Diet as prophylaxis and treatment for venous thromboembolism?

Bleeding

Regarding patients treated for VTE with standard anticoagulants, a recent meta-analysis of published RCTs showed major and fatal bleeding rates of 1.8% and 0.2%, respectively [6]. Older cohort studies report up to triple these rates [7–9]. Applying the range of reported fatal bleeding rates for VTE treatment (0.2% - 0.6%) to an estimated 300,000-1.2 million people treated for VTE worldwide per year (about half in the USA [10]), 600-7,200 people per year suffer fatal bleeds from VTE anticoagulant treatment. There are many more non-fatal major bleeds, some of which are permanently debilitating.

Anticoagulant prophylaxis for surgical patients increases the risk of major bleeding [11]. VTE prevention trials report markedly different rates of major bleeding despite similar patient populations and doses and durations of anticoagulant prophylaxis. For instance, major bleeding with enoxaparin reportedly ranged from 0.1% to 3.1% in hip arthroplasty trials and from 0.2% to 1.4% in knee arthroplasty trials. If surgical-site bleeding is included in the definition of major bleeding, the reported rates have been about 10-fold higher [12]. Major bleeding adversely affects overall mortality. In a meta-analysis of trials comparing fondaparinux with LMWHs or placebos (major bleeding incidence overall = 2.4%), the risk of death by 30 days was 7-fold higher among patients with compared to those without a major bleeding event (8.6% versus 1.7%) [13]. If the major bleeding is considered the cause of the higher death rate, 6.9% of deaths in patients with major bleeds may be attributed to the bleeding (8.6% - 1.7% = 6.9%). Consequently, deaths of about 0.166% of anticoagulated patients are arguably attributable to bleeding (0.069 × 0.024 = 0.00166). Given that at least 12 million medical and surgical patients worldwide receive prophylactic anticoagulants per year [14, 15], this means that approximately 20,000 people may die each year from complications of bleeding from prophylactic anticoagulants (0.00166 × 12 million = 19,872); many more may suffer the consequences of hypovolemia.

Efficacy

Anticoagulant therapy for VTE became established as the standard of care in the 1940s and 1950s before randomized trials were considered necessary to prove efficacy and safety. A very small RCT comparing anticoagulants versus placebo for people with clinical diagnoses of PE published in 1960 [4] has been used to justify anticoagulant therapy. However, by current scientific standards, this study is highly flawed [10, 16].

In 1962 when the U.S. Food and Drug Administration began requiring randomized controlled clinical trials (RCTs) showing efficacy before approving drugs, anticoagulation treatment of VTE was 'grandfathered in' with no rigorous efficacy trials ever required. Only three small methodologically rigorous RCTs of patients with DVTs [17–19] have compared standard anticoagulants with placebos or non-steroidal anti-inflammatory drugs. Combining the data from these trials, 6/66 patients receiving standard heparin and vitamin K inhibitors died and 1/60 unanticoagulated patients died [10]. Consequently, standard anticoagulant treatment for VTE cannot be considered evidence-based to be effective [10, 20].

Anticoagulant prophylaxis of 'high VTE risk' patients may increase fatal pulmonary emboli (FPE) due to 'rebound hypercoagulation'

Goldhaber and colleagues tracked the incidence of developing DVT or PE during or up to 30 days after hospital discharge in about 80,000 patients admitted over a two year period in Boston's Brigham and Women's Hospital. Out of 384 patients with hospital-acquired VTE, 318 (82.8%) were potential candidates for prophylaxis (i.e., they had ≥2 VTE risk factors). Of prophylaxis candidates, 170 (53%) of those with hospital-acquired VTE had received anticoagulants [21]. To estimate the influence of prophylactic anticoagulants in this study, we can use Goldhaber's USA-wide estimates of hospitalized patients that are at 'high VTE risk' -- 32% [14] or 25,600/80,000 in the Brigham and Women's Hospital study -- and the proportion of those at VTE risk who receive anticoagulant prophylaxis -- 50% [14] or 12,800/25,600 in this study. According to these estimates, 'high VTE risk' patients receiving anticoagulants in this population had a non-significant trend toward a higher incidence of VTE (OR = 1.15, 95% CI = 0.92 - 1.44) [22].

More importantly in this chart study, out of 13 deaths attributed to hospital-acquired FPE, 12 had received anticoagulant prophylaxis [21]. As above, assuming that 32% of the hospitalized patients were at risk for VTE and that 50% of all patients at risk for VTE received anticoagulants, anticoagulation prophylaxis was associated with a 12-fold increase in hospital-acquired FPE (OR: 12.0; 95% CI, 1.6-92) [22].

