Albumin as a predictive tool

Kenneth A. Kudsk, Elizabeth A. Tolley, R. Chance DeWitt, Peter G. Janu .Preoperative Albumin and Surgical Site Identify Surgical Risk for Major Postoperative Complications JOURNAL OF

PARENTERAL AND ENTERAL NUTRITION 2003 27 1-9,• Background: Although malnutrition contributes to morbidity, studies of pre- and postoperative nutrition often include

well-nourished patients unlikely to benefit from therapy and usually do not stratify by the site of surgical pathology.

• This study evaluates whether perceived preoperative markers of nutritional status recorded in charts correlates with postoperative complications and resource use in patients who receive no preoperative nutrition support and reinterprets the results of several conflicting randomized, prospective studies in this context.

• Methods: This is a retrospective cohort study of 526 surgical patients who had preoperative serum albumin levels measured and were undergoing elective esophageal, gastric, pancreaticoduodenal, or colon surgery between 1992 and 1996 who could have received preoperative nutrition but did not.

• Results: Most medical records contained inadequate analysis of preoperative nutritional status, but preoperative albumin correlated inversely with complications, length of stay, postoperative stay, intensive care unit (ICU) stay, mortality, and resumption of oral intake. Patients undergoing esophageal or pancreatic procedures sustained a significantly higher complication rate at most albumin levels, whereas colonic surgery resulted in lower complication rates at the same albumin levels. Resource use (eg, length of stay and ICU stay) related to these complication rates; esophageal and pancreatic procedures used the most resources and colon procedures used the fewest at most albumin levels. This lack of appreciation for nutritional risk and operative site can explain discrepancies in outcome noted in several randomized, prospective nutritional studies and must be applied to the design and implementation of new studies.

• Conclusions: Elective, nonemergent esophageal and pancreatic procedures performed in patients who could have had surgery delayed for preoperative nutrition, but did not, result in higher risk than colon surgery at any given level of serum albumin below 3.25 g/dL. Patient populations in trials should be stratified by operative site and by markers of nutritional status. Degree of hypoalbuminemia and other potential markers of nutritional status may explain many of the discrepancies between trials of nutrition support. Preexisting hypoalbuminemia in patients undergoing elective surgery remains underappreciated, unrecognized, and untreated in many hospitalized patients.

Cclinical and laboratory data used to

determine nutritional status and correlated with outcome.

• Depressed total lymphocyte count,• protein depletion,• low serum albumin or transferrin• history of significant preoperative weight loss

– are associated with increased postoperative complications.7–18

Body mass index, Anthropometrics percent body weight loss can be used to evaluate muscle protein and fat stores.19

Preoperative Albumin and Surgical Site Identify Surgical Risk for Major Postoperative Complications

• This study evaluates patients undergoing elective surgical procedures of the esophagus, stomach, pancreas, or colon for either benign or malignant disease to identify and quantify relationships among markers of preoperative nutritional status and postoperative complications.

• Our goal was to identify a simple, but clinically useful, preoperative indicator of the postoperative recovery from elective general surgical procedures.

• The focus of this research was to answer four questions.

• (1) Do existing medical records contain an adequate nutritional preoperative risk analysis?

• (2) Can suspected preoperative nutritional risk indicators predict the postoperative risk of major complications and associated postoperative resource use?

• (3) Does operative site (eg, esophagus, stomach, colon, or pancreas) influence postoperative outcome and resource use?

• (4) Can discrepancies in nutritional outcome studies be explained by these results?

• A prerequisite for inclusion in the analysis was that all patients had to be capable of receiving reoperative nutrition support, but did not, to exclude emergent conditions with no opportunity for preoperative nutrition support and to eliminate the effects of stress and resuscitation on serum protein levels.

Incidence of major complications by type of

surgery and preop albumin levels.

