Surgical Risk after Unilateral Lobectomy Versus Total Thyroidectomy: A Review of 47,434 Patients

Background: We reviewed the 2005-2012 ACS-NSQIP databases to evaluate factors associated with adverse events (AE) after unilateral thyroid lobectomy (UL) and total thyroidectomy (TT). Methods: All unilateral lobectomies and total thyroidectomies performed from 2005 to 2012 were identified for analysis. The cohort was characterized with respect to preoperative and demographic characteristics, complications, reoperation, and mortality. Results: 47,434 patients were identified, of which 17,584 underwent unilateral lobectomy and 29,850 underwent total thyroidectomy. On multivariable regression analysis, UL was associated with a 2.786 greater risk of returning to the OR, and a 1.377 risk of surgical complications. The increased risk of return to the OR was eliminated when controlling for patients returning to the OR for completion thyroidectomy after UL. Conclusion: NSQIP is the only dataset that is able to discern between unilateral lobectomy and total thyroidectomy to make viable comparisons in outcomes. The NSQIP dataset may be imperfect, as pertinent details of chemotherapy and radiation, and procedure-specific complications, including recurrent laryngeal nerve palsy and hypocalcemia, are not tracked. In spite of this, our findings suggest avenues for improvement in the care of thyroidectomy patients, and suggest directions for a thyroidectomy-specific outcomes database.


Introduction
The incidence and severity of thyroid cancer continues to rise. The rate for new diagnoses of thyroid cancer has increased an average of 6.4% per year over the last 10 years and mortality of thyroid cancer has increased 0.9% per year over the same period [1]. With this increasing disease burden, the need for both diagnostic and therapeutic thyroidectomy remains high.
Although thyroid surgery is a relatively safe procedure, there are a number of severe, preventable complications [2,3]. Because medical sustainability proposals link reimbursements with quality control measures, it is imperative to establish normative data by which surgeons and hospitals can be compared to their cohorts with regard to thyroidectomy outcomes. Although more than 20,000 thyroidectomies are performed every year in the United States, only a few papers have attempted to describe high-volume, multi-center outcome data for thyroid surgery [4,5].
Many factors may be weighed when considering a unilateral thyroid lobectomy (UL) versus total thyroidectomy (TT). Even in the setting of known thyroid cancer, there is not always agreement about the requirement for lobectomy or total thyroidectomy [6,7]. Furthermore, there remains discussion about the need for central and lateral neck dissection in the setting of thyroid cancer [8,9]. It

Review Article
Surgical Risk after Unilateral Lobectomy Versus Total Thyroidectomy: A Review of 47,434 Patients

Data acquisition and patient population
Data collection methods for the American College of Surgeons National Surgical Quality Improvement Program (NSQIP) registry have been previously described [11,12]. All study aspects were approved by the respective Institutional Review Boards. The 2005 to 2012 NSQIP registries were queried for all patients who were recorded to have undergone a total thyroidectomy or a unilateral lobectomy. Total thyroidectomy cases were identified by the presence of the Current Procedural Terminology (CPT) codes 60240 and 60225, and unilateral lobectomies were identified by the CPT code 60220. Patients were stratified by surgical modality.

Risk adjustment variables and outcomes
NSQIP-defined preoperative variables were compared between total thyroidectomy and unilateral lobectomy. They included demographic variables (eg, age, BMI class); lifestyle variables (eg, smoking), medical comorbidities (eg, diabetes, dyspnea, hypertension, COPD, congestive heart failure, bleeding disorders, prior angioplasty or cardiac surgery, previous stroke or transient ischemic attack, radiotherapy within 90 days of operation, chemotherapy within 30 days of operation, previous operations within 30 days of operation) and intraoperative characteristics (e.g., total operation time). Tracked 30-day adverse events (AE) were categorized as surgical complications, medical complications, and overall complications. Surgical complications included superficial, deep, organ-space surgical site infection (SSI), or wound disruption. Medical complications included deep venous thrombosis (DVT), pulmonary embolism (PE), unplanned re-intubation, ventilator dependence >48 hours, progressive renal insufficiency, acute renal failure, coma, stroke, cardiac arrest, myocardial infarction (MI), peripheral nerve injury, pneumonia, urinary tract infection (UTI), blood transfusions, and sepsis/septic shock. All AE were used as defined in the NSQIP user guide. Overall complications included all surgical and medical complications. Data for unplanned readmission, which was available in 2011 and 2012, was tracked. Readmission was defined as an unplanned readmission to the same or other hospital within 30 days of the primary or concurrent procedure. Return to the operating room within 30 days identified all major surgical procedures that required the patient to be taken to the surgical operating room for intervention of any kind.

