- Email: [email protected]
Treatment of envenomation by Echis coloratus (mid-east saw scaled viper): A decision tree
Tosicon Vol. 27, No . 10, pQ. 1101112, 19X9. Printed in Great Britin.
0041-0IOIj89 33 .00+ .00 ® 1989 Pergemon Prar pk
TREATMENT OF ENVENOMATION BY ECHIS COLORATUS (MID-EAST SAW SCALED VIPER) : A DECISION TREE Dwx
Gn.ox, ODED Sx~r ev and JocHexnx BENBA33AT
Department of Medicine, Hat9assah University Hospital, Mt Scopus, Jerusalem, Israel (Accepted jar publication 7 Aprit 1989)
Gn.ox, O. SxnL.sv and J. BENBASSAT, Treatment of envenomation by Echls coloratus (mid-east saw scaled viper) : a decision tree. Toxicon 27, 1101112, 1989.-Envenomation by Echis coloratus causes a transient hemostatic failure. Systemic symptoms, hypotension and evident bleeding are rare, with only one reported fatality. In this paper, we examine the decision to treat victims of Echis coloratus by a specific horse antiserum. The decision model considers the mortality of treated and untreated envenomation, and the side effects of antiserum treatment: fatal anaphylaxis, serum sickness and increased risk of death after a possible repeated exposure to horse antiserum in the future . The results of the analysis are not sensitive to variations in the probability of side effects of antiserum treatment. They are sensitive to variations in the risk of bleeding after envenomation, in the degree of reduction of this risk by antiserum treatment and in the risk of dying after an event of bleeding . Prompt administration of antiserum appears to be the treatment of choice if it reduces the risk of bleeding from 23.6% to 20.3% and if 1 .6% or more of the bleeding events are fatal. We conclude that presently available data support antiserum treatment of victims of Echis coloratus who present with hemostatic failure, even though the advantage imparted by this treatment appears to be small. D.
INTRODUCTION
Echis coloratus envenomation causes a transient hemostatic failure with a prolonged clotting time, hypofibrinogenemia and increased circulating fibrinogen split products (Racxrnc et al., 1962; F~vaRU et al., 1970, 1974; Dvu.nxsxi and Bmnx, 1973; YATZIV et al., 1974). Pain, bleeding or necrosis at the site of the snake bite are of variable severity . Hypotension and bleeding from other sites are rare, and there is only one death which has been reported by Mnxx (1978) and in the series of Scxur cHYxsxn-C~srEt et al. (1986) . A specific antiserum is available, however, indications for its use are uncertain. The proponents of antiserum therapy argue that it hastens hemostatic recovery (ScxuLcxxrtsxn-C~srEI, et al., 1986), that the only known fatality was not treated by antiserum (SextrLCt-nrrrstcw-Cns'rat, et al., 1986) and that envenomation by related spades, e.g. Echis carinatus, causes a severe hemostatie failure, which, if untreated by antiserum, is associated with a 7-15% mortality (WAItREL et al., 1974, 1976). On the other hand, some clinicians advocate no treatment, or treatment with heparin or antiserum only in cases with severe bleeding (F~nvnRU et al., 1974; Mnxx, 1978). They argue that the hemostatic 1105
1106
D . GILON et aJ.
failure after Echis coloratus envenomation is self-limited, much less severe than that after envenomation by E. carittatus, and is commonly followed by an uncomplicated recovery . Antiserum treatment carries a risk of anaphylaxis (MAr nsIT et al., 1986) and serum sickness (R», 1980), and may increase the risk of anaphylaxis if the patient is treated again with antiserum in the future . What is then the preferred management of Echis coloratus envenomation7 The rarity of this envenomation precludes randomized prospective trials of alternative treatments. Therefore, despite their potential inaccuracies, non-experimental methods are worth exploring. In this paper we analyse the trade-off between the risk of bleeding and death in untreated patients, and the side effects of antiserum treatment; anaphylaxis, serum sickness and an increased risk of dying after a possible repeated exposure to horse antiserum. METHODS We compared the possible outcomes of Echts coloratus envenomation, with and without antiserum treatment, in 8 previously healthy 18-year-old person, who presented with indefinite clotting time and no evident bleeding . Standard decision analysis methods were used (WmvsT~r and FnveaEtta, 1980), and the calculations were performed with the Smltree computer program (Pratt Medical Group, 1987). Sensitivity analyses were carried out to test the stability of our conclusions to variations in the probability estimates and in the assigned utilities. (aJ The decision tree The decision tree (Fig. 1) is divided into two parts, representing possible immediate (part A) and late (part B) events after envenomation . With antiserum therapy (Fig . lA, top branch) patients will either develop anaphylaxis or not, and those who do may die. Those who survive will either bleed or not, and those who do may die. Those who survive may or may not have serum sickness . With expectant management (Fig . lA, second branch) patients may bleed or not, and those who bleed may die . Within a time horizon, set arbitrarily at 25 years after recovery from envenomation, the patient may sustain a second envenomation by the same or other snakes and be treated again with horse antiserum (Fig. lB, top branch). Conceivably, this repeated treatment may carry a still higher risk of analphyhuris and death . Subjects who do not sustain a second envenomation during the first year after recovery (Fig . 1B, second branch) may die from other causes during the same year or survive, and cycle again through the risk of a second envenomation during the next year. The triangular Boolean node in Fig. 1 permits recycling for a predetermined number of years. It was asenmed that those who survive 25 years will live through their life expectancy. It was assumed that bleeding may result in either death or temporary morbidity, but no permanent morbidity . This assumption is incorrect : half of the survivors of intracerebral haemorrhage have some residual morbidity (Dot.r.aEatâ et al., 1986; C~te~mes et al., 1987) . This erroneous assumption was deliberately made in order to bias the model against antiserum treatment. We assumed also, that similarly to victims of envcnomation by E. carinatmr (Watexec et al., 1974), the severity of swelling, necrosis or permanent injury at the site of Echis coloratus envenomation are not affected by antiserum treatment. (b) Probabilities The analysis requires estimates of the probability of anaphylaxis, death and serum sickness after antiserum treatment, of death after recurrent antiserum treatment, of bleeding after treated and untreated envenomation, and of death after an event of bleeding . Arwphylaxis, death and senare sickness ajter antiserum treatment The incidence of immediate reactions to antiserum treatment ie related to the degree of antiserum purification, its dose and rate of injection (Its, 1980 ; Btr\c[ and Gutav, 1980) . Anaphylaxis occurs in 2~% of the recipients (Table 1). Fatal anaphylaxis is rare; there were no deaths among the cases surveyed in Table 1, nor among 4000 persons treated with antiserum for snakebite between 1962 and 1977 (Sur~et.eNn, 1977) . On the other band, one death (0.025%) has been attributed to antiserum from a total of 4000 persona at risk (Ram, 1980), and another three deaths (0.09%)occurred after anaphylaxis in 3300 recipients of antiserum for snake envenomation (CtnttslutvsErt, 1981). For purposes of the analysis we estimated the probability of anaphylaxis as 3 .4%,and the mortality after in-hospital anaphylaxis as 0.05% of all patients treated by antiserum, or 1 .6% of the cases with anaphylaxis .
