PART II:
Pathophysiology and Complications
Chapter 6:
Clinical Biostatistics of Safety

Cosmetic surgical procedures must be judged in three dimensions: (1) patient safety (does the procedure expose the patient to any unnecessary risks?), (2) ethical propriety (is the procedure truly in the patient’s best interest, or would the patient be better served without the procedure?), and (3) aesthetic results (are surgical outcomes beneficial to the patient?). Patient safety has precedence over the other two and all other considerations.

How can a surgeon objectively decide when a cosmetic surgical procedure is safe and when it is not? Risk-benefit analysis is one of the surgeon’s most basic and recurring activities. A comparison between two methods to prevent perioperative deep venous thromboses is a straightforward example of comparing the risks and benefits of alternative procedures. When considering drug safety, however, the analysis and decisions are not the simple “yes” or “no” results of comparing two alternatives. Drug safety is a more subtle assessment and involves a toxicologic dose-response analysis.

What is safety? Dictionary definitions of safety, such as “the state of being safe” or “freedom from injury,” have little clinical usefulness. The clinical distinction between safe and unsafe is cloudy, with no definite boundary line. In medicine and surgery the concept of safety is most appropriately defined in terms of probability and statistics.

Dose-Response Phenomenon

Tumescent liposuction has two potential sources of danger: too much lidocaine and excessive liposuction. The more lidocaine or the more liposuction, the greater is the danger. Either situation can be represented mathematically by doseresponse function (see later discussion).

What is a safe amount of lidocaine? A surgeon’s administration of more than 85 mg/kg body weight of lidocaine on numerous occasions without “serious complications” does not prove this is a safe practice. Which of the following local lidocaine doses is ethically preferable: 90 mg/kg for a single surgery, at a risk of one death in 1000 surgeries, or 45 mg/kg for each of two liposuction procedures, each performed 1 month apart, at a risk of one death in 100,000 surgeries for each procedure?

What is a safe amount of liposuction? Although a single anecdotal report can easily disprove the safety of a procedure, it requires a much greater sample size to prove safety. A surgeon’s removal of more than 5 L of supranatant fat by liposuction on numerous occasions does not prove the safety of this procedure. Which of the following is safer: a single megaliposuction of 9 L of fat, at a risk of one death in 1000 surgeries, or three liposuction procedures, each removing 3 L of fat and performed at 1-month intervals, at a risk of one death in 100,000 surgeries for each procedure?

In cosmetic surgery, especially liposuction surgery, questions about safety often are not determined by scientific method. The philosophic and conceptual aspects of safety are a challenging part of biostatistics. For example, when determining the maximum safe dose of lidocaine with tumescent liposuction, the surgeon might simply treat a few patients whose mg/kg doses of tumescent lidocaine were not predetermined, then do linear regression on mg/kg lidocaine dose versus serum lidocaine concentration. Such an approach gives the illusion of science, but it is no more helpful than a “clinical guess.”

Experimental Toxicology

The experimental design of a scientific study of toxicity using experimental animals is distinctly different from a clinical study of toxicity in human subjects. It is instructive to have a conceptual understanding of experimental toxicology using animals before discussing clinical toxicology in humans.

Observational linear regression merely provides an estimate of safe tumescent lidocaine dose. Any estimate with a large but unknown standard deviation (or variance) is of minimal clinical utility when it involves potentially lethal doses of a drug used for cosmetic surgery. The smaller the standard deviation of the estimate of a parameter’s numeric value, the greater is the accuracy of the estimate. Predictions based on observational linear regression are much less accurate than experimental linear regression.

Experimental linear regression provides more accurate results by giving the same mg/kg dose of lidocaine to a large number of experimental animals (e.g., mice) and provides good estimates with small variance. For example, a specified dose, d_i (mg/kg), results in a fatality with a probability, p_i. Thus, if 10 mice are given a subcutaneous dose (d_i) of lidocaine and observe n_i lethal outcomes, n_i/10 is an estimate of the true p_i that d_i will cause a death. The more mice that are given the same d_i, the more accurate (smaller standard deviation) will be the estimate of p_i. If 1000 mice are given the same d_i, the frequency of death as N_i/1000 is a more accurate estimate of p_i than the first estimate, n_i/10.

The conceptual framework of animal toxicology is important for an appreciation of clinical situations involving patients. A typical toxicologic study might use mice to determine the probability that any chosen dose will produce a toxic reaction. Suppose P(d) is the probability of toxic reaction T occurring at dose d (mg/kg) of lidocaine, where 0≤P(d)≤1 by definition. Although we cannot ever know the true probability that d will cause a toxic reaction, we can obtain an estimate of P(d) by counting the frequency p(d) = n/100, where n is the number of toxic reactions that occur when 100 mice are given dose d.