Statistical analysis
Chi-square tests, for categorical variables, and Student t test, for continuous variables, were used to identify differences in perioperative variables between the two groups. Significance was defined as P < 0.05. This method was then used to identify differences in overall, medical, and surgical complications. Perioperative variables with n ≥ 10 and P < 0.2 were identified as possible predictors for AE and included in a binary logistic regression which assessed the independent association of surgical modality with overall, medical, and surgical complications, return to the operating room, and unplanned readmission, with proper risk adjustment for other factors. Again, P < 0.05 was considered significant. Hosmer-Lemeshow (H-L) and c-statistics were calculated to assess model calibration and discriminatory capability, respectively. All analysis was performed using SPSS version 22 (IBM Corp, Armonk, NY).

Cohort characteristics
Between 2005 and 2012, 47,434 patients were identified for our analysis. 17,584 patients (37%) underwent UL and 29,850 patients (63%) underwent TT ( Table 1). The UL group was more likely to be male and have a history of alcohol use. Comorbidities and clinical characteristics including diabetes, smoking, dyspnea, COPD, CHF, hypertension, and steroid use were more prevalent in the TT group. Patients who underwent TT were more likely to be ASA class 3, 4, or 5, older, and have a higher BMI.

Post-operative outcomes
Univariate analysis of 30-day AE for both UL and TT were captured in Table 2. Incidence of overall and medical complications were significantly higher (p<.001) in TT. There was no significant difference between the two surgical modalities in surgical complications. Patients who underwent TT were more likely to experience pneumonia, reintubation, cardiac arrest, sepsis and septic shock. Rates of return to the OR were higher in the UL group. Rates of unplanned readmission were higher in the TT group. No significant difference in mortality rates between the two groups was detected. Table 3 details the multivariate analysis performed to evaluate the independent effect of surgical modality on 30-day AE, adjusting for other risk factors. Neither surgical modality was associated with increased overall or medical complications, or unplanned readmission rates. Of note, unilateral thyroidectomy conferred increased surgical complication risk (OR, 1.377) as well as a 2.786 greater risk of returning to the OR.

