Clinical Chemistry Link to Randox Laboratories Web Site
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Clinical Chemistry 46: 351-364, 2000;
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Submit an electronic Letter to
the Editor about this paper
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (13)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Shih, W. J.
Right arrow Articles by Stein, E. A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Shih, W. J.
Right arrow Articles by Stein, E. A.
Related Collections
Right arrow Lipids, Lipoproteins, and Cardiovascular Risk Factors
(Clinical Chemistry. 2000;46:351-364.)
© 2000 American Association for Clinical Chemistry, Inc.

Articles

Estimating the Long-Term Effects of Storage at -70 °C on Cholesterol, Triglyceride, and HDL-Cholesterol Measurements in Stored Sera

Weichung Joe Shih1,1, Paul S. Bachorik2,a, Jo A. Haga3,2, Gary L. Myers4 and Evan A. Stein5

1 Merck Research Laboratories, Rahway, NJ 07065.

2 Departments of Pediatrics and Pathology, Johns Hopkins University School of Medicine, Baltimore, MD 21287 (retired).

3 Wilford Hall Medical Center, Department of Pathology, Lackland Air Force Base, TX 78236.

4 National Center for Environmental Health, Division of Environmental Health Laboratory Sciences, Centers for Disease Control and Prevention, Atlanta, GA 30341.

5 Medical Research Laboratories, Highland Heights, KY 41076.
a Address correspondence to this author at: c/o D. Bartholomew, 11 Fox Run Rd., Falmouth, ME 04105. Fax: 207-797-8596 (primary) or 410-955-1276 (secondary); e-mail pbach{at}prodigy.net


   Abstract
Top
 Abstract
Introduction
 Materials and Methods
 Results
 Discussion
 Appendix 1
 References
 
We estimated the effects of long-term storage at -70 °C on serum total cholesterol, HDL-cholesterol, and triglycerides in specimens that had been stored for up to 7 years. These estimates were made using measurements in serial specimens collected from the placebo control group of the Air Force/Texas Coronary Atherosclerosis Prevention Study over a period of ~5 years. We compared the group means for pairs of serial specimens taken at 6- and 12-month intervals, assuming that (a) a negligible placebo effect occurred between the serial specimen pairs; (b) in the absence of storage effects, the variation in the group means would reflect only normal biological variation and would not materially affect the group means for the serial specimens; (c) any systematic changes in these group means would reflect storage-related changes; and (d) storage-related changes are cumulative, i.e., the overall changes for a given storage period are the sum of the changes during previous storage periods. We observed average decreases of 2.0% per year for total cholesterol over 7 years and 2.8% per year in triglycerides for the first 5 years. HDL-cholesterol decreased by 1.3% per year, but this change was not statistically significant. This approach may be useful for estimating storage-related changes for studies in specimens stored for a period of years and for which stability data may not be available.


   Introduction
Top
 Abstract
 Introduction
Materials and Methods
 Results
 Discussion
 Appendix 1
 References
 
It is a common and important practice, particularly in large clinical and epidemiological studies, to provide for the long-term storage of one or more aliquots of serum or plasma from study participants. Samples to be stored usually are from critical visits, e.g., baseline, randomization, annual, and final visits, and are commonly stored at ultra-low temperatures, i.e., at -70 °C or lower. Such samples can be useful if, for logistical, analytical, or economic reasons, particular measurements must be delayed or made in batch mode. They also afford a particularly cost-effective way to collect additional information on well-characterized populations that may have undergone particular interventions as new analytes of interest emerge. This approach has been used both for case-control studies (1)(2)(3) and for establishing population-based distributions of components such as apolipoproteins (4)(5). In many cases, however, the samples may be used only years after they were originally collected and to measure components for which the long-term stability during storage has not yet been determined (1)(2)(3)(4)(5). This is generally not thought a major problem in case-control, cross-sectional studies or follow-up studies because, although it may not be true in all cases (6), it can generally be assumed that any deterioration that occurs affects all of the specimens to the same extent. Ideally, the component of interest should either be absolutely stable during the storage period or the rate of deterioration should be known. In practice, this usually is not the case.

Although several investigators have provided information about the stability of HDL-cholesterol (HDL-chol)3 during low-temperature storage, such information was available only for periods of several months to ~1 or 2 years (7)(8)(9)(10)(11)(12)(13)(14)(15). In addition, by their nature such studies tend to be conducted under conditions in which specimen collection, handling, and storage are closely controlled and monitored. The extent to which they can be generalized to conditions that may obtain in large-scale, multiyear studies conducted in thousands of subjects is not known. In the present study, we analyzed total cholesterol (TC), triglycerides (TGs), and HDL-chol in serum specimens, collected from several thousand subjects in a large clinical trial, that had been stored at -70 °C for up to 7 years before analysis. This report describes the procedure we used to estimate the stability of serum TC, TGs, and HDL-chol during storage periods up to 7 years.


   Materials and Methods
Top
 Abstract
 Introduction
 Materials and Methods
Results
 Discussion
 Appendix 1
 References
 
specimen collection, storage, and analysis
The Air Force/Texas Coronary Atherosclerosis Prevention Study (AFCAPS/TexCAPS) was a placebo-controlled double-blind study to determine the efficacy of cholesterol and LDL-cholesterol (LDL-chol) lowering on the incidence of first coronary events in a population of middle-aged individuals with moderately increased TC and LDL-chol, and low HDL-chol. Subjects who were potentially eligible for the study were placed on the American Heart Association Step 1 diet for a 2- to 3-month period before randomization. They were then randomly assigned to either a placebo group or a treatment group (16). On the day of randomization, fasting venous serum was obtained from each participant, and aliquots were placed into 5-mL screw-capped polypropylene vials designed for specimen storage. These aliquots were frozen and stored at -70 °C at the AFCAPS/TexCAPS serum bank. Fig. 1 illustrates the time frames in which the baseline and follow-up specimens were collected. The baseline specimens were collected between May 1990 and February 1993. The follow-up specimens were obtained at various times after baseline, as indicated in Fig 1 . For example, the 12-month follow-up specimens were collected beginning in 1991. It can be appreciated from Fig. 1 that the longest-term follow-up samples were collected most recently and, therefore, were stored for shorter times, whereas the baseline samples had been stored the longest. None of the stored specimens had been thawed before they were analyzed.