If we select 10 different doses (d₁, d₂, d₃ . . . , d₁₀) of lidocaine and estimate the probability of a toxic reaction p(d_i) at each dose, a plot of the graph of p(d_i) will be a sigmoid, S-shaped curve (Figure 6-1).

For example, if we define T as a peak serum lidocaine level of 6 μg/ml or greater, the graph of p(d) will help us estimate the chance that any given dose will result in a serum lidocaine concentration greater than 6 μg/ml. This graph allows us to derive a quantitative definition of safety in terms of potentially toxic serum levels. Thus we might define safe to mean that less than one mouse in 100 will have a serum lidocaine concentration greater than 6 μg/ml.

Similarly, if we want to study the probability that any given dose of lidocaine will kill a mouse, we can repeat the previous study by defining T as death. In this case we might define safe dose as the dose expected to yield less than one death per 100 mice. The graph of p(d) is also used to estimate the median lethal dose (LD₅₀) for lidocaine in mice, which is the minimum dose that has a 50% probability (p = 0.5) of killing a mouse.

Human Toxicology

When researching lidocaine toxicity among humans, any study design using experimental linear regression is unlikely to find many volunteers. Because toxicity experiments using humans are unacceptable, we must glean as much information as possible by simply observing the real world.

When studying human toxicology, epidemiologic methods provide an ethical way of obtaining reasonably accurate data on dose-response phenomena. When studying a drug that might cure a fatal disease, we can ethically consider a drug dose with a 1 in 10 chance of lethal toxicity. In this setting a clinical study involving a relatively small number of patients is reasonable. In cosmetic surgical procedures, however, ethical tolerance for lethal toxicity is much lower. This is where moral philosophy interfaces with cosmetic surgery. How much of a risk of fatal toxicity are we willing to accept? Which of the following thresholds represent an acceptable probability that a given total dose of lidocaine will result in a patient’s death?

1 in 10 patients

1 in 100 patients

1 in 1000 patients

1 in 10,000 patients

1 in 100,000 patients

1 in 1 million patients

After agreeing on the level of acceptable risk, we can then discuss whether or not a given surgical procedure is “safe.”

Toxicity Estimates. Without human experiments, we cannot generate enough experimental data to construct a sigmoid dose-response curve. Nevertheless, by understanding the principles behind the curve, we can better understand how safety is defined quantitatively. The clinical approach to lidocaine toxicity in humans is based more on pragmatism and art than on precise science.

In the experimental animal study the likelihood of toxicity was estimated by calculating the frequency n/100. In this case, the experimenter specified the denominator as 100 mice, then determined the number n by giving dose d to each of 100 mice.

In humans, typically the numerator is specified as n = 1, and the denominator is determined by epidemiologic observation. For example, suppose I have never observed a patient with a potentially toxic serum lidocaine concentration of 6 μg/ml. Then, after giving a patient 60 mg/kg, I find that the patient’s serum lidocaine concentration is 6 μg/ml. In my experience with tumescent lidocaine, I have given 60 mg/kg or more to about 300 patients. The 1 in 300 is a rough estimate of the incidence of lidocaine toxicity at a dose of 60 mg/kg.

At present in the United States, it is virtually impossible to conduct epidemiologic studies of iatrogenic disasters (see Chapter 5). Not revealing such data is systematically built into the U.S. legal system, and peer review often makes it illegal to comment publicly about another surgeon’s complications. The only clinical solution is an expert’s best estimate.

The following estimates of liposuction-related surgical mortality are based on my subjective experiences as a physician with training in epidemiology and biostatistics (Table 6-1). Objective epidemiologic data would show that the true risk of death is less for the tumescent technique totally by local anesthesia but greater for the superwet technique (see Chapter 9).

The true tumescent technique for liposuction uses no intravenous (IV) fluids, with supranatant volumes of fat less than 4 L and less than 4% of the patient’s weight. Let us make the hypothetic supposition that one unreported death has been associated with tumescent liposuction (as a result of a pulmonary embolism). Considering the number of patients worldwide who have had true tumescent liposuction, 1:500,000 is probably a conservative estimate of mortality.

The superwet technique for liposuction uses systemic anesthesia, significant IV fluid infusions, and a 50% smaller volume of subcutaneous infiltration. The mere exposure to prolonged use of general anesthesia or IV sedation/analgesia has an estimated mortality rate of 1:10,000 to 1:20,000 in healthy ASA (American Society of Anesthesiologists) class I patients. A recent survey of surgeons who use the superwet technique reported that the risk of death associated with liposuction under systemic anesthesia is 1:5000.¹ Among surgeons who limit volumes of supranatant fat to less than 4 L and do not use IV infusions, the mortality should be significantly less than 1:10,000.

