SPINAL OR GENERAL ANAESTHESIA FOR SURGERY OF THE FRACTURED HIP?

Br. J. Anaesth. (1986), 58, 284-291

SPINAL OR GENERAL ANAESTHESIA FOR SURGERY OF THE FRACTURED HIP? A Prospective Study of Mortality in 578 Patients

N. VALENTIN, B. LOMHOLT, J. S. JENSEN, N. HEJGAARD AND S. KREINER

NIELS VALENTIN, MJJ., BJARNE LOMHOLT,* M.D. (Department of Anesthesia); JBRGEN STEEN JENSEN.T DR MED. S O

(Department of Orthopedic Surgery T-2); NIELS HEJGAARD.T M.D. (Department of Orthopedic Surgery T-3); Gentofte Hospital, D K 2900 Hellerup, Denmark. SVEND KKEINER, CAND. STAT., Department of Statistics, Herlev Hospital, DK 2730 Herlev, Denmark. Present addresses: •Department of Anesthesia, Hvidovre Hospital, DK 2650 Hvidovre, Denmark. fDepartment of Orthopedic Surgery U, Rigshospitalet, DK 2100 Copenhagen 0 , Denmark. Correspondence to N.V.

SUMMARY The mortality following surgical correction of upper femoral fractures was investigated in 578 patients, over the age of 50 yr, randomly allocated to receive spinal (bupivacaine) or general (enflurane or neurolept) anaesthesia. Thirty days after surgery the mortality was 6% after spinal and 8% after general anaesthesia (ns). Six months to 2 years after surgery the mortality was identical in the two groups. There were no differences with respect to ambulation and discharge. The estimated blood loss was smaller (P < 0.05) in patients receiving spinal anaesthesia. Regard/ess of the anaesthetic technique, a high short-term mortality was related to age, male sex, and trochanteric fracture, whereas excess long-term mortality was related to male sex and high ASA scores. PATIENTS AND METHODS

The investigation comprised 662 patients, 50 years of age or older, undergoing surgical repair of femoral neck or trochanteric fractures at the two departments of orthopaedic surgery at Gentofte Hospital between April 1980 and October 1982. Criteria of exclusion were: pathological fractures; neurological disorders of the lower extremities; previous fracture of the vertebral column with sequelae; anticoagulant therapy, including low-dose heparin treatment; psychiatric disorders contraindicating regional anaesthesia; and skin infection at or near the site of lumbar puncture. Provided no contraindications were present, the patient was included and randomly allocated to

Downloaded from http://bja.oxfordjournals.org/ at Michigan State University on June 18, 2015

Although poorly supported by the available literature, it is believed commonly that regional (spinal or extradural) anaesthesia carries a lesser risk than general anaesthesia. As far as we are aware, table I lists all the prospective, randomized studies on the mortality associated with surgical correction of upper femoral fractures carried out under regional or general anaesthesia. However, among the early studies, only that by McLaren, Stockwell and Reid (1978) demonstrated a significant decrease in short-term mortality with spinal anaesthesia. However, as the death rate following general anaesthesia was high (table I), particularly when compared with data from our institution (Jensen and Tandevold, 1979), the results of that particular study may not be generally applicable. Consequently, we decided to undertake an investigation of the mortality in patients undergoing the surgical correction of fractures of the proximal part of the femur under spinal or general anaesthesia. Furthermore, as no studies of long-term survival were available (before the paper by McKenzie, Wishart, and Smith, 1984) we included a similar assessment in the study.

SPINAL OR GENERAL ANAESTHESIA?

285

TABLE I. Mortality following surgery of the fractured hip. Only prospective, randomized studies are listed Source

Mean age (yr)

Observation

Anaesthesia

Mortality

General Extradural General Spinal General Spinal

12/501

86

3 months

McLaren, StockweU and Reid (1978) McLaren (1982)

76

1 month

75

1 month

80

1 month

Davis and Laurenson (1981)

80

1 month

Wickstrom, Holmberg and Stefansion (1982)

81

1 month

General Extradural

McKenzie, Wishart and Smith (1984)

74

2 weeks

General Spinal

2 months

General Spinal

McKenzie and colleagues (1980) White and Chappell (1980)

receive spinal or general anaesthesia. In all other respects the patients were treated according to the current practice in the departments involved. In most cases anaesthesia and surgery were undertaken by senior or junior registrars, general anaesthesia often by a nurse anaesthetist supervised by a registrar. Preoperative management

With the exception of patients with a stable fracture of the femoral neck which permitted weight-bearing at once, we consider patients with fracture of the hip to be emergencies which require operative management on the first possible occasion. Thus, most patients underwent their operation during the evening or night, any delay being caused by the presence of a full stomach, a need for resuscitation, or by the presence of patients with a higher priority. In a few patients, surgery was delayed for several days because of the need for more extensive preoperative treatment. Premedication

No premedication was given to poor-risk patients; in the others, premedication consisted of pethidine 25-75 mg, often supplemented with promethazine 12.5-25 mg, both administered i.m., approximately 30 min before the induction of anaesthesia.