Treatment of Envenomation by Schis coloratw
1107
IIRRnIruxLs
IIRi15ERUl
4
Rx
RLEEDIRR EYERT
SURr1Y~L
3°RYIY~L
SERVI SIWIESS
RO RLEEDIIIG EYEIIT
Ro uuPxruxls
OFATN RLEEDIIi EYERT EIPECTAIII IYIIYICEREIR
f
9
SI"Y1YI1L
RO RLEEDIRR EYEIIT
ENVENONATION
EVER
hROBABII.ITY TREE FOR TREATI~N'r OF ENVENOMATION HY Echis coloratus . The tree is subdivided into two parts indicating possible events occurring shortly after envenomation (bleeding, anaphylaxis, serum sickness ; A), and during the 25 years following recovery (second envenomation necessitating antiserum treatment, death after antiserum treatment or death from other causes ; B) . The square decision node (left, A) represents the choice . The two branches emanating from it represent the two therapeutic strategies: treatment with antiserum and no antiserum treatment. Circular nodes represent chance events which may follow each choice . The triangular (Boolean) node in part B indicates that either of the two branches may be followed, depending on the number of preceding cycles . Rectangular nodes represent possible outcomes. FIG .
1.
The incidence of serum sickness after treatment for snake envenomation has been reported as 5.5-14.3% (Table 1). In the present analysis it was assumed to be 8% . This estimate is much lower than the 80% reported for U.S. practice (ELLEIUtoRN and BARCEI .oux, 1988) . We have no ready explanation for these differences. Therefore, the impact of higher estimates of the probability of serum sickness was tested by sensitivity analysis. Risk of recurrent antiserran treatment The mean annual number of snake bites in Israel was 111 .9 during 1969-79 and 109.1 during 1979-87 (Epidemiologic Reports of the Ministry of Health of the State of Israel, 1984) in a population at risk aged 15-60
1108
D. GILON et al. TAHLe 1 . INCmsNCe oF ANAPxY~Ax>s AND sFxtrM stcaNSSS AF~e rxEATS~Nr BY xoasE AN~tt~ At risk Anaphylaxis 90% confidence n n (%) limits (%)
Reference SCHULCHYNSICA-CASret . e t al. (1986) MAt.w$T et al. (1986) RErD (1980) TxtNCA (1963) Svrr~tt,AND and LovewNa (1979) Annual Reports (1984) Total
28 35 547 100 203 151 1064
1 2 20 2 7 4 36
(3 .6) (5.7) (3.7) (2.0) (3 .4) (2.6) (3.4)
0-9.3 0-12 .2 2.4-5 .0 0-4.3 1 .3-5 .6 1 .3-3 .9 2.5 .3
Serum sickness n (%)
90% confidence 11mltS (%)
4 (14.3)
3.4-25.2
59 6 10 1 80
(10.8) (6 .0) (S .5) (0 .7) (7 .9)
8.6-13.0 2.1-9.9 2.7-8 .2 0-1.7 6.5-9.3
of roughly 1,800,000 . Therefore, the annual risk of snake bite was estimated as 110/1,800,000, or 0.006% . This estimate is probably correct for a person with an incidental exposure to snakes, but not for snake collectors or employees of snake farms. The impact of higher estimates of the risk of a second exposure to horse antiserum was tested by sensitivity analysis . We know of no estimates of the risk of anaphylaxis after repeated antiserum treatment. The average subjective consensus of three experts in emergency medicine was that a previous exposure to antiserum triples the risk of anaphylaxis after a second treatment, and doubles the risk of death. Therefore, the risk of anaphylaxis after a recurrent exposure to antiserum was estimated as 10.2% with a mortality of 3.2% of ell cases with anaphylaxis, yielding an overall mortality after recurrent antiserum treatment of 0.33% . Although not every envenomation is judged to warrant antiserum therapy, we assumed that all subjects who will sustain a second envenomation will be treated by horse antiserum, thus biasing the model still further against antiserum therapy. Risk of bleeding and death after envenomation The main problem of this analysis was the lack of reliable estimates of the risk of bleeding and death in treated and untreated victims of Echis coloratus . There are no experiments, in which similar cases were either treated by a specific antiserum or not, and then followed in order to determine their outcome in terms of frequency of bleeding and survival . Our estimates were based on the uncontrolled retrospective observations by ScxaLCEnrN.rtcA-CAS~ret. et al. (1986) on 42 patients, 28 of whom were treated by antiserum. The criteria for patient selection for treatment are unknown. It is conoeivabk, that antiserum was given to more severely sick patients, to those with more prolonged hemostatic failure, or to those who bled. Subsequently, conclusions about the eSrcacy of therapy weld be biased against antiserum tt~eatment . SCxv[,cxYNSxA-CASret, et al. (1986) reported that the hemostatic failure lasted on the average for 87 hr in untreated patients, and 75 hr in patients treated by antiserum (Table 2) . When the three patients who received antiserum only on the fifth day after envenomation were excluded, the average duration of the hemostatic failure in antiserum-treated patients appeared to be only 59 hr. SCHCTLCFIYN3KA-CASrEr. et al. (1986) reported a total 3318 patient-hours of hemostatic failure, with nine bleeding events (two from the gastrointestinal tract, five from the mucosae, one in muscles and one macroscopic hematuria), or one per 369 hr of hemostatic failure (Table 2) . In nine other patients the hemoglobin level declined TABIE 2. Rrsx OP BIEEDINO AND DiJRAT10N OF I~IIOBTAIIC FAII.URE AP1IIt 1SiVHAiO~tATION HY EChis At risk
Treatment Eckas coloratus No antiserum Antiserum Total
n
Pt-hr of hemostatic failure
Events of bleeding
n
Per hr hemostatic failure
(Scxur .CHYNSeA-CASrEt, et al., 1986) 14 1218 28 2100 42 3318 9"-18t 1 :369 "-1 :184t
Echis carinates (WAxxet, et al., 1976) No antiserum 14 Antiserum Total 14 348
"Minimal estimate; tmaximal estimate .