The extreme superwet technique for liposuction is significantly more dangerous than the true tumescent technique totally by local anesthesia. It exposes the patient to excessive volumes of liposuction (supranatant fat exceeding 4 L), excessive number of areas liposuctioned in a single day, routine infusion of multiple liters of IV fluids, use of bupivacaine instead of lidocaine, surgery of excessive duration, excessive number of ancillary surgical procedures, and excessive amounts of anesthetic agents. I estimate that approximately 100 deaths have been associated with the extreme superwet technique or with ultrasonic liposuction.

Megaliposuction combines all the high-risk factors for liposuction and inflicts them on a single, usually naive and unsuspecting patient. Its value and safety have never been established. It is an experimental procedure without ethical oversight by a human studies research committee or an institutional review board. Megaliposuction exposes patients to the greatest risks of liposuction syndromes, which involve pulmonary embolism and pulmonary edema.

Liposuction volume limits

What is the LD₅₀ for liposuction? The volume limits of safe liposuction must be defined, but it is simplistic to measure the degree of surgical risk based solely on the volume of aspirated supranatant fat. Which is more dangerous: removing 3 L of fat from a 50-kg woman or 6 L from a 100-kg woman, where each patient has the same lean body mass?

Several factors help to determine the amount of liposuction that can be safely performed in a single day of surgery (Box 6-1). Despite the lack of a well-defined threshold for safety, subjective criteria can be defined for liposuction safety. I feel more secure about safety when my patient is alert and conversing during surgery. When the patient is alert and smiling and can walk out of the surgicenter 30 minutes after liposuction, I have some confidence that the surgery has not been too aggressive. When a patient can return to work within 1 or 2 days after surgery, I am confident that I have not treated too many areas or removed too much fat.

The total volume of supranatant fat is not the only determinant of safety, but it does correlate with the length of postoperative recovery and the risk of complications. The following definitions of liposuction volumes of supranatant fat facilitate communication among surgeons:

1. Small-volume liposuction: less than 100 ml. Small-volume liposuction with tumescent local anesthesia is probably as safe as any minor dermatologic surgery. The risk from small-volume liposuction is comparable to removing an equivalent volume of fat by the excision of multiple lipomas.
2. Medium-volume liposuction: 100 to 1500 ml. With traditional liposuction, aspiration of more than 1500 ml of “stuff” was widely regarded as an indication for an autologous blood transfusion. With current tumescent liposuction, blood loss with medium-volume liposuction should be insignificant. Surgeons with limited experience should not remove more than 1500 ml of fat in a single day.
3. Large-volume liposuction: 1500 to 4000 ml. Large-volume liposuction was revolutionized with the widespread recognition of the profound hemostasis provided by tumescent vasoconstriction. A patient’s ease of recovery is inversely related to the volume of aspirated fat. In my practice, liposuction of more than 3000 ml in a single day is uncommon; liposuction of more than 4000 ml is extremely rare. The belief that “do-it-all-at-once” surgery will minimize recovery time is a fallacy. Patients who have had liposuction of 2000 ml of supranatant fat require approximately 25% of the recovery time needed by patients with liposuction of 4000 ml.
4. Extremely large-volume liposuction: 4000 to 7000 ml. Removing more than 4 L of supranatant fat in 1 day is relatively unsafe; it is safer to divide the procedure and perform it on 2 different days separated by several weeks or months. Extremely large-volume liposuction has been appropriately described as “beyond the pale”; it is “beyond bounds” and beyond the pale of safety. Surgeons who assert that extremely large-volume liposuction is safe because many other surgeons boast of doing it are instruments of the consensus gentium fallacy (see Chapter 7).
5. Megaliposuction: (more than 7000 ml). Megaliposuction is “licentious” in the sense of disregarding commonly accepted rules, deviating freely from correctness, and overstepping customary limits.

Lidocaine dose limits

To verify that 55 mg/kg yields safe peak serum lidocaine concentrations, a researcher would have to repeat the study in a large number of volunteers. This requirement makes such an achievement unrealistic even if the effort is shared by multiple researchers. If we accept a conservative estimate for a maximum safe tumescent lidocaine dose, however, a large study is probably unnecessary. Again, we can obtain considerable information from simple clinical observations.

Two fatalities have occurred in patients who inadvertently received 105 mg/kg of lidocaine, as well as excessive IV fluids. In each case the coroner diagnosed pulmonary edema and lidocaine toxicity. Therefore any estimate of the maximum safe dose of tumescent lidocaine will be significantly less than 100 mg/kg.

One surgeon (who shall remain anonymous) gave lidocaine doses of 70 to 90 mg/kg to more than 10 patients and reported that 30% experienced nausea or vomiting. Clearly this range of tumescent lidocaine is associated with an unacceptably high incidence of clinical toxicity.

