Introduction

On April 17, 2014, the US Food and Drug Administration (FDA) published a safety communication discouraging “the use of laparoscopic power morcellation during hysterectomy or myomectomy for the treatment of women with uterine fibroids.” Due to the concern of worsening outcomes for patients with occult uterine malignancy, specifically uterine leiomyosarcoma, the FDA recommended a significant change to existing surgical planning, patient consent, and surgical technique.

Uterine power morcellation allows for an efficient retrieval of large surgical specimens through small laparoscopic skin incisions. Laparoscopic and vaginal approaches to hysterectomy are generally associated with improved outcomes and decreased complications compared to abdominal approaches. However, due to an inability to ensure benign pathology and a concern for worsening the outcomes for patients with occult, aggressive malignancies, patients and surgeons are shifting away from power morcellation as a surgical tool used during hysterectomy. A recent survey of minimally invasive surgeons found that 84% have changed their surgical approach planning after the FDA communication. Of surgeons, 25% reported changing to total abdominal hysterectomy for selected patients. Although utilization of abdominal hysterectomy may decrease the specific risk of dissemination of occult malignancy, patients are at risk for increased surgical morbidity with open procedures. Consequently, patients and their surgeons may be trading the risk of one complication for another. To date, the effect that this widespread change in surgical practice has had on surgical complications has not been evaluated.

Therefore, we sought to analyze changes in surgical approach to hysterectomy and postoperative surgical complications before and after the April 2014 FDA power morcellation safety communication for patients undergoing hysterectomy for benign indications from a statewide database.

Materials and Methods

This study evaluated patients undergoing hysterectomy from the Michigan Surgical Quality Collaborative, a voluntary statewide surgical collaborative including community and academic hospitals. Data were abstracted from charts by specially trained, dedicated nurse abstractors. Patient characteristics, intraoperative processes of care, and 30-day postoperative outcomes from hysterectomy cases at member hospitals are prospectively collected by trained nurse abstractors using standard data collection instruments. To reduce sampling bias, a standardized data collection methodology is employed that uses only the first 25 cases of an 8-day cycle. Routine validation of the data is maintained by scheduled site visits, conference calls, and internal audits. Detailed methods of the registry’s data collection have been described previously.

Hysterectomies performed from Jan. 1, 2013, through Dec. 31, 2014, were included. To focus on patients with benign indications for hysterectomy, we excluded hysterectomies with an indication of malignancy, endometrial hyperplasia, or cervical dysplasia.

Our primary outcome was the change in utilization of hysterectomy surgical approach: abdominal, vaginal (including laparoscopic-assisted vaginal), and laparoscopic (including robotic-assisted) before and after April 17, 2014. Our secondary outcome was the change in incidence of major postoperative complications, readmissions, and reoperations within 30 days of surgery. Postoperative complications were classified using a database-wide classification system. Major postoperative complications included blood transfusion, vaginal cuff infection, pelvic abscess diagnosis, vaginal cuff infection, vaginal cuff dehiscence, ureteral obstruction, vesicovaginal fistula, ureterovaginal fistula, rectovaginal fistula, intestinal obstruction, central line infection, deep space surgical site infection, organ space surgical site infection, sepsis, pulmonary embolism, deep vein thrombosis requiring therapy, unplanned intubation, acute renal failure, cerebral vascular accident, myocardial infarction, cardiac arrest requiring cardiopulmonary, and death.

The primary independent variable was whether or not the surgery was performed after April 17, 2014, the day of publication of the FDA power morcellation safety communication.

To adjust for the baseline differences between the cases of the 2 time periods, several case-mix characteristics were included as covariates in the statistical models. Covariates included in the analysis were patient and surgical characteristics: age, race, body mass index, surgical indication, history of pelvic surgery, pathologist-measured uterine specimen weight; and a modified Charlson comorbidity score including: history of myocardial infarction, congestive heart failure, peripheral vascular disease, chronic obstructive pulmonary disease, stroke, diabetes, chronic kidney disease, cancer diagnosis, liver disease, and age in the score calculation. To help control for confounding by overall surgical procedure complexity, we included in the model the total concurrent operative procedure total relative value units based on codes from the Physicians’ Current Procedural Terminology Coding System, Fourth Edition .

The cost of hospital readmissions was calculated using the Michigan Value Collaborative, a voluntary statewide collaborative including community and academic hospitals that collects and risk-adjusts claims payments from a large private insurance provider and Medicare. All payments are risk-adjusted by age and medical comorbidities and standardized to published Medicare payment amounts. The episode cost of hospital readmission represents the 90-day costs for hospital readmissions for any patient who was readmitted.

Multinomial outcomes (ie, surgical approach to hysterectomy) were analyzed using tests on the equality of proportions and multinomial regression models controlling for the covariates noted previously, as well as robust SE accounting for hospital-level clustering.

Dichotomous outcomes (ie, the presence of a postoperative complication) were analyzed with χ ² tests and multivariable logistic regression models controlling for the covariates noted previously, as well as robust SE accounting for hospital-level clustering. By accounting for violations in model assumptions due to nonindependence of observations within clusters of the data, Huber-Eicker-White robust SE better reflect the collected data characteristics. Readmission episode costs are reported by their median value and interquartile range.

We reported the outcome in terms of the marginal effect: the predicted outcome probabilities when the independent variable is present or not, holding all other covariates at their known values, because effect sizes cannot be directly inferred from coefficients of logistic regression models.

To detect a 1% increase from a baseline hospital readmission rate of 3.5%, type 1 error of 5%, type 2 error of 20%, and a 2:1 ratio of number of cases before the safety communication to cases after the safety communication, a sample size of 10,540 hysterectomies was required. The study was deemed exempt by the University of Michigan Institutional Review Board-Medical. Software (STATA 13.1; StataCorp LP, College Station, TX) was used for all analyses.