View larger version (14K):
[in this window]
[in a new window]
  		Figure 1. Relationships between baseline and follow-up visits, year of venipuncture, period of analysis, and storage time.

mo, month.

In late 1996, the stored specimens were arranged in batches chronologically according to the draw date of the randomization specimen. All follow-up specimens from subjects assigned to a particular batch were included in that batch. This led to the assembly of 71 batches containing 300–450 specimens per batch. The batches were then shipped on solid CO2 and in random order to the Johns Hopkins Lipoprotein Analytical Laboratory over ~10 months. Analyses were completed within 1–2 weeks. This sample arrangement and shipment schedule was used for two reasons: (a) It was not practical to analyze the baseline and follow-up specimens in the same order and over the same period in which they had been drawn because this would have required a 5-year period of analysis. It was therefore decided to randomize the shipments. (b) All specimens from an individual were analyzed in the same analytical run to eliminate the effects of among-run variation when examining temporal changes in serial specimens from each of the study groups. For this report, we used data from paired serial measurements in specimens from the placebo group.

The relationships between the length of storage, storage effect, and identities of the visits for the specimens we used to estimate the storage effects are illustrated in Table 1 . Storage effect, Fn, where n = 1–7, signifies the effects of storage after periods of 0–1 year (F1), through 6–7 years (F7). Baseline specimens are designated X0, and were drawn during the period 1990–1993. Thus, some of the baseline specimens, those drawn in 1993, had been stored for 3–4 years before analysis. Those drawn in 1992 had been stored for 4–5 years before analysis. The remaining baseline samples had been stored for 5–6 and 6–7 years before analysis, as indicated in Table 1 . Similarly, the 12-month follow-up samples are designated X12, and were drawn between 1991 and 1994. They had been stored for periods ranging from 2–3 years (F3), to 5–6 years (F6) before analysis. The follow-up samples from the 30-month visit (X30) were drawn in 1995 and had been stored for 1–2 years before analysis (F2). The remaining follow-up samples were drawn in 1995 and 1996, and as indicated in Table 1 , had been stored for 0–1 year or 1–2 years before analysis.


View this table:
[in this window]
[in a new window]
  		Table 1. Specimens1 available for estimating the effects of long-term storage on lipid and lipoprotein measurements.

lipid and lipoprotein analysis
In the Johns Hopkins laboratory, cholesterol and TGs were measured enzymatically as described previously (17)(18). Briefly, cholesterol and TGs were analyzed on a Hitachi 704 automated clinical chemistry analyzer (Boehringer-Mannheim) using reagents supplied by the manufacturer (Cholesterol/HP, cat. no. 704036; Triglyceride/GPO, cat. no. 450028). HDL-chol was measured in the supernatant after precipitation of the apolipoprotein B-containing lipoproteins [VLDL, intermediate-density lipoprotein, LDL, and lipoprotein(a)], with heparin and MnCl2 (final concentrations, 1.3 g/L and 0.046 mol/L), and removal of excess Mn2+, as described previously (19). The laboratory was standardized for lipid and lipoprotein measurements through the CDC-National Heart, Lung and Blood Institute (NHLBI) Lipid Standardization Program (20). For cholesterol and TGs, bias and imprecision were estimated by the analysis of serum control pools at each of two concentrations every 13 samples. The control pools were obtained from Solomon Park Laboratories, Kirkland, WA, and were provided with reference values assigned by the CDC Clinical Chemistry Standardization Section (Atlanta, GA). Bias and imprecision for HDL-chol were similarly estimated using two serum control pools. The first was provided by Solomon Park Laboratories and had a CDC-assigned HDL-chol concentration of 0.509 g/L. The second had an HDL-chol concentration of 0.360 g/L, and was specially prepared for this study by Pacific Biometrics, Inc., Seattle, WA, one of the laboratories of the National Cholesterol Reference Method Laboratory Network (21). Pacific Biometrics provided the reference value for this pool.

determination of storage effects
The effects of storage on the lipid and lipoprotein measurements were estimated as follows.

First-year storage effects.
The 1-year storage effect, F1, was estimated as the weighted mean from two independent studies. The first was performed by the CDC Clinical Chemistry Standardization Section, and the data were kindly provided by Dr. Myers, CDC. In that study, analyses were performed in 25 fresh serum specimens immediately, and were repeated in aliquots of the same samples after storage at -70 °C for 1 year. The second study was conducted in the Johns Hopkins laboratory. Analyses were performed in 152 fresh serum specimens immediately and after storage at -80 °C for 13–14 months.

Storage effects for longer periods.
Using measurements in serial specimens from the AFCAPS/TexCAPS placebo control group, we estimated the storage-related changes likely to occur over various periods. The procedure required the following assumptions:


The data set from which these estimates were made is shown in Table 2 . There were a total of 3285 patients in the placebo group for which baseline specimens were available. Of these, there were 1707 patients with follow-up samples drawn 6 months apart and that had been stored 1–2 years before analysis. As seen in Table 2 , the 30 month-36 month serial paired samples were available for 229 participants, and 36 month-42 month pairs were available for 597 participants. The number of sample pairs from longer term follow-up that had been stored for 1–2 years before analysis are similarly indicated in Table 2 . Thus, six separate estimates of the 1–2 year storage effects could be made. Similarly, there were 263 participants for whom 12 month-30 month pairs were available and that had been stored for 2–3 years before analysis (Table 2 ). Estimates of the effects of storage for longer periods were made using the numbers of baseline-12 month pairs indicated in Table 2 .


View this table:
[in this window]
[in a new window]
  		Table 2. Serial specimen pairs used to calculate storage effects for samples stored 1–2 years and longer.

estimating storage effect between 1 and 2 years
We estimated the 1–2 year storage effect, F2 avg, using specimen pairs drawn from the following cells in Table 1Up : F2X30 and F1X36; F2X36 and F1X42; F2X42 and F1X48; F2X48 and F1X54; F2X54 and F1X60; and F2X60 and F1X66. In this notation, F2X30 indicates the specimens from the 30 month visit that were 1–2 years old at the time of analysis, and F1X36 indicates the specimens from the 36 month visit that were stored <=1 year when analyzed. Each pair of cells contains serial specimens from the same individuals, and each thus provided a separate estimate of the 1–2 year storage effect, F2, and its associated variance. The six estimates were calculated using 84–597 specimen pairs (Table 2Up ). These estimates were then pooled by weighting their variances to produce the average storage effect (F2 avg) between 1 and 2 years.

