Bleeding is the most common complication of treatment with heparin or warfarin. Aside from this, heparin and warfarin have had remarkably good safety records for 50 years. Nevertheless, two rare but serious nonhemorrhagic complications have been documented and widely reported: heparin-induced thrombocytopenia with thrombosis [2] and warfarin-induced skin necrosis [3]. Paradoxically, both of these complications are associated with thrombosis, the very problem that anticoagulant treatment is supposed to treat or prevent.

Heparin-induced thrombocytopenia occurs in 1% to 3% of patients who receive heparin for deep venous thrombosis [2]. It usually occurs more than 5 days after the start of heparin therapy but may occur earlier in patients who have had recent exposure to heparin. Heparin-induced thrombocytopenia is caused by heparin-dependent antibodies that bind to a complex of heparin and platelet factor 4 [4, 5]. Arterial thromboembolism, known as the white clot syndrome [6], is a potentially devastating consequence of heparin-induced thrombocytopenia that has been recognized for 20 years. More recently, the association between heparin-induced thrombocytopenia and venous thromboembolic events was shown [7].

Warfarin has been associated with a rare complication known as warfarin-induced skin necrosis [3], which involves thrombosis of small vessels and has the clinical appearance of cutaneous gangrene. The mechanism is thought to be related to a depletion of the anticoagulant vitamin K-dependent proteins C and S that occurs more rapidly than does depletion of the procoagulant vitamin-K-dependent proteins (coagulation factors II, VII, IX, X).

In this issue, Warkentin and colleagues [8] report on a series of patients with heparin-induced thrombocytopenia who developed limb gangrene after receiving warfarin. Eight patients with this complication were seen over a 15-year period among five hospitals in Hamilton, Ontario, Canada. The authors suggest that these cases of venous limb gangrene differ somewhat from classic warfarin-induced skin necrosis. They hypothesize that the prothrombotic state created by heparin-induced thrombocytopenia was aggravated by the addition of warfarin; this in turn accelerated thrombosis rather than protecting the patient from it because protein C was depleted before vitamin K-dependent procoagulant factors were depleted.

Several aspects of the reported cases are of interest. First, two of the patients with limb gangrene received large initial doses of warfarin (15 to 20 mg/d for 2 to 3 days); initial doses were not provided for five patients. Moreover, a trend was seen toward a larger median dose of warfarin administered over the first 5 days to the eight patients who developed venous limb gangrene. All eight patients had international normalized ratios greater than 3.0 (median, 5.8 [range, 3.8 to 9.4]) [8]. It is therefore possible that venous limb gangrene is related to large initial doses of warfarin, or at least to excessive doses, rather than to the use of warfarin per se. This point is illustrated by patient 3, who initially received 15 mg of warfarin per day for 3 days and in whom the protein C level was rapidly and almost completely suppressed [8].

Similarly, in patient 2, the protein C level was reduced to less than 0.01 U/mL, which is less than 2% of the lower limit of the normal range reported by the investigators' laboratory (0.65 U/mL). This finding also contrasts with the levels of protein C found in patients with heterozygous protein C deficiency (33% to 55% of normal) [9]. Warfarin has been associated with skin necrosis in patients with congenital heterozygous deficiency of protein C [3], so it is not surprising that rapid suppression of protein C to very low levels may lead or at least contribute to limb gangrene and necrosis. Second, as Warkentin and colleagues [8] point out, necrosis of the feet or hands occurs in about 10% of patients with warfarin-induced skin necrosis. Venous limb gangrene may simply be part of the spectrum of warfarin-induced skin necrosis. Third, it is possible that some patients would have developed limb gangrene even if they had not received warfarin. Venous gangrene and skin necrosis have been reported in patients with heparin-induced thrombocytopenia who did not receive warfarin [3]. This type of skin necrosis is clinically indistinguishable from warfarin-induced necrosis and has a similar histologic picture [3].

Thus, venous limb gangrene may be a combination of three thrombosis-related problems: 1) a transient prothrombotic state induced by excessive doses of warfarin and rapid depletion of protein C that is superimposed upon 2) heparin-induced thrombocytopenia (itself a prothrombotic state) [7]; both occur in the setting of 3) a primary thrombotic condition (deep venous thrombosis). Finally, it would be useful to know the duration of heparin therapy given to the five patients who were treated for deep venous thrombosis before they developed thrombocytopenia. If these patients received heparin for more than 5 days before the onset of thrombocytopenia, venous gangrene may now be more rare; current practice, supported by randomized clinical trials [10, 11], is to begin warfarin therapy on the first or second day of heparin treatment and discontinue heparin treatment on the fifth day. Furthermore, heparin-induced thrombocytopenia may become less common as low-molecular-weight heparin replaces unfractionated heparin for the initial treatment of deep venous thrombosis [7, 12-14].

What should clinicians do in response to case reports of venous limb gangrene as a complication of heparin-induced thrombocytopenia? First, they should keep the problem in perspective: This complication is rare. Clinicians should not withhold warfarin treatment from patients with proximal deep venous thrombosis who develop heparin-induced thrombocytopenia for fear of causing venous limb gangrene. Failure to provide adequate long-term anticoagulant treatment in patients with proximal deep venous thrombosis results in a 47% incidence of recurrent venous thromboembolism [15]. The rates of morbidity and mortality from recurrent venous thromboembolism if warfarin is withheld would be much greater than the rate of illness caused by the rare problem of venous limb gangrene. Insertion of a vena caval filter is an alternative to warfarin therapy in this situation. However, although a filter can prevent death from pulmonary embolism, it does not prevent morbidity from recurrent venous thrombosis and the post-thrombotic syndrome, which is strongly associated with ipsilateral recurrent venous thrombosis [16]. Second, when warfarin treatment is initiated, dosages of 5 mg/d to 10 mg/d should be used [17]. Large initial dosages (such as 20 mg/d) should be avoided. No pharmacologic rationale or clinical benefit exists for such large initial dosages, and they may be harmful. Third, the cases reported by Warkentin and colleagues [8] emphasize the value of promptly initiating warfarin therapy with heparin or low-molecular-weight heparin in patients with deep venous thrombosis; this approach provides an overlap of these treatments of 4 to 5 days, before most cases of heparin-induced thrombocytopenia develop. If heparin-induced thrombocytopenia develops before warfarin therapy has been started or before the overlap of 4 to 5 days is finished, an alternate anticoagulant may be needed, overlapping warfarin, for sufficient time to allow prothrombin and factor X to decrease to the desired levels. Options include the defibrinogenating agent ancrod [18]; the heparinoid danaparoid [19]; or a direct thrombin inhibitor, such as hirudin [20] or argatroban.

Conclusions about the advantages and disadvantages of alternate antithrombotic strategies and further recommendations for the care of patients who develop heparin-induced thrombocytopenia will require randomized clinical trials. With the development of new strategies for antithrombotic therapy, we can realistically hope that, in the near future, these catastrophic case reports will be of only historical interest.