Abstract
Background
Hemorrhage after pancreaticoduodenectomy is a potentially fatal complication. We retrospectively reviewed state-wide data to evaluate incidence, type of hemorrhage, treatment modalities, and outcomes.
Methods
Healthcare Cost and Utilization Project's Florida State Inpatient Database was queried 2007–2011 for patients undergoing pancreaticoduodenectomy. Characteristics and outcomes were compared by χ2. Multivariate logistic regression model was generated for risk of hemorrhage during index visit.
Results
Of 2548 patients, 217 (8.5%) developed post-operative hemorrhage during their index visit with 139 (64.0%) requiring angiographic, endoscopic, or operative intervention. Overall mortality during index visit was 5.7% (146) – significantly higher in those patients who had post-operative hemorrhage (24.9%) vs not (4.0%) (p < 0.0001). Mortality was significantly higher when post-operative hemorrhage occurred during the second (POD 8–14) vs first (POD 0–7) week at 15/28 vs 16/74, respectively (p = 0.007). On multivariate analysis, male sex (OR 1.56, p = 0.003), vascular resection (OR 1.88, p = 0.017), very low hospital volume (≤7 PD/year; OR 1.62, p = 0.016), and post-operative intra-abdominal/wound infection (OR 2.31, p < 0.0001) were independent predictors for risk of hemorrhage during index visit.
Conclusions
Hemorrhage following pancreaticoduodenectomy remains common, resulting in significantly increased mortality. Hemorrhage during the second post-operative week carries approximately double the mortality of early bleeding, suggesting different etiologies requiring differing treatment approaches.
Background
The mortality associated with pancreaticoduodenectomy has significantly decreased since the introduction of the Whipple procedure; however, associated morbidity remains high.1, 2, 3, 4, 5 Amongst the most dreaded post-operative complications is hemorrhage, with associated mortality rates as high as 30%.6 A consensus definition of post-pancreatectomy hemorrhage was proposed by the International Study Group of Pancreatic Surgery (ISGPS).7 With ongoing advances in less invasive techniques, including endoscopy and angiography, several studies have proposed treatment algorithms in the setting of post-operative hemorrhage.6, 8 Most of these have been limited by small sample sizes with low event rates; large studies conversely have prolonged study eras.
We retrospectively analyzed a large modern cohort of patients undergoing pancreaticoduodenectomy for the development and timing of post-operative hemorrhage, intervention type, and outcomes. The primary outcome of interest was overall mortality. Secondary outcomes of interest included length of stay (LOS) and cost. We hypothesized that patients with severe hemorrhage would have inferior outcomes, including longer LOS, higher hospital cost, and increased mortality. We further wished to determine if the timing of hemorrhage impacted outcomes.
Methods
Aim
To examine the rates and types of hemorrhage, as well as hemorrhage-associated outcomes, including overall mortality, length of stay (LOS), and cost, occurring after pancreaticoduodenectomy (Whipple procedure) during the index hospital visit and at revisits in the state of Florida.
Population
Discharge data was retrospectively reviewed for patients 18–95 years of age identified as having a pancreaticoduodenectomy by the International Classification of Disease (ICD-9) code (52.7) in the Florida State Inpatient Database (SID) or by Current Procedural Terminology (CPT) code (48150, 48152, 48153, 48154) in the Florida State Emergency Department Database (SEDD), Healthcare Cost and Utilization Project (HCUP), Agency for Healthcare Research and Quality, from 2007 to 2011.9, 10 The CPT coding system is maintained by the American Medical Association and describes medical, surgical, and diagnostic services. Patients coded as having undergone more than one Whipple procedure were excluded (N < 11). The Florida State Emergency Department Database (SEDD) was also utilized to identify patients having emergency room visits after index procedure.
Procedures and indications
All index pancreaticoduodenectomies identified were from inpatient service type with ICD-9 code 52.7. Vascular resections occurring at the time of Whipple procedure were also queried via ICD-9 procedure codes and classified as either venous, arterial, or other (see Appendix A1). Indications were queried via ICD-9 diagnosis codes and classified as malignant primary, malignant secondary, carcinoma in situ, benign disease, cystic disease, acute and chronic pancreatitis, and uncertain or unspecified behavior (see Appendix A2). For analyses, malignant disease is defined as primary and secondary malignancy and carcinoma in situ, benign disease is defined as benign and cystic disease, and pancreatitis is defined as acute and chronic pancreatitis (see Appendix A2). Of note, all diagnosis codes were evaluated for targeted procedure.
Patient and hospital characteristics
Patient characteristics of interest were sex, age, race, insurance status, Elixhauser comorbidity score (calculated by HCUP Comorbidity Software Version 3.7), and the quartile for median household income based on patient's ZIP code.11 Type of index admission included elective, urgent, emergent, and trauma, which were grouped into elective and non-elective status. Hospital characteristics included annual hospital pancreaticoduodenectomy volume, divided into quartiles comprised of very low, low, medium, and high volume. Very low volume hospitals were those in the lowest quartile performing ≤7 pancreaticoduodenectomies per year, whereas high volume hospitals were in the highest quartile performing greater than or equal to 80 pancreaticoduodenectomies per year. Low and medium volume hospitals performed greater than 7 and less than or equal to 49 and >49–80 pancreaticoduodenectomies per year, respectively.
