title
stringlengths 1
1.19k
| keywords
stringlengths 0
668
| concept
stringlengths 0
909
| paragraph
stringlengths 0
61.8k
| PMID
stringlengths 10
11
|
|---|---|---|---|---|
Primary outcome | In the adjusted model (Table Adjusted mean differences [95% confidence interval] of study outcomes based on mixed effects models | PMC10054219 |
||
Secondary staff outcomes | SECONDARY | There was no statistically significant difference in any of the secondary staff outcomes, based on adjusted mean scores at follow-up between the intervention and control group (Table Change in scores measuring how often care aides suggested ways to improve performance to their colleagues by level of enactment based on mixed effects models | PMC10054219 |
|
Resident care outcomes | No intervention effects were found for those teams working on responsive behaviors. However, the adjusted level of resident dependency significantly decreased for residents whose teams addressed mobility ( | PMC10054219 |
||
Discussion | SCOPE was a multicomponent intervention designed to facilitate the use of best practices in care for older LTC home residents. SCOPE incorporated elements of design to support sustainability and address the need for programs of research on implementation and improvement in healthcare [ | PMC10054219 |
||
Primary outcome | SCOPE was initially planned as a randomized clinical trial, attempting to take the complex nature of the system in which it was implemented into account. Using quantitative measures, SCOPE’s primary aim, improvement in the conceptual use of best practices, was not demonstrated. The intention was that teams would implement what they had learned by taking part in SCOPE, developing, and testing small changes in care practices, based upon best practices in care, which would then be spread across the unit and embedded in usual care. The outcome, Conceptual Research Use (use of best practices), was thought to best capture the essence of the change in thinking resulting from the team quality improvement collaborative and theoretically links beliefs regarding research use being predictive of actual best practice use [ | PMC10054219 |
||
Secondary outcomes—staff | SECONDARY | There were no statistically significant group differences in staff-related secondary outcomes. These findings should inform sample size calculations of future studies of this nature for example, if based on the size of effects we report using similar outcomes measures.A post hoc analysis revealed a statistically significant improvement in care aide perspectives on new ways of working on the unit (one element of OCB) and a positive relationship between SCOPE fidelity enactment and these behaviors. This last finding reinforces the importance of recent Medical Research Council guidance [In addition to the, retrospectively, smaller than needed sample size, our concurrent process evaluation [Secondly, SCOPE teams consisted of four—six members, actively working on their PDSA cycles, and attempting to improve quality of care; the team members had to spread their improvement efforts across the entire unit, ensuring that “new ways of doing” were adopted by care providers who were not part of the SCOPE team. It is possible that adoption was not spread sufficiently | PMC10054219 |
|
Secondary outcomes—residents | pain | Examination of two of the three resident care indicators from the RAI-MDS showed no difference in change between groups over the period of observation. The finding in pain assessment is however, encouraging, suggesting that teams were successful in improving the quality of care for residents in pain, but this is a single finding in need of replication. There may be several reasons for the inability to detect a change in resident outcome measures. Firstly, the SCOPE intervention was implemented over a year and most teams took time to function effectively, to design their aim statements and conduct tests of small changes to improve quality [For all outcomes and teams, SCOPE was dependent upon leadership sufficiently able to support change. Leaders were called upon to remove obstacles, provide sufficient time for QI initiatives and to allow the care aide leaders to lead the team. This was variably achieved, with some leaders adopting and maintaining a more authoritarian approach to SCOPE implementation, providing less than sufficient resources and time for teams to meet, potentially adding to work life pressures, rather than relieving them [ | PMC10054219 |
|
Limitations | SECONDARY, RECRUITMENT | We note above limitations related to the primary and secondary outcomes. There are additional important limitations. While initially intended as a randomized controlled trial with propensity matching to TREC units not participating in SCOPE, two problems were encountered that were eventually insurmountable, resulting in the quasi-experimental design reported here. Firstly, the number of LTC homes in the sampling frame was exhausted before the intended sample size was reached. Secondly, to maintain the size of our control cohort in a potential 1:1 match, we had to return homes that had declined to participate to the general pool, potentially introducing a bias to the comparator group. Despite the close working relationships established with many of the homes, and our previous experience with this intervention, recruitment was more challenging than originally anticipated [ | PMC10054219 |
|
Conclusion | In conventional terms, SCOPE was a negative study further contributing to the file drawer problem [ | PMC10054219 |
||
Acknowledgements | McLeod | We would like to thank the LTC homes and their care teams who participated in this study. We would also like to thank Don McLeod for facilitating the Learning Congresses and contributing to the development of the SCOPE materials for participants; the Quality Advisors (Carolyn Brandly, Fiona MacKenzie, Barb Stolee) for supporting the SCOPE teams and keeping them engaged; Judith Palfreyman for administrative support and the TREC data unit manager, Joseph Akinlawon for his expertise. | PMC10054219 |
|
Authors’ contributions | MH | AW led the study and led the writing of the protocol. AW, MH, LG, WB, MD, JK-S, YS, PN, and CE were all involved in the conduct of the study. AW, PN, LG, and MH developed the data analysis plan, conducted the analysis, and interpreted the results. MH drafted all tables. AW wrote the first draft of the manuscript and led revisions. AW, MH, LG, WB, MD, JK-S, YS, PN, and CE all revised the paper critically for intellectual content and approved the final version. | PMC10054219 |
|
Funding | This study was funded by a Canadian Institutes of Health Research (CIHR) Operating Grant CIHR PS 148582 Wagg and funds from the Muhlenfeld Family Trust held by Dr. Wagg. | PMC10054219 |
||
Availability of data and materials | The data used for this article are housed in the secure and confidential Health Research Data Repository (HRDR) in the Faculty of Nursing at the University of Alberta ( | PMC10054219 |
||
Declarations | PMC10054219 |
|||
Ethics approval consent to participate | This study was approved by the Research Ethics Boards of the University of Alberta (Pro00012517) and University of British Columbia (H14-03286). Operational approval was obtained from all included facilities as required. SCOPE sponsors and team members were asked for oral informed consent before participating in any primary data collection (evaluation surveys, focus groups, interviews). | PMC10054219 |
||
Consent for publication | Not applicable. | PMC10054219 |
||
Competing interests | The authors declare that they have no competing interests. | PMC10054219 |
||
References | PMC10054219 |
|||
INTRODUCTION | REGRESSION | Many trials use stratified randomisation, where participants are randomised within strata defined by one or more baseline covariates. While it is important to adjust for stratification variables in the analysis, the appropriate method of adjustment is unclear when stratification variables are affected by misclassification and hence some participants are randomised in the incorrect stratum. We conducted a simulation study to compare methods of adjusting for stratification variables affected by misclassification in the analysis of continuous outcomes when all or only some stratification errors are discovered, and when the treatment effect or treatment‐by‐covariate interaction effect is of interest. The data were analysed using linear regression with no adjustment, adjustment for the strata used to perform the randomisation (randomisation strata), adjustment for the strata if all errors are corrected (true strata), and adjustment for the strata after some errors are discovered and corrected (updated strata). The unadjusted model performed poorly in all settings. Adjusting for the true strata was optimal, while the relative performance of adjusting for the randomisation strata or the updated strata varied depending on the setting. As the true strata are unlikely to be known with certainty in practice, we recommend using the updated strata for adjustment and performing subgroup analyses, provided the discovery of errors is unlikely to depend on treatment group, as expected in blinded trials. Greater transparency is needed in the reporting of stratification errors and how they were addressed in the analysis.