8
1 :43.5
Average duration of hemoatatic failure per patient (hr)
Risk of blending (%)
87 75t-59 "
24"~7t 20"-32t
Not reported 25
57
1109
T~eatment of Envenomation by Echis coloratus
by 2-5 g/dl . If this decline was due to undetected hemorrhage, then the number of blceding episodes would increase to 18, with a fi^equency of I per 184 hr of hemostatic failure. Assuming one bleeding event per 369hr hemoatatic failure, and assuming that the number of bleeding events correlates with the duration of hemostatic failure, then the risk of bleeding may be calculated as 75/369-0.203, or 20 .3% in treated patients (for an average duration of hemostatic failure of 75 hr) and 23 .6% in untreated patients (for an average duration of hemostatic failure of 87 hr). Assuming one bleeding event per 184 hr hemoatatic failure and an average duration of hemostatic failure of 59 hr in treated patients, the risk of bleeding could be as high as 47.3% in the untreated group and 32 .1% in treated patients. We used as baseline values the lower estimates of the risk of bleeding in untreated envenomation and the lower estimate of the degree of reduction of his risk by antiserum treatment in order to bias the analysis still further against antiserum treatment. The risk of dying after an event of bleeding was estimated also on the basis of the observations by ScHUL.cHVNacA-CA9'L'EI, et al. (1986) who reported one death (I1 .1%) among the nine bleeding events (90% confidence intervals: 0-28 .4%) . It is interesting to note that this estimate is quite similar to the 0 to 33% (average-10%) mortality after an event of bleeding in patients on anticoagulant treatment (Levtxe et al., 1986). (c) Utilities Utilities, i.e . the value of each alternative outcome, were expressed in years of quality-adjusted life expectancy, obtained by subtracting from the life expectancy the estimated period of time spent off regular life routine because of complications of bleeding or antiserum therapy. For serum sickness, 0.08 years were substracted from the life expectancy; for bleeding -0.25 years; and for non-fatal anaphylaxis -0 .04. The utility of each outcome was expressed as the sum of (a) time interval from envenomation to the end of the follow-up, (b) years of expected survival after the end of the follow-up period, as derived from life-tables, and (c) adjustment for the morbidity resulting from bleeding, non-fatal anaphylaxis and serum sickness . Deaths following bleeding or anaphylaxis were given a utility of 0. Delayed deaths (i .e . deaths from other causes, or deaths after a second envenomation) were given a utility equal to the sum of (a) and (c). (dJ Summary ojprobability estimates Md SOf?CGJ Ojbias We deliberately biased the analysis against antiserum treatment by assuming : (a) that there is no long term morbidity after a non-fatal bleeding, (b) that the treated and untreated groups of patients in the non-randomized retrospective study of ScHULCxYNS6w-CASTEL et al. (1986) were comparable, (c) that every subject sustaining a second envenomation will be treated by antiserum, and (d) a low base line probability of bleeding after envenomation . These biases ensure that we did not miss a possible advantage of witholding antiserum treatment. The probability estimates used in the present analysis are summarized in Table 3. R FCLTLTS
When the baseline values were used to evaluate the two strategies for management of envenomation by Echis coloratus, the results favored antiserum treatment. The quality TABLE
3.
SUMMARY OP THE BASELINE PROBAHILIl'Y VALUES U~D IN THIS ANALYSIS
Probability of Anaphylaxis after administration of horse serum Mortality after in-hospital anaphylaxis Bleeding event after untreated Echis coloratus envenomation Bleeding event after Echis coloratus envenomation treated by antiserum Death after an event of bleeding Serum sickness after treatment with antiserum Recurrent envenomation (annual) Death after recurrent treatment with antiserum
%
RSII$e Of 8eilalüvlty 8n81yB18 (%)
3.4 1 .6
0-22 0-11
23 .6
23.6~7.2
20 .3 11 .1
16.0-32.0 0-10
8.0 0.006 0.33
2-100 0.006-0 .6 0-100
111 0
D. GILON
tt
al.
adjusted life expectancies in years were 55.96 for antiserum treatment, and 55.78 for expectant management, as compared to a 56.3 years life expectancy of an 18-year-old man. Immediate treatment with antiserum added 0.18 yr (0.33%) to the average life expectancy over the strategy of expectant therapy. The results of the sensitivity analyses indicated that the decision was not sensitive to changes in the probability of serum sickness, of second envenomation, or of death after recurrent antiserum treatment. Antiserum treatment was the preferred management even assuming an up to 22% probability of anaphylaxis, an up to 11 % mortality after anaphylaxis, a 100% probability of serum sickness, or a 100% probability of death after recurrent antiserum treatment. Also, results of the analysis were not sensitive to variations in the adjustments of the utilities of patients after bleeding and serum sickness . Antiserum treatment was the preferred management, even after subtracting from the life expectancy up to 2 years for serum sickness or up to 5 years for non-fatal anaphylaxis. The analysis was sensitive to variations in the probability of bleeding with and without antiserum treatment, and the mortality after an event of bleeding . Figure 2 is a three-way analysis of the relation between the probability of bleeding in treated and untreated envenomation and the probability of death after bleeding . The results suggest a superior utility of antiserum treatment if it is assumed that antiserum reduces the risk of bleeding from 23.6% to 20.3%,and that 1 .6% or more of the bleeding episodes will result in death. DISCUSSION
The main conclusion of this analysis is that the complications of antiserum treatment (anaphylaxis, serum sickness, postulated increase in risk after a second exposure to antiserum) have a marginal weight in the decision about the treatment of Echis coloratus envenomation . This decision depends predominantly on the probability of bleeding with and without antiserum treatment, and on the mortality after a bleeding event. The review of the literature supports the impression that the consequences of envenomation by Echis 0 .8
dc b
z
é is
G .2
0 .0 ' 0 .0
FYG . 2.