Outcomes
Main outcomes of interest were post-operative hemorrhage, additional post-operative interventions in the setting of hemorrhage, and mortality (during index inpatient visit). Secondary outcomes of interest were length of stay and cost. Post-operative hemorrhage was defined using ICD-9 diagnosis codes for visceral artery aneurysm (442.84), gastrointestinal hemorrhage (578), hematemesis (578.0), melena (578.1), unspecified hemorrhage (578.9, 459.0), and hemorrhage/hematoma post-procedure (998.1, 998.11, 998.12).
Additional post-operative interventions were queried via ICD-9 procedure codes and classified as either endoscopic, endovascular, or operative (see Appendix A3). If the diagnosis code for hemorrhage type was not defined but the patient underwent a specific procedure, such as endovascular repair of aneurysm (39.79), that individual was classified as having a post-operative aneurysm/pseudoaneurysm. The day of post-operative hemorrhage during the index visit was defined by the procedure day (N = 139). Patients who did not undergo a procedure for control of bleeding were not assigned a day of hemorrhage and were excluded from further timing analyses (N = 78). In this manuscript, severe hemorrhage refers to any hemorrhage that required additional intervention, while moderate hemorrhage refers to hemorrhage with a defined diagnosis, but no further intervention. Additional outcomes of interest queried via ICD-9 diagnosis code were post-operative wound infections and disruptions (998.59, 998.30, 998.31, 998.32).
Revisit data
Revisit data were reviewed from the Florida State Inpatient Database (SID) and the State Emergency Department Database (SEDD). Outcomes of interest included post-operative hemorrhage, additional interventions, mortality, and post-operative infections, using the same ICD-9 and CPT codes as defined above for index visits. Post-operative day of hemorrhage was defined as the difference in days between the index visit and revisit(s).
Statistical analysis
All statistical analyses were performed using SAS (version 9.4; SAS Institute, Cary, NC, USA). Univariate analyses of categorical variables were performed using chi-square tests. Certain continuous variables were divided into categories based on clinical significance, including age, index length of stay, and index median hospital cost. Hospital volumes of procedures were divided into approximate quartiles. Continuous variables were compared via Wilcoxon rank-sum tests. A multivariate logistic regression model was created for hemorrhage during the index visit following pancreaticoduodenectomy. All patient-specific and pre-operative predictors that were deemed clinically important and considered potential confounders were included in the multivariate model and include: sex, age, race, insurance status, Elixhauser score, vascular resection status, hospital volume, procedure indication, elective status, and post-operative intra-abdominal/wound infection. Of note, since there is no ICD-9 code for pancreatic fistula, intra-abdominal/wound infections was utilized as a potential surrogate for a pancreatic leak. P-values of less than 0.05 were considered significant.
Results
Patient and hospital characteristics for index visit
Two thousand five hundred forty eight admissions for radical pancreaticoduodenectomy were identified over a 5-year period. The majority of patients undergoing Whipple procedure were male (1336, 52.4%), 65 years of age or older (1480, 58.1%), white (1930, 75.8%), had an Elixhauser score of three or greater (146, 56.8%), were insured under Medicare (1430, 56.1%), and had a pre-operative diagnosis of malignancy (2014, 79.0%). Most procedures were performed electively (1999, 78.5%). The median yearly hospital volume for pancreaticoduodenectomy was 49.6 with an interquartile range of 7.2–80.2, with the highest proportion of procedures performed at high volume centers (826, 32.4%). Two hundred seventeen (8.5%) patients had hemorrhage during their index hospital visit.
Males were more likely to have hemorrhage than females (62.7% vs 37.3%, p = 0.002). A higher proportion of patients with hemorrhage had an Elixhauser score greater than or equal to 3 versus those with no hemorrhage (74.2% vs 55.1%, p < 0.0001). Procedures performed non-electively were more likely to experience hemorrhage during the index visit (29.5% vs 20.8%, p = 0.003). Very low volume centers had higher rates of hemorrhage vs not (33.2% vs 23.9%, p = 0.006) and the highest rate of hemorrhage of all other centers (see Table 1).
Table 1.