Stratified randomisation is a popular approach for randomly assigning participants to treatment groups in clinical trials.It is widely recognised that analytic methods should reflect the trial design and hence adjustment should be made for the stratification variables in the analysis of trials that use stratified randomisation.When a stratification variable is affected by measurement error, some participants may be misclassified and hence randomised in the incorrect stratum. Such errors, referred to hereafter as Several previous studies have explored the impact of stratification errors on treatment effect estimation using regression models. Ke et al.The aims of the current article are: (1) to explore the impact of stratification errors on the correlation between sample means induced by stratification; and (2) to compare methods of adjusting for stratification variables affected by misclassification in the analysis of continuous outcomes, when all or only some stratification errors are discovered, and when the treatment effect or treatment‐by‐covariate interaction effect is of interest. In Sections | PMC7614797 |
|
SIMULATION METHODS | A simulation study was conducted to investigate the impact of stratification errors on the analysis of trials that use stratified randomisation. Continuous outcomes for the For each scenario, 10 000 simulated datasets were generated and analysed using Stata 16.1 (StataCorp, College Station, Texas, USA). Simulation results were summarised using the simsum commandSummary of simulation parameters included in each simulation setting.When the covariate prevalence was 0.75, the error rate was three times higher in the more common stratum. | PMC7614797 |
||
Effect of stratification errors on the correlation between treatment groups | The previous finding that stratification induces correlation between the sample mean outcomes in the treatment groups | PMC7614797 |
||
Estimating treatment effects when all stratification errors are discovered | To compare the performance of different methods of adjusting for stratification variables when all stratification errors are discovered, outcomes were generated from model (1) with Using no additional covariates (unadjusted analysis),a main effect for a main effect for While adjusting for both Letting bias in the parameter estimate empirical standard error model‐based standard error relative percent error in the model‐based standard error coverage (the percentage of 95% confidence intervals for type I error/power (the percentage of tests of Monte Carlo standard errors were calculated for each performance measure based on equations presented elsewhere. | PMC7614797 |
||
Estimating treatment effects when only some stratification errors are discovered | To compare the performance of different methods of adjusting for stratification variables when only some stratification errors are discovered, the simulation study described in Section | PMC7614797 |
||
Estimating treatment by covariate interaction effects | SECONDARY | To explore the impact of stratification errors on treatment‐by‐covariate interaction tests, outcomes were generated from model (1) with The primary estimand of interest was A secondary estimand of interest was the marginal treatment effect | PMC7614797 |
|
Sensitivity of results to sample size and covariate prevalence | To explore the sensitivity of our simulation results to the chosen sample size and covariate prevalence, we repeated a subset of simulations (see Table | PMC7614797 |
||
SIMULATION RESULTS | PMC7614797 |
|||
Effect of stratification errors on the correlation between treatment groups | When there were no stratification errors, the correlation between the sample means in the two treatment groups was 0.10 for a strong covariate effect (Relationship between the probability of a stratification error and the correlation between the sample means in the intervention and control groups following stratified randomisation based on a covariate with a strong ( | PMC7614797 |
||
Estimating treatment effects when all stratification errors are discovered | SE | EFFECT INCREASED | All methods of analysis (unadjusted, adjusted for the randomisation strata and adjusted for the true strata) produced unbiased treatment effect estimates across all scenarios (results not shown). However, the methods differed according to other performance measures. In the reference setting when there were no stratification errors, the unadjusted analysis performed poorly as expected, producing model‐based standard errors that were too large, coverage rates that were too high, type I error rates that were too low and reduced power compared to the adjusted analysis methods. Problems worsened as the covariate effect increased. In contrast, both adjusted analysis methods performed well on all performance measures (see Supplementary Table When stratification errors occurred, the unadjusted analysis still performed poorly, albeit to a lesser extent when the error rate was high compared to low, and hence the correlation induced by stratification was reduced (as shown in Section Empirical SE across simulation scenarios comparing an unadjusted analysis, adjusting for the randomisation strata and adjusting for the true strata (which matches the updated strata in this setting) when the treatment effect is of interest and all stratification errors are discovered. The maximum Monte Carlo SE across all methods and scenarios was 0.0007.Type I error and power (%) across simulation scenarios comparing an unadjusted analysis, adjusting for the randomisation strata and adjusting for the true strata when the treatment effect is of interest and all stratification errors are discovered.
Matches the updated strata in this setting where all stratification errors are discovered. | PMC7614797 |
Estimating treatment effects when only some stratification errors are discovered | SE | The unadjusted analysis produced unbiased treatment effect estimates but performed poorly on other performance measures in all scenarios where only some stratification errors were discovered, consistent with the scenarios where all stratification errors were discovered (as expected, since this method is unaffected by whether errors are discovered or not). Adjusting for the true strata was the best performing approach, consistently producing unbiased treatment effect estimates (Supplementary Figure Type I error and power (%) across simulation scenarios comparing an unadjusted analysis, adjusting for the randomisation strata, adjusting for the true strata and adjusting for the updated strata when the treatment effect is of interest and only some stratification errors are discovered.
