Chronic hyperlactatemia in HIV-infected patients taking antiretroviral therapy

Objective: To determine the prevalence, course and risk factors for hyperlactatemia in HIV-infected patients.

Design: A prospective, longitudinal study of venous lactate concentrations over an 18-month period in 349 participants of the Western Australian HIV Cohort Study.

Results: In 516 patient-years of observation, two patients experienced severe fulminant lactic acidosis (lactate > 5 mmol/l) and hepatic steatosis attributable to nucleoside analogue reverse transcriptase inhibitors (NRTI). A further five patients with lesser elevations of lactate (2.8-4.1 mmol/l) but with symptoms of nausea or abdominal discomfort and evidence of hepatic steatosis had NRTI therapy revised, with relief of symptoms and a fall in lactate levels. Most remaining patients on highly active antiretroviral therapy (HAART) had mild, chronic, asymptomatic hyperlactatemia, with mean lactate level between 1.5 mmol/l and 3.5 mmol/l most commonly. Longitudinal data was analysed in a non-linear mixed effects growth model which indicated that average lactate levels rose after the start of HAART but tended to stabilise at low-grade elevation, with an average 0.23 mmol/l greater long term level in stavudine users compared with zidovudine users (p < 0.01). A multiple linear regression model showed that the association between stavudine and higher lactate level was not confounded by longer duration of total NRTI exposure. Risk of hyperlactatemia was not significantly associated with use of other NRTIs, protease inhibitors, non-nucleoside analogue reverse transcriptase inhibitors or multiple immunological and virological factors in multivariate analyses.

Conclusions: Chronic, compensated, asymptomatic hyperlactatemia is common in patients taking HAART. Decompensated, life-threatening lactic acidosis/hepatic steatosis is rare. Treatment with stavudine appears to be the predominant risk factor for development of chronic hyperlactatemia.

Introduction

Acute, severe lactic acidosis is a rare but potentially fatal complication of treatment with nucleoside analogue reverse transcriptase inhibitors (NRTI) in HIV-infected patients ^[1-5]. Severe lactic acidosis is typically symptomatic with nausea, vomiting, severe malaise and prostration and may occur precipitously after months or even years of NRTI treatment ^[5]. Recently, mild or moderate lactate elevation in NRTI-treated individuals has been reported ^[6-9]. The natural history of mild to moderate hyperlactatemia and its relationship to the risk of severe, life-threatening lactic acidosis is not known.

As lactate is the product of anaerobic glycolysis, hyperlactatemia in normal aerobic conditions may indicate mitochondrial dysfunction ^[10]. The mitochondrial basis of NRTI-induced hepatic steatosis, lactic acidosis and myopathy is well established, and several other adverse effects such as peripheral polyneuropathy and cardiomyopathy have also been linked to mitochondrial toxicity ^[11]. It has been hypothesized that some features of the 'lipodystrophy syndrome' are also tissue-specific mitochondrial toxicities caused by NRTI treatment ^[12-14]. As hyperlactatemia may be a correlate of mitochondrial dysfunction, we sought to determine the prevalence, longitudinal course and risk factors for the development of hyperlactatemia in a large HIV-infected population.

Methods

Venous lactate concentration was measured in 349 participants of the Western Australian HIV Cohort Study ^[15] as part of routine outpatient review from January 1999 to June 2000. Visits typically occurred at intervals of 1-3 months for all patients. Clinic phlebotomists were instructed to ensure that patients were at rest for at least 10 minutes prior to extraction of a heparinized venous blood sample without a tourniquet or fist clenching. Samples were transported immediately to an adjacent laboratory and measured in a lactate biosensor using amperometric technology within minutes of collection. If there was a delay in immediate processing, the sample was put on ice and assayed within 20 minutes.

Comprehensive demographic, clinical and laboratory data are routinely collected on all participants of the Western Australian HIV Cohort Study, which was established in 1983 ^[14,15]. The variables relevant to this study include age; sex; body mass index; HIV- and non-HIV-related illnesses, including hepatitis B and C coinfection; history of antiretroviral drugs, including reason for therapy revision; history of prophylactic and non-HIV-related medications; serial CD4 T cell counts and plasma HIV RNA concentration [HIV Amplicor (Roche, Branchburg, USA), limit of detection < 400 copies/ml until November 1999, then Roche Amplicor HIV monitor Version 1.5 Ultrasensitive, limit of detection < 50 copies/ml]; and full blood picture and serum biochemistry, including liver function tests. All laboratory-derived data are electronically downloaded into the cohort study database.