General Spinal Psoas General Spinal

7/50 j 17/601 4/56/ 8/511 5/49/ 0/20 i 0/20 1/16) 9/681

< 0.005

3/64} 9/1371 2/32 / 12/75 j 3/73/ 14/75 1 14/73/

<0.05

Anaesthesia

Spinal anaesthesia. Following the administration of fentanyl 0.05-0.1 mg i.v., the patient was turned to the lateral position with concomitant traction of the fractured leg and, using a 22-gauge spinal needle, isobaric 0.5 % bupivacaine (3-)4 ml was injected to the subarachnoid space. After 5-15 min, analgesia had usually reached the lower thoracic segments, thus permitting the patient to be positioned for surgery. Oxygen (nasal cannulae) was supplied to many patients, as well as small doses of diazepam and fentanyl to decrease anxiety. Arterial hypotension, not readily restored by increasing the rate of an infusion of physiological saline, was treated with ephedrine in doses of 10 mg i.v. or 25 mg i.m. or both. General anaesthesia. The anaesthetists were free to choose one of two anaesthetic regimens: (a) enflurane and nitrous oxide in oxygen with or without thiopentone at induction or (b) neurolept anaesthesia with droperidol, fentanyl and nitrous oxide in oxygen. In both techniques the trachea was intubated following neuromuscular blockade induced with suxamethonium or, more rarely, with gallamine. All patients having neurolept anaesthesia, and the majority of the patients receiving enflurane received gallamine to provide neuromuscular blockade during the surgical procedure.

Downloaded from http://bja.oxfordjournals.org/ at Michigan State University on June 18, 2015

Couderc and colleagues (1977)

286 Surgery and the postoperative period

Data collection and statistical methods

Before the induction of anaesthesia the anaesthetist classified the patient according to the ASA grouping (Owens, Felts and Spitznagel, 1978). During surgery blood loss was estimated and, if significant, was measured by swab weighing. In each patient the times of admission, mobilization, walking and discharge were recorded. Through the National Danish Bureau of Personal Registration the survival of the patients was studied for 10 months after the inclusion of the last patient in the study. Data were analysed using standard methods for survival analysis based on the Cox model of proportional hazard (Cox, 1972; Kalbfleich and Prentice, 1980) as well as methods for analysis of contingency tables by log-linear and graphical models (Bishop, Fienberg and Holland, 1975; Edwards and Kreiner, 1983). In addition to standard likelihood ratio and x1 tests, hypotheses were tested by the gamma rank correlation for contingency tables by Goodman and Kruskal (1979) and the partial gamma coefficient of Davis (1967). Ethical aspects

It was decided not to obtain informed consent from the patients as to participation in the study because a large proportion of patients with hip fractures are old, and mentally unable to reach a logical decision. In addition, both of the anaesthetic regimens used were standard procedures, and it was stated in the plan of the investigation that the rejection of the selected method of anaesthesia by a patient should always be taken into account. Nine patients were excluded from the study for that reason. Moreover, human research committees were not

established in Denmark at the start of this investigation. RESULTS

During the study period, a minimum of 152 patients underwent surgery for hip fracture but were not included in the study, mainly as a result of the exclusion criteria (table II). The anaesthetist preferred to use a particular technique in 10 patients who did not fall into the criteria of exclusion. Of the 662 patients admitted to the study, 84 were later excluded by the authors for the reasons listed in table II, leaving data from 578 patients for consideration. Table III demonstrates that the two anaesthetic groups were comparable with respect to age, sex, type of fracture and surgical procedure. Classification according to the ASA criteria was applied to 508 patients: there were more patients in ASA group III in the general anaesthesia group, and TABLE II. The study sample Operated during the study period Criteria of exclusion "Private" contraindications Not included—reasons unknown Study entered by Erroneous admission Criteria of exclusion present Surgery cancelled Lost to follow-up Anaesthetic Protocol not followed Rejected by patient Not feasible Available for data processing

