- Email: [email protected]
Risks of surgery for upper gastrointestinal hemorrhage: 1972 versus 1982
Risks of Surgery for Upper Gastrointestinal Hemorrhage: 1972 Versus 1982 Ben Kim, MD, New Haven; Connecticut
Hastings K. Wright, MD, New Haven, Connecticut Dennis Bordan, MD, Bridgeport, Connecticut L. Peter Fielding, MB, FRCS, Waterbury, Connecticut Robert Swaney, MD, Waterbury, Connecticut
The majority of patients with acute upper gastrointestinal bleeding are treated nonoperatively. When patients require surgical therapy, the introduction and widespread adoption of newer diagnostic and therapefftic tools during the past decade might be expected to improve the risks of surgery. However, recent studies have not shown such improvement [1,2]. To determine why, we undertook a retrospective comparison of all patients who underwent surgery for upper gastrointestinal hemorrhage in the years 1972 and 1982 to detect any significant differences over the decade.
Patients and Methods We reviewed the charts of all patients who underwent surgery for upper gastrointestinal hemorrhage in 1972and 1982 at three community hospitals and one university hospital. Those with variceal bleeding were excluded from the study. Indications for operation and choiceof operative procedure were decided by the individual attending physician. Results A total of 131 patient charts were reviewed, 66 in 1972 and 65 in 1982. The mean ages in the 2 periods were 57 and 66 years, respectively. The increase in age of 9 years was significant (p <0.05). The range of ages were similar, being 25 to 84 years in 1972 versus 22 to 96 years in 1982. The source of bleeding from discrete peptic ulcers decreased in the 10 year period from 86 percent to 62 percent of cases. Breakdown into duodenal locations versus gastric locations revealed a significant decrease in the occurrence of duodenal ulcers from 56 to 31 percent (p <0.02}. The incidence of gastric tflcer From the Departmentsof Surgery, Yale,New HavenHospital, New Haven, St. VInc,ent's Hospital, Bridgeport, St. Mary's Hospital. and Waterbury Hospital, Waterbury, Ceqnectlcut. Requests for reprints should be addressed to Ben KIm, MD, Department of Surgery, Yale University School of Medicine. 333 Cedar Street, New Haven, Connecticut, 08510. Presentedat the 65th Annual Meeting of the New EnglandSurgical Soc|ety, Dixvi|le Notch, New Hampshire,October 12-14, 1984.
474
remained almost identical being 30 percent in 1972 and 31 percent in 1982. Diffuse gastritisincreased in incidence from I percent to 18 percent (p <0.02). The incidence of tumor as a source of bleeding was 3 percent in 1972 and 14 percent in 1982. The source of bleeding was never identified in 11 percent of the patients in 1972. In 1982, this was true in only 6 percent. Operative procedures were grouped as either gastrectomies, if any form of gastric resection was involved, or vagotomy and drainage, ifno resection was involved N o significant change in the choice of operation was noted in the 10 year period. A slight insignificant trend toward fewer gastrectomies, (68 percent in 1972 to 55 percent in 1982) and more vagotomies with drainage (26 percent to 37 percent) was noted. Complication rates for rebleeding were 17 percent in 1972 and 9 percent in 1982, although the difference was not significant. Infection and postoperative myocardial infarction rates were almost identical, being 3 percent in 1972 versus 6 percent in 1982 and 3 percent in 1972 versus 2 percent in 1982, respectively. Operative mortality increased from 12 percent to 17 percent. This difference was not statisticallysignificant. The mean age of patients who died was similar, being 72 patients in 1972 versus 74 patients in 1982. In 1972, two of eight patients who died (25 percent) exsanguinated from an unidentified bleeding source. None of the patients who died in 1982 had an unidentified bleeding source (p <0.05). The cause of death in both years was most often some combination of organ failure and sepsis. W h e n admitting diagnoses were compared with final diagnoses, all eight patients who died in 1972 were admitted for bleeding. In 1982, 4 of~ll patients who died (36 percent) were admitted for another condition, and then upper gastrointestinal hemorrhage developed during their hospitalization (p <0.0I). These four patients' initial diagnoses were diverticulitis with subphrenic abscess, genitourinary
1~ ~
,I~,~I ~
Risks of Surgery for Gastrointestinal H e m o r r h a g e
TABLE I
Surgical Mortality of Upper Gastrolntesllnal Bleeding
Study Elerdit~g et al [8] Hellers and lhre [ 12] Hunt [3] KIm et al [9] Hoare [ 10] Vellacott et a l [ 1] Sandlow et a l [
tl]
NIIsson and W a h ! b e r g [2! ..............
