On 2016 Mar 20, Ryckie Wade commented:

Re: Lopez et al. “Surgical Timing and the Menstrual Cycle Affect Wound Healing in Young Breast Reduction Patients.” PRS 2016;137(2):406–10.

On the face of it, this article appears interesting, however, it contains a number of irremediable flaws.[1] Most reputable medical journals including Plastic and Reconstructive Surgery mandate compliance with the STROBE statement [2, 3], but this article does not meet such standards. Thus this article has limited use for readers and its conclusions are unsubstantiated.

Introduction

The STROBE statement recommends a clear objective and hypothesis to test (STROBE item 1). The authors did not explain their basis for studying this topic. References to ankle injuries and striae after breast augmentation are cited, but their relevance to wound healing in breast reduction is lacking. Such omissions prevent the reader’s from determining the importance of the study.

Design & Sampling

The methods section should include the population of interest, sampling strategy and eligibility criteria (STROBE items 5 and 6a). The authors did not explain their rationale for either the study design or sample size. I have no idea why they selected 49 from a potential 561 patients. It is well known that opportunistic sampling introduces bias and confounding (represented, for example, by the wide range of BMIs and different operations described in this study).

Eligibility criteria

The article lacks inclusion and exclusion criteria (STROBE item 6a). The abstract states that “studies have found an association between hormone levels and wound healing” and previous work by this group showed that the contraceptive pill affected breast striae after augmentation[4], but this time the authors excluded women taking oral contraception (who will have supraphysiological hormone levels) without any explanation. Also, they did not explain their reason for excluding smokers.

Interventions

The authors pooled women who had two different operations - breast amputation and free nipple grafting, and Wise pattern inferior pedicle reduction. These are different operations with different indications, which is obvious from the range of reduction weights (180 - 2525 grams). This heterogeneous group of women could confound comparisons. The authors could have performed subgroup analyses or excluded outliers.

Statistics

The SAMPL statement points out that statistical errors in scientific papers are long-standing, widespread, potentially serious and largely unsuspected by readers.[6] Both the International Committee of Medical Journal Editors (ICMJE)[2] and SAMPL[6] give clear guidance on manuscript preparation and recommendations of peer review from qualified persons, in order to avoid statistical errors.

By dichotomising a continuous variable (i.e. changing the continuous variable of 0-28 days from ovulation into two groups “pre-” and “post-ovulatory”) the authors sacrificed power for simplicity. Clearly, day 13 and 14 of the menstrual cycle are not the same as day 1 and 14. However, by dichotomising the data, the authors have treated the data the same. Dichotomisation is a well recognised problem because information is lost, statistical power is reduced, the risk of Type 1 errors increased, individuals juxtaposed to the cut-off point are categorised as different when in fact they may be similar (e.g., day 14 versus 15 of the menstrual cycle is arguably indifferent) and non-linearity between exposure and outcome is lost.[5] The decision to arbitrarily categorise this important continuous variable should have been addressed by the authors in the discussion (STROBE items 19 and 20).

The authors improperly used the chi square test for proportional comparisons of between-group complications and incorrectly used the term '”correlation” in reference to such tests. Two assumptions were violated (cells had counts of zero and >25% had counts <5) so resampling methods (bootstrapping) or the Fisher Exact test would have been better (incidentally, these methods yield statistically significant differences).

Outcomes

The Results section does not comply with STROBE guidance (items 13a, 13b, 13c, 14a, 14b, 16a and 17) and as the outcomes of interest were not described, their approach suggests data mining. And, in the absence of adequate between-group demographics, readers are unable to judge potential confounders. Therefore, the ‘statistically significant differences’ found are at high risk of Type 1 error because when more statistical tests are performed, the odds of chance findings increases, especially when tests are underpowered and improperly used. The authors should have only analysed outcomes of interest or generated a family wise error rate. While the method section describes a multivariate analysis those results were missing.