Patient and hospital characteristics by occurrence of post-operative hemorrhage during index admission for Whipple procedure
| No hemorrhage |
Hemorrhage |
Total |
p-value | |
|---|---|---|---|---|
| n (%) | n (%) | % | ||
| 2331 (91.5) | 217 (8.5) | 2548 | ||
| Sex | ||||
| Male | 1200 (51.5) | 136 (62.7) | 52.4 | 0.002 |
| Female | 1131 (48.5) | 81 (37.3) | 47.6 | |
| Age | ||||
| <65 | 970 (41.6) | 98 (45.2) | 41.9 | 0.311 |
| ≥65 | 1361 (58.4) | 119 (54.8) | 58.1 | |
| Race | ||||
| White | 1773 (76.1) | 157 (72.3) | 75.7 | 0.286 |
| Black | 189 (8.1) | 24 (11.1) | 8.4 | |
| Other/Unknown | 369 (15.8) | 36 (16.6) | 15.9 | |
| Elixhauser score | ||||
| 0 | 157 (6.7) | 11 (5.1) | 6.6 | <0.0001 |
| 1 | 389 (16.7) | 19 (8.8) | 16.0 | |
| 2 | 500 (21.5) | 26 (12.0) | 20.6 | |
| ≥3 | 1285 (55.1) | 161 (74.2) | 56.8 | |
| Insurance | ||||
| Private | 736 (31.6) | 71 (32.7) | 31.7 | 0.165 |
| Medicare | 1318 (56.5) | 112 (51.6) | 56.1 | |
| Medicaid | 149 (6.4) | 22 (10.1) | 6.7 | |
| Other/Uninsured | 128 (5.5) | 12 (5.5) | 5.5 | |
| Indication | ||||
| Benign/Cystic disease | 249 (10.7) | 22 (10.1) | 10.6 | 0.246 |
| Malignancy | 1846 (79.2) | 168 (77.4) | 79.0 | |
| Pancreatitis | 102 (4.4) | 16 (7.4) | 4.6 | |
| Unspecified/Unknown | 134 (5.7) | 11 (5.1) | 5.7 | |
| Income quartile by Zip codea | ||||
| Less than 25th percentile | 600 (26.3) | 62 (29.2) | 26.6 | 0.459 |
| 26–50th percentile | 713 (31.3) | 72 (34.0) | 31.5 | |
| 51–75th percentile | 634 (27.8) | 50 (23.6) | 27.4 | |
| 76th–100th percentile | 333 (14.6) | 28 (13.2) | 14.5 | |
| Vascular resection | ||||
| Yes | 134 (5.8) | 20 (9.2) | 6.0 | 0.040 |
| No | 2197 (94.2) | 197 (90.8) | 94.0 | |
| Admission type | ||||
| Elective | 1846 (79.2) | 153 (70.5) | 78.4 | 0.003 |
| Non-elective | 485 (20.8) | 64 (29.5) | 21.6 | |
| Hospital PD volume | ||||
| Very low volume | 557 (23.9) | 72 (33.2) | 24.7 | 0.006 |
| Low volume | 520 (22.3) | 52 (24.0) | 22.5 | |
| Medium volume | 481 (20.6) | 40 (18.4) | 20.4 | |
| High volume | 773 (33.2) | 53 (24.4) | 32.4 | |
56 missing.
Index admission outcomes
The median index LOS for all patients undergoing Whipple procedure was 13 days (IQR: 9–20). Those who developed post-operative hemorrhage had a significantly prolonged index LOS of 22 days (IQR: 14–40) vs 13 days (IRQ: 9–19) (p < 0.0001). Most patients (1089, 42.7%) remained in the hospital for more than two weeks following Whipple procedure, including 71.0% (154) of those who had post-operative hemorrhage. Median cost per index admission was significantly higher in patients with hemorrhage at $63 296 (IQR: $40 888–$115 733) vs $32 300 (IQR: $24 712–$45 708) (p < 0.0001). While only 29.4% (748) of visits had a cost of more than $45 000, this represented 69.0% (149) of patients with post-operative hemorrhage. 13.8% (352) developed post-operative wound/intra-abdominal infections, with 26.3% (57) identified in the hemorrhage cohort (p < 0.0001). Overall mortality during index admission was 5.7% (146). Inpatient mortality was significantly higher in the hemorrhage group at 24.9% (54) vs 4.0% (92) in the non-hemorrhage group (p < 0.0001). Morality was highest for those patients who had severe hemorrhage requiring an intervention at 30.9% (43). Please see Table 2 for more detailed index admission outcomes.
Table 2.
