Use of statins and risk of haematological malignancies: a meta-analysis of six randomized clinical trials and eight observational studies

Abstract

Aims

Statins have been suggested to prevent haematological malignancies. Several epidemiological studies have evaluated this association, while randomized controlled trials (RCTs) on cardiovascular outcomes have provided relevant data as secondary end-points. Our aim was to examine the strength of this association through a detailed meta-analysis of the studies published in peer-reviewed literature.

Methods

A comprehensive search for articles published up to December 2006 was performed, reviews of each study were conducted and data abstracted. Prior to meta-analysis, the studies were evaluated for publication bias and heterogeneity. Pooled relative risk (RR) estimates and 95% confidence intervals (CIs) were calculated using the random effects model.

Results

Fourteen studies (six RCTs, seven case–control and one cohort study) contributed to the analysis. Studies were grouped on the basis of study design, and two separate meta-analyses were conducted. There was no evidence of an association between statin use and haematological malignancies among either RCTs (RR = 0.92, 95% CI 0.72, 1.16) or the observational studies (RR = 0.83, 95% CI 0.53, 1.29). Similarly, we found no evidence of publication bias. However, high heterogeneity was detected among the observational studies.

Conclusion

Our meta-analysis findings do not support a potential role of statins in the prevention of haematological malignancies.

Introduction

Statins are some of the most widely prescribed drugs worldwide [1]. They have been developed as treatment for lowering cholesterol by inhibiting hepatic 3-hydroxy-3-methylglutaryl coenzyme A reductase [2] and have been shown to prevent cardiovascular events, not only in the patients with hypercholesterolaemia, but also in patients with a wide range of cholesterol levels in a number of large clinical trials.

Although early studies hinted that these agents might promote several malignant neoplasms at doses similar to those used in humans [3], a significantly elevated risk of cancer has not been reported in the setting of cardiovascular disease prevention. On the contrary, recentexperimental studies have suggested that statins may have chemopreventive potential against haematopoietic and lymphatic malignancies [4–6].

A number of meta-analyses have been reported in the literature, demonstrating no association between statins and cancer risk [7–11]. However, it is very unlikely that exposure such as to statins affects the incidence of all types of cancer, and increases or decreases in a specific type of cancer are likely to be masked by random variation in the effects of statins on all other types. The end-point of all cancers is therefore not very sensitive. Thus, the effect of statins on the risk of haematological malignancies remains to be determined. To address this issue, we carried out a detailed meta-analysis of the studies published on the subject in peer-reviewed literature.

Materials and methods

Search strategy

Studies were identified by a systematic literature search of Medline (1966 to December 2006) and Web of Science (1970 to December 2006) databases. Search terms included: ‘HMG-CoA reductase inhibitor(s)’ or ‘statin(s)’ combined with ‘leuk(a)emia’ or ‘lymphoma’ or ‘multiple myeloma’ or ‘cancer(s)’ or ‘neoplasm(s)’ or ‘malignancy(ies)’. The titles and abstracts of studies identified were scanned to exclude any that were clearly irrelevant. The full text of the remaining articles was read to determine whether it contained information on the topic of interest. The reference lists of articles with information on the topic were also reviewed for additional pertinent studies.

Selection criteria

The studies considered in this meta-analysis were either randomized controlled trials (RCTs) or observational studies (case–control or cohort) that evaluated exposure to statins and risk of haematological malignancies. Articles were excluded if there were insufficient published data for estimating relative risk or a confidence interval. RCTs were considered eligible if they (i) evaluated a statin therapy compared with placebo or no treatment, (ii) had no other intervention difference between the experimental and the control group, (iii) had a minimum duration of 3 years, (iv) enrolled at least 2000 participants, and (v) reported incidence of haematological malignancies during the trial.

We did not assess the methodological quality of the primary studies, since quality scoring in meta-analyses of observational studies is controversial, as it is for RCTs [12, 13], because scores constructed in an ad hoc fashion may lack demonstrated validity and results may not be associated with quality [14]. Instead, several subgroup analyses were performed.