Discussion
The NSQIP database is a unique and robust database with large volumes that can support critical analyses of risk factors with small confidence intervals. NSQIP collects data from both major academic tertiary-care medical centers as well as community hospitals, thereby capturing a broad snapshot of procedures and patients with their preoperative risks and postoperative outcomes. Because the data is prospectively collected and validated by a highly-trained and dedicated surgical clinical nurse reviewer (SCNR), the NSQIP has a large advantage over other registry and Medicare-based analyses that are primarily administrative discharge datasets without emphasis on accuracy.
Aside from decreased morbidity and cost savings, the use of TT versus UL is relevant for a number of thyroid cancer subtypes, including papillary thyroid cancer (PTC), Hurthle cell microcarcinomas, and well-differentiated thyroid microcarcinomas (WDTC). Thyroid lobectomy has been shown in previous literature to have less associated vocal and throat function morbidity, and is associated with lower cost and hospital length of stay than total thyroidectomy [17,18]. Current guidelines recommend TT for Submit your Manuscript | www.austinpublishinggroup.com PTC tumors greater and 1.0 cm, in spite of the increased associated morbidity. Adam, et al evaluated 61,775 patients with PTC from the National Cancer Database, and found no difference in survival in TT versus UL, regardless of tumor size [19]. Lee, et al made a similar conclusion in their study of 2,014 patients [20]. However, Ebina, et al found decreased survival after UL in patients > 50 years old, those with massive extrathyroidal extension, or large (>3 cm) lymph node metastasis) in a study of 1,187 patients from a single center [21]. Of note, these studies only evaluated survival and tumor recurrence, as the respective datasets were not capable of analyzing other outcomes. Ogilvie, et al, reviewed 346 patients with WDTC and found that UL was not sufficient for local tumor control in WDTC patients with tumor size between 6-10 mm, in spite of American Thyroid Association guidelines recommending UL alone [22]. Kuo, et al, found poor survival for patients with Hurthle cell microcarcinoma in a study of 22,738 patients, regardless of the extent of resection [23].
Thyroid tumors can recur as a result of positive margins, microsatellite disease, or unrecognized nerve invasion. A statistical evaluation of the presence of histologic tumor subtypes, positive surgical margins, extent of locoregional disease, need for postoperative radiation, and tumor recurrence is beyond the level of detail available in ACS-NSQIP. The dataset does have a variable for "disseminated cancer" (referring to stage IV disease), which was significantly elevated in the TT group (p < 0.001; Table 1).
In addition, preoperative diagnosis codes identified thyroid malignancy as the reason for surgery in 88% of cases, and goiter in 8% of cases, with non-specified reasons making up the remaining 4% of cases. Data analysis when stratifying versus preoperative diagnosis only demonstrated continued significant increased in return to the OR for UL in the setting of thyroid malignancy, but not goiter (data not shown). This finding likely indicates that unilateral goiter can be successfully managed with UL with proper selection, while there continues to be a significant rate of conversion from UL to TT for patients with thyroid carcinoma (in spite of patient selection) [24][25][26]. Further development of the NSQIP dataset may facilitate more granular detail with regards to indications for surgery, use of completion thyroidectomy after lobectomy, and use of adjuvant therapies (e.g., radioactive iodine, levothyroxine).
One analysis evaluated the NSQIP database for trends in thyroidectomy and parathyroid outcomes between different specialties [27]. They concluded that there was no difference   Table 3: Risk-adjusted analysis. Odds ratios reflect risk of total thyroidectomy (with respect to reference group of unilateral thyroidectomy) for the given adverse event.
Submit your Manuscript | www.austinpublishinggroup.com in outcomes or case complexity between general surgery and otolaryngology, that operative time was less for general surgery (115 versus 123 minutes), and that duration of hospital stay beyond 1 postoperative day was more frequent for otolaryngology (8% vs 5%). One weakness of their study was that their data set included 51,470 cases from general surgery and only 3,932 cases from otolaryngology, a reflection of the fact that NSQIP captures far more general surgery cases than surgical subspecialty cases. Furthermore, the percentage of thyroidectomy cases vs parathyroidectomy cases was higher for otolaryngology (86%) than general surgery (68%), which may have influenced overall average operative times and hospital stays. Our analysis did not detect any differences in total, surgical, or medical complications, reoperation, readmission, or mortality between surgical specialties performing thyroidectomy (i.e., general surgery, otolaryngology, thoracic surgery, and plastic surgery). A more valuable distinction may be surgical oncology specialization with respect to the performing surgeon; however, NSQIP does not identify this variable.
