SUBJECT FIELD: Final replies to comments

Thanks for the comments. Please find below some additional and final replies. We have declared the funding sources of this research in the article as well as our connections to IC - The Integrative Care Science Center in Sweden. For those readers interested to know more about the IC center, please follow the link [http://www.integrativecar...] that was also provided previously.

QUESTION 3
Q3: Can the authors tell us the average and median time in hospital stay for the two groups?
A3: No.
Comment: Since the authors state that they have used (digital) inpatient registry data from Stockholm County Council covering years 2005 to 2010 it seems implausible that only diagnosis and date of admission were available and not date of discharge and length of stay.

REPLY: The Stockholm County Council inpatient registry was used to be able to identify patients with the target pain diagnosis that had been to Vidarkliniken (an inpatient setting) and the controls. The study outcome was drug utilization. The study used a generic description of the integrative and conventional care management due to the fact that, in real world settings, patients' health care utilization are rarely limited to only inpatient care, but also concurrently occur in e.g. outpatient, specialist and allied health care provider settings. It may thus be misleading to only refer to one of these domains as the type of care delivered/received over a given period of time. Accordingly, the study outcome was related to index visits rather than periods of care. Notably, this intriguing fact of complex health care utilization patterns among patients is a strong incentive for future studies, where data can be extracted from several different health care registries, and the health care utilization outcomes can be presented representing this complexity (e.g. in terms of frequency, duration and costs of visits) within and between the different health care registries/providers over time, in order to more accurately describe health care delivery and utilization patterns related to integrative and conventional care.

QUESTION 4
*Q4: Table 4 shows that in the year before the control patients more often had a cancer diagnosis, but less often a psychiatric diagnosis compared to those who received ”integrated care”.
Does this not make the groups too dissimilar for comparison?
A4: ”As the controls were randomized, it may be argued that differences in background variables may be disregarded due theoretically balanced confounders.”
Comment: This must be a misunderstanding. The controls are of course drawn randomly but they are not randomized, they are matched to the cases only by pain diagnosis, sex and socioeconomic area of living. Cases and controls can systematically differ in a number of other ways, and the data given in the paper points to at least one obvious source of bias. Assuming only one cancer diagnosis per patient, the probability of the cases being given a cancer diagnosis in the year before the index visit was 9 %, but the rate of cancer in the controls was 18 %. The opposite balance was the case with psychiatric diagnoses. This suggests that the cases had less somatic disorders and more of mental health problems, leading on to question no. 5.

REPLY: Perhaps a language misunderstanding. Controls fulfilling the same target diagnosis, age, gender and socio-demographic characteristics as the cases were applied to a "randomized procedure" ("randomiserat förfarande"), i.e. randomly drawn to the control cohort, so that, at least in theory, potentially confounding and unknown background variables could be balanced within that cohort. Inherent to the selected study design (retrospective case-control) it is of course possible that cases and controls may systematically differ. However, as mentioned in the previous reply, the clinical relevance of the observed differences (e.g. the suggested assumption example of 9% vs 18%) is uncertain and, relating to question 5, supporting this there were in fact no statistically significant differences in change between the integrative and conventional care groups over time in the use of psycholeptics (please see Tables 2 and 3). Nontheless, it is of course important to consider such issues and limitations. We hope this reply can help to acknowledge that matter.

End of reply to comments.

In closing, we appreciate academic dialogues to facilitate sensible interaction and knowledge building. Every reader is of course entitled to make their own interpretation of carefully scrutinized peer reviewed scientific studies based on her/his level of research knowledge, clinical experiences and particular representative interests. We appreciate that and respect that people may have different perspectives and opinions and ask of our readers to do the same.

As stated in the methods section of the paper, retrospective registry studies have, just as other study designs, many limitations and considerations. On the other hand this type of studies also have many strong benefits including making informed estimates of clinical outcome trajectories and highlighting areas of importance for future studies.