Data extraction

Two reviewers abstracted the data independently. The following data were collected from each study: (i) publication data, first author's last name, year of publication, and country of the population studied; (ii) study design; (iii) number of subjects; (iv) relative risks (RR) and 95% confidence intervals (CIs); (v) types of haematological malignancies studied; (vi) definition of statin exposure; and (vii) control for confounding factors by matching or adjustments, if applicable.

Risk ratios and 95% CIs were calculated for each RCT by reconstructing contingency tables based on the number of subjects randomly assigned and the number of subjects with incident haematological malignancies (intention-to-treat analysis). In observational studies, the risk estimates were extracted that reflected the greatest degree of control for potential confounders. Differences in data extraction were resolved by consensus, referring back to the original article.

Statistical analysis

Included in this meta-analysis are studies reporting different measures of RR: RCTs (risk ratio), case–control studies (odds ratio) and cohort studies (rate ratio). In practice, these measures of effect yield very similar estimates of RR, since the absolute risk of haematological malignancies is very low [15].

Studies were grouped on the basis of study design, and two separate meta-analyses were conducted: one meta-analysis of RCTs and a second one of observational studies. This was done to examine consistency of results across varying study designs with different potential biases.

Summary RR estimates with their corresponding 95% CIs were derived with the method of DerSimonian and Laird [16] by the use of the assumptions of a random effects model, which incorporates both within- and between-study variation. Publication bias was assessed using the Begg and Mazumdar adjusted rank correlation test [17] and the Egger regression asymmetry test [18]. To evaluate whether the results of the studies were homogeneous, the Cochran's Q-test was used [19].

We also calculated the quantity I² that describes the percentage variation across studies that is due to heterogeneity rather than chance [20]. Negative values of I² were put equal to zero, so that I² lies between 0% (i.e. no observed heterogeneity) and 100%. High values would show increasing heterogeneity. Furthermore, the summary RR estimates derived from the two separate meta-analyses (meta-analysis of RCTs vs. meta-analysis of observational studies) were compared with a test of interaction [21].

All P-values are two-tailed. For all tests, a probability level <0.05 was considered to be statistically significant. This work was performed according to the guidelines proposed by the Meta-analysis of Observational Studies in Epidemiology (MOOSE) group [22], and the Quality of Reporting of Meta-analyses (QUOROM) recommendations for improving the quality of meta-analyses of RCTs [23]. Stata software was used for the statistical analyses (Stata Corp., College Station, TX, USA).

Results

Search results

The initial search strategy yielded 1368 records. After screening the titles and abstracts, we retrieved 46 potentially relevant manuscripts for further review. The full text was read and the reference lists were checked. Finally, 14 independent studies met the predefined inclusion criteria [24–37]. Six out of 14 were randomized, double blind, placebo-controlled trials of statins for cardiovascular outcomes [24–29], seven were case–control studies [30–32, 34–37] and one was a cohort study [33]. The number of cases with haematological malignancy ranged from 18 to 116 in the RCTs, from 24 to 2362 in the case–control studies and was 1626 in the cohort study.

All RCTs reported site-specific cancer outcomes including haematological malignancies. It was therefore possible to conduct a post hoc analysis of these trials and calculate risk ratios for haematological malignancies in an intention-to-treat analysis. All observational studies evaluated exposure to statins and risk of haematological malignancies, and were controlled for potential confounding factors (at least for age) by matching or adjustments. The publication dates of the studies included in the meta-analysis were between 1996 and 2006. Study designs, along with the estimated RRs and 95% CIs, are shown in Table 1 for the RCTs and Table 2 for the observational studies.

Table 1.