Coverage rates across simulation scenarios compared to the nominal rate of 95% comparing an unadjusted analysis, adjusting for the randomisation strata, adjusting for the true strata and adjusting for the updated strata when the treatment effect is of interest and only some stratification errors are discovered either equally or unequally across treatment groups. The maximum Monte Carlo SE across all methods and scenarios was 0.46. | PMC7614797 |
|
Estimating treatment by covariate interaction effects | When all stratification errors were discovered, using the true strata to test for an interaction effect performed well on all measures across all scenarios. This method consistently produced unbiased parameter estimates (Figure Bias in the treatment (Type I error and power (%) for the interaction test across simulation scenarios comparing an analysis based on the randomisation strata and the true strata when the treatment by covariate interaction effect is of interest, the covariate effect is strong (
Matches the updated strata in this setting where all stratification errors are discovered.When only some stratification errors were discovered, using the updated strata to test for an interaction effect was problematic in some cases. This method produced biased treatment and interaction parameter estimates (Figure Bias in the treatment ( | PMC7614797 |
||
Sensitivity of results to sample size and covariate prevalence | Reducing the sample size from 1000 to 200 resulted in larger empirical standard errors and reduced power across all methods as expected (results not shown). It had little impact on analyses based on the true strata otherwise and little impact on bias for any method. In contrast, the smaller sample size reduced but did not eliminate the undercoverage observed when estimating treatment effects using the updated strata (Supplementary Figure Increasing the covariate prevalence from 0.5 to 0.75 introduced bias and undercoverage into the treatment effect estimate based on the updated strata in the setting where errors were equal across strata but error discovery was unequal across treatment groups (Supplementary Figure | PMC7614797 |
||
EXAMPLE: THE | SE | REGRESSION, SECONDARY | To illustrate how different methods of adjusting for stratification variables perform in a real trial affected by stratification errors, we consider data from the OPTIMISE trial.Following randomisation, obstetric history was obtained from all participants' medical records, which revealed stratification errors for 23 women (3.6% of trial participants). Based on their medical records, 59% of women had no previous pregnancies. Accessing medical records prior to randomisation to determine the appropriate strata was not possible due to requirements for consent to access this information and the time delay involved in accessing paper records. The errors resulted from participants incorrectly answering the relevant question (“Have you had any pregnancies of 20 weeks or more?”) on the trial screening form, rather than data entry errors in the web‐based randomisation system. The reasons for this are unclear, since the screening question was relatively straightforward and language difficulties were not a factor for trial participants. Stratification errors were more common in one stratum, affecting only 5 (1%) women with no previous pregnancies compared to 18 (7%) women with a previous pregnancy. However, the discovery of these errors did not appear to depend on treatment group (11 errors in the intervention arm vs 12 in the control arm).For illustration, the key secondary outcome of infant birthweight was analysed using a linear regression model including the randomised treatment group and either no additional covariates (unadjusted model), the randomisation strata or the updated strata (which likely equals, or very closely matches, the true strata in this setting where medical records were reviewed for all trial participants). The stratification variable was an important predictor of this outcome, since the mean infant birthweight increased by 107 g (95% CI 21–193), or approximately 0.2 standard deviations, for women with a previous pregnancy compared to women without a previous pregnancy. The estimated treatment effects (SEs) were −79.6 g (43.9), −81.1 g (43.6) and −78.3 g (43.7) when adjusting for no covariates, the randomisation strata and the updated strata, respectively. Differences in treatment effect estimates between methods were small as expected from the simulation results, given the stratification errors were balanced across treatment groups and hence all methods are expected to produce unbiased treatment effect estimates. The unadjusted analysis produced the largest SE, consistent with simulation findings. An ad hoc approach to error discovery in OPTIMISE could have feasibly resulted in more errors being discovered in the intervention group through greater contact with these women, leading to larger differences in results depending on the chosen adjustment approach. For instance, if all errors in the intervention group had been discovered but none in the control group, an analysis adjusting for the updated strata (which likely matches the true strata for the intervention group but matches the randomisation strata for the control group) would have produced an estimated treatment effect (SE) of −82.1 g (43.7), which differs somewhat from the estimate of −78.3 g when adjustment was made for the updated strata and (presumably) all errors were discovered in both treatment groups. This result highlights the potential for the chosen method of adjustment to impact trial results. | PMC7614797 |
DISCUSSION | REGRESSION | In this article, we have used simulation studies and an example dataset to explore the impact of misclassification in a stratification variable when analysing continuous outcomes. Building on previous research,With few stratification errors, our simulations revealed little difference in performance between the different adjustment approaches and hence it is unlikely to matter which covariate is used for adjustment (the randomisation strata, true strata or updated strata) in most practical settings. Reassuringly, our results therefore raise no concerns over previous trial results based on any adjustment strategy, provided the error rate was low. Substantial differences between adjustment approaches were seen when the error rate was high, however, and the chosen approach has the potential to influence trial conclusions in this case. Our view is that the best performing method for handling stratification errors should be pre‐specified for the analysis, regardless of whether the trial ends up being affected by many errors or only one, and we provide recommendations regarding which methods to pre‐specify in Section Previous work on measurement error has shown that misclassification in a binary predictor variable leads to biased regression coefficient estimates for that predictor and possibly other predictors in a linear regression setting.Our simulations confirm previous findings that failing to adjust for stratification variables by performing an unadjusted analysis leads to biased standard errors and overcoverage, while adjusting for the randomisation strata avoids these issues.A limitation of our study is that we only considered continuous outcomes. Binary, count and time‐to‐event outcomes analysed via logistic regression, log Poisson regression and Cox proportional hazards models, respectively, have been studied previously. | PMC7614797 |
|