0.2 0 .4 0 .8 PROBABILITY OF BLEEDING WfTHOUT TREATMENT
0 .8
THam-WAY ANALYSE OF THS CHOI(~ HErWEEPi H~®IATE TRFAT1r~Mr WITH AN1'6EaUlI AND H7CPECrANT ètANAGElC~Nr OF ENVIIQOMATION HY ECIIlS COIOrOtYJ.
The analysis examines the relation between the probability of bleeding in treated (vertical axis) and untreated (horizontal axis) patients and the probability of death after a bleeding event (line a, 0% ; line b, 1% ;line c, 3%;line d, 10%) . The shaded area covers clinically relevant areas . Assuming a zero probability of death after an event of bleeding, the analysis favors no antiserum treatment . Antiserum treatment is preferred for a probability of death exceeding 1 .6% of all events of bleeding .
Treatment of Envenomation by Echis caloratus
111 1
coloratus are less severe than those of envenomation by E. carinates. The data reported by WARREL et al. (1974) imply a 57% risk of bleeding in treated Echis carinates envenomation, while the data by $CHULCHYNSKA-CasTal, et al. (1986) imply a maximal estimate of 47% risk of bleeding in untreated Echis colorates envenomation (Table 2). The differences may be related to differences in venom toxicity, or in the characteristics of the population at risk . The results of our analysis support the recommendation of Scxul.cxnvsxA-CASTEL et al. (1986) to treat Echis colorates envenomation with a specific antiserum, even though the advantage ofthis treatment is probably small. Even assuming that the probability of death is only 3% of all bleeding events, antiserum treatment of 100,000 victims would result in 610 fatal bleeding events, 54 deaths from anaphylaxis, 0.5 deaths after recurrent treatment with antiserum, and in 7815 cases of serum sickness . On the other hand, witholding treatment would result in 708 fatal bleeding events and in 0.06 deaths after treatment for second envenomation . Under these assumptions, antiserum treatment would be the preferred strategy even for an up to 0.6% annual risk of second envenomation, i.e. 100 times higher than our baseline estimate . The uncertainty regarding the treatment of Echis coloratus envenomation is due to the lack of precise estimates of the risk of bleeding and death after a bleeding event. Furthermore, there are no data regarding the efficacy of other modes of treatment, e.g. replacement therapy by fresh frozen plasma and procoagulants. In the absence of data, clinical decisions must be based on assumption . We feel that the recommendation of antiserum treatment for every case of Echis coloratus envenomation with hemostatic failure is based on reasonable assumptions, which are consistent with presently available observations . REFERENCES Annual Epidemiologic n:ports, Ministry of Health, State of Israel (1984). Bt .ectc, R. E. and Guxx, R. A. (1980) Hypersensitivity reactions associated with Botulinal antitoxin . Am . J. Med. 69, 567-570. G~~.g, D. R., Noxxts, J. W., Suxvrna., B. L. and li~cFmastu, V. C. (1987) Prognosis of acute stroke. Neurology 37, 221-225. CtiRB17AN~N, P. A. (1981) Snakebite and use of antivenin in southern Africa. S.A . Med. J. 59, 934-938. Dota,ar.RO, S., Rosy, A. S. and Ftst~e, D. (1986) A new look to the natural history and clinical features of intracerebral haemorrhage . A clinical-C'T scan correlation. Gerontology 32, 211-216. DVII .APfSICY, A. and B~wtx, H. (1973) Hypofibrinogenemia after Echis coloratus bite in man. Acta Haemat . 49, 123-127. EI .IENHOR.N, M. J. and B~c~.oux, D. G. (Eds) (1988) Medical Toxicology New York : Elsevier. F~nv~au, M., Mexxv, N., HBtsHgo, C. and E~xs~to, S. (1970) Defibrination following Echis coloratus bite in men. Isr . J. med. Sei. 6, 720-725. F~wxu, M., Mexxv, N., Hr o, C. and E~xanta, S. (1974) The natural course of defibrination syndrome caused by Echis coloratus venom in man. Throneb. Diathes . Haemorr. 31, 42028. L~xE, M. N., Rye, G. and Hnes~t, J. (1986) Hemmorhagic complications of long-term anticoagulant therapy. Chest (Suppl.) Sg, 16-25. MST, P., W~xxst., D. A . and C~x~t~ttNUxtcx, P. (1986) Prediction, prevention and mechanism of early (anaphylactic) antiserum reaction in victims of snake bite. Br . Med. J. 292, 17-20. M~xx, G. (1978) Echis coloratus bites in Israel: an evaluation of specific andgerum use on the base of 21 cages of snake bite . Toxicol. Eur . Res. 1, 36370. RECHNtC, J., Tsee~rreva®ea, P., Cesrsa, J., MOAOZ, C. and Dta VxII"s, A. (1962) Afibrinogenemia and thrombocytopenia in guinea pigs following injections of Echis coloratus venom. Blood 20, 73749. Rte, H. A. (1980) Antivenom reactions and efficacy. lancet 1, 1024-1025. Scxtn.cxvxsw+-Cesrer., H., Dvu wsxSr, A. and Kmrxex, A. (1986) Echir coloratus bites. Clinical evaluation of 42 patients . A retrospective study. Isr. J. med. Sci. 22, 880-884. $U1~RLAND, S. K. (1977) Acute untoward reactions to antiserums. Med. J. Rust . 2, 841,842.
1112
D. GILON et al.
Svrrmer wrro, S. K. and IAVPRINa, K. E. (1979) Antivenoms. Use and adverse reactions over a 12-month period in Australia and Papua New Guinea . Med. J. dust. 2, 67174. Tauvcw, F. G. (1963) The treatment of snake bite. Med. J. dust . 1, 27280. Wwßxet., D. A., Dwvmsox, N. M., O~ou, L. D., Pore, H. M., Wwixnvs, B. J., Gxwwooo, B. M. and Rrrn, H. A. (1974) Bites by the saw scaled viper (Echis carinatus). Trial of two specific antiserums . Br. Med. J. 4, 43740. Wwxxet,, D. A., Pore, H. M. and PxmvzicE, C. R. M. (1976) Disseminated intravascular coagulation caused by the carpet viper (Echis carinatus): trial of Heparin. Br. J. Haemat . 33, 33342. Ywruv, S., Mwivtar, N., Rrrct~, J. and Rvs.~t., J. (1974) The inductions of afibrinogenemia by Echis coloratus snake bite . J. trop. Med. Hyg. Tl, 136-143.