The full study cohort comprised 349 active patients with 1379 lactate measurements between January 1999 and June 2000. Though lactate measurements done prior to January 1999 were recorded in the cohort database, they were not used in any analysis in this study. Highly active antiretroviral therapy (HAART) was defined as two NRTIs (stavudine or zidovudine with either lamivudine or didanosine) and a protease inhibitor (PI) or two NRTIs and abacavir or two NRTIs and a non-nucleoside reverse transcriptase inhibitor (NNRTI). Five patients who presented during the study period with acute symptomatic severe lactic acidosis were described separately and excluded from subsequent analyses. Cross-sectional analysis was restricted to those patients who were antiretroviral naive or those who had been on their current HAART regimen for at least 30 days at the time of their lactate measurement. Patients who were not on any current therapy (but were not treatment naive) or those not on HAART as defined above were not included in these analyses. The multiple linear regression model and non-linear mixed model of lactate concentrations over time were further restricted to only HAART-treated patients.

A review of the medical records of all patients who had HAART revised during the study period was undertaken to confirm reason(s) for revision, as recorded in the study database.

Statistical analyses

Cross-sectional comparisons and linear mixed model analyses were carried out using the SAS statistical package (SAS Institute, Cary, North Carolina, USA). In cross-sectional analysis, the mean of the consecutive lactate measurements taken while the patient was on current therapy was used as a single value for each individual. ANOVA was used to assess significant differences in the averages between treatment groups. For the longitudinal analysis, both mixed effects linear and non-linear models of the growth in lactate measurements were used. When considering time from HAART, a non-linear growth model of the form lactate = a [1 - b exp(-kt)] was fitted using the S-PLUS statistical package ^[16]. The coefficients a and b in this model were assumed to be random, varying about a population average according to the individual, but k was taken to be fixed across individuals; t was time. The model allows individual variation in initial lactate measurements and in the value at which the measurements may level off over time. Estimates of the population averages of these parameters for each treatment regimen allowed comparison of the trends using standard large-sample Wald tests. Linear mixed models comparing effects after the initial increase in lactate levels were fitted using SAS.

Results

Patient characteristics and distribution of current antiretroviral treatment is shown in Table 1.

Table 1

In 516 patient-years of observation and routine lactate measurement, five patients, three male and two female, had lactate elevation above 5 mmol/l (laboratory healthy reference range 0.3-1.3 mmol/l). One male patient had non-Hodgkin's lymphoma with widespread hepatic metastases and multiorgan failure. Another male patient had advanced AIDS, refractory cytomegalovirus encephalomyelitis and cardiomyopathy. Neither of these patients were taking antiretroviral therapy and it was thought that the lactic acidosis was a consequence of their underlying illnesses. The third male patient had a history of alcohol abuse, alcohol-induced hepatitis and had one lactate level of 9.3 mmol/l after acute alcohol intoxication. He was not compliant with his prescribed antiretroviral therapy during this time. His lactate levels subsequently declined rapidly to < 2.00 mmol/l. The two remaining patients, both female, experienced typical NRTI-induced lactic acidosis/hepatic steatosis, giving an incidence rate of 3.9 per 1000 person-years. In one, symptoms of nausea and abdominal discomfort began 6 months after commencing stavudine and nelfinavir and were rapidly progressive. Venous lactate concentration was 6.4 mmol/l at diagnosis but was 1.6 mmol/l 1 month earlier and was 1.9 mmol/l 5 months earlier (when asymptomatic). The other female patient had been taking stavudine, lamivudine and nelfinavir for 1 year. Acute-onset severe malaise, vomiting, abdominal pain and clinical hepatomegaly prompted urgent lactate measurement, which showed the patient had a lactate concentration of 8.2 mmol/l. There was also massive hepatomegaly on abdominal ultrasound. No prior lactate measurements were available for this patient; however, the calculated anion gap in serum samples taken a month before presentation with fulminant lactic acidosis was within normal range. Both patients with severe NRTI-induced lactic acidosis/hepatic steatosis had serum bicarbonate < 20 mmol/l and had NRTI permanently discontinued. The multiple lactate values of all five patients with severe lactic acidosis were removed from all subsequent analyses.