RECOMMENDATIONS FOR PRACTICE | SECONDARY, APPENDIX | Stratification errors can occur in any trial that uses stratified randomisation. We encourage researchers to consider the possibility of stratification errors at all stages of the trial and provide 10 general recommendations for addressing these errors during the design, conduct, analysis and reporting of the trial in Table Recommendations for addressing stratification errors by trial stage in trials using stratified randomisation.Implement strategies to minimise the risk of stratification errors, such as clearly defining stratification variables, avoiding stratification variables that are likely to be subject to high levels of misclassification, training staff to carefully check the stratification variables before randomising participants, and designing randomisation systems that are easy to use.Pre‐specify how stratification errors will be addressed in the primary analysis and any planned secondary or subgroup analyses in the statistical analysis plan.If the discovery of errors is unlikely to be related to treatment group (eg, in a blinded trial), then (i) adjust for the updated strataIf the discovery of errors could plausibly be related to treatment group (eg, in an unblinded trial involving more contact with intervention participants), then (i) adjust for the randomisation strataImplement systematic strategies to identify stratification errors across all participants, such as cross‐checking stratification variables against alternate data sources (where available).Thoroughly document any stratification errors that are identified, including how they occurred, to assist with staff training and inform the design of future trials.Record the updated strataExamine the number of stratification errors that were identified overall and broken down by treatment group, stratum and their combination.Address stratification errors in the planned analyses using the approach(es) pre‐specified in the statistical analysis plan.Consider whether any additional, unplanned analyses should be performed, based on the number and pattern of stratification errors that were identified in the trial (eg, a sensitivity analysis excluding participants affected by stratification errors).Report the number of stratification errors that were identified by treatment group and stratum, potentially in a Supplementary Appendix S1, or state that no such errors were identified.Indicate whether the randomisation strataThe In most trials, the discovery of stratification errors is unlikely to be related to treatment group. For example, in blinded trials the research staff typically have similar amounts of contact with participants in each group, providing similar opportunities for errors to be discovered in each group. In this case, we recommend adjusting for the updated strata in the primary analysis for all trial outcomes (Recommendation 2a) due to its power advantages. Adjusting for the randomisation strata should be considered in a sensitivity analysis, at least for the primary outcome, to explore the impact of any stratification errors on the trial results. If subgroup analyses by the stratification variable are planned, we recommend using the updated strata to define the subgroups and perform the interaction test in order to minimise any bias in the subgroup effect estimates.An exception to these recommendations should be made when the discovery of stratification errors could plausibly be related to treatment group. This can occur, for example, in unblinded trials involving substantially greater contact with trial participants assigned to one treatment arm and hence providing a greater opportunity for stratification errors to be discovered among these participants. In this setting, we recommend adjusting for the randomisation strata in the primary analysis for all trial outcomes (Recommendation 2b) to avoid the risk of introducing bias into the treatment effect estimates. Adjusting for the updated strata should be performed in a sensitivity analysis, at least for the primary outcome, to explore the impact of any stratification errors on the trial results, while acknowledging the results of this sensitivity analysis could be biased. The risk of bias can be judged by examining and reporting the number of stratification errors that were identified by treatment group, with bias expected when the discovery of stratification errors is more common in one group. If subgroup analyses by the stratification variable are planned, we recommend using the randomisation strata to define the subgroups and perform the interaction test. Although this approach will likely be biased and this should be acknowledged, less bias is expected compared to using the updated strata in this setting.Existing analysis guidelines state that ‘if one or more factors are used to stratify the design, it is appropriate to account for those factors in the analysis’ | PMC7614797 |
|
CONCLUSION | Stratification is a useful tool for achieving balanced treatment groups on important baseline covariates, but greater attention should be given to the risk of stratification errors and the implications these errors have for the trial analysis. In most trials, the discovery of stratification errors is unlikely to depend on treatment group and hence using the updated strata, where the errors that have been discovered are corrected, is preferable to using the randomisation strata for adjustment and performing subgroup analyses. In trials where the discovery of stratification errors could plausibly depend on treatment group, analyses should be based on the randomisation strata. Given the potential for stratification errors to introduce bias into treatment and subgroup effect estimates, greater transparency is needed in the reporting of stratification errors and how they were addressed in the analysis. | PMC7614797 |
||
FUNDING INFORMATION | This research was supported by a Centre of Research Excellence grant from the Australian National Health and Medical Research Council (ID 1171422), to the Australian Trials Methodology (AusTriM) Research Network. BCK and TPM are funded by the UK MRC, grants MC_UU_00004/07 and MC_UU_00004/09. KJL and TRS are supported by the Australian National Health and Medical Research Council (KJL Career Development Fellowship ID 1127984; TRS Emerging Leadership Grant ID 1173576). | PMC7614797 |
||
Supporting information | PMC7614797 |
|||
ACKNOWLEDGEMENTS | FORBES | The authors would like to thank Professor Andrew Forbes and other members of the Australian Trials Methodology (AusTriM) Research Network for useful feedback on this work. Open access publishing facilitated by The University of Adelaide, as part of the Wiley ‐ The University of Adelaide agreement via the Council of Australian University Librarians. | PMC7614797 |
|
DATA AVAILABILITY STATEMENT | The simulated data used to generate the findings presented in this article can be generated using the Stata code provided as supplementary material. The data from the OPTIMISE trial example are not publicly available due to privacy restrictions but can be requested by contacting the second author. | PMC7614797 |
||
REFERENCES | PMC7614797 |
|||
Background | ovulatory dysfunction | The aim of this study was to compare the efficacy of the combination of clomiphene citrate (CC) and letrozole to that of CC alone in inducing ovulation in infertile women with ovulatory dysfunction. | PMC10647029 |
|
Methods | Anovulatory infertility | ADVERSE EVENTS, ANOVULATORY INFERTILITY, SECONDARY | A randomized controlled trial was conducted at a single academic medical center between November 2020 and December 2021. Anovulatory infertility females, aged 18 to 40, were evenly distributed by a computer-generated block of four into two treatment groups. A “combination group” received a daily dose of CC (50 mg) and letrozole (2.5 mg), while a “CC-alone group” received a daily dose of CC alone (50 mg). The study medications were administered on days 3 through 7 of menstrual cycle. The primary outcome was the ovulation rate, defined by serum progesterone levels exceeding 3 ng/mL at the mid-luteal phase. The secondary outcomes were ovulation induction cycle characteristics, endometrial thickness, conception rate, and adverse events. | PMC10647029 |
Results | One hundred women (50 per group) were enrolled in the study. The mean age was not significantly different in both groups: 31.8 years in the combination group and 32.4 years in the CC-alone groups ( | PMC10647029 |
||
Conclusions | ovulatory dysfunction | Our study found no significant difference between the combination of CC and letrozole and CC alone in inducing ovulation in infertile women with ovulatory dysfunction in one cycle. The small number of live births precluded any meaningful statistical analysis. Further studies are needed to validate and extend our findings beyond the scope of the current study. | PMC10647029 |
|
Trial registration | The study was registered at | PMC10647029 |
||