0041-0IOIj89 33 .00+ .00 ® 1989 Pergemon Prar pk
TREATMENT OF ENVENOMATION BY ECHIS COLORATUS (MID-EAST SAW SCALED VIPER) : A DECISION TREE Dwx
Gn.ox, ODED Sx~r ev and JocHexnx BENBA33AT
Department of Medicine, Hat9assah University Hospital, Mt Scopus, Jerusalem, Israel (Accepted jar publication 7 Aprit 1989)
Gn.ox, O. SxnL.sv and J. BENBASSAT, Treatment of envenomation by Echls coloratus (mid-east saw scaled viper) : a decision tree. Toxicon 27, 1101112, 1989.-Envenomation by Echis coloratus causes a transient hemostatic failure. Systemic symptoms, hypotension and evident bleeding are rare, with only one reported fatality. In this paper, we examine the decision to treat victims of Echis coloratus by a specific horse antiserum. The decision model considers the mortality of treated and untreated envenomation, and the side effects of antiserum treatment: fatal anaphylaxis, serum sickness and increased risk of death after a possible repeated exposure to horse antiserum in the future . The results of the analysis are not sensitive to variations in the probability of side effects of antiserum treatment. They are sensitive to variations in the risk of bleeding after envenomation, in the degree of reduction of this risk by antiserum treatment and in the risk of dying after an event of bleeding . Prompt administration of antiserum appears to be the treatment of choice if it reduces the risk of bleeding from 23.6% to 20.3% and if 1 .6% or more of the bleeding events are fatal. We conclude that presently available data support antiserum treatment of victims of Echis coloratus who present with hemostatic failure, even though the advantage imparted by this treatment appears to be small. D.
INTRODUCTION
Echis coloratus envenomation causes a transient hemostatic failure with a prolonged clotting time, hypofibrinogenemia and increased circulating fibrinogen split products (Racxrnc et al., 1962; F~vaRU et al., 1970, 1974; Dvu.nxsxi and Bmnx, 1973; YATZIV et al., 1974). Pain, bleeding or necrosis at the site of the snake bite are of variable severity . Hypotension and bleeding from other sites are rare, and there is only one death which has been reported by Mnxx (1978) and in the series of Scxur cHYxsxn-C~srEt et al. (1986) . A specific antiserum is available, however, indications for its use are uncertain. The proponents of antiserum therapy argue that it hastens hemostatic recovery (ScxuLcxxrtsxn-C~srEI, et al., 1986), that the only known fatality was not treated by antiserum (SextrLCt-nrrrstcw-Cns'rat, et al., 1986) and that envenomation by related spades, e.g. Echis carinatus, causes a severe hemostatie failure, which, if untreated by antiserum, is associated with a 7-15% mortality (WAItREL et al., 1974, 1976). On the other hand, some clinicians advocate no treatment, or treatment with heparin or antiserum only in cases with severe bleeding (F~nvnRU et al., 1974; Mnxx, 1978). They argue that the hemostatic 1105
1106
D . GILON et aJ.
failure after Echis coloratus envenomation is self-limited, much less severe than that after envenomation by E. carittatus, and is commonly followed by an uncomplicated recovery . Antiserum treatment carries a risk of anaphylaxis (MAr nsIT et al., 1986) and serum sickness (R», 1980), and may increase the risk of anaphylaxis if the patient is treated again with antiserum in the future . What is then the preferred management of Echis coloratus envenomation7 The rarity of this envenomation precludes randomized prospective trials of alternative treatments. Therefore, despite their potential inaccuracies, non-experimental methods are worth exploring. In this paper we analyse the trade-off between the risk of bleeding and death in untreated patients, and the side effects of antiserum treatment; anaphylaxis, serum sickness and an increased risk of dying after a possible repeated exposure to horse antiserum. METHODS We compared the possible outcomes of Echts coloratus envenomation, with and without antiserum treatment, in 8 previously healthy 18-year-old person, who presented with indefinite clotting time and no evident bleeding . Standard decision analysis methods were used (WmvsT~r and FnveaEtta, 1980), and the calculations were performed with the Smltree computer program (Pratt Medical Group, 1987). Sensitivity analyses were carried out to test the stability of our conclusions to variations in the probability estimates and in the assigned utilities. (aJ The decision tree The decision tree (Fig. 1) is divided into two parts, representing possible immediate (part A) and late (part B) events after envenomation . With antiserum therapy (Fig . lA, top branch) patients will either develop anaphylaxis or not, and those who do may die. Those who survive will either bleed or not, and those who do may die. Those who survive may or may not have serum sickness . With expectant management (Fig . lA, second branch) patients may bleed or not, and those who bleed may die . Within a time horizon, set arbitrarily at 25 years after recovery from envenomation, the patient may sustain a second envenomation by the same or other snakes and be treated again with horse antiserum (Fig. lB, top branch). Conceivably, this repeated treatment may carry a still higher risk of analphyhuris and death . Subjects who do not sustain a second envenomation during the first year after recovery (Fig . 1B, second branch) may die from other causes during the same year or survive, and cycle again through the risk of a second envenomation during the next year. The triangular Boolean node in Fig. 1 permits recycling for a predetermined number of years. It was asenmed that those who survive 25 years will live through their life expectancy. It was assumed that bleeding may result in either death or temporary morbidity, but no permanent morbidity . This assumption is incorrect : half of the survivors of intracerebral haemorrhage have some residual morbidity (Dot.r.aEatâ et al., 1986; C~te~mes et al., 1987) . This erroneous assumption was deliberately made in order to bias the model against antiserum treatment. We assumed also, that similarly to victims of envcnomation by E. carinatmr (Watexec et al., 1974), the severity of swelling, necrosis or permanent injury at the site of Echis coloratus envenomation are not affected by antiserum treatment. (b) Probabilities The analysis requires estimates of the probability of anaphylaxis, death and serum sickness after antiserum treatment, of death after recurrent antiserum treatment, of bleeding after treated and untreated envenomation, and of death after an event of bleeding . Arwphylaxis, death and senare sickness ajter antiserum treatment The incidence of immediate reactions to antiserum treatment ie related to the degree of antiserum purification, its dose and rate of injection (Its, 1980 ; Btr\c[ and Gutav, 1980) . Anaphylaxis occurs in 2~% of the recipients (Table 1). Fatal anaphylaxis is rare; there were no deaths among the cases surveyed in Table 1, nor among 4000 persons treated with antiserum for snakebite between 1962 and 1977 (Sur~et.eNn, 1977) . On the other band, one death (0.025%) has been attributed to antiserum from a total of 4000 persona at risk (Ram, 1980), and another three deaths (0.09%)occurred after anaphylaxis in 3300 recipients of antiserum for snake envenomation (CtnttslutvsErt, 1981). For purposes of the analysis we estimated the probability of anaphylaxis as 3 .4%,and the mortality after in-hospital anaphylaxis as 0.05% of all patients treated by antiserum, or 1 .6% of the cases with anaphylaxis .