814 patients 84 10 58

152 patients 662 patients

4 13 3 2 19 9 34

84 patients 578 patients

TABLE III. Demographic and surgical data of patients

Age(yr) Median Quartiles Range Sex (M/F) Fracture Femoral neck Trochanteric Operation Internal fixation Hemiarthroplasty

Spinal

Genera]

79 72 85 50-100 58/223

79 72 85 53-94 59/238

129 152

128 169

173 108

192 105

Downloaded from http://bja.oxfordjournals.org/ at Michigan State University on June 18, 2015

The surgical procedures were internal fixation for trochanteric fractures and for femoral neck fractures in younger patients, whereas femoral neck fractures in disabled and in very old patients were treated with Moore's hemiarthroplasty. In the postoperative period, mobilization and weight bearing were attempted from the 2nd or 3rd day. Antiembolic stockings were provided, but anticoagulant therapy was not given routinely. Antibiotics were not used. The postoperative rehabilitation programme included chest physiotherapy.

BRITISH JOURNAL OF ANAESTHESIA

SPINAL OR GENERAL ANAESTHESIA?

287

TABLE IV. ASA classification, n = 508; P < 0.02

YA Spinal

30-

1 Healthy 71 (28%) 2 Mild systemic disease 121 (47%) 3 Severe systemic disease, not 42(17%) incapacitating 4 Incapacitating systemic disease, a constant threat to life 21 (8%) Totals 255

Spinal anaesthesia

General I

69(27%) 113(45%) 63(25%)

8(3%) 253

20-

: :

0

200

300 400 Blood loss (ml)

750

FIG. 1. Estimated blood loss during hip fracture surgery related to the anaesthetic regimen. P < 0.001.

undergoing spinal anaesthesia. Neither this difference nor the apparently different courses of the survival curves during the first 30 days were statistically significant. The 95 % confidence limits of the difference in mortality were±4%, meaning that there is a probability of 95 % that the gain in survival from using spinal instead of

100

• - - - • Spinal anaesthesia • — • General anaesthesia

10 15 20 Time after surgery (days)

25

FIG. 2. Survival during the first 30 days after hip fracture surgery related to the anaesthetic regimen. P > 0.05.

Downloaded from http://bja.oxfordjournals.org/ at Michigan State University on June 18, 2015

more of group IV in the spinal anaesthesia group (P < 0.02) (table IV). A correction for this uneven distribution was made when comparing the mortality rates. With the possible exception of ASA group IV, higher ASA scores were related to greater age (P < 0.005). Internal fixation was a more time-consuming procedure than arthroplasty (51% and 35%, respectively, of cases lasting more than 60 min (P < 0.005)), but there was no difference between the two anaesthetic groups in this respect. Figure 1 demonstrates that the estimated blood loss—when recorded—was less during spinal than during general anaesthesia (P < 0.001). Bleeding was more pronounced in procedures of long duration (P < 0.005), and thus, indirectly, in patients requiring internal fixation. Short-term survival is shown in figure 2. At 30 days the mortality in the general anaesthesia group was 8.0%, compared with 6.0% in those patients

I General anaesthesia

288

BRITISH JOURNAL OF ANAESTHESIA

6

9

12

15

18

21

24

DISCUSSION

Time after surgery (months) FIG. 3. Survival during the first 2 yr after hip fracture surgery related to the anaesthetic regimen. P > 0.05.

general anaesthesia was between +6 and —2%. There was a significantly (P < 0.05) lower mortality among females (4.7% v. 9.1 % in males) and among patients with femoral neck fractures (4.7 % v.9.1% among patients with trochanteric fractures (P < 0.05)). In addition, mortality increased with increasing age (P < 0.005). The technique of general anaesthesia had no influence on mortality.

100

a 4 o •

Expected - Combined

3

6

9

12

15

18

ASAI ASA II ASA III ASA IV

21

24

Time after surgery (months)

FIG. 4. Survival during the first 2 yr after hip fracture surgery related to the preoperative ASA classification. P < 0.0005.

The number of patients not admitted to or excluded from the study may seem large. The criteria of exclusion were very restrictive; however, with the exception of "pathological fractures" (15 patients), the conditions contraindicating entrance to the study were not life-threatening, and there is hardly any reason to believe that these contraindications have markedly affected the results of the study. As stated in table II, 34 patients were excluded because the method of anaesthesia was "not feasible". With a few exceptions this term covers

100

Expected Combined

3

6

9

12

15

o Females • Males

18

21

24

Time after surgery (months) FIG. 5. Survival during the first 2 yr after hip fracture surgery related to sex. P < 0.02.