An autopsy study by Lindblad and colleagues [23] from Malmo, Sweden corroborated the Goldhaber study. From a population of 31,238 post-operative patients from the 1980s, it found that 27/30 patients with autopsy-proven FPE had received post-op prophylactic anticoagulants. The authors did not report the proportion of 'high VTE risk' surgical patients in their hospital receiving anticoagulant prophylaxis. To provide an approximation of the degree of increased risk of FPE related to anticoagulant prophylaxis in this autopsy study from a defined clinical population, we can conservatively assume that all Malmo surgical patients had 'high VTE risk' and again use Goldhaber's estimate that about 50% of those at risk received anticoagulant prophylaxis [14]. This translates to about 15,619 patients with anticoagulants and the same number without. Compared with patients not receiving anticoagulant prophylaxis, the Lindblad autopsy data show the estimated FPE rate in anticoagulated patients is nine-fold higher (OR: 9.0; 95% CI, 2.7-29.6).

Since many Malmo surgical patients would have been at 'low VTE risk' and fewer than 50% of those at 'high VTE risk' may have received anticoagulants in the 1980s, the FPE rate associated with anticoagulant prophylaxis could well have been considerably higher. Combining the FPE data from Goldhaber and Lindblad yields a very conservative estimated increased FPE risk associated with anticoagulant prophylaxis of 9.75 fold (OR, 9.75; 95% CI, 3.5 - 27.3). Combining these studies, 35/43 cases can be attributed to 'rebound hypercoagulation' (i.e., 39/43 FPE patients had received anticoagulation prophylaxis versus 4/43 with no anticoagulation: 39 - 4 = 35).

Surgery is associated with a substantial systemic and local activation of the coagulation and fibrinolytic systems. Post-operative prophylactic anticoagulants significantly mitigate the stimulation of these systems. However, following the discontinuation of prophylactic anticoagulants, a second wave of activation of markers of the coagulation and fibrinolytic systems continues for up to 35 days after surgery (e.g., plasma TAT and D-dimer [24]). Cundiff has suggested that 'rebound hypercoagulation' after stopping anticoagulants causing restimulation of coagulation and fibrinolysis may account for this marked increase in FPE risk associated with anticoagulant treatment [25] and prophylaxis [26]. Given that at least 12 million medical and surgical patients worldwide receive prophylactic anticoagulants per year [14, 15], an estimated 5,000 to 40,000 people per year die of 'rebound hypercoagulation' (i.e., 12,000,000 (hospitalized people/year with anticoagulant prophylaxis) × 35/28,419 (excess risk for fatal PE per Goldhaber and Lindblad studies) = 14,779; 95% CI, 5,305 - 41,381).

Marked reduction in FPE risk over time unrelated to anticoagulants

Given (1) the incidence of major and fatal bleeding from anticoagulants for prophylaxis and treatment of VTE, (2) the efficacy data for both that have been called into question, and (3) the evidence for previously unrecognized and largely uncounted deaths from 'rebound hypercoagulability'; reconsideration of the evidence-basis of anticoagulants for treatment and prophylaxis of VTE is in order.

Historical data

In Norway from 1940 to 1944, intake of meat, whole milk, cream, margarine, cheese, eggs, and fruit decreased while people increased their intake of fish, cod liver oil, skimmed milk, whole grain bread, potatoes, and fresh vegetables. The rate of post-operative VTE decreased markedly during the Second World War in Norway followed by a marked increase after the war [34].

During the Second World War, people in Norway, Sweden, Switzerland, Germany, Finland, and Denmark had significantly reduced intake of food from animal sources. However, only Denmark showed no decrease in vascular disease mortality. In Denmark alone, there was no significant reduction in consumption of dairy fats and eggs [35].

The autopsy incidence of FPE over time in Heidelberg, Germany showed a clear relationship between pulmonary embolism and wartime conditions. The lowest incidence of FPE, expressed as a percentage of all hospitalized patients, was registered during the post-Second World War years with a relative and absolute minimum between 1945 and 1949. The 1947 value (0.04%) was lower than 1932 (0.45%) or 1955 (0.38%) [36] (Fig. 1).

In Vienna after the First World War, FPE accounted for less than 0.5% of deaths versus 2.5% in the early 1930s. Again, in the late 1940s, incidence of FPE at autopsy was <1% versus almost 8% by the early 1970s [30] (Fig. 2).

These historical studies have limitations but suggest that the high-complex-carbohydrate, low-fat diet associated with war-time food rationing and perhaps increased exercise may have markedly reduced the tendency to form thrombi and/or lessened the consequences of those that do form. Judging from the autopsy data, the effects of these lifestyle influences on VTE risk had a rapid onset and offset, and wartime conditions afforded substantial protection against VTE, especially FPE.