Treatment of Envenomation by Schis coloratw
1107
IIRRnIruxLs
IIRi15ERUl
4
Rx
RLEEDIRR EYERT
SURr1Y~L
3°RYIY~L
SERVI SIWIESS
RO RLEEDIIIG EYEIIT
Ro uuPxruxls
OFATN RLEEDIIi EYERT EIPECTAIII IYIIYICEREIR
f
9
SI"Y1YI1L
RO RLEEDIRR EYEIIT
ENVENONATION
EVER
hROBABII.ITY TREE FOR TREATI~N'r OF ENVENOMATION HY Echis coloratus . The tree is subdivided into two parts indicating possible events occurring shortly after envenomation (bleeding, anaphylaxis, serum sickness ; A), and during the 25 years following recovery (second envenomation necessitating antiserum treatment, death after antiserum treatment or death from other causes ; B) . The square decision node (left, A) represents the choice . The two branches emanating from it represent the two therapeutic strategies: treatment with antiserum and no antiserum treatment. Circular nodes represent chance events which may follow each choice . The triangular (Boolean) node in part B indicates that either of the two branches may be followed, depending on the number of preceding cycles . Rectangular nodes represent possible outcomes. FIG .
1.
The incidence of serum sickness after treatment for snake envenomation has been reported as 5.5-14.3% (Table 1). In the present analysis it was assumed to be 8% . This estimate is much lower than the 80% reported for U.S. practice (ELLEIUtoRN and BARCEI .oux, 1988) . We have no ready explanation for these differences. Therefore, the impact of higher estimates of the probability of serum sickness was tested by sensitivity analysis. Risk of recurrent antiserran treatment The mean annual number of snake bites in Israel was 111 .9 during 1969-79 and 109.1 during 1979-87 (Epidemiologic Reports of the Ministry of Health of the State of Israel, 1984) in a population at risk aged 15-60
1108
D. GILON et al. TAHLe 1 . INCmsNCe oF ANAPxY~Ax>s AND sFxtrM stcaNSSS AF~e rxEATS~Nr BY xoasE AN~tt~ At risk Anaphylaxis 90% confidence n n (%) limits (%)
Reference SCHULCHYNSICA-CASret . e t al. (1986) MAt.w$T et al. (1986) RErD (1980) TxtNCA (1963) Svrr~tt,AND and LovewNa (1979) Annual Reports (1984) Total
28 35 547 100 203 151 1064
1 2 20 2 7 4 36
(3 .6) (5.7) (3.7) (2.0) (3 .4) (2.6) (3.4)
0-9.3 0-12 .2 2.4-5 .0 0-4.3 1 .3-5 .6 1 .3-3 .9 2.5 .3
Serum sickness n (%)
90% confidence 11mltS (%)
4 (14.3)
3.4-25.2
59 6 10 1 80
(10.8) (6 .0) (S .5) (0 .7) (7 .9)
8.6-13.0 2.1-9.9 2.7-8 .2 0-1.7 6.5-9.3
of roughly 1,800,000 . Therefore, the annual risk of snake bite was estimated as 110/1,800,000, or 0.006% . This estimate is probably correct for a person with an incidental exposure to snakes, but not for snake collectors or employees of snake farms. The impact of higher estimates of the risk of a second exposure to horse antiserum was tested by sensitivity analysis . We know of no estimates of the risk of anaphylaxis after repeated antiserum treatment. The average subjective consensus of three experts in emergency medicine was that a previous exposure to antiserum triples the risk of anaphylaxis after a second treatment, and doubles the risk of death. Therefore, the risk of anaphylaxis after a recurrent exposure to antiserum was estimated as 10.2% with a mortality of 3.2% of ell cases with anaphylaxis, yielding an overall mortality after recurrent antiserum treatment of 0.33% . Although not every envenomation is judged to warrant antiserum therapy, we assumed that all subjects who will sustain a second envenomation will be treated by horse antiserum, thus biasing the model still further against antiserum therapy. Risk of bleeding and death after envenomation The main problem of this analysis was the lack of reliable estimates of the risk of bleeding and death in treated and untreated victims of Echis coloratus . There are no experiments, in which similar cases were either treated by a specific antiserum or not, and then followed in order to determine their outcome in terms of frequency of bleeding and survival . Our estimates were based on the uncontrolled retrospective observations by ScxaLCEnrN.rtcA-CAS~ret. et al. (1986) on 42 patients, 28 of whom were treated by antiserum. The criteria for patient selection for treatment are unknown. It is conoeivabk, that antiserum was given to more severely sick patients, to those with more prolonged hemostatic failure, or to those who bled. Subsequently, conclusions about the eSrcacy of therapy weld be biased against antiserum tt~eatment . SCxv[,cxYNSxA-CASret, et al. (1986) reported that the hemostatic failure lasted on the average for 87 hr in untreated patients, and 75 hr in patients treated by antiserum (Table 2) . When the three patients who received antiserum only on the fifth day after envenomation were excluded, the average duration of the hemostatic failure in antiserum-treated patients appeared to be only 59 hr. SCHCTLCFIYN3KA-CASrEr. et al. (1986) reported a total 3318 patient-hours of hemostatic failure, with nine bleeding events (two from the gastrointestinal tract, five from the mucosae, one in muscles and one macroscopic hematuria), or one per 369 hr of hemostatic failure (Table 2) . In nine other patients the hemoglobin level declined TABIE 2. Rrsx OP BIEEDINO AND DiJRAT10N OF I~IIOBTAIIC FAII.URE AP1IIt 1SiVHAiO~tATION HY EChis At risk
Treatment Eckas coloratus No antiserum Antiserum Total
n
Pt-hr of hemostatic failure
Events of bleeding
n
Per hr hemostatic failure
(Scxur .CHYNSeA-CASrEt, et al., 1986) 14 1218 28 2100 42 3318 9"-18t 1 :369 "-1 :184t
Echis carinates (WAxxet, et al., 1976) No antiserum 14 Antiserum Total 14 348
"Minimal estimate; tmaximal estimate .