Downloaded from http://bja.oxfordjournals.org/ at Michigan State University on June 18, 2015

Figure 3 demonstrates the long-term survival related to anaesthetic regimen: there was no difference between the two anaesthetic groups. The annual mortality in a sample of the Danish population matching the study sample with respect to age and sex is very close to 10%. An increase in mortality was found for about 9 months after the operation (fig. 4). The increase in long-term mortality was related to ASA scores (P < 0.0005), and sex (P < 0.02) (figs 4 and 5). Information about the postoperative course is given in table V. It is evident that the two anaesthetic groups did not differ with respect to times of ambulation or discharge.

Spinal anaesthesia General anaesthesia

SPINAL OR GENERAL ANAESTHESIA?

289

TABLE V. Duration (days) from surgery to ambulation (chair or walking) and discharge. * Never left bed: 30 patients; -f never walked: 55 patients; %diedin hospital: 25 patients

Ambulation (chair)* Spinal General Ambulation (walking)t Spinal General Discharge^ Spinal General

Median

Range

2 2

0-64 0-33

1 1

2 3

3 3

0-66 0-34

2 2

5 6

10 11

1-158 1-368

4 4

19 19

blockade (28%). A better preservation of blood flow to the lower limb during spinal anaesthesia and a reduced risk of thromboembolism was suggested as the explanation of the lower mortality in this group of patients. Similar results were reported by McKenzie, Wishart and Smith (1984) who found a mortality of 4.1% 2 weeks after surgery for hip fracture under spinal anaesthesia compared with 16% mortality when general anaesthesia (spontaneous respiration with halothane and nitrous oxide in oxygen following induction of anaesthesia with Althesin) was used. However, 8 weeks after surgery the mortality was close to 20% in both groups, and 1 year after surgery both groups had been decreased by approximately one-third. There is no obvious single feature to explain why we were not able to reproduce the findings of the two Glasgow studies. The most striking difference between these studies and ours is the much lower overall mortality among our patients. Only a small part of this difference is explained by the fact that we included patients undergoing hemiarthroplastic procedures. The omission of these patients from our study would have caused an overall short-term and long-term mortality approximately 2 % higher. The ASA classification of the patients of McKenzie, Wishart and Smith (1984) was not stated, whereas the patients of McLaren, Stockwell and Reid (1978) were estimated to be in a poorer condition than our patients, about 75 % belonging to ASA group III or IV. It is not likely that the entire difference from our ASA rating (24% belonging to ASA groups III and IV) can be attributed to inter-observer differences in rating (Owens, Felts and Spitznagel, 1978). However, the annual background mortality in Glasgow (9 %

Downloaded from http://bja.oxfordjournals.org/ at Michigan State University on June 18, 2015

difficulties in performing lumbar puncture. Although a 10% failure rate is rather high, it should be remembered that the majority of these anaesthetics was administered during the evening and night by junior anaesthetists. Compared with other studies, our patients appear to be typical in regard to age (table I) and sex distribution (Fitts et al., 1959; Manpel, Marzulli and Boley, 1961; Colbert and O'Muircheartaigh, 1976; Jensen and Tendevold, 1979; Davis and Laurenson, 1981). They were evenly distributed between the two anaesthetic groups with respect to age, sex and type of fracture and operation (table III). As mentioned, the skew distribution with respect to ASA groups was corrected during the calculation of anaesthesiarelated mortality. Although the data from the first postoperative weeks suggested a difference in mortality between the two anaesthetic groups (fig. 2), we did not find support for our hypothesis that spinal anaesthesia would carry a lower surgical mortality than general anaesthesia in patients with upper femoral fractures. Therefore, it is hardly surprising that the long-term survival was similar in the two groups (fig. 3). In most of the prospective, randomized studies listed in table I, general and regional anaesthesia were not found to be significantly different with respect to survival. An objection common to these studies is the small size of the sample and, thus, a high risk of committing type 2 errors. In contrast to these studies, McLaren, Stockwell and Reid (1978) and McLaren (1982) reported a considerably lower mortality 4 weeks after surgery of the fractured hip under spinal anaesthesia with Althesin sedation (7%) than following general anaesthesia with Althesin, nitrous oxide in oxygen and neuromuscular