A 2012 NSQIP report showed that patients undergoing either thyroidectomy or parathyroidectomy, or both had a 30-day mortality of 0.08%, 0.16%, and 0.2%, respectively. They identified mean length of stay (LOS) values at 1.1 days, 1.1 days, and 1.4 days, respectively. Congestive heart failure (CHF), dependent functional status, dialysis dependence, and chronic corticosteroid use were significantly associated with increased LOS and postoperative adverse events [28]. Preoperative pneumonia and new exacerbation of CHF are well documented comorbidities that affect postoperative pulmonary complications [29,30]. Incentive spirometry, head of bed elevation, ambulation, and oral hygiene have been shown to decrease these complications [31,32].
A 2014 NSQIP study assessed how various preoperative comorbidities and intraoperative variables were tied to thyroidectomy outcomes [33]. In an analysis of 38,577 patients they found that risk factors independently associated with morbidity after thyroidectomy included hypertension, diabetes, advanced age greater than 70 years, COPD, and dialysis. Interestingly they found that the surgical approach was related to the rate of return to the operating room. As compared with a partial thyroidectomy, patients undergoing a total thyroidectomy were 73% less likely to return to the operating room and those undergoing substernal thyroidectomy were 37% less likely to return to the operating room. We similarly found an increased utilization of TL in patients older than 70 years of age (10.6% versus 10%, p = 0.021).
In our study the 30-day mortality was 0.03% for UL and 0.06% for TT. This mortality rate compares well with mortality rates of other procedures such as pancreatectomy (8.3%), coronary artery bypass graft (3.5%), craniotomy (10.7%), and repair of aortic aneurysm (3.9%) [34]. The morbidity rates found in our study were also low: 1.14% for UL and 1.64% for TT. This underscores the relative safety profile of thyroidectomy procedures. Thyroidectomy surgeons and hospitals should use these data as a benchmark to compare their own rate of complications and for quality improvement initiatives.
Interestingly, UL was significantly more likely than TT to be associated with surgical complications and return to OR (Table 3). With regard to surgical complications, there was a greater incidence of wound dehiscence and deep SSI in the UL group, indicating that some unidentified factor contributes to increased surgical  Table 4: Diagnosis codes associated with return to operating room and readmission.
Unilateral refers to unilateral lobectomy; total refers to total thyroidectomy. complications in this cohort. It is possible that the challenge of partial thyroidectomy through a limited skin incision results in greater soft tissue trauma and consequent AE, although the techniques for UL versus TT are nearly identical. While the surgical extent of TT is greater, it may be that the technical sophistication of UL is more, in that it entails organ-sparing of half of the thyroid gland. Of note, when we separated surgical complications by type (i.e., superficial/ deep/ organ space wound infection, wound dehiscence, etc) and performed additional multivariable analysis for TT versus UL, no one specific complication type was significantly elevated in the UL group. NSQIP is limited in that it does not track surgeon experience (which has been shown to decrease the risk of AE after UL or TT). However, it is possible to track the reason for reoperation and/or readmission. After analyzing causes for reoperation and readmission, we found that over 70% of reoperations and readmissions were for cancer, neoplasm, and/or goiter (Table 4). These cases are thus attributable to completion thyroidectomy, and not reoperation or readmission for postoperative complications. After excluding these cases from statistical analysis, there were no longer any significant differences in UL versus TT groups for reoperation and/or readmission (1.9% vs 1.9%, p=.91 for reoperation; 2.5% vs 3.1%, p=.52 for readmission). Based on this preliminary analysis, it appears as though the vast majority of reoperations and readmissions are thus attributable to the need for completion thyroidectomy after UL, and not any specific AE requiring surgical intervention. This finding would suggest that UL does not confer any specific risk for reoperation greater than TT, outside of the possible need for completion thyroidectomy. Thus, future research may be useful identifying which patients will go on to ultimately require completion thyroidectomy, versus any specific modification of the techniques for UL. Similarly, we found a higher rate of readmission on univariate analysis for TT, as opposed to UL. However, regression analysis demonstrated that modality choice does not confer increased risk of readmission, which would suggest that there is some other reason that readmission rates are higher for TT. Given that this outcome is partially, but not totally ameliorated when accounting for completion thyroidectomies, it suggests that there are other, unidentified factors contributing to these results.