Supplementary Information | The online version contains supplementary material available at 10.1186/s12905-023-02773-7. | PMC10647029 |
||
Keywords | PMC10647029 |
|||
Background | anovulation, Polycystic Ovary Syndrome II, PCOS, infertility, Infertility | INFERTILITY, ANOVULATION | Infertility is defined as the inability to conceive through regular intercourse without contraception for 12 months in women under 35 and 6 months in women 35 or older [Medical induction of ovulation is a primary treatment for anovulation, particularly in patients with PCOS. Clomiphene citrate (CC), a selective estrogen receptor modulator, is commonly used for this condition [Letrozole has been proposed as a first-line treatment for ovulation induction in PCOS [Several studies have compared the effects of CC and letrozole on infertility, particularly their impact on ovulation induction. The Pregnancy and Polycystic Ovary Syndrome II trial found a significantly higher live birth rate (27.5% vs. 19.1%; | PMC10647029 |
Methods | PMC10647029 |
|||
Study design and overview | Infertility | INFERTILITY | This study employed a randomized controlled trial design. It was conducted at the Infertility and Reproductive Biology Unit of the Department of Obstetrics and Gynecology, Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok, Thailand, from November 2020 to December 2021. Eligible participants were randomly allocated to a “combination group” or a “CC-alone group” after Institutional Review Board approval. Those assigned to the combination group received a daily dose of CC (50 mg) and letrozole (2.5 mg), whereas participants in the CC-alone group were administered a daily dose of CC alone (50 mg) [ | PMC10647029 |
Randomization and blinding | ONCOLOGY, WITHDRAWAL BLEEDING | The randomization scheme in this study was computer-generated using blocks of four, with group assignments concealed in sealed envelopes. The sonographer was blinded to the assignments. Participants were randomized in a 1:1 ratio to receive a daily dosage of CC (50 mg; Ovamit, Remedica Ltd, Limassol, Cyprus) in combination with letrozole (2.5 mg; Fresenius Kabi Oncology Ltd, Kolkata, India) or CC alone (50 mg) daily. Both treatment regimens were taken from days 3 through 7 of the menstrual cycle. Patients with long menstrual cycles were prescribed an oral progestogen to induce withdrawal bleeding. The progestogens administered were medroxyprogesterone acetate (10 mg daily; Provera; Pfizer, New York, NY, USA) or norethisterone acetate (5 mg daily; Primolut N; Bayer Thai Co Ltd, Bangkok, Thailand) for 7–10 days. | PMC10647029 |
|
Study procedures | ADVERSE EFFECTS | At the start of their menstrual cycle, participants were instructed to contact the investigator to arrange the ovulation induction schedule. The allocated treatment medication regimen was taken from days 3 through 7 of one menstrual cycle. Home urinary luteinizing hormone (LH) tests were performed twice daily, in the morning and at night, starting on cycle day 12 until a positive result was obtained or until cycle day 21 if the results continued to be negative. Patients were instructed to send pictures of the urinary LH test results to the researcher to confirm the results. Regular intercourse, performed two to three times per week, was recommended starting on cycle day 12 and on the day of the positive urinary LH test. Transvaginal ultrasound was performed on day 12 to day 14 of the cycle by a single operator throughout the project, and follicular growth and endometrial thickness were recorded. Serum progesterone levels were obtained 7 days after a positive urinary LH test or on cycle day 21 or day 22 in cases with negative urinary LH. A urine pregnancy test was performed 7 days after an ovulatory serum progesterone level or on cycle day 35 if there was no confirmation of ovulation and no menstrual period. Women with a positive urine pregnancy test were scheduled for a transvaginal ultrasound to confirm pregnancy 2 to 3 weeks later. A questionnaire was used to elicit information about any adverse effects of the study medications during the study period. | PMC10647029 |
|
Outcome measures | The primary outcome was the ovulation rate, defined as a mid-luteal progesterone level greater than 3 ng/mL [ | PMC10647029 |
||
Sample size calculation and statistical analyses | The sample size calculation was informed by Meija et al. [The statistical analyses were performed using PASW Statistics for Windows, version 18.0 (SPSS Inc, Chicago, IL, USA). The study conducted an intention-to-treat analysis involving all randomized participants and a per-protocol analysis restricted to those who followed the designated treatment. Descriptive statistics were used to describe patient and cycle characteristics. Continuous data were presented as the means ± standard deviations or medians (ranges) for normally distributed and nonnormally distributed data, respectively, whereas categorical data were expressed as numbers and percentages. For categorical data, Pearson chi-squared, Yates’ continuity correction, or Fisher’s exact test were performed to compare the proportions between two groups. The independent Student’s t-test was used for normally distributed continuous data, while the Mann-Whitney U-test was used for nonnormally distributed continuous data for comparing the mean or median, respectively, between two groups. Absolute differences were computed by substracting percentages of the combination group’s outcomes from percentages of the CC-alone group’s outcomes, while the rate ratio was computed by percentages of the combination group’s outcomes divided by percentages of the CC-alone group’s outcomes. The level of statistical significance was set at | PMC10647029 |
||
Discussion | Abdominal bloating, congenital anomalies, PCOS, infertility, ovulatory dysfunction | ADVERSE EFFECTS, MINOR, ANOVULATORY, SIDE EFFECT | This study aimed to compare the efficacy of a combination of CC and letrozole versus CC alone for ovulation induction in infertile women with ovulatory dysfunction. A previous study reported that women with PCOS had a significantly higher ovulation rate with combination therapy than with letrozole alone [From a theoretical standpoint, the local action of letrozole and the central influence of CC were anticipated to synergistically enhance ovulation induction. However, our study found comparable ovulation rates for the combination and CC-alone groups. Furthermore, the dose of CC was identical in both groups of our study, and a previous study reported a higher ovulation rate in a combination group than in a letrozole-alone group [Numerous studies have investigated the individual effects of letrozole and CC on the ovulation rate; however, studies examining the combined effect of these medications are limited. A study by Hajishafiha et al. investigated the combination of letrozole and CC in PCOS patients and reported a follicle development rate of 82.9% in the combination group. However, that study included patients who were refractory to CC or letrozole, and ovulation confirmation using follicle size alone (as was done by Hajishafiha et al.) is not always reliable [Previous meta-analyses have reported significant heterogeneity among randomized controlled trials investigating the ovulation rates achieved with letrozole and CC in the PCOS population, although letrozole significantly increased the ovulation rate compared to CC [The side effects of the 2 groups in our study were comparable, and all adverse effects from medication were minor and tolerable. Abdominal bloating was the most common side effect reported in both groups. We also found no evidence of congenital anomalies in the live births from either group. Taken together, these results indicate that both treatment regimens were safe and well tolerated.The observations made in this study represent a significant contribution to the literature. Notably, our study is the first to compare the efficacy of ovulation induction achieved with the combination of CC and letrozole versus CC alone in anovulatory women. Our study’s randomized controlled trial design minimized confounding factors and neutralized baseline characteristics between groups. Additionally, mid-luteal serum progesterone clearly defined the primary