Surgical Mortality (% )
Mean Age (yr)
Patients (n)
Operated (%)
Inclusive Years
Protocol"
9 11
51 ?
47 38
20 25
73-78 72-74
No Yes
11.5 14.4
60 50
206
34
72-82
Yes Yes
12
s7
"86"
:::
i2
No
17 14-26 21 24 26 40 41
66 57 66 68 50 48 63
65 36 140 105 15 5 29
2"3" 33 21 19 7 11
82 72-74 75-77 78 -80 74 74 77-78
No No No No No No No
" In Hunt's series protocols wore followed for patients who had peptic ulcers only.
tract sepsis, and trauma (two of the patients}. The two trauma patients bled as the result of stress ulceration.
Comments In the decade from 1972 through 1982, many innovations have been added to our diagnostic and therapeutic armamentarium, including fiberoptic endoscopy, histamine-2 antagonists, intensive care units, angiographic control of bleeding, and total parenteral nutrition. Therefore, it seemed logical to expect a decrease in tile surgical risk for upper gastrointestinal bleeding. Our data do not support this contention. Mortality rate increased from 12 to 17 percent. This difference was not statistically significant. However, closer inspection reveals some striking differences between the two patient groups. First, in the 10 year period, the average age of our patients increased by 9 years. Previous studies have noted an increased risk in patients over the age of 60 years [2,3]. In 1972, 48 percent were older than 60 years of age. By 1982, this figure had increased to 74 percent. Second, there was a dramatic decrease in the percentage of patients with duodenal ulcers, from 56 percent to 31 percent. The incidence of diffuse gastritis increased from 1.5 percent to 18 percent and that of gastric ulcer remained unchanged. Thus, the more favorable discrete duodenal ulcer was supplanted by nondiscrete bleeding gastritis, which carries a higher mortality rate [1,4-7]. Third, although 25 percent of the patients from 1972 who died underwent blind gastrectomy and exsanguinated from an unidentified source of bleeding, none of the patients from 1982 died this way. Finally, in 1982, bleeding developed in 30 percent of the patients during hospitalization for other conditions, whereas it developed in only 11 percent of the patients in
Volume 149, Aprf! 1985
1972. Many of these patients had prolonged, complex hospitalizations and thus, are at higher risk for the development of complications because of their compromised condition, often from trauma, sepsis, or underlying malignancy. How does our data compare with those of others? In Table I, we have summarized the surgical mortality rates for nonvariceal bleeding from several studies published over the 10 year period. At first it is disconcerting to note the great variation from the low 9 percent mortality rate of Elerding et al [8] to the high 41 percent rate of Nilsson and Wahlberg [2]. Several patterns can be discerned from the data. In the series of Elerding et al [8], Hellers and Ihre [12], Hunt [3] and Kim et al [9], the mortality rates ranged from 9 percent to 14 percent. The average age of these patients was under 60 years. Three of these four studies set up a specialized unit or followed a set management protocol. All four studies had high operative rates, being between 20 and 35 percent for all patients with upper gastrointestinal bleeding. By contrast, Vellacot et al [I], Nilsson and Wahlberg [2], and Sandlow et al [11] had higher mortality rates (21 to 41 percent), but their population was older, there was no fixed management protocol, and a smaller proportion of patients with upper gastrointestinal bleeding underwent operation. Our mortality rate of 17 percent in a patient population with an average age above that in any of the reported series compares favorably with the other published series. In conclusion, when patients require surgery for upper gastrointestinal bleeding, their chances for a successful outcome remain similar to those of a decade ago. However, there appears to be emerging a subset of older patients with a high likelihood of bleeding from a nondiscrete site during hospitalization for another condition. We suggest that the next major decrease in surgical
475
Kim et al
risk will require the identification and aggressive m a n a g e m e n t of this new patient population. Summary
A retrospective comparison was u n d e r t a k e n to