Conclusions Both the STROBE and ICMJE statements[2,3] recommend cautious interpretation of observational data, giving careful consideration to: potential sources of bias and confounding variables, the original objectives, limitations of the study design and execution, multiplicity of analyses, relationship to results from other studies and external validity (STROBE items 19, 20 and 21). Therefore, the conclusion that “wound healing is affected by the menstrual cycle” seems unfounded.

Even if this research were of sterling quality, I am concerned it represents little practical value in light of the question “it is reasonable and feasible to organise surgery around a menstrual cycle?”.

Mr Ryckie G. Wade MBBS MClinEd MRCS FHEA

NIHR Academic Clinical Fellow in Plastic Surgery, Leeds General Infirmary, UK

References

1. Lopez, Mariela M., Alexander Chase Castillo, Kyle Kaltwasser, Linda G. Phillips, and Clayton L. Moliver. 2016. “Surgical Timing and the Menstrual Cycle Affect Wound Healing in Young Breast Reduction Patients.” Plastic and Reconstructive Surgery 137(2):406–10.
2. International Committee of Medical Journal Editors (ICMJE). Journals Following the ICMJE Recommendations. Available at http://www.strobe-statement.org/" target="_blank">http://www.strobe-statement.org/
3. Vandenbroucke, Jan P. et al. 2007. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: Guidelines for Reporting. Annals of Internal Medicine 147(8):573–78. Checklists available at http://www.strobe-statement.org
4. Tsai R, Castillo A, Moliver C. Breast striae after cosmetic augmentation. Aesthet Surg J. 2014;34:1050–1058
5. Douglas G Altman and Patrick Royston. 2006. The Cost of Dichotomising Continuous Variables. British Medical Journal 332(7549):1080.
6. Lang, TA and DG Altman. 2013.“Statistical Analyses and Methods in the Published Literature: The SAMPL Guidelines. Science Editors’ Handbook 29–32. Available at http://inaspauthoraid.stage.aptivate.org

On 2016 Mar 20, Ryckie Wade commented:

Re: Lopez et al. “Surgical Timing and the Menstrual Cycle Affect Wound Healing in Young Breast Reduction Patients.” PRS 2016;137(2):406–10.

On the face of it, this article appears interesting, however, it contains a number of irremediable flaws.[1] Most reputable medical journals including Plastic and Reconstructive Surgery mandate compliance with the STROBE statement [2, 3], but this article does not meet such standards. Thus this article has limited use for readers and its conclusions are unsubstantiated.

Introduction

The STROBE statement recommends a clear objective and hypothesis to test (STROBE item 1). The authors did not explain their basis for studying this topic. References to ankle injuries and striae after breast augmentation are cited, but their relevance to wound healing in breast reduction is lacking. Such omissions prevent the reader’s from determining the importance of the study.

Design & Sampling

The methods section should include the population of interest, sampling strategy and eligibility criteria (STROBE items 5 and 6a). The authors did not explain their rationale for either the study design or sample size. I have no idea why they selected 49 from a potential 561 patients. It is well known that opportunistic sampling introduces bias and confounding (represented, for example, by the wide range of BMIs and different operations described in this study).

Eligibility criteria

The article lacks inclusion and exclusion criteria (STROBE item 6a). The abstract states that “studies have found an association between hormone levels and wound healing” and previous work by this group showed that the contraceptive pill affected breast striae after augmentation[4], but this time the authors excluded women taking oral contraception (who will have supraphysiological hormone levels) without any explanation. Also, they did not explain their reason for excluding smokers.

Interventions

The authors pooled women who had two different operations - breast amputation and free nipple grafting, and Wise pattern inferior pedicle reduction. These are different operations with different indications, which is obvious from the range of reduction weights (180 - 2525 grams). This heterogeneous group of women could confound comparisons. The authors could have performed subgroup analyses or excluded outliers.