Outcomes by occurrence of post-operative hemorrhage during index admission for Whipple procedure
| No hemorrhage |
Hemorrhage |
Total |
p-value | |
|---|---|---|---|---|
| n (%) | n (%) | % | ||
| 2331 (91.5) | 217 (8.5) | 2548 | ||
| LOS | ||||
| <10 days | 733 (31.5) | 23 (10.6) | 29.7 | <0.0001 |
| 10–14 days | 663 (28.4) | 40 (18.4) | 27.6 | |
| >14 days | 935 (40.1) | 154 (71.0) | 42.7 | |
| Median LOS (IQR) | 13 (9–19) | 22 (14–40) | <0.0001 | |
| Total cost per admission | ||||
| <$30 000 | 1006 (43.2) | 18 (8.3) | 40.3 | <0.0001 |
| $30 000–45 000 | 723 (31.1) | 49 (22.7) | 30.4 | |
| >$45 000 | 599 (25.7) | 149 (69.0) | 29.4 | |
| Median Index Cost (IQR) | 32 300 (24 712–45 708) | 63 296 (40 888–115 733) | <0.0001 | |
| Post-op wound/abd infection | ||||
| Yes | 295 (12.7) | 57 (26.3) | 13.8 | <0.0001 |
| No | 2036 (87.3) | 160 (73.7) | 86.2 | |
| Mortality | 92 (4.0) | 54 (24.9) | 5.7 | <0.0001 |
Index hemorrhage cohort
Two hundred seventeen (8.5%) patients developed post-operative hemorrhage during their index visit. The most common type was coded as “post-operative hemorrhage” at 49.3% (107), followed by “unspecified hemorrhage” type at 16.1% (35), and “post-operative hematoma” at 12.9% (28). Hematemesis (14, 6.5%), melena (19, 8.8%), and pseudoaneurysm (14, 6.5%) were rare descriptors of hemorrhage. As shown in Fig. 1, 64.1% (139) underwent an intervention. Of those, 64.7% (90) required only a single intervention, with the majority performed via endoscopic approach (43, 47.8%). The rates of undergoing multiple procedures were the same regardless of whether the first intervention was endoscopic or angiographic, with 27/70 vs 11/28, respectively, in both groups requiring at least one additional procedure. For those patients undergoing a procedure, post-operative hemorrhage was most likely to occur during the first post-operative week (74, 53.2%), with hemorrhage occurring on post-operative day zero in 19 patients. However, mortality was highest when hemorrhage occurred during the second post-operative week (post-operative days 8–14) with a rate of 15/28 (p = 0.008). The predominant type of intervention performed did not differ based on the timing of hemorrhage. Endoscopic interventions were more commonly used as first line intervention during the first (33/74), second (15/28), and after the second (23/37) post-operative weeks.
Figure 1.
a) Post-operative hemorrhage and intervention type during index visit. b) Timing of post-operative hemorrhage and mortality during index visit
Univariate and multivariate analysis for post-operative hemorrhage
On univariate logistic regression, significant unadjusted predictors for post-operative hemorrhage during the index admission included male gender, vascular resection, very low hospital volume (versus high volume), post-operative wound/intra-abdominal infection, and non-elective procedure status (versus elective). Unadjusted odds ratios (OR) are presented in Table 3. On multivariate analysis, depicted in Fig. 2, significant predictors for post-operative hemorrhage were male gender, vascular resection, very low hospital volume, and post-operative wound/intra-abdominal infection. The C statistic for the multivariate model was 0.68.
Table 3.
Unadjusted model of post-operative hemorrhage during index Whipple hospitalization
| Unadjusted OR | 95% CI | p-value | |
|---|---|---|---|
| Sex | |||
| Male | 1.582 | 1.188–2.108 | 0.002 |
| Female | Ref | ||
| Age | |||
| <65 | Ref | ||
| ≥65 | 0.865 | 0.654–1.145 | 0.311 |
| Race | |||
| White | Ref | ||
| Black | 1.434 | 0.910–2.260 | 0.121 |
| Other | 1.102 | 0.754–1.610 | 0.616 |
| Insurance | |||
| Private | Ref | ||
| Medicaid | 1.531 | 0.919–2.548 | 0.102 |
| Medicare | 0.881 | 0.646–1.202 | 0.424 |
| Other/Uninsured | 0.972 | 0.512–1.843 | 0.930 |
| Elixhauser | |||
| 0 | Ref | ||
| 1 | 0.697 | 0.324–1.499 | 0.356 |
| 2 | 0.742 | 0.359–1.536 | 0.422 |
| ≥3 | 1.788 | 0.950–3.367 | 0.072 |
| Vascular resection | |||
| No | Ref | ||
| Yes | 1.665 | 1.018–2.722 | 0.042 |
| PD hospital volume | |||
| Very low | 1.885 | 1.301–2.732 | 0.001 |
| Low | 1.458 | 0.979–2.172 | 0.063 |
| Medium | 1.213 | 0.792–1.857 | 0.375 |
| High | Ref | ||
| Indication | |||
| Benign | Ref | ||
| Malignant | 1.030 | 0.648–1.638 | 0.900 |
| Pancreatitis | 1.775 | 0.896–3.518 | 0.100 |
| Unspecified | 0.929 | 0.437–1.974 | 0.848 |
| Admission type | |||
| Elective | Ref | ||
| Non-elective | 1.592 | 1.170–2.167 | 0.003 |
| Post-op infection | |||
| No | Ref | ||
| Yes | 2.459 | 1.775–3.407 | <0.0001 |
Bold text represents significant values (odds ratio, interval and p-value).
Figure 2.
Adjusted predictors of post-operative hemorrhage during index visit for Whipple procedure. Adjusted odds ratios with 95% confidence intervals. LCL: lower confidence limit; OR: odds ratio; UCL: upper confidence limit
Revisits
One thousand seventy two (42.1%) patients had at least one hospital or emergency department revisit for any reason following pancreaticoduodenectomy within 90 days. Of those, 62 (5.8%) were for a diagnosis of hemorrhage. Of the two hundred seventeen patients who developed post-operative hemorrhage during index Whipple procedure, 76 (35.0%) had at least one revisit for any reason within 90 days, with less than eleven (<14.5%) patients having a revisit for hemorrhage.