I now recommend lidocaine doses of 50 mg/kg or less, with a strict maximum of 55 mg/kg. I have measured the peak serum lidocaine concentration in approximately 20 patients after tumescent lidocaine doses in the range of 55 to 65 mg/kg. All patients had peak serum concentrations of less than 3.5 μg/ml. In addition, we have treated more than 400 patients with doses of approximately 55 mg/kg without clinical toxicity.

In one patient, 60 mg/kg of tumescent lidocaine with liposuction was associated with an episode of clinical toxicity (nausea and disorientation). Serum lidocaine concentration approximately 12 hours after infiltration was 6.1 μg/ml. In this case a probable lidocaine drug interaction with sertraline (Zoloft) may have been mediated by competitive inhibition of the hepatic microsomal enzyme cytochrome P450 3A4 (CYP3A4). Because serum lidocaine levels greater than 6.0 μg/ml are considered potentially toxic, we must assume that a safe maximum lidocaine dosage is less than 60 mg/kg.

Another patient had three serial tumescent liposuction procedures over 1 year. At the first and second surgery the patient received 58.3 and 49.7 mg/kg of lidocaine, respectively, without incident. At the third surgery, however, the lidocaine dose of 55.3 mg/kg produced a 3-hour episode of disorientation, confusion, short-term memory loss, and ataxia. The patient was taken to an emergency room, and the serum lidocaine concentration was 5.0 μg/ml. Interestingly, the day before surgery, the patient had completed a 10-day course of the antibiotic clarithromycin (Biaxin), which inhibits CYP3A4.

Based on these experiences, I would recommend that lidocaine doses be minimized and, again, should never intentionally exceed 50 mg/kg. Furthermore, drugs that interfere with CYP3A4 should be discontinued 1 or 2 weeks before surgery (see Chapter 18).

Finally, the maximum safe dose of tumescent lidocaine should be determined with serial measurements of blood concentration when there is no concurrent liposuction. Sequential serum lidocaine concentrations measured in the same patient after equal tumescent lidocaine doses are approximately 20% lower with liposuction than without liposuction. Because some disruption might prevent completion of liposuction after infiltration of lidocaine, the surgeon would not want to give a relatively high dose of tumescent lidocaine and then be obligated to do liposuction in order to avoid lidocaine toxicity.

To reiterate, my current estimate for a safe maximum tumescent lidocaine concentration is 50 mg/kg. Greater dosages, as high as 55 mg/kg, should be avoided.

Summary

One case report is often sufficient to prove a procedure is unsafe. An anecdotal report of 100 cases without complications, however, is insufficient to prove a procedure is safe. A huge sample size is necessary to prove safety.

When is it reasonable to be enthusiastic about a new procedure? If a new procedure has advantages in terms of increased safety and improved clinical results, a cautious test of the new technique is warranted. When tumescent liposuction was first introduced, the advantages in safety were the elimination of significant surgical bleeding and the elimination of risks of general anesthesia. Surgeons were correct to withhold judgment, however, until the tumescent doses of lidocaine were shown to be safe based on extensive clinical experience.

When is it reasonable to be skeptical about a new procedure? Clearly, if a procedure offers no cosmetic advantages over an existing technique and is potentially more dangerous, skepticism is in order. The mere absence of unfavorable reports does not prove safety. For example, internal ultrasonic liposuction has yet to be proved safe. The dismal European experience with ultrasonic-assisted liposuction (UAL) should be a warning not to jump on the U.S. media bandwagon that has promoted UAL (see Chapter 29).

Reference

1. Glazer FM, de Jong RH: Fatal outcomes from liposuction: census survey of cosmetic surgeons, Plast Reconstr Surg 105: 436-446, 2000.

Figure 6-1 Estimating median lethal dose (LD₅₀), the minimal dose that can be expected to kill approximately 50% of treated individuals. In other words, LD₅₀ is minimum dose that has a 50% chance, or a probability p = 0.5, of killing a mouse. In this graph, half the mice can be expected to have a fatal reaction to a dose equivalent to d₆. Thus LD₅₀ = d₆. Analogously, minimal dose that can be expected to kill 10% of mice is d₂.

BOX 6-1 Factors in Determining Safety of Liposuction

Number of areas treated

Volume of supranatant fat removed

Percent of body fat removed

Ratio of body weight to the weight of fat removed

Dosage (mg/kg) of lidocaine

Volume of intravascular fluid infused

Duration of surgical procedure

TABLE 6-1 Estimates of Liposuction-related Mortality Rates
Technique*	Mortality Estimate
True tumescent	1:500,000
Superwet	1:5000
Extreme superwet	1:1000
Megaliposuction	1:100
*See text for descriptions.