In their August 1, 2013 reply letter, the authors characterize as a misunderstanding the conclusion of a 2012 meta-analysis (1) -- using many of the same databases and with one of the same co-authors as the instant study -- that only non-routine bed-sharing is associated with an increased risk of SIDS. In this way they dismiss a critical limitation of their own study, one which calls fundamentally into question the soundness of the suggestion that the medical profession "should take a more definite stand against bed sharing." In fact no evidence has been produced against the possibility that primary bed-sharing, which is to say an arrangement in which the baby routinely spends most of every night in the parental bed, might be the safest sleeping arrangement of all. In the absence of such evidence, it is a terrible disservice to warn non-smoking parents against establishing the bed-sharing routine from birth.

The 2012 meta-analysis cited two papers -- a 1993 article based on New Zealand data (2) and a 2009 article based on the German 1998-2001 SIDS Study (3) -- for the proposition that "routine bed-sharing did not increase the risk of SIDS." The authors now say that those two studies in fact prove the opposite, that bed-sharing on the last night is a risk factor whether or not it is routine.

The New Zealand study found with respect to non-smoking mothers that "the measure of bed sharing in the last sleep had no increase in risk," so it hardly supports the present claims. As for the 2009 German paper, a closer examination reveals acute methodological limitations with respect to the question of routine vs. non-routine bedsharing. Those limitations explain how two subsequent interpretations could reach opposite conclusions on this essential point. They further illustrate dramatically the fact that warnings against primary bed-sharing to the present day lack evidentiary foundation.

The problem arises from the classification of families where parents occasionally allow the infant into the parental bed as "usual" bed- sharers. Among parents who flex their sleeping arrangement this way, certainly one reason for taking the infant into the parental bed would be for comforting or settling when something seems "wrong." One might then see elevated instances of SIDS occurring in the parental bed not because the sleep location was a causative factor, but because that something "wrong" elevated both the probability of the infant being taken into the parental bed and the risk of SIDS.

The only way to avoid this methodological problem would be to break out families which engage in "primary" bed-sharing, which is to say families in which the infant is in the parental bed for most of every night.

According to the 2009 paper on the German SIDS Study data set, of 333 cases only 26 families responded "parental bed" when asked what bed the child usually used in the last 4 weeks. But the researchers created an additional division of families into two groups, those who stated that their infants were never in the parental bed, and those who stated that they were in the parental bed either "sometimes" or "every night." There were 146 families of SIDS cases coded as being in the latter category. (Among the "sometimes/every night" control families, less than a quarter had the infant in the parental bed at the time of the reference sleep, emphasizing the fact that this grouping was in no way composed of primary bed-sharers.)

It was this larger group of 146 families, the group whose choice to take the infant into the parental bed on the last night might have been influenced by the perception of something wrong, which somewhat exceeded the control group in the percentage of infants sharing a bed during their last sleep, thereby giving rise to the finding about increased risk for "bed sharing usual and bed shared during last sleep." Among the smaller group of 26 families who reported the parental bed as the child's primary place of sleep, there was no increase seen in the risk of SIDS.

This latter result seems the more relevant one on the question of "routine" bed-sharing. At the very least it appears we have come a long way down the path of warning parents against the family bed without having resolved one of the most fundamental issues. The current study continues this unfortunate pattern.

Elizabeth S.Bernstein

Bisbee, Arizona, United States

REFERENCES

1. Vennemann MM, Hense H-W, Bajanowski T, et al. Bed sharing and the risk of sudden infant death syndrome: Can we resolve the debate? J Pediatr 2012;160: 44-8.

2. Scragg R, Mitchell EA, Taylor BJ, Stewart AW, Ford RP, Thompson JM, et al. Bed sharing, smoking, and alcohol in the sudden infant death syndrome. New Zealand Cot Death Study Group. BMJ 1993;307:1312-8.

3. Vennemann MM, Bajanowski T, Brinkmann B, Jorch G, Sauerland C, Mitchell EA. Sleep environment risk factors for sudden infant death syndrome: the German Sudden Infant Death Syndrome Study. Pediatrics 2009;123:1162-70.

Conflict of Interest:

None declared

I was interested to read Callander et al's recent article entitled 'Chronic health conditions and poverty: a cross-sectional study using a multidimensional poverty measure'. While the study is of high quality, I would argue that the authors have been remiss to exclude the possibility that it is just as likely that poverty is impacting on health, and not just poorer health leading to increased poverty. Given that the Survey of Disability, Ageing and Carers is cross-sectional in nature, it is not possible to distinguish the temporal nature of the relationship between economic circumstances and health. Different disciplines may approach the questions of the association between health and (socio)economic circumstances from different perspectives. For example, as a health inequalities researcher I would typically focus on health as the outcome, but when using cross-sectional data it would be important to acknowledge that 'health selection' is a valid and alternative, or more likely complimentary, explanation of the poverty-health association. It is important to acknowledge this limitation, especially when recommending policy measures.

