Loading...
Please wait, while we are loading the content...
Similar Documents
Author ' s response to reviews Title : Delay factors in the diagnosis and treatment of sympomatic colorectal cancer
| Content Provider | Semantic Scholar |
|---|---|
| Author | Sánchez-Calavera, María A. Costa-Alcaraz, Ana M. Segura, Josep Maria Guiu |
| Copyright Year | 2012 |
| Abstract | 1. This needs considerable revision Methods: The measures need to be described more clearly detailing which is the outcome and which are the covariates Cox analysis (and measures for it) should all be described in the methods section. We have changed it The phrase “median symptom diagnosis interval” does not really make sense unless SDI is first defined in the methods section SDI has been defined in the methods section of the fist manuscript Results “1rst” should be first DONE Main paper Methods 2. This is not a descriptive study as it involves statistical modeling and analysis. It is a cross-sectional study where patient interviews as combined with data from patient notes. More detail is needed on how the primary care measurements were extracted. Was this coded data or free text notes? Were the dates of symptom presentation and diagnosis validated – if so how and if not, can you cite refs that check this? This is important – as explained in ref 53? We changed descriptive by cross-sectional in the abstract and methods section Data in clinical records was text notes We already have not validated date of symptom presentation in primary care and hospital care .But at the moment, we are working in a paper to compare differences in diagnosis interval when using our different data sources to establish symptom onset date (patient, primary care and patient). As expressed in the previous manuscript date of diagnosis was recorded from histology reports (1st positive cancer finding from colonoscopy or surgery) from hospital. This was not obtained from GPs records nor from patient. 3. I note that in the original proposal for this work you planned to compare patient-reported and GP-reported symptoms? Although you cite other authors who have done this, there is no mention of this here? This should be reported. This is a broad study with different objectives as can be assessed in the protocol publication. We are planning to publish concordance between the different 3 sources in a next paper. We have done a short comment in the discussion. 4. For the Cox model, were the assumptions of proportionality tested and if so what was the result? This needs to be reported. (Alternatively a Poisson model may be more appropriate) We have changed a little bit the explanation of survival analysis. Including results of proportional test would be long and we thing would not give essential information. A multivariate Cox proportional hazard model with time-dependent covariates were performed to allow covariates change their values over time. For example a patients has a value of hemogram zero until the time the hemogram is done, and 1 after that moment. We consider this model fits well with this type of variables as suggested by Lloyd D, et al. Time-dependent covariates in the Cox proportional-hazards regression model. Ann Rev Public Health 1999;20:145-57. We thank the suggestion of using a Poisson model is a good alternative to analyze Time-to-event data with time dependant covariate. Both approach are different and we think yield to similar results. We test the PH assumption, variables which PH cannot be assumed include all intrinsic time-dependent covariates (covariates that changes overtime). 5. Many different types of variables are mentioned, but little detail of how some these were obtained, and/or aggregated, is provided in either this paper or the protocol paper. Some variables e.g. number of visits number of symptoms are only mentioned in the results but are not defined in the Methods. Some have different names in different tables e.g. “ Referral mention of CRC” in table 4 is renamed as “CRC suspicion in referral letter” in table 6. (I assume these refer to the same thing, or is this a new variable introduced for the Cox regression?) Is the same variable, we have unified the terms about the variable. Much more detail (and consistency) is needed. Some of this detail could be given in an appendix if appropriate. A new table with more details about variables has been included in an appendix Results 6. The (bivariate) tables are well presented, although some variables that are shown in Table 6 seem to be missing – e.g. abdominal occlusion, CRC suspicion why is this? Please include, together with the numbers involved. We apologize but it was a term confusion, abdominal occlusion is abdominal obstruction and Referral mention of CRC is CRC suspicion in referral letter 7. The result that family history of cancer is positively associated with delay seems surprising and counter-intuitive. This result needs checking and further investigation. I note that 45% of patients reported this family history of cancer, which seems rather high – although again this variable is not described, so I can only assume it must relate to non-immediate family and all types of cancer (including very common benign skin cancers?) another measure that needs more detailed explanation. There were all type of cancers (not benign skin cancer) we concretely asked for parents, children, spouse, brothers or sisters, any other family member and friends. There are few studies on delay and family history of cancer but when looking to breast cancer screening, some studies show women more reluctant to participate in screening if there is previous familiar or quittances with cancer because fear of some cancer findings is higher. See appendix table for variable definitions. 8. The results of the Cox regression are very unclear and I‟m not really sure why these are included as so little is said about the results. This is the first time symptom duration has been mentioned – presumably you mean symptom diagnosis interval? We have included some changes in the results of Cox regression to better clarify and eliminated symptom duration. After multivariate analysis, factors independently associated with a longer SDI and STI in CRC patients were: female gender, not visiting the doctor when he/she felt the first symptoms, the number of GP visits for CRC symptoms before referral, the absence of CRC suspicion in the GP referral letter and not performing the investigations prescribed by hospital doctors. Abdominal occlusion is related to a shorter treatmet time interval but not with diagnosis interval (Table 6)” 9. In Table 6, all the p-values should be given (NS is not sufficient – do you mean N/A (not applicable)? If so explain why this is so. It is not possible to have a p value of <0.000 We changed <0.000 for <0.001 and eliminated NS, and p values given Statistical Analysis 10. Please explain what you mean by “In order to avoid outliers”? In order to avoid symptoms that could not be related to CCR , no more than a 24-month interval between symptom and diagnosis was accepted 11. The sentence “Most of our patients have to wait median of 4 months” doesn‟t make sense and is incorrect. By definition, half the patients have a value greater or equal to the median. The correct statement is “half (or 50% of) the patients have to wait at least 4 months”. Changed by After experiencing first CRC symptoms, half of the patients have to wait at least 4 months until diagnosis |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://static-content.springer.com/openpeerreview/art:10.1186%2F1471-2407-13-87/12885_2012_3720_AuthorComment_V2.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
