Is the Schizophrenia Mortality Study in The Lancet CREDIBLE?

A study purporting to analyze mortality rates of 66, 881 schizophrenia patients in Finland (1973 to 2005) was published in the prestigious journal, The Lancet. [Abstract below]

The study has received much media attention because the authors claim—contrary to well documented previous reports about spiraling mortality rates among schizophrenia patients treated between the 1970s and 2000 [1, 2, 3, 4, 5, 6]—that the use of the second generation antipsychotic drugs in Finland was associated with lower mortality rate compared to no drug treatment.

What’s more, the authors claim that Clozaril (clozapine) among all antipsychotics was associated with the lowest number of deaths, and its restricted use should therefore be reassessed for use a first-line treatment. Unless the overall treatment and services provided to schizophrenia patients in Finland is unique and especially protective—which the authors do not suggest—their claimed findings are belied by a consistent body of evidence.

To whit, an eight year study of FDA MedWatch adverse drug reports (between 1998 to 2005) found that Clozapine was linked to 3,277 deaths, making it the third most dangerous prescribed drug in the U.S. Clozapine was also linked to over 4,300 adverse drug events that led to disability or required serious medical intervention. [7]

A critical analysis by psychiatrist, Grace Jackson, MD (below), identifies fatal flaws in the study design and numerous methodological artifacts that introduced bias which minimized the detection of drug-related mortality.

1. To begin with, the claim that “no drug use” resulted in the HIGHEST mortality was misleading inasmuch as the non-drug users were, in fact, NOT “antipsychotic naïve” patients. The patients assigned to the “No Drug Use” group were simply those individuals who did not use antipsychotics as outpatients between 1996 and 2006.

2. In the Analysis of Cumulative Drug Use, the “cumulative” periods of exposure was restricted to 8.6 year average duration of drug use, and EXCLUDED the drugs used by patients PRIOR to the year 1996. Given the fact that the “start” date for patient entry was 1973, some patients in this study may have received antipsychotic drugs between 1973 and 1996 (23 years) without having these 23-years of drug use COUNTED in the final analysis.

What reasonable explanation is there for the 23-year data exclusion?

3. The investigators failed to disclose ABSOLUTE #s of DEATHS per drug—no information was given about the total numbers of patients dying in relation to the use of each drug. 

Inasmuch as a mandatory Clozaril database documents the exact number of patients prescribed Clozaril, it is possible–and would have been especially informative–to compare the actual number of Clozaril-related deaths with the actual number of patients exposed to the drug.

4. Reporting of mortality events in terms of PERSON YEARS of EXPOSURE had the effect of “masking” the actual importance of drug-related toxicity by “drowning out” the signal of toxicity provided by those individuals who died sooner.

Inexplicably, the investigators excluded any deaths which occurred AFTER the first 48 hours of a new hospital admission (e.g., if a current Clozaril user died during day #4 of hospitalization, this death was excluded).

The authors’ conclusions favoring the long-term use of second-generation antipsychotics, and Clozaril in particular, as Dr. Jackson demonstrates, are not supported by their study design: indeed, their "favorable" conclusions about patients’ mortality were based upon numerous confounders (methodological tricks and problems) which mitigated the detection of actual drug-associated mortality.

By way of analogy:
This study design was akin to comparing rates of lung cancer among former smokers who had smoked cigarettes for as long as 23 years (1973 through 1996) to those who were relatively new smokers (1996 to 2006: average duration of smoking = 8.6 years). Lung Cancer rates were excluded for any individual whose cancer emerged prior to 1996. In response to the finding that lung cancer rates among “non-smokers’ (people no longer smoking between 1996 and 2006) were higher than current smokers, the tobacco industry falsely claimed that “current smoking protects against lung cancer.”

Although the free abstract / title page (which is read by most) indicates that this study was funded by the Finnish government, the principle authors of the study have substantial financial ties to companies that market antipsychotic drugs, having served as expert witnesses / lecturers / and or consultants:

The principle investigator, Dr. Tiihonen, has served as a consultant to Lundbeck, Organon, Janssen-Cilag, EH Lilly, and Bristol-Myers Squibb, and has received fees for giving expert opinions to Bristol-Myers Squibb and GlaxoSrnithKlme, and lecture fees from Janssen-Cilag, Bristol Myers-Squibb, Eli Lilly, Pfizer, Lundbeck, GlaxoSmithKline, and Astra Zeneca. JT has not served as consult, received any fees or had interest with any pharmaceutical company manufacturing, marketing, or selling clozapine.