Incidence of major complications bysite of surgery and preop albumin level

Albumin preop level gr/dL

• Table II also demonstrates the incidence of complications at each albumin level by operative site compared with the combined data. Because of the small number of patients in several of the low albumin categories, those categories were combined for each operative site and compared for the incidence of major complications. In patients with an albumin 2.75 g/dL(Categories 1 to 3), patients undergoing pancreatic surgery had significantly more major complications than those undergoing stomach surgery (p .03) or colonic surgery ( p .003). For patients with an albumin 3.25 g/dL (Categories 1 to 4), pancreatic surgery resulted in a higher major complication rate than those undergoing stomach ( p .03) or colon surgery ( p .003), whereas esophageal surgery resulted in significantly more major complications than patients undergoing colon surgery ( p .02). With an albumin 3.75 g/dL (Categories 1 to 5), esophageal surgery resulted in

• significantly more major complications than stomach ( p .04) or colon ( p .005) surgery, whereas pancreas surgery resulted in significantly more complications than colon surgery ( p .007).

Mortality by type of surgery and preopalbumin level

Mortality by type of surgery and preopalbumin level

Mortality

• Mortality by site of operation and degree of hypoalbuminemia is reported in Table III. Mortality was exceptionally low (1% to 5%) with a preoperative albumin above 3.25 g/dL, regardless of operative site.

• As albumin levels dropped below 3.25 g/dL, mortality rate progressively increased up to 20% to 30% in the 2 lowest albumin categories. There were no significant differences in mortality rate between the 4 operative sites if albumin levels were 3.25 g/dL.

Postoperative days for patients withand without complications by preop

albumin level

• Albumin cutoff value :3.25 gr/lt

PrognosticNutrition Index(PNI)

• Ingenbleek Y, Carpentier YA: A prognostic and inflammatory nutritional index scoring critically ill patients. Int J Vitam Nutr Res55:91–101, 1985

• Vehe KL, Brown RD, Kuhl DA, et al: The prognostic inflammatory and nutritional index in traumatized patients receiving enteralnutrition support. J Am Coll Nutr 10:355–363, 1991

Biblio on the value of albumin as a statistical predictor of postoperative risk.

• 9. Hickman DM, Miller RA, Rombeau JL, et al: Serum albumin and body weight as predictors of postoperative course in colorectal cancer. JPEN 4:314–316, 1980

• 10. Buzby GP, Mullen JL, Matthews DC, et al: Prognostic nutritional index in GI surgery. Am J Surg 139:160–167, 1980

• 11. Reinhardt GF, Myscofski JW, Wilkens DB, et al: Incidence and mortality of hypoalbuminemic patients in hospitalized veterans. JPEN 4:357–359, 1980

• 12. Mullen JL, Gertner MG, Buzby GP, et al: Implication of malnutrition in the surgical patient. Arch Surg 114:121–125, 1979

• 13. Pikul J, Sharpe MD, Lowndes R, et al: Degree of preoperative malnutrition is predictive of postoperative morbidity and mortality in liver transplant recipients. Transplantation 57:469– 472, 1994

• 14. Daley J, Khuri SF, Henderson W, et al: Risk adjustments of the postoperative morbidity rate for the comparative assessment of the quality of surgical care. Results of the National Veterans Affairs Surgical Risk Study. J Am Coll Surg 185:328–340, 1997

• 15. Collins TC, Daley J, Henderson WH, et al: Risk factors for prolonged length of stay after major elective surgery. Ann Surg 230:251–259, 1999

• 16. Khuri SF, Daley J, Henderson W, et al: The Department of Veterans Affairs’ NSQIP. The first national, validated, outcomebased, risk-adjusted, and peer-controlled program for the measurement and enhancement of the quality of surgical care. Ann Surg 228:491–507, 1998

• 17. Arora NS, Rochester DF: Respiratory muscle strength and maximal voluntary ventilation in undernourished patients. Am Rev Re spir Dis 126:5–8, 1982

• Khuri SF, Daley J, Henderson W, et al: Risk adjustment of the postoperative mortality rate for the comparative assessment of the quality of surgical care: Results of the National Veterans Affairs Surgical Risk Group. J Am Coll Surg 185:315–327, 1997

FINE

Complications by hospital system withand without albumin stratification

Serious Complications by Hospital System

• Patients were categorized into three hospital systems: public hospital, Veterans Hospital, and private hospitals (Table IV). There were significantly fewer serious complications in the private hospitals than in the VA or public hospitals ( p .05). However, when patients were stratified by albumin (3.25 or 3.25 g/dL), all differences disappeared, reflecting the disproportionate number of hypoalbuminemicat-risk patients in the Veterans Administration and publichospital systems

Postoperative days for patients withand without complications by preop

albumin level

Postoperative stay

• POS—Combined Data

• POS correlated with the presence or absence of complications.Complications (versus no complications) significantly increased POS in all categories except for the highest and lowest albumin groups (Category 1 and 7, Table V). Failure to achieve statistical significance in the lowest albumin level (Category 1) was primarily caused by the 31% mortality rate (Table III; 4 of 13 patients) that was associated with an extremely short postoperative stay before death. The average POS in surviving patients with complications in category 1 was 70.0 6.6 days compared with 27.2 5.2 days ( p .0001) in patients with no complications.