A further five patients in the cohort had revision of NRTI by their physician during the study period because of symptoms that were thought to be related to moderate hyperlactatemia (2.8-4.1 mmol/l) and/or hepatic steatosis. All patients had nausea and/or abdominal discomfort, abnormal liver function tests and an appearance suggestive of hepatic steatosis on abdominal ultrasound. All five had their stavudine replaced by zidovudine or abacavir, with relief of symptoms and subsequent fall in lactate levels, though two patients had transient rises to 4.2 mmol/l at 9 months later and 3.5 mmol/l at 11 months later, respectively. No patients in the cohort had revision of HAART because of asymptomatic hyperlactatemia detected on study monitoring tests.

There was an average of 4.2 consecutive lactate measurements taken per patient during the 18-month observation period. The mean of consecutive lactate measurements for each individual was calculated. Of the 344 patients who never had a lactate level > 5 mmol/l, 39.8% had a mean lactate concentration between 1.5 and 2.5 mmol/l, 4.4% between 2.5-3.5 mmol/l and none between 3.5 and 5 mmol/l during the study period. However, 65.0% had a lactate concentration above 1.5 mmol/l, 18.3% above 2.5 mmol/l and 5.7% above 3.5 mmol/l on at least one occasion.

Mean lactate concentrations on current therapy were compared between various HAART regimens, stratified by drugs and drug classes that are used in mutually exclusive groups (see Table 2). In an analysis of combined groups, all patients treated with stavudine-containing HAART (with either a PI or a NNRTI) had a higher mean lactate concentration (n = 140; 1.65 mmol/l) compared with those on zidovudine-containing HAART (n = 101; 1.45 mmol/l) and those who were antiretroviral naive (n = 73; 1.34 mmol/l) (P < 0.05, ANOVA). There was no significant difference between zidovudine-treated and treatment-naive patients. Mean lactate concentrations associated with use of lamivudine versus didanosine and PI versus NNRTI versus abacavir (matched for concurrent NRTI) were not significantly different. Similarly there were no significant differences between those on different individual PIs or individual NNRTIs (data not shown).

Table 2

The longitudinal profile of lactate levels was analysed and plotted using the non-linear mixed effects growth model described earlier. Figure 1 shows that population average curves of lactate values rise from time of starting a HAART regimen. Patients were stratified into stavudine- or zidovudine-treated groups and data were included only for patients who did not change their treatment regimen. There was no significant difference in the average lactate values at the commencement of HAART (P = 0.67; combined estimate 0.74 mmol/L, SE = 0.24 mmol/l). However, there was a significant difference between the estimated long-term population average lactate concentrations (1.63 mmol/l in stavudine users, 1.40 mmol/l in zidovudine users: difference 0.23 mmol/l; P < 0.01). Individual averages varied about these population averages with standard deviations of 0.47 mmol/l for those using stavudine and 0.44 mmol/l for zidovudine.

Fig. 1

In view of the time taken for equilibration of lactate concentration after starting HAART, a multiple linear mixed model with individual specific coefficients was used to identify the predictors of venous lactate concentration with the data values restricted to only those obtained after at least 9 months of therapy. This negated possible effects of rising unstable levels of lactate in the early months of HAART. Multiple potentially predictive immunological and viral factors were examined in the model. Treatment with stavudine was associated with an estimated 0.36 mmol/l higher lactate concentration compared with zidovudine (P = 0.003), consistent with the finding from the growth model. However, no other variables were found to be significantly associated with chronic hyperlactatemia, including use of PI or NNRTI, age, sex, AIDS, duration of AIDS and HIV infection, number of past opportunistic infections, nadir CD4 cell count, peak plasma HIV RNA concentration, hepatitis B or hepatitis C coinfection, mean corpuscular volume and serum alanine aminotransferase. Adjustment for duration of total NRTI exposure or duration of past zidovudine use in stavudine users did not abrogate the results of the analysis, indicating that the increased risk associated with stavudine use was not confounded by the longer history of NRTI exposure in a proportion of stavudine users.