Assuming the additive model indicated above, and using the first member of the first pair of cells listed above:

where F2X30 is the observed value after 1–2 years storage; X30 is the true value in the 30 month specimens when they were originally drawn; and F2a is the storage effect. Similarly, for the second member of the pair:

and the difference in the observed values in the pair is given as:

Assuming no systematic placebo effect during the period (i.e., participants are in the steady state), (X30 - X36) will average zero, and the following relationship should hold:

In other words, an estimate of F2a, the 1–2 year storage effect, can be obtained by adding the estimate of F1 to the average of the observed differences in the paired group:

(1)

The variance of F2a avg is given by:

(2)

The same procedure was used to derive additional estimates of F2 from the other five pairs of cells listed above.

These calculations produced six estimates of the 1–2 year storage effect (F2a avg through F2f avg) and their associated variances. The weighted average of these estimates was taken as the best overall estimate of the 1–2 year storage effect, F2 avg (see Appendix).

estimating storage effect between 2 and 3 years
The specimens in cells F3X12 and F2X30 (Table 1Up ) constituted the only pair available for calculation of F3 avg, the estimate of the storage effect between 2 and 3 years. In calculations analogous to those above, F3 avg and its associated variance were calculated as follows, using data from 263 specimen pairs (Table 2Up ):

(3)

where F2 avg and Var (F2 avg) are those calculated above. Note that baseline specimens were not used for estimates of storage effects for up to 3 years of storage (Table 2Up ).

estimating storage effect between 3 and 4 years and for longer periods
The calculations of the storage effects for longer periods (F4 to F7) were made similarly, using paired serial specimens from the cell pairs indicated in Tables 1Up and 2Up , except that data from the Lipid Research Clinics (LRC) Program (23) were used to adjust for a possible small placebo-diet effect between the baseline and 12-month follow-up specimens:

(4)



One cell pair was available for each of the storage times between 3 and 7 years (Table 1Up ).

estimating true values
The regression equations relating storage effects, F1 avg to F7 avg estimated as described above, to storage time were then determined by the weighted least-squares method, using the inverse of the variance-covariance matrix of F1 avg to F7 avg as the weight (see Appendix). The regression-fitted storage effects were then used to estimate the true values for the specimens using the following rearrangement of Eq. 1Up above:

(5)
where Fn avg (regr) is the storage effect for storage period n as calculated from the regression equation for the particular analyte under consideration. Note that if values decrease with storage, then Fn avg (regr) will be a negative number, and the true value will be higher than the value observed after storage time n.


   Results
Top
 Abstract
 Introduction
 Materials and Methods
 Results
Discussion
 Appendix 1
 References
 
analytical drift
The results of TC, TG, and HDL-chol measurements made in serum control pools in the Johns Hopkins laboratory during the period in which the stored specimens were analyzed are shown in Table 3 . Bias was calculated with respect to CDC reference values. The average percentages of bias (and CVs) for the three primary measurements were as follows: for TC, -0.8% (1.8%); for TG, -3.3%, (1.8%); for HDL-chol, the bias (CV) was 0.0% (3.8%) at 0.509 g/L. The absolute bias (CV) was -0.011 g/L (0.019 g/L) for HDL-chol at 0.36 g/L, and the overall percentage of bias was -1.5% for HDL-chol. Fig. 2 illustrates the stability of the measurements as a function of time during the 11-month period of analysis. Fig. 2A shows the results of linear regression analyses for TC measurements in the serum control pools with normal or increased concentrations (Table 3 ) used for this study. The analytical drift for TC was ~0.02 g/L from the beginning to the end of the analytical period. Analytical drift for TG (Fig. 2B ) was ~0.01 g/L, and that for HDL-chol was approximately the same (Fig. 2C ). The small amount of drift that occurred was in the positive direction for all three analytes.


View this table:
[in this window]
[in a new window]
  		Table 3. Analysis of lipid and HDL-chol in serum control pools.1



View larger version (18K):
[in this window]
[in a new window]
  		Figure 2. Quality-control sera analyzed during the period of analysis of stored sera from the AFCAPS/TexCAPS study.

Regression lines and 95% confidence intervals are shown for normal and increased concentration control pools analyzed for TC (A), TGs (B), and for low and normal concentration HDL-chol pools (C).

storage effects
In the description that follows, we use the term "storage effect" to refer to the combined effects of freezing, per se, and any subsequent changes that occurred during storage.

Estimates of the storage-related changes during the first 1-year period were based on measurements made in 25 specimens at CDC and 152 specimens at Johns Hopkins. The means (SD) and percentages of change for TC and HDL-chol are shown in Table 4 . In both studies, there were no significant changes in TC. HDL-chol did not change significantly in the CDC study, and decreased by 0.009 g/L in the Johns Hopkins study. The change in TGs as estimated from the Johns Hopkins data was used to calculate the F1 avg value for TG; similar data for TGs were not available from the CDC study. The values of F1 in Table 4 were used to calculate a weighted mean storage effect (F1 avg) for specimens that had been stored for 1 year before analysis.


View this table:
[in this window]
[in a new window]
  		Table 4. Stability of TC and HDL-chol in specimens stored for 1 year at or below -70 °C.

From the specimen groups available, it was possible to calculate six different estimates of F2 for each of the measured components. For TC, the estimates ranged from -0.0485 to 0.0364 g/L. Five of the six estimates were negative, and the weighted estimate for cholesterol was -0.0370 g/L (Table 5 ). For HDL-chol, all the estimates were negative, and the weighted estimate was -0.0068 g/L. The weighted average storage effect for TG was -0.0566 g/L (Table 5 ). The 95% confidence intervals for the weighted average storage effects are also shown in Table 5 .


View this table:
[in this window]
[in a new window]
  		Table 5. Estimates of 1–2 year storage effects for lipids and lipoproteins.