determine if the risks of undergoing surgery for nonvariceal upper gastrointestinal hemorrhage had changed between 1972 and 1982. In 1982, patients were on the average 9 years older, there was a significant decrease in bleeding from duodenal ulcers compared with 1972 data, gastric ulcer rates remained unchanged, and diffuse gastritis occurred more frequently in 1982. Mortality and morbidity rates showed no significant differences; however, the patient population did change with the emergence of older patients, in whom bleeding developed after hospitalization for other reasons. These patients comprised 30 percent of the 1982 s t u d y population. If further improvements in surgical t r e a t m e n t of upper gastrointestinal hemorrhage are to occur, these patients must be identified and aggressively managed. References
1. Vellacott KD, Dronfield MW, Atktnson M, Langman MJS. Comparison of surgical and medical managementof bleeding peptic ulcers. Br Med J 1982;284:15.1a-15.2b. 2. Nilsson F, WahlbergJ. Survival and emergency surgery in upper gastrointestinal bleeding. Acta Chir Scand 1981;147: 555-9. 3. HuntPS. Surgical managementot bleeding chronic peptic ulcer. Ann Surg 1984; 199:44-50. 4. Himal HS, Perrault C, Mzabi R. Upper gastrointestinal hemorrhage: aggressive managementdecreases mortality. Surgery 1978;84:448-54. 5. Larson DE, Famell MB. Upper gastrointestinal hemorrhage. Mayo CIIn Proc 1983;58:371-87. 6. Hubert JP, Kieman PD, Welch JS, ReMIne WH, Beahrs OH. The surgical management of bleeding stress ulcers. Ann Surg 1980; 19 t:627-79. 7. Schrock TK. Does endoscopy affect the surgical approach to the patient with upper gastrointestinal bleeding? Dig Dis Sci 1981;26:27-30. 8. Elerding SC, Moore EE, Woiz JR, Norton LW, Outcome of operations for upper gastrointestinal tract bleeding. Arch Surg 1980; 115:1473-7. 9. Kim U, Rudick J, Aufses AH. Surgical management of acute upper ~jastrotntestlnalbleeding. Value of early diagnosis and prompt surgical intervention. Arch Surg 1978;113:14447. 10. Hoare AM. Comparative study between endoscopy and radiology in acute upper gastrointestinal hemorrhage. Br Med
J 1975;1:27-30. 11. Sandlow I_J, Becker GH, Spellberg MA et al. A prospective randomized study of the management of upper gastrointestinal hemorrhage. Am J Gastroenterol 197:282-9. 12. HellersG, Ihre T. lmpact of change to early diagnosis and surgery in major upper gastrointestinal bleeding. Lancet t975:1250-1.
Discussion
Jerome S. Abrams (Burlington, VT): I thought perhaps the Society might be interested in a study that was carried out a year ago by one of our former students, Joel LaFleur, working with Dr. Gordon Page of this Society, in which they looked at perforated ulcers during this same time period, 1972 to 1982, and came up with 92 patients. Basically, what they found was that the average age for the patients with perforated ulcers was 56 years, with an average age of a third of them being over 70 years. This is quite different from what most of us have previously experienced in cases of perforated ulcer. Ninety percent occurred at the duodenum or at the pylorus, and only I0 percent were pure gastric perforations, Only a third of these patients had previous symptoms. Probably the most important single thing LaFleur and Page found was that almost 45 percent of these patients had been receiving significant antiinftammatory drugs (20 percent were receiving aspirin and another 20 percent nonsteroidal mltiinflammatory drugs) and by the latter half of the study, a quarter were receiving some kind of steroid. As you might guess, the mortality rate of those who were not operated on, usually because the surgeon was too smart to take them to the operating room, was 63 percent; in those who underwent surgery, the mortality rate was 16 percent. It made little difference whether they were plicated, resected, or had vagotomy and pyloroplasty. This represents a significant change in the picture that most of us have of duodenal ulcer. Dr. Kim, did you study these two time periods to see what percentage of your patients may have been receiving any nonsteroidal or steroidal antiinflammatory drugs? Ben Kim (closing): I'd like to firstthank Dr. Ackroyd for his comments; I agree with all of them. Dr. Abrams, regarding the nonsteroidal antiinflammatory drugs, I do have some data that I did not include. Looking at the subpopulation of patients from Yale-New Haven Hospital, 16 percent of them were receiving the newer nonsteroidal antiinflammatory drugs in 1982; none were receiving them in 1972. However, the number of patients taking aspirin decreased and that of patients taking steroids remained about the same. These numbers were too small to derive statistical significance.