Discussion

Under normal conditions in aerobic tissues, the pyruvate generated by glycolysis is largely metabolised oxidatively in mitochondria. Lactate is the end-product of cytoplasmic glucose metabolism and lactate production is favored in anaerobic conditions. At rest, lactate concentration in venous blood reflects the equilibrium maintained between the rate of lactate production in several metabolically active tissues (in turn dependent on mitochondrial function and cellular oxygen levels) and the rate of lactate utilization by liver and renal cortex ^[17]. Lactate concentration is normally maintained within a narrow physiological range, with constant (basal) turnover estimated at 1 mEq/min (1 mmol/min). However, under various conditions of lactate excess, hepatic and renal clearance of lactate may be augmented and other organs that are normally lactate producers, such as skeletal muscle, can become important lactate consumers. Hence large (up to 100-fold) acute elevations in lactate after vigorous exercise are rapidly re-equilibrated, indicating the efficiency of lactate regulation, which is thought to be part of a generalized homeostatic system regulating organic acids via systemic pH ^[17].

It is possible that continuous and chronic excess of lactate could lead to a progressive increase in the 'set-point' of homeostatic control, as in the case of other tightly regulated biological molecules. Though the majority of patients in this study had lactate levels above the reference range at some time during the observation period, average levels did not stay at 3.5-5 mmol/l in any patient. Chronic mild hyperlactatemia with lactate concentrations of 1.5-3.5 mmol/l, maintained over many months, was the most common pattern of hyperlactatemia observed. The longitudinal data confirm that lactate concentrations do rise overall from the pretreatment baseline but are largely maintained at only mildly elevated levels in the long term. Presumably, homeostatic mechanisms are preserved enough to compensate, at least partially, for lactate excess in most patients. The primary cause(s) of this net rise in lactate - increased production and/or reduced clearance - is not known. A study of exercise physiology has shown that skeletal muscle is probably not a major source of excess lactate production in NRTI-treated HIV-infected patients, as their oxidative phosphorylation, lactate/pyruvate production and lactate clearance following exercise was not significantly different from matched controls ^[18].

In contrast to the common form of mild stable hyperlactatemia, severe lactic acidosis appears to indicate an extreme, decompensated metabolic state in which homeostasis is lost completely. Notably, NRTI-induced severe lactic acidosis is almost always accompanied by massive hepatic steatosis and frequently by hepatic failure ^[1-5]. The liver has limited lactate clearance capacity in these circumstances and hence becomes a net lactate producer. The extent of mitochondrial toxicity in hepatocytes may be the key determinant of risk of decompensation into severe lactic acidosis in NRTI-treated patients ^[19].

The predominant risk factor for developing chronic hyperlactatemia was the use of stavudine compared with zidovudine; this was found in three separate analyses and after adjustment for the potential confounding effect of duration of past and total NRTI exposure. Moreover, the change from stavudine to zidovudine in five patients with moderate hyperlactatemia may make our estimates of the differences between the two drugs conservative. Differences between zidovudine and stavudine are most easily detected in observational cohort data, as they are never used together. The independent contribution of concurrent agents in HAART may be difficult to determine. However, we found no significant difference in risk of chronic hyperlactatemia between other largely 'mutually exclusive' or competing antiretroviral drugs, namely lamivudine versus didanosine, PI versus NNRTI versus abacavir, individual PI and individual NNRTI. This suggests that these agents have differences that are too small for our analyses to detect, have no contribution to hyperlactatemia or all have exactly equivalent effects. It is possible that we did not have the power to detect the contribution of agents used by only a small number of patients, such as didanosine, abacavir and efavirenz.