8
1 :43.5
Average duration of hemoatatic failure per patient (hr)
Risk of blending (%)
87 75t-59 "
24"~7t 20"-32t
Not reported 25
57
1109
T~eatment of Envenomation by Echis coloratus
by 2-5 g/dl . If this decline was due to undetected hemorrhage, then the number of blceding episodes would increase to 18, with a fi^equency of I per 184 hr of hemostatic failure. Assuming one bleeding event per 369hr hemoatatic failure, and assuming that the number of bleeding events correlates with the duration of hemostatic failure, then the risk of bleeding may be calculated as 75/369-0.203, or 20 .3% in treated patients (for an average duration of hemostatic failure of 75 hr) and 23 .6% in untreated patients (for an average duration of hemostatic failure of 87 hr). Assuming one bleeding event per 184 hr hemoatatic failure and an average duration of hemostatic failure of 59 hr in treated patients, the risk of bleeding could be as high as 47.3% in the untreated group and 32 .1% in treated patients. We used as baseline values the lower estimates of the risk of bleeding in untreated envenomation and the lower estimate of the degree of reduction of his risk by antiserum treatment in order to bias the analysis still further against antiserum treatment. The risk of dying after an event of bleeding was estimated also on the basis of the observations by ScHUL.cHVNacA-CA9'L'EI, et al. (1986) who reported one death (I1 .1%) among the nine bleeding events (90% confidence intervals: 0-28 .4%) . It is interesting to note that this estimate is quite similar to the 0 to 33% (average-10%) mortality after an event of bleeding in patients on anticoagulant treatment (Levtxe et al., 1986). (c) Utilities Utilities, i.e . the value of each alternative outcome, were expressed in years of quality-adjusted life expectancy, obtained by subtracting from the life expectancy the estimated period of time spent off regular life routine because of complications of bleeding or antiserum therapy. For serum sickness, 0.08 years were substracted from the life expectancy; for bleeding -0.25 years; and for non-fatal anaphylaxis -0 .04. The utility of each outcome was expressed as the sum of (a) time interval from envenomation to the end of the follow-up, (b) years of expected survival after the end of the follow-up period, as derived from life-tables, and (c) adjustment for the morbidity resulting from bleeding, non-fatal anaphylaxis and serum sickness . Deaths following bleeding or anaphylaxis were given a utility of 0. Delayed deaths (i .e . deaths from other causes, or deaths after a second envenomation) were given a utility equal to the sum of (a) and (c). (dJ Summary ojprobability estimates Md SOf?CGJ Ojbias We deliberately biased the analysis against antiserum treatment by assuming : (a) that there is no long term morbidity after a non-fatal bleeding, (b) that the treated and untreated groups of patients in the non-randomized retrospective study of ScHULCxYNS6w-CASTEL et al. (1986) were comparable, (c) that every subject sustaining a second envenomation will be treated by antiserum, and (d) a low base line probability of bleeding after envenomation . These biases ensure that we did not miss a possible advantage of witholding antiserum treatment. The probability estimates used in the present analysis are summarized in Table 3. R FCLTLTS
When the baseline values were used to evaluate the two strategies for management of envenomation by Echis coloratus, the results favored antiserum treatment. The quality TABLE
3.
SUMMARY OP THE BASELINE PROBAHILIl'Y VALUES U~D IN THIS ANALYSIS
Probability of Anaphylaxis after administration of horse serum Mortality after in-hospital anaphylaxis Bleeding event after untreated Echis coloratus envenomation Bleeding event after Echis coloratus envenomation treated by antiserum Death after an event of bleeding Serum sickness after treatment with antiserum Recurrent envenomation (annual) Death after recurrent treatment with antiserum
%
RSII$e Of 8eilalüvlty 8n81yB18 (%)
3.4 1 .6
0-22 0-11
23 .6
23.6~7.2
20 .3 11 .1
16.0-32.0 0-10
8.0 0.006 0.33
2-100 0.006-0 .6 0-100
111 0
D. GILON
tt
al.
adjusted life expectancies in years were 55.96 for antiserum treatment, and 55.78 for expectant management, as compared to a 56.3 years life expectancy of an 18-year-old man. Immediate treatment with antiserum added 0.18 yr (0.33%) to the average life expectancy over the strategy of expectant therapy. The results of the sensitivity analyses indicated that the decision was not sensitive to changes in the probability of serum sickness, of second envenomation, or of death after recurrent antiserum treatment. Antiserum treatment was the preferred management even assuming an up to 22% probability of anaphylaxis, an up to 11 % mortality after anaphylaxis, a 100% probability of serum sickness, or a 100% probability of death after recurrent antiserum treatment. Also, results of the analysis were not sensitive to variations in the adjustments of the utilities of patients after bleeding and serum sickness . Antiserum treatment was the preferred management, even after subtracting from the life expectancy up to 2 years for serum sickness or up to 5 years for non-fatal anaphylaxis. The analysis was sensitive to variations in the probability of bleeding with and without antiserum treatment, and the mortality after an event of bleeding . Figure 2 is a three-way analysis of the relation between the probability of bleeding in treated and untreated envenomation and the probability of death after bleeding . The results suggest a superior utility of antiserum treatment if it is assumed that antiserum reduces the risk of bleeding from 23.6% to 20.3%,and that 1 .6% or more of the bleeding episodes will result in death. DISCUSSION
The main conclusion of this analysis is that the complications of antiserum treatment (anaphylaxis, serum sickness, postulated increase in risk after a second exposure to antiserum) have a marginal weight in the decision about the treatment of Echis coloratus envenomation . This decision depends predominantly on the probability of bleeding with and without antiserum treatment, and on the mortality after a bleeding event. The review of the literature supports the impression that the consequences of envenomation by Echis 0 .8
dc b
z
é is
G .2
0 .0 ' 0 .0
FYG . 2.