The discussion section has been abbreviated due to space limitations. The full text can be obtained from the reprints author listed on the title page.
476
The AmericanJournalof Surgery
Hastings K. Wright, MD, New Haven, Connecticut Dennis Bordan, MD, Bridgeport, Connecticut L. Peter Fielding, MB, FRCS, Waterbury, Connecticut Robert Swaney, MD, Waterbury, Connecticut
The majority of patients with acute upper gastrointestinal bleeding are treated nonoperatively. When patients require surgical therapy, the introduction and widespread adoption of newer diagnostic and therapefftic tools during the past decade might be expected to improve the risks of surgery. However, recent studies have not shown such improvement [1,2]. To determine why, we undertook a retrospective comparison of all patients who underwent surgery for upper gastrointestinal hemorrhage in the years 1972 and 1982 to detect any significant differences over the decade.
Patients and Methods We reviewed the charts of all patients who underwent surgery for upper gastrointestinal hemorrhage in 1972and 1982 at three community hospitals and one university hospital. Those with variceal bleeding were excluded from the study. Indications for operation and choiceof operative procedure were decided by the individual attending physician. Results A total of 131 patient charts were reviewed, 66 in 1972 and 65 in 1982. The mean ages in the 2 periods were 57 and 66 years, respectively. The increase in age of 9 years was significant (p <0.05). The range of ages were similar, being 25 to 84 years in 1972 versus 22 to 96 years in 1982. The source of bleeding from discrete peptic ulcers decreased in the 10 year period from 86 percent to 62 percent of cases. Breakdown into duodenal locations versus gastric locations revealed a significant decrease in the occurrence of duodenal ulcers from 56 to 31 percent (p <0.02}. The incidence of gastric tflcer From the Departmentsof Surgery, Yale,New HavenHospital, New Haven, St. VInc,ent's Hospital, Bridgeport, St. Mary's Hospital. and Waterbury Hospital, Waterbury, Ceqnectlcut. Requests for reprints should be addressed to Ben KIm, MD, Department of Surgery, Yale University School of Medicine. 333 Cedar Street, New Haven, Connecticut, 08510. Presentedat the 65th Annual Meeting of the New EnglandSurgical Soc|ety, Dixvi|le Notch, New Hampshire,October 12-14, 1984.
474
remained almost identical being 30 percent in 1972 and 31 percent in 1982. Diffuse gastritisincreased in incidence from I percent to 18 percent (p <0.02). The incidence of tumor as a source of bleeding was 3 percent in 1972 and 14 percent in 1982. The source of bleeding was never identified in 11 percent of the patients in 1972. In 1982, this was true in only 6 percent. Operative procedures were grouped as either gastrectomies, if any form of gastric resection was involved, or vagotomy and drainage, ifno resection was involved N o significant change in the choice of operation was noted in the 10 year period. A slight insignificant trend toward fewer gastrectomies, (68 percent in 1972 to 55 percent in 1982) and more vagotomies with drainage (26 percent to 37 percent) was noted. Complication rates for rebleeding were 17 percent in 1972 and 9 percent in 1982, although the difference was not significant. Infection and postoperative myocardial infarction rates were almost identical, being 3 percent in 1972 versus 6 percent in 1982 and 3 percent in 1972 versus 2 percent in 1982, respectively. Operative mortality increased from 12 percent to 17 percent. This difference was not statisticallysignificant. The mean age of patients who died was similar, being 72 patients in 1972 versus 74 patients in 1982. In 1972, two of eight patients who died (25 percent) exsanguinated from an unidentified bleeding source. None of the patients who died in 1982 had an unidentified bleeding source (p <0.05). The cause of death in both years was most often some combination of organ failure and sepsis. W h e n admitting diagnoses were compared with final diagnoses, all eight patients who died in 1972 were admitted for bleeding. In 1982, 4 of~ll patients who died (36 percent) were admitted for another condition, and then upper gastrointestinal hemorrhage developed during their hospitalization (p <0.0I). These four patients' initial diagnoses were diverticulitis with subphrenic abscess, genitourinary
1~ ~
,I~,~I ~
Risks of Surgery for Gastrointestinal H e m o r r h a g e
TABLE I
Surgical Mortality of Upper Gastrolntesllnal Bleeding
Study Elerdit~g et al [8] Hellers and lhre [ 12] Hunt [3] KIm et al [9] Hoare [ 10] Vellacott et a l [ 1] Sandlow et a l [
tl]
NIIsson and W a h ! b e r g [2! ..............