Materials and Methods

This study evaluated patients undergoing hysterectomy from the Michigan Surgical Quality Collaborative, a voluntary statewide surgical collaborative including community and academic hospitals. Data were abstracted from charts by specially trained, dedicated nurse abstractors. Patient characteristics, intraoperative processes of care, and 30-day postoperative outcomes from hysterectomy cases at member hospitals are prospectively collected by trained nurse abstractors using standard data collection instruments. To reduce sampling bias, a standardized data collection methodology is employed that uses only the first 25 cases of an 8-day cycle. Routine validation of the data is maintained by scheduled site visits, conference calls, and internal audits. Detailed methods of the registry’s data collection have been described previously.

Hysterectomies performed from Jan. 1, 2013, through Dec. 31, 2014, were included. To focus on patients with benign indications for hysterectomy, we excluded hysterectomies with an indication of malignancy, endometrial hyperplasia, or cervical dysplasia.

Our primary outcome was the change in utilization of hysterectomy surgical approach: abdominal, vaginal (including laparoscopic-assisted vaginal), and laparoscopic (including robotic-assisted) before and after April 17, 2014. Our secondary outcome was the change in incidence of major postoperative complications, readmissions, and reoperations within 30 days of surgery. Postoperative complications were classified using a database-wide classification system. Major postoperative complications included blood transfusion, vaginal cuff infection, pelvic abscess diagnosis, vaginal cuff infection, vaginal cuff dehiscence, ureteral obstruction, vesicovaginal fistula, ureterovaginal fistula, rectovaginal fistula, intestinal obstruction, central line infection, deep space surgical site infection, organ space surgical site infection, sepsis, pulmonary embolism, deep vein thrombosis requiring therapy, unplanned intubation, acute renal failure, cerebral vascular accident, myocardial infarction, cardiac arrest requiring cardiopulmonary, and death.

The primary independent variable was whether or not the surgery was performed after April 17, 2014, the day of publication of the FDA power morcellation safety communication.

To adjust for the baseline differences between the cases of the 2 time periods, several case-mix characteristics were included as covariates in the statistical models. Covariates included in the analysis were patient and surgical characteristics: age, race, body mass index, surgical indication, history of pelvic surgery, pathologist-measured uterine specimen weight; and a modified Charlson comorbidity score including: history of myocardial infarction, congestive heart failure, peripheral vascular disease, chronic obstructive pulmonary disease, stroke, diabetes, chronic kidney disease, cancer diagnosis, liver disease, and age in the score calculation. To help control for confounding by overall surgical procedure complexity, we included in the model the total concurrent operative procedure total relative value units based on codes from the Physicians’ Current Procedural Terminology Coding System, Fourth Edition .

The cost of hospital readmissions was calculated using the Michigan Value Collaborative, a voluntary statewide collaborative including community and academic hospitals that collects and risk-adjusts claims payments from a large private insurance provider and Medicare. All payments are risk-adjusted by age and medical comorbidities and standardized to published Medicare payment amounts. The episode cost of hospital readmission represents the 90-day costs for hospital readmissions for any patient who was readmitted.

Multinomial outcomes (ie, surgical approach to hysterectomy) were analyzed using tests on the equality of proportions and multinomial regression models controlling for the covariates noted previously, as well as robust SE accounting for hospital-level clustering.

Dichotomous outcomes (ie, the presence of a postoperative complication) were analyzed with χ ² tests and multivariable logistic regression models controlling for the covariates noted previously, as well as robust SE accounting for hospital-level clustering. By accounting for violations in model assumptions due to nonindependence of observations within clusters of the data, Huber-Eicker-White robust SE better reflect the collected data characteristics. Readmission episode costs are reported by their median value and interquartile range.

We reported the outcome in terms of the marginal effect: the predicted outcome probabilities when the independent variable is present or not, holding all other covariates at their known values, because effect sizes cannot be directly inferred from coefficients of logistic regression models.

To detect a 1% increase from a baseline hospital readmission rate of 3.5%, type 1 error of 5%, type 2 error of 20%, and a 2:1 ratio of number of cases before the safety communication to cases after the safety communication, a sample size of 10,540 hysterectomies was required. The study was deemed exempt by the University of Michigan Institutional Review Board-Medical. Software (STATA 13.1; StataCorp LP, College Station, TX) was used for all analyses.

Results

There were 18,299 hysterectomies that met the inclusion criteria, with 2753 cases excluded due to malignant or premalignant surgical indications and an additional 174 cases excluded for missing covariate data ( Supplemental Figure 1 ). Of the included hysterectomies, 9597 were performed prior to the FDA power morcellation safety communication and 5775 were performed afterward. Groups were similar in terms of demographics and perioperative characteristics ( Table 1 ).

Table 2 and Figure 1 show the comparison of hysterectomies by route before and after the FDA power morcellation safety communication. Overall, utilization of all minimally invasive routes of hysterectomy decreased from 77.1% before to 75.4% after the safety communication. After adjustment for small differences in case-mix characteristics, laparoscopic hysterectomies decreased by 4.1% ( P = .005), while abdominal and vaginal hysterectomies increased by 1.7% and 2.4% ( P = .119 and P = .012), respectively. Laparoscopic supracervical hysterectomies, a technique that frequently employs power morcellation, decreased by 59% after the safety communication ( P < .001). Figure 2 and Supplemental Figure 2 show the monthly rates of hysterectomy by approach and monthly rates of supracervical hysterectomy by approach during the study period.

Stay updated, free articles. Join our Telegram channel

Join