0.2 0 .4 0 .8 PROBABILITY OF BLEEDING WfTHOUT TREATMENT
0 .8
THam-WAY ANALYSE OF THS CHOI(~ HErWEEPi H~®IATE TRFAT1r~Mr WITH AN1'6EaUlI AND H7CPECrANT ètANAGElC~Nr OF ENVIIQOMATION HY ECIIlS COIOrOtYJ.
The analysis examines the relation between the probability of bleeding in treated (vertical axis) and untreated (horizontal axis) patients and the probability of death after a bleeding event (line a, 0% ; line b, 1% ;line c, 3%;line d, 10%) . The shaded area covers clinically relevant areas . Assuming a zero probability of death after an event of bleeding, the analysis favors no antiserum treatment . Antiserum treatment is preferred for a probability of death exceeding 1 .6% of all events of bleeding .
Treatment of Envenomation by Echis caloratus
111 1
coloratus are less severe than those of envenomation by E. carinates. The data reported by WARREL et al. (1974) imply a 57% risk of bleeding in treated Echis carinates envenomation, while the data by $CHULCHYNSKA-CasTal, et al. (1986) imply a maximal estimate of 47% risk of bleeding in untreated Echis colorates envenomation (Table 2). The differences may be related to differences in venom toxicity, or in the characteristics of the population at risk . The results of our analysis support the recommendation of Scxul.cxnvsxA-CASTEL et al. (1986) to treat Echis colorates envenomation with a specific antiserum, even though the advantage ofthis treatment is probably small. Even assuming that the probability of death is only 3% of all bleeding events, antiserum treatment of 100,000 victims would result in 610 fatal bleeding events, 54 deaths from anaphylaxis, 0.5 deaths after recurrent treatment with antiserum, and in 7815 cases of serum sickness . On the other hand, witholding treatment would result in 708 fatal bleeding events and in 0.06 deaths after treatment for second envenomation . Under these assumptions, antiserum treatment would be the preferred strategy even for an up to 0.6% annual risk of second envenomation, i.e. 100 times higher than our baseline estimate . The uncertainty regarding the treatment of Echis coloratus envenomation is due to the lack of precise estimates of the risk of bleeding and death after a bleeding event. Furthermore, there are no data regarding the efficacy of other modes of treatment, e.g. replacement therapy by fresh frozen plasma and procoagulants. In the absence of data, clinical decisions must be based on assumption . We feel that the recommendation of antiserum treatment for every case of Echis coloratus envenomation with hemostatic failure is based on reasonable assumptions, which are consistent with presently available observations . REFERENCES Annual Epidemiologic n:ports, Ministry of Health, State of Israel (1984). Bt .ectc, R. E. and Guxx, R. A. (1980) Hypersensitivity reactions associated with Botulinal antitoxin . Am . J. Med. 69, 567-570. G~~.g, D. R., Noxxts, J. W., Suxvrna., B. L. and li~cFmastu, V. C. (1987) Prognosis of acute stroke. Neurology 37, 221-225. CtiRB17AN~N, P. A. (1981) Snakebite and use of antivenin in southern Africa. S.A . Med. J. 59, 934-938. Dota,ar.RO, S., Rosy, A. S. and Ftst~e, D. (1986) A new look to the natural history and clinical features of intracerebral haemorrhage . A clinical-C'T scan correlation. Gerontology 32, 211-216. DVII .APfSICY, A. and B~wtx, H. (1973) Hypofibrinogenemia after Echis coloratus bite in man. Acta Haemat . 49, 123-127. EI .IENHOR.N, M. J. and B~c~.oux, D. G. (Eds) (1988) Medical Toxicology New York : Elsevier. F~nv~au, M., Mexxv, N., HBtsHgo, C. and E~xs~to, S. (1970) Defibrination following Echis coloratus bite in men. Isr . J. med. Sei. 6, 720-725. F~wxu, M., Mexxv, N., Hr o, C. and E~xanta, S. (1974) The natural course of defibrination syndrome caused by Echis coloratus venom in man. Throneb. Diathes . Haemorr. 31, 42028. L~xE, M. N., Rye, G. and Hnes~t, J. (1986) Hemmorhagic complications of long-term anticoagulant therapy. Chest (Suppl.) Sg, 16-25. MST, P., W~xxst., D. A . and C~x~t~ttNUxtcx, P. (1986) Prediction, prevention and mechanism of early (anaphylactic) antiserum reaction in victims of snake bite. Br . Med. J. 292, 17-20. M~xx, G. (1978) Echis coloratus bites in Israel: an evaluation of specific andgerum use on the base of 21 cages of snake bite . Toxicol. Eur . Res. 1, 36370. RECHNtC, J., Tsee~rreva®ea, P., Cesrsa, J., MOAOZ, C. and Dta VxII"s, A. (1962) Afibrinogenemia and thrombocytopenia in guinea pigs following injections of Echis coloratus venom. Blood 20, 73749. Rte, H. A. (1980) Antivenom reactions and efficacy. lancet 1, 1024-1025. Scxtn.cxvxsw+-Cesrer., H., Dvu wsxSr, A. and Kmrxex, A. (1986) Echir coloratus bites. Clinical evaluation of 42 patients . A retrospective study. Isr. J. med. Sci. 22, 880-884. $U1~RLAND, S. K. (1977) Acute untoward reactions to antiserums. Med. J. Rust . 2, 841,842.
1112
D. GILON et al.
Svrrmer wrro, S. K. and IAVPRINa, K. E. (1979) Antivenoms. Use and adverse reactions over a 12-month period in Australia and Papua New Guinea . Med. J. dust. 2, 67174. Tauvcw, F. G. (1963) The treatment of snake bite. Med. J. dust . 1, 27280. Wwßxet., D. A., Dwvmsox, N. M., O~ou, L. D., Pore, H. M., Wwixnvs, B. J., Gxwwooo, B. M. and Rrrn, H. A. (1974) Bites by the saw scaled viper (Echis carinatus). Trial of two specific antiserums . Br. Med. J. 4, 43740. Wwxxet,, D. A., Pore, H. M. and PxmvzicE, C. R. M. (1976) Disseminated intravascular coagulation caused by the carpet viper (Echis carinatus): trial of Heparin. Br. J. Haemat . 33, 33342. Ywruv, S., Mwivtar, N., Rrrct~, J. and Rvs.~t., J. (1974) The inductions of afibrinogenemia by Echis coloratus snake bite . J. trop. Med. Hyg. Tl, 136-143.