Only one pair of specimen groups was available for each of the longer storage periods; therefore, only a single estimate for Fn avg could be calculated for each. The values of F3 avg through F7 avg for each of the analytes are given in Table 6 . Because the several estimates of F1 and F2 varied somewhat, we assumed that there would also be some inherent variation in the single estimates available for longer term storage effects. For this reason, we felt that the most reliable estimates of the storage effects for any particular storage time would be those calculated from the weighted regressions relating the magnitudes of the cumulative storage effects to storage time. Fig. 3 illustrates the regressions for storage times 1–2 years and longer, and Table 7 provides the regression parameters for the regression lines shown in Fig. 3 . For cholesterol, the values of Fn avg became progressively more negative, and these changes were highly correlated with storage time; r = 0.91 (Fig. 3A ). The slope of the regression line in Fig. 3A indicates that cholesterol values decayed by ~0.071 g/L per year over the period indicated in Fig. 3A . In contrast, there was no significant correlation between change in HDL-chol values and storage period (Fig. 3B and Table 7 ).


View this table:
[in this window]
[in a new window]
  		Table 6. Estimates of storage effects for lipids and lipoproteins for specimens stored >2 years.



View larger version (13K):
[in this window]
[in a new window]
  		Figure 3. Cumulative storage effects showing the weighted regressions and 95% confidence intervals for TC (A), HDL-chol (B), and TGs (C) for storage periods exceeding 1 year.

The equations for the regression lines are given in Table 7Up .


View this table:
[in this window]
[in a new window]
  		Table 7. Relationships between the effects of storage at -70 °C and storage time.

The pattern for TGs was different. Overall, there was no significant correlation between the storage effect and storage time. However, inspection of Fig. 3CUp revealed that Fn avg became more negative with storage time for up to 4–5 years. After this period, however, Fn avg tended to increase. Overall, the observations in Fig. 3CUp were described by a cubic equation:



The regression equations were calculated without using the 0–1 year time point because the 0–1 year data were gathered in separate experiments rather than from the stored AFCAPS/TexCAPS specimens. However, inclusion of this point did not significantly affect the regression patterns because little deterioration was observed after storage for 1 year.

In Table 8 , the lipid and lipoprotein values observed for the baseline specimens in the placebo group before and after correction for the storage effects are compared using the regression equations in Table 7Up . For TC, the mean concentration after adjustment for storage effects was 2.346 g/L, or 7.8% higher than the unadjusted mean (2.177 g/L). Because average storage time for all baseline specimens combined was 3.8 years, the decay in TC values amounted to ~2% per year. The adjusted mean for TGs was 1.904 g/L, or 10.8% higher than the unadjusted mean, suggesting an overall decay rate of ~2.8% per year. The adjusted HDL-chol value was 0.019 g/L higher than the unadjusted mean, but this cannot be considered a significant change because the regression equation used for this calculation showed no significant relationship between storage effect and storage time for HDL-chol.


View this table:
[in this window]
[in a new window]
  		Table 8. Baseline lipid and lipoprotein concentrations in stored specimens from the AFCAPS/TexCAPS control group after correction for storage effects.1


   Discussion
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
Appendix 1
 References
 
In this report, we present an approach to determining the effects of long-term storage at -70 °C on TC, TG, and HDL-chol values measured in serum. This approach involved determining the mean differences in serial specimens obtained 6–12 months apart in groups of specimens that had been stored for various periods before analysis. We assumed that the storage effects were additive, i.e., that adding the change that occurred over a 1-year period in specimens that had been stored 1–2 years to the change that occurred over a 1-year period in specimens that had been stored 2–3 years provides an estimate of the change expected to occur between 1 and 3 years. We used this approach because only a single aliquot of each specimen was available from each study subject, and it was not possible to measure storage effects directly in separate aliquots of the same specimens after various storage periods.

The procedure we used depends on several reasonable assumptions. The first is that the study subjects were in the steady state, i.e., the difference in TC, TG, or HDL concentrations in serial specimens from each patient primarily reflect normal biological variation. To this end, we used measurements made in baseline and follow-up specimens from the placebo control group of the AFCAPS/TexCAPS study. The second assumption was that because of the 3-month diet run-in, any placebo treatment effect would have already occurred and would therefore be negligible in the periods between serial specimens. Under steady-state conditions, the concentration of a particular analyte in the second specimen of each serial specimen pair would be expected to exceed its concentration in the first specimen of the pair in approximately one-half of the subjects and to be lower than the first serial value in the other half. Thus, for a group of subjects in the steady state, there should be no systematic change in the group mean; or stated more precisely, any differences between the group means for serial specimens would expectedly be small and, among the various times studied, should occur in both directions with equal frequency.

These assumptions are supported by previous studies, e.g., the Helsinki Heart Study (24)(25), which used a 3- to 6-week dietary run-in before baseline. There was a 7% reduction in TC and no change in HDL-chol between initiation of diet treatment and baseline, but only minimal and random changes in the placebo group means from baseline to 5 years (24), particularly in the patients who were most comparable to those in the present study, i.e., those with increased LDL-chol and normal TG values (25). The LRC Coronary Primary Prevention Trial also used a 3-month dietary run-in in which subjects were placed on a moderate cholesterol-lowering diet before the baseline visit (23). In the placebo group, there was a 4.3% decrease in TC during this period. Between baseline and 12 months, however, there was a further decrease of only 1.3% in TC. Similarly, there was a 1.6% (7 mg/L) decrease in HDL-chol during the diet run-in, but no significant change between baseline and 12 months. TGs decreased by 3.2% in response to diet, but by 12 months they were 5.9% higher than baseline (23). The Cholesterol-Lowering Atherosclerosis Study also reported a 4.5% decrease in TC, no change in HDL-chol, and an 8% decrease in TGs in the diet-treated placebo control group between baseline and 2 years (26). However, the time course of the TC and TG changes cannot be ascertained for that study (a) because the data were not reported in this way; and (b) because "baseline" was defined as the average for the first three screening visits although diet treatment was initiated at the second screening visit (26).

The NHLBI Type II Coronary Intervention Study was conducted in men with high LDL-chol (above the 90th percentile) and coronary heart disease (27). The authors of that study reported decreases in TC, TGs, and HDL-chol of 7%, 15%, and 3.5%, respectively, after treatment with a low-cholesterol, low-saturated fat diet. However, there were no further significant changes in TC and HDL-chol between the postdiet baseline and the averages for 5 years of follow-up, although the 5-year average for TGs was ~23% higher than the postdiet baseline. Again, the time course of the diet-induced changes was not entirely clear because the prediet baseline was defined as the mean of two screening visits, and the postdiet baseline as the mean of three postdiet visits at 2-month intervals. The dietary effects appeared to have occurred sometime during the first 6 months of dietary treatment (27).

