, as a rerequisite to articiate was to be a regular customer, refusal rate may be assumed to be low. (c) Why did recruitment sto at 551 atients? For reasons of funding? Considerations of statistical ower? Other reasons. Initially, the aim was to recruit 600 atients: each investigator (n=75) was to include 8 atients. This number was urely emirical and not based on any ower calculation. 579 atients comleted self-questionnaires. 11. Re identification of co-morbid diagnoses, more information is needed as well: Which drug theraies disensed were considered? Only those by the harmacist recruiting the atient? Indeed, it was only drugs of the studied categories disensed in the harmacy during the ast 12 months before inclusion. These data were obtained from comuterised records of the Pharmacy. Nonetheless, only regular customers of harmacies were included (and at least 12 months of medication recorded in the harmacy) and most of them were treated for a chronic disease. Although the risk of missing any drugs disensed in other harmacies cannot be formally excluded, it may be exected to be low. Only those disensed at the visit within which recruitment took lace? Any medications of studied categories disensed to the atients within the 12 months receding inclusion were taken into account, and not only inclusion visit. 17
12. Skewness is not a sufficient reason for dichotomizing quantitative scores for dimensions 1 and 2 (for instance, one could use transformations and conduct analyses of covariance). Indeed, normalization of these dimensions would have been desirable. However, transformations tested (log, square-root, square, exonential) did not ermit to normalize the data, which led us to dichotomisation of the scores. Also, further on in the text the authors do reort results on the quantitative scores (. 8, sentences 2 and 3). Thus, more justification is needed for dichotomization. Finally, choice of the lowest quartile has to be justified more substantively. We wanted to focus on the redicting factors of oor quality of life rather than a quality of life lower than the median value which may have been less secific. This led us to choose the 25% ercentiles for quantitative scores instead of median value. This was an emirical comromise, as a more selective threshold could have resulted in a substantial loss of statistical ower. This threshold enabled us to identify numerous correlates for these scores (Tables 3-5 of the revised manuscrit). 13.. 10, Discussion, third sentence: The items used to assess behaviours do not justify such a conclusion. We tend to disagree with the reviewer. Our results suggest that few atients endeavoured to make efforts in diet and hysical activity to imrove their health, which is in line with the conclusions of revious authors. 14.. 10, Discussion, sentence after citation 19: This is incorrect: As reorted in Table 4, moderately obese atients OSQOL did NOT significantly differ from those with BMI 28-29.9 lease do not overstate findings! We aologize this is a misunderstanding. We have rehrased the sentence as follows: Patients QOL significantly decreased with severe obesity (BMI 35 g/m²) for 3 dimensions of the OSQOL questionnaire (Table 5). 15. Tables: (a) Table 1: More information on age distribution should be given in the text (range etc.) Age (mean=62) ranged from 27 to 86 (25 th -75 th ercentiles: 54-72). As a consequence our samle mainly consisted of middle aged /elderly overweight and obese atients. The range and the median age were already mentioned. 25-75 th ercentiles have been added in the revised version (Results section). (b) Missing data has a dash ( - ) under Social economic status. This should be deleted, and Social economic status should be changed into Socio-economic status. We aologize for this tyo that we have corrected. (c) Tables 1, 3 + 4: In BMI-categories, < 30 should be 28-29.9, and [30 35[ should be 30 34.9. We have corrected these oints in the different tables. We have also made the same changes for age categories. (d) Tables 1, 3 + 4: Given that the Three and more -number of co-morbid diagnoses grou (which by the way should read Three or more - thank you for noticing this oint that we have corrected.) comrises almost 60% of the samle, a finer-grained analysis should be ossible here (e.g. 0-1, 2, 3, 4 or more ). We have recomuted the statistical relationshi between OSQOL scores and the number of co-morbid diagnoses with the more detailed stratification suggested by the reviewer. The results suggested that no major changes were observed after detailing this category (Table below for information). Physical dimension noticeably decreased in case of > 2 co-morbid diagnoses with no real further imairment beyond this threshold. Similar conclusions can be reached for vitality, while the other two dimensions remain non significant. The corresonding aragrah in the result section has been amended as follows: Finally, the number of co-morbid diagnoses had a significant imact on hysical functioning and vitality dimensions, notably beyond two associated diagnoses (Table 3). When the category 3 or more co-morbid diagnoses was detailed into 3 and 4 or more, these conclusions were not affected (data not shown). 18