Conflict of Interest:

None declared

In their proposed study protocol of Adherence to ART among HIV positive patients in North West Ethiopia, Bezabhe et al properly noted the scarcity of information on determinants of long-term adherence to ART medications in Ethiopian context. Given the importance of adherence to long term treatment outcomes, understanding its determinants can be beneficial to improve the quality of healthcare provision to patients and minimising cost of treatment. We examined this behaviour sometime ago among HIV/AIDS patients in south and central Ethiopia.1

Similar to what proposed by the authors of this protocol we employed multiple methods for investigating determinants and prevalence rate of adherence to ART. However, we found that using some of the methods was challenging. It would be interesting if the investigators present their detail plan on how to handle the shortcomings of each technique for measuring adherence. For instance, the protocol lacks clarity about the proposed pill count technique. There are two different techniques for counting pills, namely announced and unannounced pill count. However, the authors did not specify which pill count technique to use for the study. From our experience, using announced pill count (i.e., where patients informed to bring their medicine container on their date of appointment) might be challenging in resource limited settings. One of such a challenge is the technique cannot identify medicines that were shared among patients. Studies in developed countries where there is better health care access reported higher prevalence rates of prescription medicine sharing among family members, friends, or acquaintances2-4; particularly this might be a concern for HIV positive partners who are on the same type of medication. I could imagine that sharing medicine may be a common practice in Ethiopia where most of HIV patients are from lower socioeconomic background and might have difficulty to pay for transportation to collect their medicines from health care facilities.

The other challenge with pill count technique is pill dumping, this is particularly important for the proposed study, it seems from the explanation given the authors will use pharmacists as data collectors; this is a bit worrisome for me. The study participants might feel being spied by their health care providers about their medication use behaviours, and they might dump some of their unused pills before their medical appointment date. Thus, I suggest to the authors to use neutral data collectors (preferably those that don't have involvement in healthcare provision for the study cohort). In our past study we tried to handle the situation by using unannounced pill count technique; however, it was costly as it required us to go to patients' home and getting their consent to count their unused pills.

References

1. Beyene KA, Gedif T, Gebre-Mariam T, et al. Highly active antiretroviral therapy adherence and its determinants in selected hospitals from south and central Ethiopia. Pharmacoepidemiol Drug Saf 2009;18:1007-15.

2. Petersen EE, Rasmussen SA, Daniel KL, et al. Prescription medication borrowing and sharing among women of reproductive age. J Womens Health 2008;17(7):1073-80.

3. Goldsworthy R, Schwartz N, Mayhorn C. Beyond Abuse and Exposure: Framing the Impact of Prescription-Medication Sharing. Am J Public Health 2008;98(6):1115-21

4. Goulding E, Murphy M, Di Blasi Z. Sharing and borrowing prescription medication: a survey of Irish college students. Ir J Med Sci 2011;180(3):687-90.

Conflict of Interest:

None declared

To the editor, We read with interest the article entitled "Patient safety and estimation of renal function in patients prescribed new oral anticoagulants for stroke prevention in atrial fibrillation: a cross-sectional study" by MacCallum et al. (1). Although this article shows well the difficulties encountered by physicians in prescribing the new oral anticoagulants dabigatran and rivaroxaban in the elderly and/or in patients having a decline in their renal function, we would like to discuss the questionable methods and the interpretation of the results. First of all, the authors compared an estimation of patients' Creatinine Clearance (CrCl) calculated with the Cockcroft and Gault (CG) (2) formula with an estimation of the Glomerular Filtration Rate (eGFR) calculated with the abbreviated or 4-variable MDRD equation (Modification of Diet in Renal Disease study equation). However, the estimated CrCl with the CG formula gives results expressed in mL/min while the aMDRD equation calculates the eGFR in mL/min/1.73 m2 (3). Raw results of the calculations can thus not be compared directly. Such comparisons require conversion into the same units. If the body surface area (BSA) of the patients did not differ significantly from 1.73 m2, this has limited consequences. However, there are no data that may answer this key issue. Secondly, the authors used the correction for ethnicity in the aMDRD formula, as required. However, the correction factor in the aMDRD equation has been validated for afro-americans. Other studies showed that Afro- europeans do not require the same correction factor and possibly no correction factor at all (4). The study being conducted in the London area, patients were most probably Afro-Europeans rather than Afro- Americans. This point is of importance since it results in an overestimation of the eGFR of around 20% for Afro-Europeans. Thirdly, it is recommended that drug dosage adjustment must be performed according to the actual renal function of the patients, according to their actual BSA, thus in mL/min and not in mL/min/1.73 m2 (5). Indeed, in the study of MacCallum et al., the results obtained by the MDRD formula expressed in mL/min/1.73 m2 should thus not be used to determine the appropriate dosage adjustment of dabigatran and rivaroxaban. Finally, in the present study, 0.9% of the patients receiving dabigatran are considered misclassified by the aMDRD formula. Besides, the proportions of patients receiving rivaroxaban and considered misclassified or requiring a reduced dose according to the use of the MDRD formula were 0.1% and 4.5% respectively. How can the authors claim that the use of the MDRD formula gives a wrong classification? The authors outlined the fact that all biochemistry laboratories in UK provide an MDRD-derived eGFR (mL/min/1.73 m2) in keeping with national guidance. Indeed, the United- States National Kidney Foundation's Kidney Disease Outcomes Quality Initiative and the international group Kidney Disease: Improving Global Outcomes (KDOQI-KDIGO) emphasized that the CG formula was developed before the standardization of creatinine assays and cannot be re-expressed for use with standardized creatinine assays. Besides, they also reminded that the MDRD Study equation was developed in 1999 and is currently recommended for eGFR reporting in adults by the National Kidney Disease Education Program (NKDEP) and by the Department of Health in the United Kingdom (6). Therefore, considering the CG formula as a reference is not acceptable in 2013. Besides, it has been shown that, especially in the elderly (age over 65), the MDRD formula had a better accuracy and a better performance than the CG formula (7-9). In fact, 74% of the patients included in this study were older than 65, and thus the results of CG calculations are simply false. We agree however on the fact that dosing guidelines for these drugs are expressed according to the CG formula, as specified in the drugs' Summary of Product Characteristics (SmPC). It is of a major importance to keep in mind that drug dosage adjustment must be performed according to the patients' renal function, and that this latter must be estimated, if not measured, using the most appropriate method depending on the patient. In this study, given the age of the patients, the aMDRD equation is the method of choice to calculate the eGFR of the patients. Dosage adjustment must be determined according to the result of this calculation, once the result has been converted into ml/min, using the actual BSA of the patient. This is the official international recommendation from the KDIGO, publicly released 2 years ago (10). This study thus presents with several serious limitations and bias. We strongly disagree with its methodology and conclusions. It is of an utmost importance that the readers be aware of these limitations and do not take into account the conclusions of the authors, which lack robustness.

References: (1) MacCallum PK, Mathur R, Hull SA, Saja K, Green L, Morris JK, Ashman N. Patient safety and estimation of renal function in patients prescribed new oral anticoagulants for stroke prevention in atrial fibrillation: a cross- sectional study. BMJ Open 2013; 3(9): e003343. (2) Cockcroft DW, Gault MH. Prediction of creatinine clearance from serum creatinine. Nephron. 1976;16(1):31-41. (3) Levey AS, Bosch JP, Lewis JB, Greene T, Rogers N, Roth D. A more accurate method to estimate glomerular filtration rate from serum creatinine: a new prediction equation. Modification of Diet in Renal Disease Study Group. Ann Intern Med 1999; 130(6): 461-470. (4) Flamant M, Boulanger H, Azar H, Vrtovsnik F. [Plasma creatinine, Cockcroft and MDRD: validity and limitations for evaluation of renal function in chronic kidney disease]. Presse Med 2010; 39(3): 303-11. (5) Stevens LA, Levey AS. Use of the MDRD study equation to estimate kidney function for drug dosing. Clin Pharmacol Ther 2009; 86(5): 465-7. (6) Stevens PE, Levin A; Kidney Disease: Improving Global Outcomes Chronic Kidney Disease Guideline Development Work Group Members. Evaluation and management of chronic kidney disease: synopsis of the kidney disease: improving global outcomes 2012 clinical practice guideline. Ann Intern Med 2013; 158(11): 825-30. (7) Cirillo M, Anastasio P, De Santo NG. Relationship of gender, age, and body mass index to errors in predicted kidney function. Nephrol Dial Transplant 2005; 20(9): 1791-8. (8) Michels WM, Grootendorst DC, Verduijn M, Elliott EG, Dekker FW, Krediet RT. Performance of the Cockcroft-Gault, MDRD, and new CKD-EPI formulas in relation to GFR, age, and body size. Clin J Am Soc Nephrol. 2010; 5(6): 1003-9. (9) Froissart M, Rossert J, Jacquot C, Paillard M, Houillier P. Predictive performance of the modification of diet in renal disease and Cockcroft- Gault equations for estimating renal function. J Am Soc Nephrol 2005; 16(3): 763-73. (10) Matzke GR, Aronoff GR, Atkinson AJ Jr, Bennett WM, Decker BS, Eckardt KU, Golper T, Grabe DW, Kasiske B, Keller F, Kielstein JT, Mehta R, Mueller BA, Pasko DA, Schaefer F, Sica DA, Inker LA, Umans JG, Murray P. Drug dosing consideration in patients with acute and chronic kidney disease-a clinical update from Kidney Disease: Improving Global Outcomes (KDIGO). Kidney Int 2011; 80(11): 1122-37.

Conflict of Interest:

None declared