Discussion
Using a large state database with information on revisits we demonstrated that hemorrhage following pancreaticoduodenectomy remains a serious and common complication, occurring at a rate of 8.5% during the index admission with high associated mortality of over 6 times that of patients without hemorrhage. Hemorrhage occurring post-operatively is often severe, with 64% of patients requiring intervention and over 35% requiring more than one procedure. We found that regardless of the timing of hemorrhage, the most common intervention type utilized first was endoscopic. Hemorrhage post-operatively also resulted in significantly increased lengths of stay and hospital costs. Furthermore, patients with severe bleeding during the second post-operative week had twice the mortality of those who developed hemorrhage during the first post-operative week. Significant predictors for development of hemorrhage during the index operative visit included male sex, vascular resection, very low hospital volume, and post-operative intra-abdominal/wound infection.
Complications following pancreatectomy have remained relatively stable in recent years5; however, without a consensus definition of hemorrhage, the previously reported incidence has varied broadly based on inclusion and severity criteria. It was not until 2007 that the International Study Group of Pancreatic Surgery (ISGPS) published a set of criteria to define postpancreatectomy hemorrhage (PPH).7 Previously published rates of moderate and severe hemorrhage following pancreaticoduodenectomy and pancreatectomy are 3.0–7.5% and 5.7%, respectively, with associated mortality ranging from 11 to 30.7%.6, 8, 12, 13, 14, 15, 16 Rates of hemorrhage were significantly higher (27.4–29.1%) with more false positives when ISGPS mild, grade A, hemorrhage was included.17, 18, 19 However, one study from Singapore did not detect a change in the incidence of PPH when the new ISGPS criteria were applied, which may be a reflection of previous inclusion requirements.20
Previous work has used varying cutoffs to discriminate between early and late hemorrhage, attempting to distinguish the potential etiologies and propose treatment algorithms. Two large studies of 559 and 1669 patients undergoing pancreatectomy defined early bleeding as occurring within the first week and first five days post-operatively, respectively.8, 12 However, with the new ISGPS guidelines, timing was divided into early versus late, with early occurring less than 24 h post-operatively and late occurring more than 24 h post-operatively.7
Endoscopic, angiographic, and operative interventions have previously been described for control of post-operative hemorrhage, but unclarity distinguishing between the underlying etiology and the most appropriate intervention remains. A study validating the new ISGPS grading criteria noted that the majority of patients presented with late bleeding (>24 h post-operatively) with most occurring between post-operative days 10–20 secondary to pseudoaneurysm development requiring operative control.14 A meta-analysis evaluating delayed hemorrhage after pancreaticoduodenectomy found no difference between angiography and operative interventions for achieving complete hemostasis and mortality.6 Multiple studies have evaluated the role of angiographic intervention for control of delayed post-operative hemorrhage, particularly in the setting of a pseudoaneurysm/major arterial bleeding, and found high success rates.13, 15, 19, 21, 22, 23, 24, 25, 26 In visceral artery pseudoaneurysms, angiography was associated with a fourth of the mortality (10% vs 43%) of operative intervention, shorter operating time, lower blood loss, and shorter intensive care unit stay, with an overall 87% success rate at achieving hemostasis.25 However, studies assessing first line endoscopic intervention for control of moderate and severe hemorrhage found lower success rates.16
We performed the largest contemporary review of hemorrhage after pancreaticoduodenectomy. This study is the first, to our knowledge, to evaluate the association between timing of hemorrhage, intervention type, and mortality. Our data demonstrated nearly a two-fold increase in mortality resulting from hemorrhage during the second post-operative week compared to the first. This suggests that not all late hemorrhages, defined by the ISGPS to occur after 24 h post-operatively, are equivalent and perhaps consideration should be given to further subclassifications. The greatest utility of the current ISGPS early vs late cutoff is to distinguish between post-operative bleeding secondary to surgical technique/failure to achieve hemostasis versus delayed post-operative complications such as pancreatic fistula, anastomotic leak, and pseudoaneurysm development.
The largest review available prior to our study noted a similar mortality rate of 47% when hemorrhage occurred on or after the sixth post-operative day for patients undergoing any type of pancreatectomy; however, the onset of bleeding in this cohort ranged from post-operative day six to thirty-two with mortality strongly associated with bleeding related to vascular pathology.8 Additional studies have also cited similarly high mortality rates in the setting of pseudoaneurysms.27, 28, 29 The comparable mortality rates between these studies and ours suggest a similar underlying etiology. We found that endoscopy was most commonly utilized for control of any bleeding, including during the second post-operative week. Prior research has noted low hemostasis success rates of 25% for those undergoing endoscopic intervention first in the setting of severe (ISPGS grade C) hemorrhage.16 In addition, our data demonstrate that the majority of patients who develop moderate/severe hemorrhage will do so during their index visit at a rate of 8.5%, while only 2% will develop initial hemorrhage within 90 days of discharge. In other words, only 19% of patients who develop postpancreatectomy hemorrhage will do so after discharge, which is lower than the rate of 39% found in a smaller study.30 These results, in combination with prior data, suggest that there may be under-detection of pseudoaneurysms, particularly during the second post-operative week, resulting in moderate/severe hemorrhage that may not respond well to endoscopic intervention. Consideration should be given to angiographic or operative intervention as first line therapy in patients who develop post-operative hemorrhage during the second week.