Randomized, double-blind, placebo-controlled trials included in the meta-analysis

				Incident haematological malignancies
Study	Agent	No. of subjects	Duration (years)	Statin group	Placebo group	RR (95% CI)	Reported outcome
4S [24]*	Simvastatin	4444	Median: 10.4	17 of 2221	19 of 2223	0.90 (0.47, 1.72)	Incident haematological malignancies
ALERT [25]	Fluvastatin	2094	Mean: 5.1	11 of 1045	18 of 1049	0.61 (0.29, 1.29)	Incident haematological malignancies
HPS [26]	Simvastatin	20536	Mean: 5.0	64 of 10 269	52 of 10 267	1.23 (0.85, 1.77)	Incident haematological malignancies
LIPID [27]	Pravastatin	9014	Mean: 8.0	37 of 4512	52 of 4502	0.71 (0.47, 1.08)	Incident lymphomas and leukaemias
AFCAPS [28]	Lovastatin	6605	Mean: 5.2	12 of 3304	11 of 3301	1.09 (0.48, 2.47)	Incident lymphomas
CARE [29]	Pravastatin	4159	Mean: 4.8	8 of 2081	10 of 2078	0.80 (0.32, 2.02)	Incident lymphomas and leukaemias

RR, Relative risk (risk ratio); CI, confidence interval.

^*Numbers in parentheses, reference citation.

Table 2.

Observational studies included in the meta-analysis

Study	Study location	Study design	All subjects	HM cases	RR (95% CI)	Control for potential confounders*	Type of HM studied
Fortuny et al. 2006 [30]†	Czech Rep., France, Germany, Ireland, Italy and Spain	C-C	4568	2362	0.61 (0.45, 0.84)	1–3	Incident lymphoma
Iwata et al. 2006 [31]	Japan	C-C	1100	221	2.24 (1.37, 3.66)	1, 2, 4–6	Incident lymphoma and myeloma
Landgren et al. 2006 [32]	USA	C-C	870	179	0.4 (0.2, 0.8)	1, 7–9	Incident myeloma
Friis et al. 2005 [33]	Denmark	Cohort	334 754	1626	0.88 (0.60, 1.29)	1, 2, 10–13	Incident haematological malignancies
Graaf et al. 2004 [34]	The Netherlands	C-C	20105	93	0.28 (0.06, 1.30)	1, 2, 12–22	Incident lymphoma
Zhang et al. 2004 [35]	USA	C-C	1318	601	0.5 (0.4, 0.8)	1, 9, 23, 24	Incident non-Hodgkin lymphoma
Blais et al. 2000 [36]	Canada	C-C	264	24	2.17 (0.38, 12.36)	1, 2, 4, 18, 25, 26	Incident lymphoma
Traversa et al. 1998 [37]	Italy	C-C	2222	202	1.5 (0.8, 2.6)	1, 2	Incident leukaemia

HM, Haematological malignancy; RR, relative risk; CI, confidence interval.

^*1, age; 2, gender; 3, country; 4, year of visit; 5, serological status for antihepatitis B surface antigens; 6, serological status for antihepatitis C virus antibodies; 7, race; 8, education; 9, body mass index; 10, calendar period; 11, use of cardiovascular drugs; 12, use of nonsteroidal anti-inflammatory drugs; 13, use of hormone replacement therapy; 14, geographical region; 15, duration of follow-up; 16, diabetes mellitus; 17, prior hospitalizations; 18, chronic disease score; 19, chronic use of diuretics; 20, chronic use of angiotensin-converting enzyme inhibitors; 21, chronic use of calcium channel blockers; 22, use of other lipid-lowering therapy; 23, menopausal status; 24, family history of non-Hodgkin lymphoma in first-degree relatives; 25, previous neoplasm; 26, use of fibric acids.

^†Numbers in parentheses, reference citation.

Meta-analysis of RCTs

Six large RCTs contributed to the analysis [24–29]. A total of 46 852 individuals (females ∼21%) participated in these trials: 23 432 in treatment groups and 23 420 in placebo groups. The participants had a mean age of 61 years at enrolment and a mean follow-up of approximately 6.1 years. A total experience of 286 000 person-years was reached.