These studies all revealed greater or lesser effects of diet treatment but suggested that the major changes were seen within 3 months or so. Three of the studies indicated no significant changes in TC or HDL after a dietary run-in of 3–6 months, or any further changes from the 12-month follow-up visits on. In view of these reports, our assumption of no placebo effects on TC and HDL-chol between postdiet baseline and 1 year, therefore, appears to be reasonable, and in any event would not have affected our estimates of storage effects up to 3 years because data from the baseline visits were not used for those calculations.

Baseline data were used to calculate longer term storage effects, however. Because the LRC data suggested that residual dietary effects might occur between the postdiet baseline and 12 months (23), we used the LRC findings to allow for this before estimating the storage effects for period longer than 3 years. After this adjustment, any tendency for the mean concentrations of the second specimens of the various serial specimen pairs to decrease with storage was interpreted to be primarily attributable to storage-related changes, provided that the laboratory measurements assays themselves remained stable. Furthermore, based both on the LRC (23) and NHLBI Type II (27) studies, if a continuous placebo effect had occurred for TGs over the entire follow-up period, the effect should have been to increase rather than decrease TG concentrations.

We attempted to detect storage-related changes in another way. As discussed above, in the steady state, normal within-individual biological variation is observed as random fluctuations in the measured values in serial specimens after adjustment for the analytical component of variation, and is described by the coefficient of biological variation, CVb, for that individual. It might be expected that sample deterioration would produce an increase in the apparent CVb because the fluctuations in serial samples would tend to be in one direction rather than equally distributed about the subject mean. Therefore, the observation of such an increase compared with CVb determined in fresh specimens would have tended to support the occurrence of storage-related changes. We examined this for cholesterol and found a median apparent CVb of 7.14%, or ~14% higher than the generally accepted average value of 6.5% (28). Although this difference was in the expected direction, the calculation neither supports nor refutes the occurrence of storage-related changes because estimates of CVb using fresh specimens have varied more than this, from ~5% to 8.5% in different studies (29)(30)(31)(32)(33)(34)(35)(36)(37)(38).

The patterns observed in the present study indicated rather consistent downward changes for TC and TGs; these changes occurred in TC over the entire storage period, and in TGs for the first 4–5 years. For several reasons, these changes could not be attributed to analytical trends occurring during the period of analysis. First, the TC and TG measurements were quite stable during the period in which the stored specimens were analyzed. Indeed, the minor drifts seen in Fig. 2Up for these measurements were in the opposite direction to the changes observed in the stored specimens. Second, the order of analysis of the 71 batches of stored specimens had been randomized specifically for the purpose of minimizing the effects of laboratory drift. Third, to minimize the effects of analytical drift further, all serial specimens from a particular subject were analyzed in the same analytical run. Finally, although the six separate estimates for the 1–2 year storage effects for each analyte varied somewhat, these variations appeared to be random. The systematic decreases observed for TC and TGs could therefore be attributed to storage-related deterioration of the specimens themselves.

The patterns of deterioration of TC and TGs were similar in magnitude over the first 5 years. After longer periods, the loss of TGs slowed. The reasons for the deterioration are not entirely clear, but the storage-related hydrolysis of TGs would not explain the observations. Such hydrolysis would have produced partial glycerides and glycerol, which would have increased the TG blank. Because we did not measure TG blanks, however, partial glycerides and free glycerol would have been measured as TGs, and unblanked TG values would not be expected to change during storage. Indeed, it had been specifically decided not to analyze free glycerol or partial glycerides in this study because of the potential for overestimating storage effects by such TG degradation products that may have been generated from TGs during storage.

Storage appeared to affect primarily the non-HDL lipoproteins in serum because there were no consistent storage effects for HDL-chol. It is necessary to consider the extent to which the present findings may be generalized to the long-term stability of banked sera from other studies. In the present study, the appearance of the stored serum specimens after they had been thawed also suggested storage-related changes. Approximately 45% of the specimens stored for 1–2 years were turbid. Sixty-five to 90% of the specimens stored for longer periods were turbid, but there was no consistent relationship between storage time and the occurrence of turbid specimens. In approximately one-third of the turbid specimens, a floating TG-rich layer began to accumulate after the specimens had been allowed to stand undisturbed for a short period. The mean TG concentration for these specimens was <2.00 g/L, well under the 4.0–5.0 g/L concentration at which fresh specimens usually are turbid. In our experience, the large proportion of turbid specimens, even at the shortest storage time (1–2 years), was unusual. The reasons underlying their turbidity were not clear because, to our knowledge, the specimens had not thawed during storage. We cannot rule out, however, the possibility that temperature fluctuations may have occurred without overt thawing. Whatever the cause, the behavior of these specimens can be considered as part of the overall effect of storage on the specimens collected in this particular study. Furthermore, it raises the question of whether the stability of banked specimens from different studies can be taken for granted simply because they may have been stored under similar conditions.

In the present study, we conducted a small pilot study in specimens that formed lipid layers. We detected decreases in the TG concentrations in the infralayer region after the specimens had been allowed to stand for at least 1 h. Because of the tendency for a substantial proportion of the specimens to form layers, we attempted to minimize the possibility of the analyzer aspirating nonhomogeneous samples during the ~30-min period required for the analysis of TC and TGs by thoroughly mixing the samples immediately before beginning the analytical run. Despite this precaution, however, this particular storage effect may not have been completely overcome in all specimens.