There are several limitations to this study. As in all large database work, the availability of ICD-9 (and CPT) codes is limited. We could not investigate certain clinically relevant complications, such as post-operative pancreatic fistula, or distinguish deep intra-abdominal collections versus superficial wound infections; this could result in potential under-estimation of pancreatic fistula rates when using post-operative infections as a surrogate.31 There is also the possibility of miscoding or undercoding of information; randomized controlled trials at times demonstrate higher rates of complications following pancreatic surgery compared to retrospective studies.32, 33 Lack of patients' clinical picture with retrospective database work can result in an inability to determine the appropriateness of procedures. In addition, we may have under-reported the number of mild post-operative bleeds (ISPGS grade A), since we cannot quantify the number of blood units transfused. This also limited our ability to detect the presence of sentinel bleeds, which may occur in as many as 46% of patients in the setting of a pseudoaneurysm.25 In addition, patients who did not receive an intervention were not able to be included in the analysis of timing of post-operative hemorrhage. However, the power of the large database allowed us to review recent treatment trends occurring over a span of just five-years. Furthermore, while retrospective studies cannot establish causal relationships, but only associations, they provide an invaluable resource in reviewing current practices and outcomes.
This is the largest review to date and the first to evaluate the timing of hemorrhage and its association with mortality and intervention type. Our data is adequately powered to detect a number of differences between the groups, allowing identification of significant predictors for the development of hemorrhage. Independent predictors included male sex, vascular resection, very low hospital volume (≤7 pancreaticoduodenectomies per year), and post-operative intra-abdominal/wound infection. Increased odds of hemorrhage were previously reported for patients with pancreatic fistula/leak, infectious clinical signs, and bile drainage.8, 18, 34 Therefore, patients with multiple risk factors should be monitored more closely in the post-operative period and a sentinel bleed should be taken seriously with consideration of further diagnostics.
The rate of postpancreatectomy hemorrhage has remained relatively common and stable over the past two decades with the associated mortality remaining staggeringly high based on available work. We found a particularly high mortality of over half (15/28) during the second post-operative week, a rate similar to that seen in patients with pseudoaneurysm.27, 28 It is possible that bleeding secondary to pseudoaneurysm development is under recognized and may be a more common occurrence than previously described during the second post-operative week, requiring more angiographic or operative intervention.
All late hemorrhage occurring after the first 24 h post-operatively does not appear to be equal, suggesting differing underlying etiologies. Future studies should further evaluate the timing of hemorrhage and intervention type. Patients at increased odds of hemorrhage should be monitored closely and early diagnostic evaluation should be considered at the first sign of bleeding.
Previous communication
Presented at 57th Annual Meeting of the Society for Surgery of the Alimentary Tract during Digestive Diseases Week in May, 2016, San Diego, CA.
Conflict of interest
The authors have no relevant conflicts of interest.
Appendix A1.
| Vascular resection | ICD-9-CM code |
|---|---|
| Venous | 38.07, 38.37, 38.47, 38.67, 38.7, 39.32 |
| Arterial | 38.06, 38.34, 38.36, 38.46, 38.66, 39.31 |
| Other | 00.40, 39.30, 39.57, 39.58, 39.59, 39.91 |
Appendix A2.
| Etiology/Pathology | ICD-9-CM code |
|---|---|
| Malignant primary | 152.0, 156.0, 156.1, 156.2, 156.8, 156.9, 157, 157.0, 157.1, 157.2, 157.3, 157.4, 157.8, 157.9, 209.01, 209.25, 209.29, 209.30 |
| Malignant secondary | 197.4, 197.8, 209.71, 209.72, 209.79 |
| Carcinoma in situ | 230.9 |
| Benign | 209.41, 209.69, 211.2, 211.5, 211.6, 211.7 |
| Cyst | 577.2 |
| Pancreatitis | 577.0, 577.1 |
| Unspecified behavior | 235.5, 239.0, 576.2, 576.8, 576.9, 577.8, 577.9 |
Appendix A3.