Quartiles

290

mortality during the first year has been reduced from 27% to 19%, suggesting the possibility of a beneficial effect of early ambulation and discharge. In conclusion, the existing evidence indicates that, under certain conditions, short-term survival may be better when surgical correction is performed under spinal anaesthesia. However, when general anaesthesia has a low mortality, spinal anaesthesia will probably not offer additional advantages. In all circumstances a reduction of short-term mortality by spinal anaesthesia is of limited interest as there are no indications of an increased long-term survival. At present, regional anaesthesia seems superior to general anaesthesia with respect to the occurrence of deep vein thrombosis (Davis, Quince and Laurenson, 1980; Hendolin, Mattila and Poikolainen, 1981; Modig et al., 1983) and the amount of blood loss (Davis and Laurenson, 1981; Hendolin, Mattila and Poikolainen, 1981; Modig et al., 1983), although only in some types of surgery. In patients with recent myocardial infarction, Reiz and colleagues (1982) found a lower incidence of myocardial ischaemia and possibly a lower re-infarction rate with thoracic extradural plus light general anaesthesia than with neurolept anaesthesia. In addition, it has been found that regional anaesthesia suppresses some endocrine (review by Kehlet, 1984) and immunological responses to surgical trauma (Rem, Brandt and Kehlet, 1980; Hole and Unsgaard, 1983), but—as previously stated—a possible beneficial effect in terms of survival or recovery, or both, remains to be proven. The fact that differences in survival following spinal or general anaesthesia could not be verified—or ruled out—in the present study comprising almost 600 patients, does not encourage further studies of this type. Future investigations of relative anaesthetic risk should probably focus on well-defined risk groups and study indices which are more sensitive than mortality. ACKNOWLEDGEMENTS We thank Max og Anna Friedmanns Legat, Dansk kirurgisk selskabs Jubilsumsfond, Astra Ltd, Abbott Laboratories Ltd and Janssenphaitna Ltd for financial support. Secretarial assistance, including invaluable help in the administration of the large amount of data, has been given by Susanne Dahlberg. REFERENCES Bishop, Y. M. M., Fienberg, S. E., and Holland, P. W. (1975). Discrete Multivariate Analysis. Cambridge: MIT Press.

Downloaded from http://bja.oxfordjournals.org/ at Michigan State University on June 18, 2015

according to McKenzie, Wishart and Smith (1984)) is very similar to the background mortality of 10% in our area. Thus, any real difference in ASA scoring would depend on the acute condition rather than pre-existing disability, suggesting the possibility of improving the ASA scoring by preoperative resuscitation. It should be added that anaesthesia-related differences in mortality were not found in any ASA class. It is important that a low initial mortality is not lost during the subsequent months. In our group of patients there was no increase in mortality after approximately 9 months, and the difference in mortality at 2 months between the present material and the study of McKenzie, Wishart and Smith (1984) was not only maintained throughout the first year, but increased by about 4 %. The observation that short-term mortality depends on age and sex is in accordance with previous work, including the paper by Jensen and Tondevold (1979) who studied 1592 patients with hip fracture admitted to our hospital during the period 1971-77. The present study also showed that trochanteric fracture per se had a higher mortality than fracture of the femoral neck. Whereas Jensen and Tandevold attributed this difference to the greater age of the former patients, a recent refinement of the statistical methods is probably responsible for the different conclusions. A longer duration of osteosynthesis was sufficient to explain the greater amount of bleeding during that operation when compared with hemiarthroplasty. In accordance with Davis and Laurenson (1981) and with authors studying other types of surgery (Hendolin, Mattila and Poikolainen, 1981; Modig et al., 1983) we also found less blood loss with regional than with general anaesthesia, possibly as a result of the lower arterial pressure. Since it was often noted by the ward staff that the patients were doing better after spinal anaesthesia, we had expected to find earlier ambulation and discharge in that group. However, as was the case in the study of McKenzie, Wishart and Smith (1984), the two anaesthetic groups were similar in this respect. According to table V, the patients of the present study had a much shorter period in hospital (10—11 days) than the patients of McKenzie, Wishart and Smith (1984) (approximately 40 days). Since the investigation of Jensen and Tondevold (1979) in patients treated at Gentofte Hospital between 1971 and 1977, the duration of the stay in hospital has been more than halved. At the same time,

BRITISH JOURNAL OF ANAESTHESIA

SPINAL OR GENERAL ANAESTHESIA?