• POS is similar in patients who recovered without complications in the combined populations except at

• the lowest albumin level (Table V). In patients with complications, POS was influenced both by the albumin level (Table V) and the operative site. POS after gastric surgery reflected combined data from the 4 operative sites, but colon surgery resulted in a shorter POS in pooled data (data not shown).

• Patients undergoing pancreatic surgery in the lowest 2 albumin categories have increased POS compared with combined data. At all points, POS for esophageal surgery with an albumin of 3.75 g/dL or less was significantly greater than the combined mean.

• Because mortality can influence POS, and mortality increases in lower categories, a separate analysis excluding deaths was performed in patients with complications (Table V). This analysis showed very little difference in average POS with and without death except in the lowest category, where the average POS increased to 70.0 6.6 days from 34.1 4.8 days.

ICU stays for patients with and withoutcomplications by preop albumin levels

ICU Stay

• ICU Stay—Combined Data• Mean ICU stay reflected the changes in POS and gradually increased with

decreasing preoperative albumin level (Table VI). As expected, development of complications significantly increased ICU stay. Exclusion of deaths had a greater influence on ICU stay than POS in the lowest albumin categories. By excluding deaths, ICU stay decreased from 13.8 1.6 to 7.9 1.8 days ( p .01) in Category 4, 15.2 2.0 to 9.8 1.9 days ( p .01) in Category 3, and 21.5 2.5 to 12.3 2.5 days (p .001) in Category 2, but increased from 32.9 3 to 48.3 3.9 days ( p .001) in the most hypoalbuminemic (Category 1) patients. Although the incidence of ICU stay seemed to be relatively stable across all categories with an albumin 2.25 g/dL (ie,

• Categories 3 to 7), this was because of the disproportionate prolonged ICU stays after esophageal surgery at these higher albumin levels. Without these patients,

• ICU stay gradually decreased as albumin level increased. The prolonged ICU stay in the lowest categories could be accounted for primarily by patients undergoing pancreatic and esophageal surgeries.

Postoperative stay

ICU stay

NPO days

• Our study shows that both preoperative serum albumin and operative site affect the complication rate, which subsequently influences POS and ICU stay as markers of resource use

• stepwise drops in preoperative albumin increases the risk of major complications and increases resource use as measured by postoperative stay and ICU stay.

• In addition, the patients with the highest risk by albumin had the longest delays in resuming oral intake, which compounds the problem of progressive, unremitting loss of lean tissue and need for further resource use.

• We specifically limited the study to stable patients who could have tolerated preoperative delay for nutrition support to minimize gross confounding factors that depress serum albumin levels, such as altered hydration, changes in permeability, and the interleukin (IL)-6 response to inflammation with its effect on hepatocellular protein production.23,24 Whether albumin levels were due purely to malnutrition or to a combination of other factors, such as alcohol abuse, hepatic disease, or other non-nutritional etiologies, was beyond the scope of this study. These variables seem less relevant to the current analysis because decreasing albumin—regardless of cause—was associated with increasing risk of complications.

• Although extreme hypoalbuminemia by itself was associated with increased ICU and postoperative stay,complications were the greatest predictor of resource use such as postoperative stay, ICU stay, and NPO days.

• Composite numbers, however, mask some important factors. For example, combined data from all surgical sites underestimates the real effect of severe hypoalbuminemia on length of stay because many patients die early.

• Composite data also mask the influence of surgical site on recovery. Patients undergoing colectomy (especially) and gastrectomy tolerated greater degrees of hypoalbuminemia than esophagectomy or pancreatectomy patients. This may relate to the anatomic importance of the serosa and submucosal layers to anastomotic strength because serosa is absent in the esophagus and the pancreas.