Statistics

The SAMPL statement points out that statistical errors in scientific papers are long-standing, widespread, potentially serious and largely unsuspected by readers.[6] Both the International Committee of Medical Journal Editors (ICMJE)[2] and SAMPL[6] give clear guidance on manuscript preparation and recommendations of peer review from qualified persons, in order to avoid statistical errors.

By dichotomising a continuous variable (i.e. changing the continuous variable of 0-28 days from ovulation into two groups “pre-” and “post-ovulatory”) the authors sacrificed power for simplicity. Clearly, day 13 and 14 of the menstrual cycle are not the same as day 1 and 14. However, by dichotomising the data, the authors have treated the data the same. Dichotomisation is a well recognised problem because information is lost, statistical power is reduced, the risk of Type 1 errors increased, individuals juxtaposed to the cut-off point are categorised as different when in fact they may be similar (e.g., day 14 versus 15 of the menstrual cycle is arguably indifferent) and non-linearity between exposure and outcome is lost.[5] The decision to arbitrarily categorise this important continuous variable should have been addressed by the authors in the discussion (STROBE items 19 and 20).

The authors improperly used the chi square test for proportional comparisons of between-group complications and incorrectly used the term '”correlation” in reference to such tests. Two assumptions were violated (cells had counts of zero and >25% had counts <5) so resampling methods (bootstrapping) or the Fisher Exact test would have been better (incidentally, these methods yield statistically significant differences).

Outcomes

The Results section does not comply with STROBE guidance (items 13a, 13b, 13c, 14a, 14b, 16a and 17) and as the outcomes of interest were not described, their approach suggests data mining. And, in the absence of adequate between-group demographics, readers are unable to judge potential confounders. Therefore, the ‘statistically significant differences’ found are at high risk of Type 1 error because when more statistical tests are performed, the odds of chance findings increases, especially when tests are underpowered and improperly used. The authors should have only analysed outcomes of interest or generated a family wise error rate. While the method section describes a multivariate analysis those results were missing.

Conclusions Both the STROBE and ICMJE statements[2,3] recommend cautious interpretation of observational data, giving careful consideration to: potential sources of bias and confounding variables, the original objectives, limitations of the study design and execution, multiplicity of analyses, relationship to results from other studies and external validity (STROBE items 19, 20 and 21). Therefore, the conclusion that “wound healing is affected by the menstrual cycle” seems unfounded.

Even if this research were of sterling quality, I am concerned it represents little practical value in light of the question “it is reasonable and feasible to organise surgery around a menstrual cycle?”.

Mr Ryckie G. Wade MBBS MClinEd MRCS FHEA

NIHR Academic Clinical Fellow in Plastic Surgery, Leeds General Infirmary, UK

References

1. Lopez, Mariela M., Alexander Chase Castillo, Kyle Kaltwasser, Linda G. Phillips, and Clayton L. Moliver. 2016. “Surgical Timing and the Menstrual Cycle Affect Wound Healing in Young Breast Reduction Patients.” Plastic and Reconstructive Surgery 137(2):406–10.
2. International Committee of Medical Journal Editors (ICMJE). Journals Following the ICMJE Recommendations. Available at http://www.strobe-statement.org/" target="_blank">http://www.strobe-statement.org/
3. Vandenbroucke, Jan P. et al. 2007. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: Guidelines for Reporting. Annals of Internal Medicine 147(8):573–78. Checklists available at http://www.strobe-statement.org
4. Tsai R, Castillo A, Moliver C. Breast striae after cosmetic augmentation. Aesthet Surg J. 2014;34:1050–1058
5. Douglas G Altman and Patrick Royston. 2006. The Cost of Dichotomising Continuous Variables. British Medical Journal 332(7549):1080.
6. Lang, TA and DG Altman. 2013.“Statistical Analyses and Methods in the Published Literature: The SAMPL Guidelines. Science Editors’ Handbook 29–32. Available at http://inaspauthoraid.stage.aptivate.org