outcome, which was unaffected by operator and participant biases. To minimize interobserver variation, a single operator blinded to the study medications performed all transvaginal ultrasonographic investigations.Several limitations should be acknowledged in this study. First, the low pregnancy rate may limit the generalizability of our findings, as we did not evaluate all possible infertility factors that could have contributed to the low rate. Second, while we advised participants on the optimal timing of intercourse, we could not determine the degree of adherence to these instructions. Third, our study’s treatment and follow-up durations were relatively short, as we evaluated only one treatment cycle in each patient. Finally, the letrozole-only group for ovulation induction was not included in this study. Therefore, more treatment groups and long follow-up periods are needed to assess and compare the cumulative ovulatory rate between these ovulation induction regimens. As we know that live birth is the most meaningful outcome, the ovulation rate was a reasonable primary outcome for evaluating the efficacy of ovulation induction agents. This is because the rate minimizes confounding factors related to infertility factors beyond ovulation. | PMC10647029 |
Conclusions | ovulatory dysfunction | Our study found no significant difference in the ovulation rates of infertile women with ovulatory dysfunction. Specifically, the rates achieved with a combination of CC and letrozole were not significantly different from those achieved with CC 50 mg alone in one cycle. However, the low number of live births precluded statistical analysis. According to the limitations mentioned above, future studies with larger sample sizes and extended follow-up periods or equivalence trials are needed to evaluate the cumulative ovulatory rates and dose-defining efficacy of these ovulation-induction regimens. | PMC10647029 |
|
Acknowledgements | The authors gratefully acknowledge the patients who generously agreed to participate in this study and Dr Sasima Tongsai for her assistance with the statistical analyses. | PMC10647029 |
||
Authors’ contributions | P.C. was responsible for project development, data collection, and manuscript editing. S.T. was responsible for manuscript writing and data analysis. I.T. and P.L. were responsible for manuscript review and editing. All authors read and approved the final manuscript. | PMC10647029 |
||
Funding | This study was supported by a grant from the Siriraj Research Development Fund of the Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok, Thailand. | PMC10647029 |
||
Data availability | The datasets used and/or analyzed during the current study are not publicly available due to the confidentiality of participants’ data and the difficulty of organizing the raw data to be suitable for publication; however, they are available from the corresponding author on reasonable request. | PMC10647029 |
||
Declarations | PMC10647029 |
|||
Ethics approval and consent to participate | The study was approved by the Siriraj Institutional Review Board and all participants provided written informed consent to participate before enrollment. All methods were performed in accordance with the Declaration of Helsinki. | PMC10647029 |
||
Consent for publication | Not applicable. | PMC10647029 |
||
Competing interests | The authors declare no competing interests. | PMC10647029 |
||
Conflict of interest | All authors declare that there are no personal or professional conflicts of interest and no financial support from the companies that produce and/or distribute the drugs, devices, or materials described in this report. | PMC10647029 |
||
References | PMC10647029 |
|||
Backgroud | malignancy, gastric gastrointestinal stromal tumors, GISTs | REGRESSION | To predict the malignancy of 1–5 cm gastric gastrointestinal stromal tumors (GISTs) by machine learning (ML) on CT images using three models - Logistic Regression (LR), Decision Tree (DT) and Gradient Boosting Decision Tree (GBDT). | PMC10327391 |
Methods | 231 patients from Center 1 were randomly assigned into the training cohort (n = 161) and the internal validation cohort (n = 70) in a 7:3 ratio. The other 78 patients from Center 2 served as the external test cohort. Scikit-learn software was used to build three classifiers. The performance of the three models were evaluated by sensitivity, specificity, accuracy, positive predictive value (PPV), negative predictive value (NPV) and area under the curve (AUC). Diagnostic differences between ML models and radiologists were compared in the external test cohort. Important features of LR and GBDT were analyzed and compared. | PMC10327391 |
||
Results | GBDT outperformed LR and DT with the largest AUC values (0.981 and 0.815) in the training and internal validation cohorts and the greatest accuracy (0.923, 0.833 and 0.844) across all three cohorts. However, LR was found to have the largest AUC value (0.910) in the external test cohort. DT yielded the worst accuracy (0.790 and 0.727) and AUC values (0.803 and 0.700) in both the internal validation cohort and the external test cohort. GBDT and LR performed better than radiologists. Long diameter was demonstrated to be the same and most important CT feature for GBDT and LR. | PMC10327391 |
||
Conclusions | GISTs | ML classifiers, especially GBDT and LR with high accuracy and strong robustness, were considered to be promising in risk classification of 1–5 cm gastric GISTs based on CT. Long diameter was found the most important feature for risk stratification. | PMC10327391 |
|
Supplementary Information | The online version contains supplementary material available at 10.1186/s12880-023-01053-y. | PMC10327391 |
||
Keywords | PMC10327391 |
|||
Backgroud | gastric mesenchymal tumors, tumor, neoplasms, hemorrhage, GISTs, necrosis, [Machine learning | TUMOR, ULCERATION, GASTROINTESTINAL STROMAL TUMORS, NEOPLASMS, DISEASE, HEMORRHAGE, GIST, NECROSIS, ONCOLOGY, METASTASES | Gastrointestinal stromal tumors (GISTs) are neoplasms that arise from Cajal cells of the gastrointestinal tract mesenchyme [The National Institutes of Health (NIH) classification system has been proposed to stratify the risk of GISTs. Currently, the modified NIH risk stratification criteria and the latest Chinese consensus guidelines (2017 Edition) of the Chinese Society of Clinical Oncology (CSCO) Expert Committee on GIST divide GISTs into very low, low, intermediate, and high risk groups according to tumor size, location, mitotic index, and whether the tumor ruptures [Contrast-enhanced CT (CE-CT) scan can clearly show not only the anatomical structure of gastric mesenchymal tumors, but also the internal and peripheral information of the lesion, including tumor density, necrosis, ulceration, hemorrhage, blood vessels, as well as invasion of surrounding tissues, lymph node metastases, and distant metastases [Plenty of studies predicting the risk stratification of GISTs based on CT imaging have been reported [Machine learning (ML) algorithm provides the possibility of mining valuable data that have significant and intricate connections among enormous data items. ML algorithms have been applied to disease identification, differential diagnosis and prognosis analysis with outstanding performance and promising prospect [ | PMC10327391 |
Methods | PMC10327391 |
|||
Patient selection | gastric GISTs, tumor | TUMOR | This retrospective study was approved by the ethics committee of Tongde Hospital of Zhejiang Province and the need for informed consent was waived (Approval No. 2021-040). Patients with gastric GISTs from two centers (Center 1: Tongde Hospital of Zhejiang Province, Center 2: Zhejiang Hospital) from January, 2012 to September, 2022 were enrolled in this research. The criteria for inclusion were as follows: (a) patients with complete CT images (including unenhanced, arterial and portal venous phase images) within 15 days before surgery; (b) solitary and primary lesions; (c) lesions without neoadjuvant treatment; (d) lesions larger than 1 cm and smaller than 5 cm in the long diameter. (e) patients with detailed clinical data (including age, gender, clinical symptoms and tumor markers). The inclusion and exclusion of patients are shown in Fig.