There is always a limitation as to how much information a scientific paper can hold and the final content is a result of the extensive peer review process each paper goes through before being considered scientifically accurate, relevant, unbiased and accepted for publication. Accordingly, some of the main reasons why we are content with this study is that it has provided a detailed clinical trajectory of drug prescriptions for pain patients that has received anthroposophic integrative care. This holds true regardless of controls. In addition to that it has provided an informed, albeit not causative, comparative understanding of drug prescriptions for a cohort of pain patients receiving conventional care. The study has also highlighted important areas of consideration for future studies including the generation of potential hypothesis and research questions for prospective clinical trials.

This will be our final comment. Please respect that. We have already exceeded what is to be expected in terms of outreach and various communication (e.g. social media/blogs, newspaper debate, email, phone and the PLOS forum). We thank for the attention and the opportunity to discuss our study. We hope that the academic dialogue here at the PLOS forum has been constructive and wish all readers a pleasant summer.

Sincerely,
T Sundberg
Corresponding author of the article

ORIGINAL COMMENTS

I sincerely thank the authors for trying to answer my questions concerning the limitations of the study design.

Since the authors of the paper seem to imply that I have not fully declared my competing interests (”present/past connections to certain interest groups, pharmaceutical companies, research or business activities”), I have made a longer list of memberships etc. that I presume they might regard as potential sources of bias on my part.

Having done that, I feel free to inform readers that the Integrative Care Science Centre seems closely linked to the anthroposophic movement. It has it’s postal adress at the anthroposophic hospital Vidarkliniken, together with the Swedish Association of Anthroposophic Physicians LAOM. Part of the funding for this study comes from four European foundations all connected to the anthroposophic movement (while the Jochnick foundation has no such connection known to me).

The answers from the authors were quite long, but as far as I can tell their essence is this:
Q1: When the use of analgesics was already so low, why was this parameter chosen as the main outcome measured?
A1: The outcome measures were defined a priori.

Q2: The use of non-prescribed analgesics sold OTC [without prescription] was not recorded in the study. How can we then know if the patients’ use of analgesics changed at all?
A2: We can’t.

Q3: Can the authors tell us the average and median time in hospital stay for the two groups?
A3: No.

Comment: Since the authors state that they have used (digital) inpatient registry data from Stockholm County Council covering years 2005 to 2010 it seems implausible that only diagnosis and date of admission were available and not date of discharge and length of stay.

Q4: Table 4 shows that in the year before the control patients more often had a cancer diagnosis, but less often a psychiatric diagnosis compared to those who received ”integrated care”.
Does this not make the groups too dissimilar for comparison?
A4: ”As the controls were randomized, it may be argued that differences in background variables may be disregarded due theoretically balanced confounders.”

Comment: This must be a misunderstanding. The controls are of course drawn randomly but they are not randomized, they are matched to the cases only by pain diagnosis, sex and socioeconomic area of living. Cases and controls can systematically differ in a number of other ways, and the data given in the paper points to at least one obvious source of bias. Assuming only one cancer diagnosis per patient, the probability of the cases being given a cancer diagnosis in the year before the index visit was 9 %, but the rate of cancer in the controls was 18 %. The opposite balance was the case with psychiatric diagnoses. This suggests that the cases had less somatic disorders and more of mental health problems, leading on to question no. 5.

Q5: The amount of prescribed psycholeptics did not change significantly after ”integrated care”. Do the authors have any thoughts about this?
A5: ”…perhaps patients in both cohorts had pain related mental issues such as sleeping problems or depressive symptoms requiring longer periods of drug use compared to analgesics”.

Q6: Was it recorded how the use of anthroposophic medication (both homeopathic and non-homeopathic) changed in the two groups?
A6: No.

I again thank the authors for taking my questions seriously, and I hope they will find time to clarify their position to my comments above regarding their answers to questions 3 and 4.

Competing interests declared:
Member of The Swedish Pediatric Association
Member of The Swedish Medical Association
Member of The Swedish Society of Medicine
Member of The Swedish Skeptics (Vetenskap och Folkbildning)
Skeptical medical blogger at www.dagensmedicin.se
I have no financial ties to any pharmaceutical company, but I am involved in phase-3 clinical trials (the economic compensation for this benefits my employer, the regional primary care health service).
I have previously been a member of the board of a Norwegian company providing radiological diagnostic services.