Surgical Mortality (% )
Mean Age (yr)
Patients (n)
Operated (%)
Inclusive Years
Protocol"
9 11
51 ?
47 38
20 25
73-78 72-74
No Yes
11.5 14.4
60 50
206
34
72-82
Yes Yes
12
s7
"86"
:::
i2
No
17 14-26 21 24 26 40 41
66 57 66 68 50 48 63
65 36 140 105 15 5 29
2"3" 33 21 19 7 11
82 72-74 75-77 78 -80 74 74 77-78
No No No No No No No
" In Hunt's series protocols wore followed for patients who had peptic ulcers only.
tract sepsis, and trauma (two of the patients}. The two trauma patients bled as the result of stress ulceration.
Comments In the decade from 1972 through 1982, many innovations have been added to our diagnostic and therapeutic armamentarium, including fiberoptic endoscopy, histamine-2 antagonists, intensive care units, angiographic control of bleeding, and total parenteral nutrition. Therefore, it seemed logical to expect a decrease in tile surgical risk for upper gastrointestinal bleeding. Our data do not support this contention. Mortality rate increased from 12 to 17 percent. This difference was not statistically significant. However, closer inspection reveals some striking differences between the two patient groups. First, in the 10 year period, the average age of our patients increased by 9 years. Previous studies have noted an increased risk in patients over the age of 60 years [2,3]. In 1972, 48 percent were older than 60 years of age. By 1982, this figure had increased to 74 percent. Second, there was a dramatic decrease in the percentage of patients with duodenal ulcers, from 56 percent to 31 percent. The incidence of diffuse gastritis increased from 1.5 percent to 18 percent and that of gastric ulcer remained unchanged. Thus, the more favorable discrete duodenal ulcer was supplanted by nondiscrete bleeding gastritis, which carries a higher mortality rate [1,4-7]. Third, although 25 percent of the patients from 1972 who died underwent blind gastrectomy and exsanguinated from an unidentified source of bleeding, none of the patients from 1982 died this way. Finally, in 1982, bleeding developed in 30 percent of the patients during hospitalization for other conditions, whereas it developed in only 11 percent of the patients in
Volume 149, Aprf! 1985
1972. Many of these patients had prolonged, complex hospitalizations and thus, are at higher risk for the development of complications because of their compromised condition, often from trauma, sepsis, or underlying malignancy. How does our data compare with those of others? In Table I, we have summarized the surgical mortality rates for nonvariceal bleeding from several studies published over the 10 year period. At first it is disconcerting to note the great variation from the low 9 percent mortality rate of Elerding et al [8] to the high 41 percent rate of Nilsson and Wahlberg [2]. Several patterns can be discerned from the data. In the series of Elerding et al [8], Hellers and Ihre [12], Hunt [3] and Kim et al [9], the mortality rates ranged from 9 percent to 14 percent. The average age of these patients was under 60 years. Three of these four studies set up a specialized unit or followed a set management protocol. All four studies had high operative rates, being between 20 and 35 percent for all patients with upper gastrointestinal bleeding. By contrast, Vellacot et al [I], Nilsson and Wahlberg [2], and Sandlow et al [11] had higher mortality rates (21 to 41 percent), but their population was older, there was no fixed management protocol, and a smaller proportion of patients with upper gastrointestinal bleeding underwent operation. Our mortality rate of 17 percent in a patient population with an average age above that in any of the reported series compares favorably with the other published series. In conclusion, when patients require surgery for upper gastrointestinal bleeding, their chances for a successful outcome remain similar to those of a decade ago. However, there appears to be emerging a subset of older patients with a high likelihood of bleeding from a nondiscrete site during hospitalization for another condition. We suggest that the next major decrease in surgical
475
Kim et al
risk will require the identification and aggressive m a n a g e m e n t of this new patient population. Summary
A retrospective comparison was u n d e r t a k e n to