The literature concerning the effects of long-term storage on serum or plasma HDL-chol measurements is limited and somewhat contradictory. Nanjee and Miller (10) evaluated HDL-chol in 93 EDTA plasma specimens stored at -70 °C and in liquid nitrogen for up to 14 months. They found no significant differences, suggesting that storage at the lower temperature conferred no advantage over storage at -70 °C. The HDL-chol values for fresh aliquots of the same specimens were not reported, however. Bachorik et al. (7) found that the mean HDL-chol in 106 EDTA plasma specimens decreased ~2.5% after 1 month and ~5% after 1 year at -70 °C, compared with the measurements in fresh samples, and a similar change was observed in a CDC serum control pool that was analyzed with the specimens. In a recent study, Kronenberg et al. (15) measured HDL-chol in eight EDTA plasma samples before and after storage for various periods up to 1 year at -80 °C, and found no significant time-related changes in HDL-chol. Bausserman et al. (14) similarly reported that HDL-chol measured in serum with the heparin-Mn2+ method was stable for up to 1 year at -70 °C but decreased by ~13% in the subsequent 6 months. These authors also found that HDL-chol, as measured with a dextran sulfate method, decreased by ~5% after storage for 1 month at the same temperature (14), suggesting that the observed storage effects may be method dependent. Changes in HDL-chol appear to be more easily observed in specimens stored at -20 °C (8)(9)(10)(11)(12)(39), apparently because of storage-related changes in the precipitability of HDL and the apolipoprotein B-containing lipoproteins (39). The findings of these studies generally suggest that if HDL-chol measurements are to be delayed, the specimens are best stored at ultra-low temperatures (i.e., -70 °C or lower). Although the storage-related changes observed at -20 °C might also be expected to occur at lower temperatures, they would occur much more slowly. Indeed, most studies have indicated that HDL-chol is stable for at least 1 year at ultra-low temperatures. Our findings are consistent with this conclusion and further suggest that serum HDL-chol may remain stable for at least 5 years.

TC and TGs are generally thought to be stable in stored specimens, although again, the available data in fresh serum or plasma are limited. Demacker and Jansen (8) reported that TC was stable for up to 10 months in sera stored at -20 °C. Stokes et al. (9) found no change in TG concentrations and only a slight increase in TC in 60 serum specimens that had been stored for up to 4.5 months at -15 °C. Wood et al. (40) reported that TC was stable in EDTA plasma stored for up to 2 years at -15 °C. TC and TGs in EDTA plasma were also reported to be stable at both -20 and -70 °C for up to ~14 months when compared with aliquots of the same specimens stored in liquid nitrogen for the same periods (10). Again, fresh specimen values were not reported in this study. Kronenberg et al. (15) reported slight increases in TC and TGs in eight EDTA-plasma specimens after storage for 2 years at -80 °C. Overall, TC and TGs appear to be relatively stable in specimens that have been stored for 1 or 2 years. Our findings are consistent with this. In the present study, however, we also examined longer storage periods in larger numbers of specimens than were used in most of the studies cited above. We found changes consistent with slight but continuous storage-related decreases in TC over 7 years and in TGs over 5 years. On the basis of the regression equations relating cumulative storage effect to storage time, these decreases amounted to ~1.7% per year for TC, and ~2% per year for TGs.

In summary, we estimated the effects of storage at -70 °C for periods up to 7 years on TC, TGs, and HDL-chol, based on measurements in serial specimens taken 6 months to 1 year apart from a total of >3000 study subjects and stored for various periods before analysis. The findings indicated that the three analytes were stable when stored at -70 oC for up to 1 year. After longer periods, there were no consistent storage-related changes in the group mean HDL-chol concentration. There were small but consistent decreases in the group mean TC and TG concentrations, consistent with deterioration of the specimens after longer periods of storage. The degree to which these findings can be quantitatively generalized to other clinical studies is not entirely clear, but the findings suggest that banked sera from different studies should not be assumed to be equally stable despite having been stored under similar conditions. The procedure we used, however, should provide a useful approach for others to estimate long-term storage effects on these or other analytes for which such information may unavailable, available only for relatively short storage periods, or available only for conditions that may differ from those that obtain in any particular study. Storage effects reflect not only the stability of the analyte of interest when handled and stored under ideal conditions, they are the overall consequences of how the specimens in a particular study were actually collected, handled, and stored. This might be expected to vary from study to study, depending on several factors, such as staff training and turnover, unavoidable delays in sample processing, conditions of storage, equipment failures, and other factors. The procedure we propose provides storage-effect estimates that would reflect the cumulative effect of whatever factors operated, and can give study specific estimates of sample deterioration during storage. However, when starting a study, steps should be taken during the planning phase to monitor the stability of specimens for as long as they are likely to be used. It is prudent to assume that the banked specimens in a particular study may in fact deteriorate with time even if the particular analytes of interest are thought to be stable. For example, cholesterol itself is fairly stable, but it is associated with carrier proteins of lesser stability. If the carrier protein is altered such that it precipitates or no longer solubilizes cholesterol, it may be difficult to obtain a representative sample for measurement, which can affect the accuracy of the measurements. For this reason, it is advisable to conduct training sessions for venipuncture technicians, individuals charged with preparing and shipping the specimens to the laboratory, and the laboratory technicians who must receive, handle, and analyze the specimens. The procedures and logistics to be used in the study should be included in a detailed study protocol for specimen acquisition, handling, storage, shipment, and analysis. The training sessions should include a discussion of factors that can affect the stability of specimens in general and of the particular analytes of interest. In addition, it is advisable to monitor the stability of banked specimens in a random subset of specimens that have been stored in aliquots sufficient to be monitored prospectively at intervals for as long as they are likely to be used. This should be done for analytes initially included in the study and those under consideration for measurement in the future. Of course, it is not always possible to predict which analytes may be measured several years later, and in such cases the procedure described in the present report may be useful. Furthermore, information from sample-monitoring protocols conducted in concert with large-scale clinical trials would be useful for developing standard sample handling and storage procedures for long-term studies in which specimens will be banked.


   Appendix 1
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 Appendix 1
References
 
i. formula for obtaining estimate of F2 as a weighted average of F2a avg through F2f avg
The weights were functions of the variances and covariances of F2a avg through F2f avg. The variance of F2a avg was given in Eq. 7 and similarly for the variances of F2b avg through F2f avg from other pairs. From Eq. 6, the covariances among F2a avg through F2f avg were equal to the variance of F1 avg, which was obtained from Table 4Up for TC (0.0000035 g/L2), HDL (0.00000169 g/L2), and TGs (0.0000271 g/L2). Let {Sigma}-1 denote the inverse of the variance-covariance matrix of F2a avg through F2f avg. The weighted average was F2 avg = WaF2a avg + ... + WfF2f avg, where the i-th weight Wi = (i-th row sum of {Sigma}-1)/(sum of all elements in {Sigma}-1).

ii. formula for weighted least-squares regression (41)
From formulas for estimating F2 avg through F7 avg, the covariance of Fiavg and Fj avg, for 1 <= i <= j <= 7, equals the variance of Fi avg, which was obtained from Tables 4–6Up Up Up . Let W denote the inverse of the variance-covariance matrix of F1 avg to F7 avg. To regress Fi avg on i for 1 <= i <= 7, let Y be the column vector of F1 avg to F7 avg and X = [e, Z], where e is the column vector of ones, and Z is the column vector of time from 1 to 7 (years). The vector of the regression coefficients from weighted least-squares regression was b = (X'WX)-1 X'WY.