Four trials [24, 25, 27, 29] reported a lower risk of haematological malignancies in the treatment group, whereas the other two trials [26, 28] reported a higher risk (Table 1). None was statistically significant. Meta-analysis of all six reports showed no evidence for an association between statin treatment and risk of haematological malignancies. The overall rate of haematological malignancies was 0.64% in the statin group (149 incident cases) and 0.69% in the placebo group (162 incident cases). The association of statin use with haematological malignancies was not statistically significant (RR = 0.92, 95% CI 0.72, 1.16). The Cochran's Q-test had a P-value of 0.38 [Q = 5.3 on 5 degrees of freedom (d.f.)] and the corresponding quantity I² was 6%, both indicating very little variability between studies that cannot be explained by chance. The P-values for the Begg's and the Egger's tests were P = 0.99 and P = 0.48, respectively, both suggesting that an assumption of no publication bias is reasonable. After stratifying the data in two subgroups (lipophilic [24–26, 28]vs. lipophobic statins [27, 29]), no statistically significant association was found between lipophilic or lipophobic statins and risk of haematological malignancies (Table 3). Figure 1 shows the RRs and 95% CIs from the individual trials and the pooled results.

Table 3.

Results of meta-analysis

			Tests of homogeneity			Tests of publication bias
	No. of studies	Pooled effect estimate RR (95% CI)	Q-value (d.f.)	P-value	I2	Begg's P-value	Egger's P-value
All studies	14	0.85 (0.64, 1.12)	45.0 (13)	<0.001	71%	0.91	0.70
RCTs	6	0.92 (0.72, 1.16)	5.3 (5)	0.38	6%	0.99	0.48
RCTs of lipophilic statins	4	1.04 (0.79, 1.37)	3.0 (3)	0.40	0%	0.73	0.26
RCTs of lipophobic statins	2	0.72 (0.49, 1.06)	0.1 (1)	0.82	0%	0.99	–
Observational studies	8	0.83 (0.53, 1.29)	38.1 (7)	<0.001	82%	0.54	0.66
Case–control studies	7	0.82 (0.47, 1.41)	37.5 (6)	<0.001	84%	0.76	0.66
Cohort studies	1	0.88 (0.60, 1.29)	–	–	–	–	–

RR, Relative risk; CI, confidence interval; d.f., degrees of freedom, RCT, randomized controlled trial.

Figure 1.

Results from individual studies and meta-analyses. The relative risk and 95% confidence interval for each study are displayed on a logarithmic scale. Pooled estimates are from a random effects model

Meta-analysis of observational studies

Seven case–control studies [30–32, 34–37] and one cohort study [33] evaluated exposure to statins and risk of haematological malignancies. The meta-analysis encompassed these eight studies in a total of 365 201 individuals, of whom 5308 had a haematological malignancy. This time, the Cochran's Q-test had a P-value of <0.001 (Q = 38.1 on 7 d.f.) and the quantity I² was 82%, both indicating high heterogeneity between the studies (Table 3). In contrast, the P-values for the Begg's and the Egger's tests were P = 0.54 and P = 0.66, respectively, both suggesting a low probability of publication bias. Statin use did not significantly affect the risk of haematological malignancies (RR = 0.83, 95% CI 0.53, 1.29) (Table 3). Figure 1 shows the RRs and 95% CIs from the individual studies and the pooled results.

When the analysis was restricted to the seven case–control studies (exclusion of the cohort study [33]), the results did not substantially change (RR = 0.82, 95% CI 0.47, 1.41; Cochran's P < 0.001 and I² = 84%; Begg's P = 0.76 and Egger's P = 0.66).

Overall analysis

We compared the summary RR estimates derived from the two separate meta-analyses with a test of interaction [22]. The difference between estimates of statin effect on the risk of haematological malignancies, in RCTs and observational studies, was not statistically significant (Z = 0.3, P = 0.7).

A combined analysis of RCTs and observational studies was also performed. The Cochran's Q-test had a P-value of <0.001 (Q = 40.0 on 13 d.f.) and the quantity I² was 71%, both indicating high heterogeneity between the 14 studies (Table 3). In contrast, the P-values for the Begg's and the Egger's tests were P = 0.91 and P = 0.70, respectively, both suggesting a very low probability of publication bias. Statin use did not significantly affect the risk of haematological malignancies (RR = 0.85, 95% CI 0.64, 1.12).