| Intervention type | ICD-9-CM code |
|---|---|
| Endoscopic | 44.13, 44.43, 45.13, 45.16, 45.23, 45.24, 45.42 |
| Angiographic | 39.79, 44.44, 88.42, 88.47 |
| Operative | 39.98, 54.11, 54.12, 54.19 |
References
- 1.Whipple A.O., Parsons W.B., Mullins C.R. Treatment of carcinoma of the ampulla of Vater. Ann Surg. 1935;102:763–779. doi: 10.1097/00000658-193510000-00023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Cameron J.L., Pitt H.A., Yeo C.J., Lillemoe K.D., Kaufman H.S., Coleman J. One hundred and forty-five consecutive pancreaticoduodenectomies without mortality. Ann Surg. 1993;217:430–435. doi: 10.1097/00000658-199305010-00002. discussion 5-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.DeOliveira M.L., Winter J.M., Schafer M., Cunningham S.C., Cameron J.L., Yeo C.J. Assessment of complications after pancreatic surgery: a novel grading system applied to 633 patients undergoing pancreaticoduodenectomy. Ann Surg. 2006;244:931–937. doi: 10.1097/01.sla.0000246856.03918.9a. discussion 7-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.McPhee J.T., Hill J.S., Whalen G.F., Zayaruzny M., Litwin D.E., Sullivan M.E. Perioperative mortality for pancreatectomy: a national perspective. Ann Surg. 2007;246:246–253. doi: 10.1097/01.sla.0000259993.17350.3a. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Simons J.P., Shah S.A., Ng S.C., Whalen G.F., Tseng J.F. National complication rates after pancreatectomy: beyond mere mortality. J Gastrointest Surg. 2009;13:1798–1805. doi: 10.1007/s11605-009-0936-1. [DOI] [PubMed] [Google Scholar]
- 6.Limongelli P., Khorsandi S.E., Pai M., Jackson J.E., Tait P., Tierris J. Management of delayed postoperative hemorrhage after pancreaticoduodenectomy: a meta-analysis. Arch Surg. 2008;143:1001–1007. doi: 10.1001/archsurg.143.10.1001. discussion 7. [DOI] [PubMed] [Google Scholar]
- 7.Wente M.N., Veit J.A., Bassi C., Dervenis C., Fingerhut A., Gouma D.J. Postpancreatectomy hemorrhage (PPH): an International Study Group of Pancreatic Surgery (ISGPS) definition. Surgery. 2007;142:20–25. doi: 10.1016/j.surg.2007.02.001. [DOI] [PubMed] [Google Scholar]
- 8.Yekebas E.F., Wolfram L., Cataldegirmen G., Habermann C.R., Bogoevski D., Koenig A.M. Postpancreatectomy hemorrhage: diagnosis and treatment: an analysis in 1669 consecutive pancreatic resections. Ann Surg. 2007;246:269–280. doi: 10.1097/01.sla.0000262953.77735.db. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.HCUP, State Inpatient Databases (SID) 2007–2011. Healthcare cost and utilization project (HCUP): agency for healthcare research and quality, Rockville, MD.www.hcup-us.ahrq.gov/sidoverview.jsp Available from: [PubMed] [Google Scholar]
- 10.HCUP, State Emergency Department Databases (SEDD) 2007–2011. Healthcare cost and utilization project (HCUP): agency for healthcare research and quality, Rockville, MD.www.hcup-us.ahrq.gov/seddoverview.jsp Available from: [Google Scholar]
- 11.HCUP, Comorbidity Software . 2008. Healthcare cost and utilization project (HCUP): agency for healthcare research and quality, Rockville, MD.www.hcup-us.ahrq.gov/toolssoftware/comorbidity/comorbidity/jsp [cited 2015]. Available from: [PubMed] [Google Scholar]
- 12.Rumstadt B., Schwab M., Korth P., Samman M., Trede M. Hemorrhage after pancreatoduodenectomy. Ann Surg. 1998;227:236–241. doi: 10.1097/00000658-199802000-00013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Blanc T., Cortes A., Goere D., Sibert A., Pessaux P., Belghiti J. Hemorrhage after pancreaticoduodenectomy: when is surgery still indicated? Am J Surg. 2007;194:3–9. doi: 10.1016/j.amjsurg.2006.08.088. [DOI] [PubMed] [Google Scholar]
- 14.Rajarathinam G., Kannan D.G., Vimalraj V., Amudhan A., Rajendran S., Jyotibasu D. Post pancreaticoduodenectomy haemorrhage: outcome prediction based on new ISGPS clinical severity grading. HPB. 2008;10:363–370. doi: 10.1080/13651820802247086. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Khalsa B.S., Imagawa D.K., Chen J.I., Dermirjian A.N., Yim D.B., Findeiss L.K. Evolution in the treatment of delayed postpancreatectomy hemorrhage: surgery to interventional radiology. Pancreas. 2015;44:953–958. doi: 10.1097/MPA.0000000000000347. [DOI] [PubMed] [Google Scholar]
- 16.Grutzmann R., Ruckert F., Hippe-Davies N., Distler M., Saeger H.D. Evaluation of the International Study Group of Pancreatic Surgery definition of post-pancreatectomy hemorrhage in a high-volume center. Surgery. 2012;151:612–620. doi: 10.1016/j.surg.2011.09.039. [DOI] [PubMed] [Google Scholar]
- 17.Welsch T., Eisele H., Zschabitz S., Hinz U., Buchler M.W., Wente M.N. Critical appraisal of the International Study Group of Pancreatic Surgery (ISGPS) consensus definition of postoperative hemorrhage after pancreatoduodenectomy. Langenbecks Arch Surg. 2011;396:783–791. doi: 10.1007/s00423-011-0811-x. [DOI] [PubMed] [Google Scholar]