Dr. Haukka has served as a consultant to Janssen-Cilag.
Dr. Niskanen has served as consultant to Eli Lilly and Bristol Myers-Squibb, and has received fees for giving expert opinions to Bristol Myers-Squibb, and lecture fees from Janssen-Cilag, Bristol Myers-Squibb, Eli Lilly, and Astra Zeneca.

References:

1. Maria G. Morgan, Paul J. Scully, Hanafy A. Youssef, Anthony Kinsella,John M. Owens, John L. Waddington, Prospective analysis of premature mortality in schizophrenia in relation to health service engagement: a 7.5-year study within an epidemiologically complete, homogeneous population in rural Ireland, Psychiatry Research 117 (2003) 127–135.

2. MATTI JOUKAMAA, MARKKU HELIOVAARA, PAUL KNEKT,HELIO« VAARA, ARPO AROMAA, RAIMO RAITASALO and VILLE LEHTINEN.Schizophrenia, neuroleptic medication and mortality, British Journal of Psychiatry (2 0 0 6) 1 8 8 : 1 2 2-1 2 7.

3. John Bola, Medication-Free Research in Early Episode Schizophrenia: Evidence of Long-Term Harm? Schizophrenia Bulletin vol. 32 no. 2 pp. 288–296, 2006

4. Martin Harrow, PhD, and Thomas H. Jobe, MD. Factors Involved in Outcome and Recovery in Schizophrenia Patients Not on Antipsychotic Medications: A 15-Year Multifollow-Up Study, The Journal of Nervous and Mental Disease • Volume 195, Number 5, May 2007.

5. Sukanta Saha, MSc, MCN; David Chant, PhD; John McGrath, MD, PhD, FRANZCP A Systematic Review of Mortality in Schizophrenia: Is the Differential Mortality Gap Worsening Over Time? Arch Gen Psychiatry. 2007;64(10):1123-1131

6. Morbidity and Mortality in People with Serious Mental Illness . National Association of State Mental Health Program Directors (NASMHPD) Oct. 2006. Tenth in a Series of Technical Reports