metabolic response to surgery. Update and perspectives. Acta Anaesthesiol. Scand., 28, 125. Manpel, J., MarzuUi, V., and Boley, S. J. (1961). The fractured hip, a complication of aging. Arch. Surg., 82, 474. McKenzie, P. J., Wishart, H. Y., Dewar, K. M. S., Gray, J., and Smith, G. (1980). Comparison of the effects of spinal anaesthesia and general anaesthesia on postoperative oxygenation and perioperative mortality. Br.J. Anaesth., 52,49. Smith, G. (1984). Long-term outcome after repair of the fractured neck of femur. Br. J. Anaesth., 56, 581. McLaren, A. D. (1982). Mortality studies. A review. Reg. Anaesth. (Suppl.), 7, S172. StockweU, M. C , and Reid, V. C. (1978). Anaesthetic techniques for surgical correction of fractured neck of femur. Anaesthesia, 33, 10. Modig, J., Borg, T., Karlstrom, G., Maripuu, E., and Sahlstedt, B. (1983). Thromboembolism after total hip replacement: Role of epidural and general anesthesia. Anesth. Analg., 62, 174. Owens, W. D., Felts, J. A., and Spitznagel, E. L. (1978). ASA physical status classifications: A study of consistency of ratings. Anesthesiology, 49, 239. Reiz, S., Balfors, E., Sorensen, M. B., Higgmark, S., and Nyhman, H. (1982). Coronary hemodynamic effects of general anesthesia and surgery. Modification by epidural analgesia in patients with ischemic heart disease. Reg. Anesth. (Suppl.), 7, 58. Rem, J., Brandt, M. R., and Kehlet, H. (1980). Prevention of postoperative lymphopenia and granulocytosis by epidural analgesia. Lancet, 1, 283. White, J. W. C , and Chappell, W. A. (1980). Anaesthesia for surgical correction of fractured femoral neck. A comparison of three techniques. Anaesthesia, 35, 1107. Wickstrom, I., Holmberg, I., and Stefansson, T. (1982). Survival of female geriatric patients after hip fracture surgery. A comparison of 5 anaesthetic methods. Acta Anaesthesiol. Scand., 26, 607.

Downloaded from http://bja.oxfordjournals.org/ at Michigan State University on June 18, 2015

Colbert, D. S., and O'Muircheartaigh, I. (1976). Mortality after hip fracture and assessment of some contributory factors. Irish J. Mtd. Set., 145, 44. Couderc, E., Mange, F., Duvaldestin, P., and Desmonts, J. M. (1977). Resultats comparatifs de l'anesthesie generale et peridurale chez le grand vieillard dans la chirurgie de la hauche. Anesth. Analg. Rian., 34, 987. Cox, D. R. (1972). Regression models and life tables. J. Roy. Stal. Soc. B, 30, 248. Davis, J. A. A. (1967). A partial coefficient for Goodman and Kruskal's gamma. J. Am. Stat. Assoc., 62, 189. Davis, F. M., and Laurenson, V. G. (1981). Spinal anaesthesia or general anaesthesia for emergency hip surgery in elderly patients. Anaesth. Intern. Care, 9, 352. Quince, M., and Laurenson, V. G. (1980). Deep vein thrombosis and anaesthetic technique in emergency hip surgery. Br. Med. J., 281, 1528. Edwards, D., and Kreiner, S. (1983). The analysis of contingency tables by graphical models. Biometrika, 70,553. Fitts, W. T. jr, Lehr, H. B., Scior, S., and Roberts, B. (1959). Life expectancy after fracture of the hip. Surg. Gynecol. Obstet., 108, 7. Goodman, L. A., and Kruskal, W. H. (1979). Measures of Association Cross Classifications. New York: Springer Verlag. Hendolin, H., Manila, M. A. K., and Poikolainen, E. (1981). The effect of lumbar epidural analgesia on the development of deep vein thrombosis of the legs after open prostatectomy. Acta Chir. Scand., 147, 425. Hole, A., and Unsgaard, G. (1983). The effect of epidural and general anaesthesia on lymphocyte functions during and after major orthopaedic surgery. Acta Anaesthesiol. Scand., 27, 135. Jensen, J. S., and Tandevold, E. (1979). Mortality after hip fractures. Acta Orthop. Scand., 50, 161. Kalbfleich, J. D., and Prentice, R. L. (1980). The Statistical Analysis of Failure Time Data. New York: Wiley. Kehlet, H. (1984). Epidural analgesia and the endocrine-

291