- 18.Ricci C., Casadei R., Buscemi S., Minni F. Late postpancreatectomy hemorrhage after pancreaticoduodenectomy: is it possible to recognize risk factors? JOP. 2012;13:193–198. [PubMed] [Google Scholar]
- 19.Darnis B., Lebeau R., Chopin-Laly X., Adham M. Postpancreatectomy hemorrhage (PPH): predictors and management from a prospective database. Langenbecks Arch Surg. 2013;398:441–448. doi: 10.1007/s00423-013-1047-8. [DOI] [PubMed] [Google Scholar]
- 20.Tan W.J., Kow A.W., Liau K.H. Moving towards the New International Study Group for Pancreatic surgery (ISGPS) definitions in pancreaticoduodenectomy: a comparison between the old and new. HPB. 2011;13:566–572. doi: 10.1111/j.1477-2574.2011.00336.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Lovecek M., Havlik R., Kocher M., Vomackova K., Neoral C. Pseudoaneurysm of the gastroduodenal artery following pancreatoduodenectomy. Stenting for hemorrhage. Wideochir Inne Tech Maloinwazyjne. 2014;9:297–301. doi: 10.5114/wiitm.2011.38178. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Robinson K., Rajebi M.R., Zimmerman N., Zeinati C. Post-pancreaticoduodenectomy hemorrhage of unusual origin: treatment with endovascular embolization and the value of preoperative CT angiography. J Radiol Case Rep. 2013;7:29–36. doi: 10.3941/jrcr.v7i4.1254. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Damle A., Clemenzi-Allen A., Jabbour N., Shah S.A. Rare cause of delayed upper gastrointestinal bleeding after pancreaticoduodenectomy. JOP. 2012;13:222–225. [PubMed] [Google Scholar]
- 24.Ellison E.C. Evidence-based management of hemorrhage after pancreaticoduodenectomy. Am J Surg. 2007;194(1):10–12. doi: 10.1016/j.amjsurg.2006.12.034. [DOI] [PubMed] [Google Scholar]
- 25.Ding X., Zhu J., Zhu M., Li C., Jian W., Jiang J. Therapeutic management of hemorrhage from visceral artery pseudoaneurysms after pancreatic surgery. J Gastrointest Surg. 2011;15:1417–1425. doi: 10.1007/s11605-011-1561-3. [DOI] [PubMed] [Google Scholar]
- 26.Schafer M., Heinrich S., Pfammatter T., Clavien P.A. Management of delayed major visceral arterial bleeding after pancreatic surgery. HPB. 2011;13:132–138. doi: 10.1111/j.1477-2574.2010.00260.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Sanjay P., Fawzi A., Fulke J.L., Kulli C., Tait I.S., Zealley I.A. Late post pancreatectomy haemorrhage. Risk factors and modern management. JOP. 2010;11:220–225. [PubMed] [Google Scholar]
- 28.Miura F., Asano T., Amano H., Yoshida M., Toyota N., Wada K. Management of postoperative arterial hemorrhage after pancreato-biliary surgery according to the site of bleeding: re-laparotomy or interventional radiology. J Hepatobiliary Pancreat Surg. 2009;16:56–63. doi: 10.1007/s00534-008-0012-3. [DOI] [PubMed] [Google Scholar]
- 29.Nakahara O., Takamori H., Ikeda O., Kuroki H., Ikuta Y., Chikamoto A. Risk factors associated with delayed haemorrhage after pancreatic resection. HPB. 2012;14:684–687. doi: 10.1111/j.1477-2574.2012.00518.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Correa-Gallego C., Brennan M.F., D'Angelica M.I., DeMatteo R.P., Fong Y., Kingham T.P. Contemporary experience with postpancreatectomy hemorrhage: results of 1,122 patients resected between 2006 and 2011. J Am Coll Surg. 2012;215:616–621. doi: 10.1016/j.jamcollsurg.2012.07.010. [DOI] [PubMed] [Google Scholar]
- 31.Parikh J.A., Beane J.D., Kilbane E.M., Milgrom D.P., Pitt H.A. Is American College of Surgeons NSQIP organ space infection a surrogate for pancreatic fistula? J Am Coll Surg. 2014;219:1111–1116. doi: 10.1016/j.jamcollsurg.2014.08.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Keck T., Wellner U.F., Bahra M., Klein F., Sick O., Niedergethmann M. Pancreatogastrostomy versus pancreatojejunostomy for RECOnstruction after PANCreatoduodenectomy (RECOPANC, DRKS 00000767): perioperative and long-term results of a multicenter randomized controlled trial. Ann Surg. 2016;263:440–449. doi: 10.1097/SLA.0000000000001240. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Diener M.K., Seiler C.M., Rossion I., Kleeff J., Glanemann M., Butturini G. Efficacy of stapler versus hand-sewn closure after distal pancreatectomy (DISPACT): a randomised, controlled multicentre trial. Lancet. 2011;377:1514–1522. doi: 10.1016/S0140-6736(11)60237-7. [DOI] [PubMed] [Google Scholar]
- 34.Tien Y.W., Lee P.H., Yang C.Y., Ho M.C., Chiu Y.F. Risk factors of massive bleeding related to pancreatic leak after pancreaticoduodenectomy. J Am Coll Surg. 2005;201:554–559. doi: 10.1016/j.jamcollsurg.2005.05.007. [DOI] [PubMed] [Google Scholar]