   Acknowledgments
 
This work was supported in part by a contract from Merck and Co., Inc., Whitehouse Station, NJ. We acknowledge the expert technical assistance of the following individuals at Johns Hopkins: Robert Arehart, John Cras, Gale Sherman, and Ella Levy, who analyzed the specimens; and Kathleen L. Lovejoy, who coordinated laboratory activities, provided the required laboratory computer programming, and assisted with certain data analyses. We also are also indebted to Alexandra Langendorfer, Joyce Gray, Deborah Shapiro, Leonard Oppenheimer, and Polly Beere, all with Merck and Co., for overall study coordination and management support. We thank Professor John De Cani, Chairman, AFCAPS/TexCAPS Data, Safety, and Monitoring Board for helpful comments during the preparation of this manuscript. Finally, we are also indebted to the anonymous reviewers of this manuscript for their incisive and constructive criticism of the work.


   Footnotes
 
1 Current address: Division of Biometrics, University of Medicine and Dentistry of New Jersey, Piscataway, NJ 08854

2 Current address: Department of Pathology, Luke Air Force Base Hospital, 56 MDSS/SGSC, 7219 Litchfield Road, Luke AFB, AZ 85309.

3 Nonstandard abbreviations: HDL-chol and LDL-chol, HDL- and LDL-cholesterol; TC, total cholesterol; TG, triglyceride; AFCAPS/TexCAPS, Air Force/Texas Coronary Atherosclerosis Prevention Study; NHLBI, National Heart, Lung and Blood Institute; and LRC, Lipid Research Clinics.


   References
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 Appendix 1
 References
 