A major weakness of the NSQIP database is that it was designed to capture data for all surgeries and therefore some procedure-specific information is not collected. NSQIP does not specifically track the most critical complications for which monitoring are recommended following thyroidectomy. For instance, it may not be representative to use "need for transfusion" as a metric to assess postoperative hemorrhage, or hematoma requiring evacuation. Some authors have used the information on reoperation as a surrogate for the development of neck hematoma [35]. While the reported morbidity rate (0.1%) is accurate based on the data recorded, this number could be an underrepresentation of the true morbidity rate associated with postoperative bleeding. For example, while 0.4% of patients were listed as having a 30-day surgical complication, 2.2% of patients had a reoperation within 30 days of the index procedure, thus suggesting a discrepancy (as most reoperations are due to surgical complications).
Neither the incidence of nerve injury specific to thyroid surgery (recurrent laryngeal nerve and superior laryngeal nerve), nor the use of intraoperative neuromonitoring is specifically tracked in the NSQIP. While a systematic review with meta-analysis showed no statistically significant difference in the incidence of recurrent laryngeal nerve palsy when using intraoperative nerve monitoring versus visualization alone during thyroidectomy, the technique of intraoperative neuromonitoring in thyroid surgery has high accuracy, specificity, sensitivity and negative predictive value in excluding postoperative recurrent laryngeal nerve palsy, and is very commonly used in thyroid surgery [36,37].
Hypocalcemia is another feared complication of thyroid surgery. The incidence of postoperative hypocalcemia (transient or permanent) following thyroidectomy ranges from 0% to 83%, with the highest incidence in patients with total thyroidectomy for cancer (28%), and those with subtotal thyroidectomy for thyrotoxicosis (23%). Conversely, its incidence is lowest in patients having subtotal thyroidectomy for other diseases (1.5%) and lobectomy (0%) [38]. Baldassarre, et al studied the incidence of hypocalcemia after thyroidectomy in 119,567 patients [39]. The overall incidence of hypocalcemia was 5.5%, with significant increases in patients undergoing total thyroidectomy (9.0%) and total thyroidectomy with bilateral neck dissection (23.4%), as opposed to thyroid lobectomy alone (1.9%). In their study, hypocalcemia resulted in an additional 1.47 inpatient hospital days, on average. There is additional debate over which postoperative labs to obtain (PTH versus ionized calcium levels), the timing of postoperative laboratory testing, and doses of supplemental calcium to administer in the setting of postoperative hypocalcemia. NSQIP dataset does not directly record this data, which is relevant to thyroid surgery. For patients readmitted within 30 days, the reason for readmission was hypocalcemia in 33%. Finally, NSQIP does not allow comparisons between high volume and low volume surgeons as well as high volume and low volume institutions; therefore conclusions cannot be made regarding surgical outcomes among these groups. Each of these factors has been found to be associated with mortality, complication rates and length of stay for different surgical procedures [40].
There are practical implications from the current study with regards to choice of operative techniques. Specifically, patients undergoing thyroid lobectomy should be counseled that they have a small but significantly increased risk of generic surgical complications (specifically, wound dehiscence), and return to the operating room (presumably for completion thyroidectomy). With regard to the latter item, this appears to be primarily relegated to patients with malignant thyroid disease, as opposed to unilateral benign disease (e.g., thyroid goiter, toxic nodule). Notably, these risks are only applicable to the generic complications outlined above, and not recurrent laryngeal nerve injury or hypocalcemia. Both of these latter adverse events appear to be significantly increased in patients undergoing total thyroidectomy, based on previously-published literature on this subject. Regardless of these limitations, the NSQIP data are important for patient informed consent, and quality improvement processes.

Conclusion
The morbidity and mortality of thyroid surgery is relatively low, and compares favorably with other surgical procedures. Generic surgical complications (specifically wound dehiscence) are increased with respect to UL as opposed to TT. The increased risk of reoperation seen in UL as compared to TT is likely attributable to the need for completion thyroidectomy in these patients with malignant