Flowchart shows inclusion and exclusion criteria for this study | PMC10327391 |
CT examination | All patients underwent abdominal CE-CT examination using two 64-slice spiral CT scanners (Siemens, Forchheim, Germany or Philips Medical Systems, Cleveland, OH, USA). The parameters of CT imaging were set as follows: for Siemens, 120 kV tube voltage, 150–250 mA tube current, 0.5 s tube rotation time, 64 × 0.6 mm detector collimation, 350 × 350 mm field of view, 5 mm section thickness and 1-1.25 mm reconstruction interval; for Philips, 120 kV tube voltage, 200–250 mA tube current, 0.5 s tube rotation time, 64 × 0.625 mm detector collimation, 350 × 350 mm field of view, 5 mm section thickness and 1-1.5 mm reconstruction interval. Subsequently, arterial phase (delay 30–40 s) and the portal venous phase (delay 60–70 s) images were obtained after 2 mL/kg of iodinated contrast medium was injected intravenously at a rate of 3 ml/s. | PMC10327391 |
||
Image analysis | tumor, calcification, necrotic, necrosis | NECROSIS, TUMOR, ENLARGED LYMPH NODES, NECROTIC, ULCERATION | Two radiologists (Reviewer 1 with 6 years and Reviewer 2 with 13 years of experience in abdominal imaging) independently reviewed CT images and reached final conclusions by consensus without knowledge of the surgical and pathological information of every patient. The determined CT imaging features included (a) the CT attenuation values (Hounsfield units, HU) in unenhancement phase (CTU), (b) in arterial phase (CTA) and (c) in venous phase (CTV) of the tumor, (d) degree of enhancement in arterial phase (DEAP) and (e) in portal venous phase (DEPP), (f) enhanced potentiality in arterial phase (EPa) and (g) in portal venous phase (EPv), (h) long diameter (LD), (i) short diameter (SD), (j) the ratio of long diameter to short diameter (LD/SD), (k) contour (round; oval; irregular), (l) necrosis (yes or no), (m) calcification (yes or no), (n) surface ulceration (yes or no), (o) intratumoral angiogenesis (yes or no) and (p) peripheral enlarged lymph node (LN) (yes or no). The CT attenuation value was measured by drawing the region of interest (ROI) on the same axial slice of the tumor avoiding vessels, calcification, and the necrotic regions. DEAP or DEPP was obtained by subtracting CTU from CTA or CTV respectively. EPa or EPv was equal to DEAP or DEPP divided by CTU. Enlarged lymph node was considered present if the shortest axis diameter of lymph node was more than 10 mm. Some of the CT features referred to our previous report [ | PMC10327391 |
Machine learning | Scikit-learn software was used to build three classifiers-DT, GBDT and LR. The detailed methodology is described on the website of official documentation ( | PMC10327391 |
||
Grid search strategy for selecting optimal parameters | In order to find the optimal parameters of three models, the grid search strategy in scikit-learn software was used. In the grid search process, 5-fold Cross-Validation (CV) was used to evaluate model performance. Meanwhile, the accuracy was used as an evaluation metric to maximize model performance. The detail of grid search method is described in the model selection module on the website of official documentation ( | PMC10327391 |
||
Logistic regression (LR) | LR is the most conventional approach to measure the relationship between discrete response variables and several covariates by estimating probabilities. It can be written as: The final optimal parameters of LR were set as follows: C = 100, random_state = 12, penalty = ’l1’, solver = ’liblinear’. Other parameter factors were set as default in sklearn software module. | PMC10327391 |
||
Decision tree (DT) | DT, as a binary method, can be used to classify data by calculating their characteristics. Decision nodes, branches and leaves are the three main components of DT. DT starts with a node and extends to many branches and child nodes, finally to leaves. The criterion used in our model is the Gini’s Diversity Index, which is a measure of node impurity. The standard CART algorithm, implemented using sciki-learn library in Python, was applied to build decision tree.The parameters set in the DT were: random_state = 0, max_features = 6, max_depth = 6, criterion = ’gini’. Other parameters were set as default in sklearn software module. | PMC10327391 |
||
Gradient boosting decision tree (GBDT) | GBDT is an ensemble classifier based on bootstrap sampling, which aims to improve the generalization ability and robustness by combining the prediction results of multiple base learners (i.e. weak decision trees). The weight is adjusted with iteration, so that the higher weight will be assigned to the data poorly classified. Totally 15 weak decision trees were created in GBDT model in this study (e.g. a tree is showed in Fig The following parameter factors were used for GBDT: learning_rate = 0.1, max_depth = 8, random_state = 0, min_samples_leaf = 2. Other parameters were also set as default in the sklearn software. | PMC10327391 |
||
Performance comparison between radiologists and models | Diagnostic performance differences between the three ML models and the two radiologists were compared in the external test cohort. Before performance comparisons, intra-class correlation coefficients (ICCs) were calculated to assess agreement between the two reviewers. | PMC10327391 |
||
Feature variable analysis | gastric GISTs | GBDT and LR showed excellent diagnostic efficiency in predicting risk classification of gastric GISTs on account of the high accuracy and strong robustness. LR is well known for determining the beneficial features to support decision by linear analysis, since the result is easy to explain. Firstly, significant CT features were determined by univariate analysis. Secondly, variable with | PMC10327391 |
|
Statistical analysis | high-grade malignancy, malignancy, SD | P-P plots and Q-Q plots were used to assess normal distribution of data. Continuous distributed data were showed as mean ± SD, and categorical variables were expressed as n (%). Univariate analysis using t test or Mann-Whitney U test for continuous variables and Fisher’s exact test for categorical variables were performed to compare CT features between the low-grade malignancy and high-grade malignancy groups. Variables with | PMC10327391 |
|
Results | PMC10327391 |
|||
Clinical characteristics of patients | gastric GISTs, GISTs, tumors | TUMORS | 231 patients (109 men and 122 women; mean age, 59.47 ± 10.13 years) from Center 1 and 78 patients (41 men and 37 women; mean age, 62.69 ± 10.78 years) from Center 2 were included in our series. The training cohort enrolled 161 patients with gastric GISTs consisting of 47 high-risk tumors and 114 low-risk ones. 70 patients with GISTs (21 high-risk tumors and 49 low-risk ones) and 78 patients with GISTs (16 high-risk tumors and 62 low-risk ones) constituted the internal validation cohort and the external test cohort, respectively.Details of the clinical characteristics of three cohorts are shown in Table | PMC10327391 |