Prospective observational studies of diet and lifestyle factors associated with VTE

In the "Longitudinal Investigation of Thromboembolism Etiology (LITE) " prospective study, hazard ratios (95% CIs) of VTE incidence across quintiles of fruit and vegetable intake were: 1.0 (reference: lowest quintile), 0.73 (0.48 to 1.11), 0.57 (0.37 to 0.90), 0.47 (0.29 to 0.77), and 0.59 (0.36 to 0.99) with P trend = 0.03 [32]. The fruit and vegetable intake in the lowest quintile, 2.0 servings per day, was far less than recommended by the United States Center for Disease Control (i.e. >5 servings per day for most people) [37]. The highest quintile averaged 6.7 servings per day. Meat intake was a predictor of VTE risk in LITE (HRs of VTE across quintiles of red and processed meat intake--1.0 (lowest quintile), 1.24 (0.78 to 1.98), 1.21 (0.74 to 1.98), 1.09 (0.64 to 1.87), and 2.01 (1.15 to 3.53) with the P trend = 0.02 [32]). Since fruit/vegetable intake in LITE correlated negatively with meat intake (r = - 0.28), the two most influential dietary variables may have acted synergistically on VTE risk.

In contrast, the Iowa Women's Health Study (IWHS) [38], to date the only other large prospective study of diet related to VTE risk, found no associations of VTE risk with intake of fruits/vegetables, meat, fish, or other foods. It also found no significant associations with dietary patterns or individual nutrients. The IWHS found an association of daily alcohol consumption with lowered VTE risk, whereas LITE only affirmed that adjusting for alcohol consumption did not diminish the strength of the correlations between diet and VTE [32]. The following differences between the LITE and IWHS studies may account for the discrepancies:

Only women were surveyed in the IWHS (99% white women); the LITE study included relatively fewer women (55%) and more non-whites (27%).

LITE's dietary assessment had an interviewer-administered food frequency questionnaire (FFQ), considered more precise than the self-administered IWHS questionnaire [32].

While the IWHS FFQ assessment was done only once and related to the subsequent 19 year (mean 13 year) follow-up for VTE incidence, the LITE FFQ assessment was done twice, six years apart (r ranged from 0.49 to 0.56 between the two assessments). This allowed the LITE diet assessment to be based on a cumulative average dietary intake. The designers of the FFQs used in both studies found that using the cumulative averages, in general, yielded stronger associations than utilizing either baseline diet or the most recent diet alone [39].

Therefore, the LITE study findings relating fruit and vegetable intake to reduced VTE risk and meat intake to increased VTE risk have more scientific validity than the IWHS finding of no dietary factor influences on VTE. The LITE findings are also consistent with the historical observations (above).

VTE risk related to lipids, inflammation, and hemostatic parameters

A meta-analysis of VTE case-control studies found that VTE patients had significantly lower HDL cholesterol levels and higher triglycerides but no differences in total cholesterol or LDL-cholesterol [40]. Prospective studies of associations between lipid levels and VTE risk are conflicting but possibly suggest that VTE risk is reduced when serum levels of HDL cholesterol are higher and triglyceride lower [41].

In LITE, VTE was not correlated with C-reactive protein or white cell count [42]. However, LITE did not differentiate between idiopathic versus secondary VTE cases. Luxembourg and colleagues reported a case-control study comparing idiopathic VTE with risk-associated VTE patients (post-op, etc.) and controls. Idiopathic VTE patients had significantly higher levels of high-sensitivity C-reactive protein (hs-CRP) than secondary VTE patients (mean 1.8 versus 1.5, P = 0.05), who had significantly higher levels than controls (mean 1.5 versus 1.2, P = 0.02) [43].

Case-control and prospective studies show correlations of VTE with serum levels of factor VIII [43–47]. The VTE case control study by Luxembourg and colleagues found significantly higher fibrinogen levels in patients with idiopathic VTE than those with risk-associated VTE (median: 331 versus 299 mg/dl, p = 0.004) [43]. Controls had levels similar to risk-associated VTE patients (median: 302 mg/dl). In LITE, factor VIII levels and von Willebrand factor levels correlated significantly with VTE (P for trend in quartiles <0.0001 for both) but fibrinogen levels did not. LITE did not assess platelet aggregation or indices of fibrinolysis.

American Heart Association (AHA) step 1 and step 2 diets

AHA step 1 and step 2 diets recommend plenty of fruits and vegetables, lean meat and two servings of fish per week [50]. A meta-analysis of randomized trials of these diets versus regular diets (27 trials with more than 30,000 patient years of follow-up) shows no significant reduction of overall mortality (RR: 0.98, 95% CI 0.86 to 1.12), or cardiovascular disease mortality (RR: 0.91, 95% CI 0.77 to 1.07) [51, 52].

There is no evidence that the AHA step 1 or step 2 diets would reduce VTE risk any more than overall cardiovascular risk, so they would not be good low VTE risk diet candidates.