The risk profile of hyperlactatemia mirrors that of some features of the 'lipodystrophy syndrome' in our cohort. The possible mitochondrial basis of lipodystrophy is, in part, predicated on the observation that use of NRTI is a sufficient condition, and an independent risk factor, for the development of progressive subcutaneous fat wasting, 'buffalo humps', breast enlargement and focal lipomatoses in HIV-infected patients ^[14,20-22]. In our previous analysis of the predictors of progressive subcutaneous wasting in the Western Australian HIV Cohort, longer duration of NRTI therapy conferred a significantly greater risk of developing both clinically apparent fat wasting and lower subcutaneous limb fat measured by dual energy X-ray absorptiometry (DEXA). Within the NRTIs, stavudine was consistently associated with a greater risk of fat wasting compared with zidovudine ^[14]. Though stavudine is also the predominant risk factor for chronic hyperlactatemia, a causal relationship between chronic hyperlactatemia (as a correlate of mitochondrial toxicity) and lipodystrophy is not proven by these findings. It is possible that lipodystrophy and hyperlactatemia are both associated with stavudine but are pathogenically completely distinct or reflect different tissue-specific effects of stavudine and other NRTIs. Our previous study also suggested that fat wasting may be related to an interaction between PI and NRTI drugs, either by distinct but additive mechanisms or by PI-mediated exacerbation of a slower (mitochondrial) NRTI toxicity ^[14]. The lack of any detectable influence of PI on lactate concentrations argues against the latter possibility.

It is unlikely that the population distribution of lactate values is skewed to a higher range only because more measurements were taken in those with very high lactate levels. All patients who ever experienced a lactate level > 5 mmol/l were highly symptomatic and had lactate measures taken outside of the routine study protocol. These patients' lactate values were removed from all subsequent analyses, leaving predominantly asymptomatic patients who had routine lactate measurements only. In any case, the multiple measurements taken in individuals were appropriately taken into account in the analyses either via averaging or through the longitudinal random effects (mixed) models.

The subsequent stable course of chronic hyperlactatemia, on average , is unlikely to be a result of revision of therapy in those with rising lactate concentrations, as asymptomatic hyperlactatemia alone did not cause therapy revision in any patient. However, the incidence of fulminant lactic acidosis/hepatic steatosis induced by NRTI may be underestimated in our cohort as five patients with moderate symptomatic hyperlactatemia had their (stavudine) therapy changed. It is possible that these patients would have progressed to severe fulminant lactic acidosis/hepatic steatosis if therapy were not revised. Of note, a preceding gradual rise in plasma lactate levels or in anion gap while asymptomatic was not detected in the two patients who had severe NRTI-induced lactic acidosis/hepatic steatosis. The data suggest that chronic hyperlactatemia found by routine testing in asymptomatic patients at 1-3 monthly intervals has poor sensitivity for, and given the rarity of severe lactic acidosis/hepatic steatosis, very poor positive predictive value for identifying cases of future symptomatic lactic acidosis/hepatic steatosis.

This longitudinal study shows that acute, severe NRTI-induced lactic acidosis/hepatic steatosis is a rare event in a centre where there is a high degree of awareness of this condition. A venous lactate concentration that is > 5 mmol/l indicates a state of widespread cellular energy deficit and metabolic decompensation. Uncommonly, patients may present with less severe symptoms, hepatic steatosis and only mild to moderate hyperlactatemia (2.8-4.1 mmol/l in our study) and tolerate a revision, rather than cessation, of NRTI. It is not certain if this represents an earlier phase in the natural history of severe lactic acidosis in some patients ^[7]. However, the fulminant form may still occur very precipitously, so diagnosis ultimately relies on having a high index of suspicion about suggestive symptoms in any patient taking a NRTI. In contrast, the incidental finding of mild to moderate hyperlactatemia without symptoms in patients taking HAART does not necessarily portend a progression to severe decompensated lactic acidosis. Based on the 516 patient-years of observation in this study, asymptomatic hyperlactatemia is stable (maintained at an average of 1.5-3.5 mmol/l) in the vast majority of patients without revision of therapy.

The mechanism by which chronic hyperlactatemia occurs in patients on HAART is not known but may be a correlate of tissue-specific mitochondrial toxicities of NRTI. Further studies to elucidate both the tissue source(s) of plasma lactate in NRTI-treated patients and the lactate clearance mechanisms may help to unravel the biological significance of chronic hyperlactatemia in this setting.

Acknowledgements

We are grateful to Associate Professor Martyn French for helpful comments on the manuscript, John Blennerhasset for supervision of lactate assays and all the staff and patients of the Department of Clinical Immunology, Royal Perth Hospital who participated in this study.