determine if the risks of undergoing surgery for nonvariceal upper gastrointestinal hemorrhage had changed between 1972 and 1982. In 1982, patients were on the average 9 years older, there was a significant decrease in bleeding from duodenal ulcers compared with 1972 data, gastric ulcer rates remained unchanged, and diffuse gastritis occurred more frequently in 1982. Mortality and morbidity rates showed no significant differences; however, the patient population did change with the emergence of older patients, in whom bleeding developed after hospitalization for other reasons. These patients comprised 30 percent of the 1982 s t u d y population. If further improvements in surgical t r e a t m e n t of upper gastrointestinal hemorrhage are to occur, these patients must be identified and aggressively managed. References
1. Vellacott KD, Dronfield MW, Atktnson M, Langman MJS. Comparison of surgical and medical managementof bleeding peptic ulcers. Br Med J 1982;284:15.1a-15.2b. 2. Nilsson F, WahlbergJ. Survival and emergency surgery in upper gastrointestinal bleeding. Acta Chir Scand 1981;147: 555-9. 3. HuntPS. Surgical managementot bleeding chronic peptic ulcer. Ann Surg 1984; 199:44-50. 4. Himal HS, Perrault C, Mzabi R. Upper gastrointestinal hemorrhage: aggressive managementdecreases mortality. Surgery 1978;84:448-54. 5. Larson DE, Famell MB. Upper gastrointestinal hemorrhage. Mayo CIIn Proc 1983;58:371-87. 6. Hubert JP, Kieman PD, Welch JS, ReMIne WH, Beahrs OH. The surgical management of bleeding stress ulcers. Ann Surg 1980; 19 t:627-79. 7. Schrock TK. Does endoscopy affect the surgical approach to the patient with upper gastrointestinal bleeding? Dig Dis Sci 1981;26:27-30. 8. Elerding SC, Moore EE, Woiz JR, Norton LW, Outcome of operations for upper gastrointestinal tract bleeding. Arch Surg 1980; 115:1473-7. 9. Kim U, Rudick J, Aufses AH. Surgical management of acute upper ~jastrotntestlnalbleeding. Value of early diagnosis and prompt surgical intervention. Arch Surg 1978;113:14447. 10. Hoare AM. Comparative study between endoscopy and radiology in acute upper gastrointestinal hemorrhage. Br Med
J 1975;1:27-30. 11. Sandlow I_J, Becker GH, Spellberg MA et al. A prospective randomized study of the management of upper gastrointestinal hemorrhage. Am J Gastroenterol 197:282-9. 12. HellersG, Ihre T. lmpact of change to early diagnosis and surgery in major upper gastrointestinal bleeding. Lancet t975:1250-1.
Discussion
Jerome S. Abrams (Burlington, VT): I thought perhaps the Society might be interested in a study that was carried out a year ago by one of our former students, Joel LaFleur, working with Dr. Gordon Page of this Society, in which they looked at perforated ulcers during this same time period, 1972 to 1982, and came up with 92 patients. Basically, what they found was that the average age for the patients with perforated ulcers was 56 years, with an average age of a third of them being over 70 years. This is quite different from what most of us have previously experienced in cases of perforated ulcer. Ninety percent occurred at the duodenum or at the pylorus, and only I0 percent were pure gastric perforations, Only a third of these patients had previous symptoms. Probably the most important single thing LaFleur and Page found was that almost 45 percent of these patients had been receiving significant antiinftammatory drugs (20 percent were receiving aspirin and another 20 percent nonsteroidal mltiinflammatory drugs) and by the latter half of the study, a quarter were receiving some kind of steroid. As you might guess, the mortality rate of those who were not operated on, usually because the surgeon was too smart to take them to the operating room, was 63 percent; in those who underwent surgery, the mortality rate was 16 percent. It made little difference whether they were plicated, resected, or had vagotomy and pyloroplasty. This represents a significant change in the picture that most of us have of duodenal ulcer. Dr. Kim, did you study these two time periods to see what percentage of your patients may have been receiving any nonsteroidal or steroidal antiinflammatory drugs? Ben Kim (closing): I'd like to firstthank Dr. Ackroyd for his comments; I agree with all of them. Dr. Abrams, regarding the nonsteroidal antiinflammatory drugs, I do have some data that I did not include. Looking at the subpopulation of patients from Yale-New Haven Hospital, 16 percent of them were receiving the newer nonsteroidal antiinflammatory drugs in 1982; none were receiving them in 1972. However, the number of patients taking aspirin decreased and that of patients taking steroids remained about the same. These numbers were too small to derive statistical significance.
The discussion section has been abbreviated due to space limitations. The full text can be obtained from the reprints author listed on the title page.
476
The AmericanJournalof Surgery