Model evaluation | gastric GISTs, Confusion | REGRESSION | Results of diagnostic performance of LR, DT and GBDT are described in Table
Diagnostic performance analysis of LR, DT and GBDT modelsLR, Logistic regression; DT, Decision tree; GBDT, Gradient boosting decision tree; AUC, area under the curve; CI, confidence interval, NPV, negative predictive value; PPV, positive predictive value
Receiver operating characteristic (ROC) curves of three models to predict the risk stratification of gastric GISTs. (
Confusion matrixes of LR ( | PMC10327391 |
Performance comparison between radiologists and models | ICC of 0.83 indicated that the agreement between two radiologists was good. Table
Results of radiologists’ diagnostic performance in the external test cohortAUC, area under the curve; CI, confidence interval, NPV, negative predictive value; PPV, positive predictive value | PMC10327391 |
||
Discussion | tumor, GISTs tumors, malignancy, gastric GISTs, GISTs, high-grade malignant GISTs | REGRESSION, TUMOR, GIST | To our best knowledge, this is the first research on the prediction of malignancy in gastric GISTs by machine learning classifiers. In addition, our report focuses on GISTs tumors of 1–5 cm in the gastric, which is different from studies that include large-size GISTs located in various sites of the gastrointestinal tract. This study has the largest sample size among relevant studies, so the reliability of the results can be guaranteed. Various qualitative and quantitative variables extracted from CT signs were inputted into LR, DT and GBDT models. The results of model evaluation were different but not inconsistent among the three cohorts. For the training cohort, GBDT had the greatest sensitivity, specificity, accuracy, PPV, NPV and AUC among the three models. What’s more, GBDT gained the largest AUC in the training and internal validation sets and performed best in all three cohorts in terms of accuracy, although the AUC was not as good as LR in the external test cohort. Furthermore, GBDT and LR showed better performance than the two radiologists. However, the performance of DT was not as outstanding as GBDT and LR. Therefore, GBDT and LR were suggested to be promising ML models for CT-based risk classification prediction of gastric GISTs due to the high accuracy and strong robustness.GBDT, an ensemble method based on bootstrap sampling, was demonstrated to be a favorable algorithm with high predictive efficiency, as reported in various previous researches [When it comes to feature variable selection, LD was found to be the only common CT feature between LR and GBDT that distinguished for high-grade malignant GISTs in this study. Several studies using multivariate logistic regression analysis revealed that the size of GISTs was the only independent risk factor for differentiation of the high-grade malignant GISTs [Artificially determined CT imaging features were used as the input variables for the three ML models to predict the risk classification of gastric GISTs in this study, and great results were obtained, especially in GBDT and LR. Compared with the diagnostic ability of radiologists, ML achieved more promising results, which may have a guiding prospect for doctors in daily diagnostic work, especially for the junior ones. It may promisingly provide theoretical and practical support for texture analysis or deep learning since ML may play an important role in feature selection.There are some limitations in our study. First, our sample size was small for ML. ML classifiers can highlight their advantages in the context of large data, amounts of predictor variables or complex relationship. Second, four risk grades were finally divided into two, so the results were unable to meet the requirement of each risk classification. Simple ML model cannot meet the needs of predicting four risk levels, but the convolutional neural network can, which puts the next step of research on the agenda. Third, only three simple ML models were implemented in our research, including the classic LR. We will try other more complex ML models to assess the risk stratification, such as random forest, support vector machine, k-nearest neighbors, etc. Fourth, radiomics, which transforms medical images into high-dimensional data by extracting tumor’s shape, intensity, and texture features, has recently shown great potential in aiding clinical decision making. Developing CT-based radiomics models for GIST risk stratifcation will be a future work. | PMC10327391 |
Conclusions | GISTs | In summary, GBDT and LR showed outstanding performance with high accuracy and strong robustness in the risk assessment of gastric GISTs less than 5 cm on CT imaging. The long diameter of lesion was found to be the most important feature for risk stratification. | PMC10327391 |
|
Electronic supplementary material | Below is the link to the electronic supplementary material.
Supplementary Material 1
Supplementary Material 2 | PMC10327391 |
||
Acknowledgements | Not applicable. | PMC10327391 |
||
Authors’ contributions | CZ, JW, HJY designed the research. JW, HJY provided the administrative support for this research. CZ, YY, BLD, ZHX, FMZ were responsible for data collection. CZ, YY, BLD, ZHX, FMZ were responsible for data analysis and interpretation. CZ wrote the first draft. All authors reviewed the analyses and drafts of this manuscript and approved its final version. | PMC10327391 |
||
Funding | This research was supported by Zhejiang Provincial Natural Science Foundation of China under Grant No.LGF21H030004. | PMC10327391 |
||
Data Availability | All datasets presented and analyzed in this study were interpreted and provided by the corresponding author. | PMC10327391 |
||
Declarations | PMC10327391 |
|||
Ethics approval and consent to participate | This retrospective study was approved by ethics committee of Tongde Hospital of Zhejiang Province (Approval No: 2021-040). Written informed consent was waived by the ethics committee of Tongde Hospital of Zhejiang Province. All methods were carried out in accordance with relevant guidelines and regulations. | PMC10327391 |
||
Consent for publication | Not applicable. | PMC10327391 |
||
Competing interests | The authors declare no competing interests. | PMC10327391 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.