  1. Wald NJ, Law M, Watt HC, Wu T, Bailey A, Johnson AM, et al. Apolipoproteins and ischaemic heart disease: implications for screening. Lancet 1994;343:75-79. [ISI][Medline] [Order article via Infotrieve]
  2. Ridker PM, Hennekins CH, Stampfer MJ. A prospective study of lipoprotein(a) and the risk of myocardial infarction. JAMA 1993;:2195-2199. [ISI][Medline] [Order article via Infotrieve]
  3. Cremer P, Nagel D, Labrot B, Mann H, Muche R, Elster H, Seidel D. Lipoprotein(a) as predictor of myocardial infarction in comparison to fibrinogen, LDL cholesterol and other risk factors: results from the prospective Gottingen Risk Incidence and Prevalence Study (GRIPS). Eur J Clin Investig 1994;24:444-453.
  4. Contois JH, McNamara JR, Lammi-Keefe CJ, Wilson PWF, Massov T, Schaefer EJ. Reference intervals for plasma apolipoprotein A-I determined with a standardized commercial immunoturbidimetric assay: results from the Framingham Offspring Study. Clin Chem 1996;42:507-514. [Abstract/Free Full Text]
  5. Contois JH, McNamara JR, Lammi-Keefe CJ, Wilson PWF, Massov T, Schaeffer EJ. Reference Intervals for plasma apolipoprotein B determined with a standardized commercial immunoturbidimetric assay: results from the Framingham Offspring Study. Clin Chem 1996;42:515-523. [Abstract/Free Full Text]
  6. Kafonek S, Bachorik PS, Kwiterovich PO. Measurement of apolipoprotein B in plasma stored at -70 °C. Clin Chem 1989;35:380-383. [Abstract/Free Full Text]
  7. Bachorik PS, Walker R, Kwiterovich PO, Jr. Determination of high density lipoprotein cholesterol in human plasma stored at -70 °C. J Lipid Res 1982;23:1236-1242. [Abstract]
  8. Demacker PN, Jansen AP. Stability of high-density-lipoprotein cholesterol in frozen sera. Clin Chem 1983;29:567-568. [Free Full Text]
  9. Stokes YM, Salmond CE, Carpenter LM, Welby TJ. Stability of total cholesterol, high-density-lipoprotein cholesterol, and triglycerides in frozen sera. Clin Chem 1986;32:995-999. [Abstract/Free Full Text]
  10. Nanjee MN, Miller NE. Evaluation of long-term frozen storage for measurement of high-density lipoprotein and its subfractions by precipitation. Clin Chem 1990;36:783-788. [Abstract/Free Full Text]
  11. Reckless JPD, Betteridge DJ, Wu P, Payne B, Galton DJ. Effect of storage on plasma-lipoproteins. Lancet 1977;2:350.[ISI][Medline] [Order article via Infotrieve]
  12. Kahn SR, Elkeles RS. Effect of storage on the precipitation procedure for the measurement of HDL subclasses in serum. Clin Chim Acta 1984;144:59-63. [ISI][Medline] [Order article via Infotrieve]
  13. Matthew A, Duggan PF. Stability of high-density-lipoprotein subfractions in stored plasma. Clin Chem 1988;34:425-426. [Free Full Text]
  14. Bausserman LL, Saritelli AL, Milosavljevic D. High-density lipoprotein subfractions measured in stored serum. Clin Chem 1994;40:1713-1716. [Abstract/Free Full Text]
  15. Kronenberg F, Lobentanz EM, Konig P, Utermann G, Dieplinger H. Effect of sample storage on the measurement of lipoprotein[a], apolipoproteins B and A-IV, total and high density lipoprotein cholesterol and triglycerides. J Lipid Res 1994;35:1318-1328. [Abstract]
  16. Downs JR, Beere PA, Whitney E, Clearfield M, Weis S, Rochen J, et al. Design and rationale of the Air Force/Texas Coronary Atherosclerosis Prevention Study (AFCAPS/TexCAPS). Am J Cardiol 1997;80:287-293. [ISI][Medline] [Order article via Infotrieve]
  17. Sempos CT, Cleeman JI, Carroll MD, Johnson CL, Bachorik PS, Gordon DJ, et al. Prevalence of high blood cholesterol among US adults. JAMA 1993;269:3009-3014. [Abstract]
  18. Bachorik PS, Lovejoy KL, Carroll MD, Johnson CL. Apolipoprotein B and AI distributions in the United States, 1988–1991: results of the National Health and Nutrition Examination Survey III (NHANES III). Clin Chem 1997;43:2364-2378. [Abstract/Free Full Text]
  19. Bachorik PS, Walker RE, Virgil DG. High density lipoprotein cholesterol in heparin-MnCl2 supernates determined with the Dow enzymic method after precipitation of Mn+2 with HCO3-. Clin Chem 1984;30:839-842. [Abstract/Free Full Text]
  20. Myers GL, Cooper GR, Winn CL, Smith SJ. The Centers for Disease Control-National Heart, Lung and Blood Institute Lipid Standardization Program. An approach to accurate and precise lipid measurements. Clin Lab Med 1989;9:105-135. [ISI][Medline] [Order article via Infotrieve]
  21. National Cholesterol Education Program Working Group on Lipoprotein Measurement. Recommendations on lipoprotein measurement. NIH Publication No. 95-3044. Bethesda: NIH, 1995..
  22. Kafonek SD, Derby CA, Bachorik PS. Biological variability of lipoproteins and apolipoproteins in patients referred to a lipid clinic. Clin Chem 1992;38:864-872. [Abstract/Free Full Text]
  23. . Lipid Research Clinics Program. The Lipid Research Clinics Coronary Primary Prevention Trial results. I. Reduction in incidence of coronary heart disease. JAMA 1984;251:351-374. [Abstract]
  24. Heikki M, Elo O, Haapa K, Heinonen OP, Heinsalmi P, Helo P, et al. Helsinki Heart Study: primary-prevention trial with gemfibrozin in middle-aged men with dyslipidemia. New Engl J Med 1987;317:1237-1245. [Abstract]
  25. Manninen V, Elo MO, Frick H, Haapa K, Heinonen OP, Heinsalmi P, et al. Lipid alterations and decline in the incidence of coronary heart disease in the Helsinki Heart Study. JAMA 1988;260:641-651. [Abstract]
  26. Blankenhorn BH, Nessim SA, Johnson RL, Sanmarco ME, Azen SP, Cashin-Hemphill L. Beneficial effects of combined Colestipol-niacin therapy on coronary atherosclerosis and coronary venous bypass grafts. JAMA 1987;257:3233-3240. [Abstract]
  27. Brensike JF, Levy RI, Kelsey SF, Passamani ER, Richardson JM, Loh IK, et al. Effects of therapy with cholestyramine on progression of coronary arteriosclerosis: results of the NHLBI Type II Coronary Intervention Study. Circulation 1984;69:313-324. [Abstract/Free Full Text]
  28. National Cholesterol Education Program Laboratory Standardization Panel. Recommendations for improving cholesterol measurement. A report from the Laboratory Standardization Panel of the National Cholesterol Education Program. NIH Publication No. 90-2964. Bethesda: NIH, 1990..
  29. Kafonek SD, Derby CA, Bachorik PS. Biological variability of lipoproteins and apolipoprotein in patients referred to a lipid clinic. Clin Chem 1992;38:864-872.
  30. Fraser CG, Cummings ST, Wilkinson SP, Neville RG, Knox JDE, Ho O, et al. Biological variability of 26 clinical chemistry analytes in elderly people. Clin Chem 1989;35:783-786. [Abstract/Free Full Text]
  31. Mjos OD, Rao SN, Bjoru L, Henden T, Thelle DS, Forde OH, et al. A longitudinal study of the biological variability of plasma lipoproteins in healthy young adults. Atherosclerosis 1979;34:75-81. [ISI][Medline] [Order article via Infotrieve]
  32. Harris EK, Kanofsky P, Shakarji G, Cotlove E. Biological and analytic components of variation in long term studies of serum constituents in normal subjects. II. Estimating biological components of variation. Clin Chem 1970;16:1022-1027. [Abstract]
  33. Brown SA, Boerwinkle E, Kashanian FK, Swanson N, Patsch W. Variation in concentrations of lipids, lipoproteins lipids, and apolipoproteins A-I and B in plasma from healthy women. Clin Chem 1990;36:207-210. [Abstract/Free Full Text]
  34. Marcovina SM, Gaur VP, Albers JJ. Biological variability of cholesterol, triglyceride, low- and high-density lipoprotein cholesterol, lipoprotein(a), and apolipoproteins A-I and B. Clin Chem 1994;40:574-578. [Abstract/Free Full Text]
  35. Warnick GR, Albers JJ. Physiological and analytical variation in cholesterol and triglycerides. Lipids 1976;11:203-208. [ISI][Medline] [Order article via Infotrieve]
  36. Young DS, Harris EK, Cotlove E. Biological and analytical components of variation in long term studies of serum constituents in normal subjects. Results of a study designed to eliminate long-term analytic deviations. Clin Chem 1971;17:403-410. [Abstract]
  37. Cotlove E, Harris EK, Williams GZ. Biological and analytic components of variation in long terms studies of serum constituents in normal subjects. III. Physiological and medical implications. Clin Chem 1970;16:1028-1032. [Abstract]
  38. Ford RP. Essential data derived from biological variation for establishment and use of lipid analyses. Ann Clin Biochem 1989;26:281-285.
  39. Bachorik PS, Walker R, Brownell KD, Stunkard AJ, Kwiterovich PO, Jr. Determination of HDL-cholesterol levels in stored human plasma samples. J Lipid Res 1980;21:608-616. [Abstract]
  40. Wood PD, Bachorik PS, Albers JJ, Stewart CC, Winn C, Lippel K. Effects of sample aging on total cholesterol values determine by the automated ferric chloride-sulfuric acid and Liebermann-Burchard procedures. Clin Chem 1980;26:592-597. [Abstract/Free Full Text]
  41. Weisberg S. Applied linear regression, 2nd ed. New York: John Wiley & Sons, 1985:324pp..



The following articles in journals at HighWire Press have cited this article:


Home page
 Endocr Relat CancerHome page
A. E Cust, R. Kaaks, C. Friedenreich, F. Bonnet, M. Laville, A. Tjonneland, A. Olsen, K. Overvad, M. U. Jakobsen, V. Chajes, et al.
Metabolic syndrome, plasma lipid, lipoprotein and glucose levels, and endometrial cancer risk in the European Prospective Investigation into Cancer and Nutrition (EPIC)
Endocr. Relat. Cancer, September 1, 2007; 14(3): 755 - 767.
[Abstract] [Full Text] [PDF]