Discussion

There is a long-standing debate concerning the association between use of statins and cancer. During preclinical and clinical development of statin-class drugs, animal studies have shown an increased risk of malignant neoplasms in rodents exposed to statins at doses similar to those administered to humans [3]. Interestingly, the commonly used statin pravastatin caused malignant lymphomas in mice at doses that ranged from 0.5 to fivefold the maximum recommended dose for humans [3]. In contrast, several recent mechanistic, in vivo and observational studies have suggested that statins may actually have chemopreventive potential against cancer at various sites [38], including haematopoietic and lymphatic tissue [4–6, 30].

Meta-analysis serves as a valuable tool for studying rare and unintended effects of a treatment. It extends prior randomized and nonrandomized studies by permitting synthesis of data and providing more stable estimates of effect. This meta-analysis of published studies does not provide evidence that statin use is associated with a substantially decreased or increased risk of neoplastic diseases of the haematopoietic and lymphatic tissue. Furthermore, our findings are in line with recent meta-analyses on the association between use of statins and other site-specific cancers. Likewise, they indicate that statin use does not substantially affect respiratory [11], colorectal [39] or breast cancer risk [40]. However, since our meta-analysis pulled all haematological malignancies, the possibility of a beneficial (or harmful) effect of statins on a specific haematological malignancy cannot be totally excluded.

When meta-analysis of published literature is performed, consideration of study bias is critical. Existence of a bias in favour of publication of statistically significant results is well documented [41–43]. However, the likelihood of important selection or publication bias in our results is small. During the identification and selection process, no article was excluded because of methodological characteristics or any subjective quality criteria, and the Begg's as well as the Egger's test revealed no relation between the estimate of RR and study size. We are therefore confident that important publication bias due to preferential publication of large studies with significant findings is unlikely to have occurred.

On the contrary, the tests of heterogeneity indicated very high variability between the observational studies included in our analysis. A likely explanation is that these observational studies, whether they showed a beneficial or a harmful effect of statin use on the risk of haematological malignancies, all appear to suffer from limitations resulting from selection bias, information bias and unaccounted confounding. As is usually the case in any epidemiological study, there are limits to what can be done to reduce unmeasured or unknown sources of bias.

Limitations of this meta-analysis stem from the limitations of the primary studies included in the analysis. The first meta-analysis included a group of trials of statins for cardiovascular outcomes, which reported the incidence of haematological malignancies. The examined populations varied and the risk of haematological malignancies was about 1.1 in 1000 per year, which could make it difficult to detect any effects, beneficial or harmful. Treatment and follow-up times were an average of only 6.1 years, which might be considered a short period on which to base definite conclusions compared with the latency time between the initiation and the clinical detection of a malignancy.

Detection or surveillance of adverse events, such as malignancy, may have varied considerably between trials, for many reasons, including differences in the threshold of patients and physicians to report; and the mode of data collection (in particular, active vs. passive surveillance for harm). Furthermore, the fact that occurrence of haematological malignancies was not the primary objective of these trials might have affected the detection rate, but this factor would probably have affected both arms of the trials equally, which would bias the results towards the null.

On the other hand, the second meta-analysis included observational studies that lacked the experimental random allocation of the intervention necessary to test exposure–outcome hypotheses optimally. They may also suffer from the ‘healthy volunteer’ bias, which is an important limitation of observational studies. These studies were also different in terms of study design and definitions of drug exposure.

Although randomized and observational designs sometimes produce equivalent results [44–46], systematic reviews have found that they often give different results, and that the difference is in all directions [47]. In our case, it is noteworthy that the findings were similar in both meta-analyses of RCTs and observational studies, although the primary studies had varying study designs with different potential biases. This fact strengthened our confidence in the validity of our results.

In conclusion, meta-analysis of existing data does not support a potential role of statins in prevention of haematological malignancies. However, given the high and growing prevalence of statin use, it is important to monitor the experience with statins for extended follow-up periods, to identify potential effects in the longer term. Until then, physicians need to be vigilant in ensuring that use of statins remains restricted to the approved indications.

Competing interests: None declared.