Mediterranean diet (MD)

While the MD does not significantly benefit serum lipids, blood pressure, or body mass index, it reduced overall cardiovascular disease risk by 70% in the "Lyon Diet Heart Study " [53].

Vegetarian diets

Epidemiological studies show that people consuming vegetarian diets (lacto-ovo, lacto, ovo, or not otherwise specified (NOS)) and vegan diets (no meat, dairy, or eggs) have lower overall vascular disease incidence than omnivores [54–58].

A lacto-ovo vegetarian diet contains plenty of fruits and vegetables and no meat, largely conforming to the LITE prospective data regarding a low VTE risk diet [32]. However, the lack of decreased vascular disease risk in Denmark during and after the Second World War (see above) [35] might be considered a point in favor of a vegan diet.

A study from Rotterdam assessed the relationship of dietary fat and fiber with coagulation factor VII in 3,007 elderly men and women subjects. Total fat and saturated fat intake were significantly associated with factor VIIc only in women. Fiber intake was inversely associated with factor VIIc in both men and women [59].

Mediterranean, vegetarian or vegan diets could all be reasonable choices for the experimental arm of non-inferiority trials with anticoagulants for prophylaxis and treatment of VTE. Given the challenges to patients in changing diets and the likelihood that RCTs would show that anticoagulation prophylaxis and treatment for VTE are themselves ineffective as suggested in the background section above, RCTs of VTE patients should allow considerable flexibility in the diet interventions so that they are not burdensome.

Example--RCT for VTE prophylaxis of hospitalized medical inpatients: Low VTE risk diet versus standard anticoagulants

For a hypothetical example anticoagulant prophylaxis RCTs, consider acutely ill patients admitted to hospital medical wards. American Association of Chest Physician (ACCP) Guidelines suggest that such patients receive LMWH, low dose heparin, or fondaparinux as VTE prophylaxis¹¹. We may use total bleeding (minor, major, and fatal bleeding) as the primary safety endpoint. A recent Cochrane Review of prophylactic anticoagulation for medical patients found that bleeding occurred in 117/2,405 (4.9%) of anticoagulated patients and 62/2,404 (2.6%) of untreated patients [87].

Subjects would be recruited from patients admitted to acute medical services who have VTE risk factors qualifying them for anticoagulant prophylaxis according to AACP Guidelines [11].

Subjects meeting the inclusion/exclusion criteria would be randomized to receive either a standard anticoagulant per AACP Guidelines (i.e., LMWH, low dose heparin, or fondaparinux) or a low-VTE-risk diet regimen. The diet (Mediterranean, vegetarian or vegan per the patient's choice) would be recommended for the entire duration of the trial, but compliance would be left to the discretion of the patient and treating physician. This flexibility in the diet arm assures that the diet intervention would not be a burden to patients. Flexibility in administering the low VTE diet is reasonable, since the RCTs are in large part testing the safety and efficacy of standard anticoagulation prophylaxis and treatment, and the low VTE diet would not be expected to do harm.

All outcomes will be recorded into the patients' medical records. In addition, follow-up RCT-related visits or calls as appropriate will be conducted approximately every 10 days up to 60 days after anticoagulation prophylaxis is completed, to determine if any of the primary or secondary safety or efficacy outcomes had occurred. Compliance with the low VTE risk diet will also be assessed, asking each patient to estimate his/her compliance on a scale of 0 - 100, with '0' representing completely non-compliant and '100' representing completely compliant. If follow-up takes place before discharge from the hospital, the information will be entered into the patient's medical record and separate data collection forms; if the follow-up takes place after discharge, the information will be recorded on to separate data collection forms.

The data collection forms will be assessed by study monitors at each participating medical center and reviewed by the responsible investigator at each medical center before anonymized data are sent to the study PI.

As noted above, we would use the overall death rate as the primary efficacy endpoint. Since the aforementioned Cochrane Review showed that prophylactic anticoagulation does not significantly reduce FPE rates or overall deaths in acute medical patients (for anticoagulation versus placebo, FPE: 0.94 [0.82, 1.07] and overall deaths: RR 0.95 [0.83, 1.07] [87]), a 50% non-inferiority tolerance is reasonable. The death rate in this Cochrane review was about 5% with or without anticoagulant prophylaxis. So with anticoagulant prophylaxis, the primary safety endpoint (overall bleeding) and the primary efficacy endpoint (overall mortality) would each be expected to occur in about 5% of patients. Expecting that the absence of anticoagulation and a low VTE diet can reduce the incidence of the total bleeding safety endpoint down to 2.5% with no change in efficacy, about 1,400 subjects would be required in each treatment group in order to have 80% power to meet both co-primary hypotheses (superior safety and non-inferior efficacy).