• Patients sustaining complications after gastric or colonic surgery have the same or shorter length of stay than the average patient noted in the combined group. Patients undergoing pancreas or esophageal surgery, however, have higher rates of complications and higher resource use at lower albumin levels, especially below 3.25 g/dL.

• Failure to stratify by site of operation in outcome studies could bias inferences about resource use. If one treatment arm includes more patients undergoing esophageal or pancreatic surgery and fewer patients with colon surgery, outcomes will likely be biased by such selection. Similarly, a treatment arm unbalanced in favor of colon or gastric surgery would tend to bias outcome data toward more favorable outcomes regardless of the effectiveness or ineffectiveness of the therapy.

• One could argue that patients undergoing colon resection should not be randomized into prospective studies unless albumin levels are below 2.75 g/dL, whereas patients undergoing esophageal surgery should be included if albumin levels drop below 3.75 g/dL, which by current standards, is the lower range of normal.

• The apparent cutoff rate at which resource use increases after gastrectomy or pancreatic surgery is approximately 3.25 g/dL, similar to that of the combined group.

• The outcome of these 3 studies is consistent. Wellnourished patients undergoing elective operations by well-trained, competent surgeons are unlikely to benefit from pre- or postoperative nutrition because the likelihood of malnutrition-related complications is very low.

• However, when patients with preexisting hypoalbuminemia undergo surgery on high-risk sites (the esophagus and pancreas), potential effectiveness of therapy on outcome was observed. These results are in concordance with randomized, prospective studies of trauma patients undergoing celiotomy and randomized to enteral, parenteral, or no feeding postoperatively25,26 and the VA cooperative study.2 Enteral nutrition primarily benefits the more severely injured patient population, and patients with lesser degrees of trauma require no specialized nutrition support25 because they recover with few, if any, complications. In the VA study, benefits from preoperative nutrition was only seen in the severely malnourished at-risk patients.

• Complex therapy with its associated risks administered to a patient who will not benefit from that therapy only exposes the patient to risk without hope of benefit

• Nutrition support had not been shown to affect mortality in clinical trials. The current study establishes that to do so, any prospective evaluation must be lim-FIG. 2. ited to patients at apparent risk of mortality. Patients undergoing gastrointestinal resection have a very lowmortality rate if their preoperative albumins are 3.25 g/dL, and their inclusion in a study design to examine nutrition and mortality only dilutes any potential to determine that answer. In particular, such a study should be restricted to patients with albumin levels2.25 g/dL, because mortality ranges between 21% and 31% in that select group. Although not a large patient population in any hospital system, it is a group readily identifiable using this simple preoperative test.If nutrition support reduced mortality by 50%, a total of 474 patients with albumins 3.25 g/dL (182 of 576 patients in our study) would be needed in preoperative nutrition arm or no nutrition arm to reach statistical significance at a power of 80%. By limiting inclusion to patients with an albumin 2.25 g/dL (104 of 556 or 19.7% of our patients), 160 patients would be required in each arm with a resultant (and probably unnecessary) p value of .002 if all patients completed the study. A study that included a disproportionate number of well-nourished patients (such as our entire patient population) requires 904 patients/group toachieve the same statistical significance, ie, a huge number is required for this determination.

• These observations have implications for both the individual clinician and healthcare systems. To compare the outcome of a surgeon whose practice is heavily weighted to more difficult (and probably more malnourished) patients needing surgery of the esophagus or pancreas to a practice limited to colorectal surgery is inappropriate because the basic risk for complications is very dissimilar between the 2 practices. Similarly, 2 hospital systems in the series had a higher complication rate compared with the private institutions, but these discrepancies completely disappeared when outcomes were stratified by the degree of hypoalbuminemia. Resource planning should assure that adequate resources are allocated if a system or practice is heavily burdened with a high-risk patient population.

• Postoperative stay (A), intensive care unit (ICU) stay (B), and nothing by mouth (NPO) days (C) remained relatively stable in the patients who recovered without complications, despite albumin level, except in the most hypoalbuminemic patients (open bars: patients with no complications; shaded bars: patients with complications).

• In patients with an albumin 3.25 g/dL, postoperative stay, ICU stay, and NPO days increased slightly with complications. As patients became more hypoalbuminemic, differences between patients with and without complications increased dramatically. The exception was the lowest albumin group who had short stays because of death after their complications.