7. Thomas J. Moore, AB; Michael R. Cohen, RPh, MS, ScD; Curt D. Furberg, MD, PhD  Serious Adverse Drug Events Reported to the Food and Drug Administration, 1998-2005, Arch Internal Medicine, 2007;167(16):1752-1759

~~~~~~~~~~~~~~~~~~~~

The Lancet, Early Online Publication, 13 July 2009
11-year follow-up of mortality in patients with schizophrenia: a population-based cohort study (FIN11 study). Prof Jari Tiihonen MD Corresponding Author, Prof Jouko Lönnqvist MD, Prof Kristian Wahlbeck MD, Prof Timo Klaukka MD, Prof Leo Niskanen MD, Antti Tanskanen PhLic,  Jari Haukka PhD

Summary
Background
The introduction of second-generation antipsychotic drugs during the 1990s is widely believed to have adversely affected mortality of patients with schizophrenia. Our aim was to establish the long-term contribution of antipsychotic drugs to mortality in such patients.

Methods
Nationwide registers in Finland were used to compare the cause-specific mortality in
66, 881 patients versus the total population (5•2 million) between 1996, and 2006, and to link these data with the use of antipsychotic drugs. We measured the all-cause mortality of patients with schizophrenia in outpatient care during current and cumulative exposure to any antipsychotic drug versus no use of these drugs, and exposure to the six most frequently used antipsychotic drugs compared with perphenazine use.

Findings
Although the proportional use of second-generation antipsychotic drugs rose from 13% to 64% during follow-up, the gap in life expectancy between patients with schizophrenia and the general population did not widen between 1996 (25 years), and 2006 (22•5 years). Compared with current use of perphenazine, the highest risk for overall mortality was recorded for quetiapine (adjusted hazard ratio [HR] 1•41, 95% CI 1•09—1•82), and the lowest risk for clozapine (0•74, 0•60—0•91; p=0•0045 for the difference between clozapine vs perphenazine, and p<0•0001 for all other antipsychotic drugs). Long-term cumulative exposure (7—11 years) to any antipsychotic treatment was associated with lower mortality than was no drug use (0•81, 0•77—0•84). In patients with one or more filled prescription for an antipsychotic drug, an inverse relation between mortality and duration of cumulative use was noted (HR for trend per exposure year 0•991; 0•985—0•997).

Interpretation
Long-term treatment with antipsychotic drugs is associated with lower mortality compared with no antipsychotic use. Second-generation drugs are a highly heterogeneous group, and clozapine seems to be associated with a substantially lower mortality than any other antipsychotics. Restrictions on the use of clozapine should be reassessed.

Funding
Annual EVO Financing (Special government subsidies from the Ministry of Health and Welfare, Finland).

======================================================

What The Study Actually Did and Found by Grace Jackson, MD

This was an investigation which involved five separate steps. Goal was to compare
All-Cause Mortality in relation to exposures (current vs. cumulative) to antipsychotic drugs.

Step #1 Identification of Schizophrenia Cases (inpatient)
Researchers used the National Hospital Discharge Register to identify all individuals hospitalized for schizophrenia at any time between 1/1/73 and 12/31/04. n = 66,881

Step #2 Identification of Outpatient Use of Antipsychotics 1996-2006
Next, investigators used Finland’s national pharmacy database in order to identify OUTPATIENT purchases of SEVEN antipsychotic drugs between the years 1/1/96 and 2006.

Included: oral perphenazine, oral haloperidol, thioridazine, clozapine, olanzapine, oral risperidone, quetiapine

Excluded: all other neuroleptics; all other psychiatric drugs (researchers provided zero information about the use of any other medications)

As this research project sought to evaluate MORTALITY RATES associated with CURRENT drug use and CUMULATIVE drug use, the methodologies for evaluating those features were as follows:

In the Analysis of Current Drug Use (mortality rates associated with CURRENT drugs), four categories of patients were created:

NO DRUG USE = no use of antipsychotics between 1996 and 2006
other drug use = “rarely used” [no explicit definition provided]
polypharmacy = mixed antipsychotic drug use between 1996 and 2006
monotherapy = limited to 3 first-generation and 4 second-generation antipsychotics

In the Analysis of Cumulative Drug Use (mortality rates associated with cumulative duration of drug use), the “cumulative” periods of exposure EXCLUDED drug use which occurred PRIOR to the year 1996.

Durations of drug use were NOT obtained by reviewing medical records or patient charts.
Rather, the Duration of Drug Use was estimated by converting outpatient pharmacy data into Defined Daily Doses (DDD) where DDD = DEFINED DAILY DOSE

From the World Health Organization’s website:
“The DDD is the assumed average maintenance dose per day for a drug used for its main indication in adults. A DDD will only be assigned for drugs that already have an ATC code.

It should be emphasised that the defined daily dose is a unit of measurement and does not necessarily reflect the recommended or prescribed daily dose. Doses for individual patients and patient groups will often differ from the DDD and will necessarily have to be based on individual characteristics (e.g. age and weight) and pharmacokinetic considerations.

Drug consumption data presented in DDDs only give a rough estimate of consumption and not an exact picture of actual use. DDDs provide a fixed unit of measurement independent of price and formulation enabling the researcher to assess trends in drug consumption and to perform comparisons between population groups.”

In the text of their article, Tiihonen et al disclosed that the average age of the patients at the beginning of the 11-year window of follow-up (1996) was age 51.
The average duration of antipsychotic drug use (1996 to 2006) was 8.6 years.

It is essential to appreciate the fact that this 8.6 year duration of drug use EXCLUDED the drugs used by patients prior to 1996. Given the fact that the “start” date for patient entry was 1973, some patients in this study may have received antipsychotic drugs between 1973 and 1996 (23 years) without having these 23-years of drug use COUNTED in the final analysis.

Step #3 Delineation of Mortality (all causes)
Using data from a national statistical database (Statistics Finland), mortality statistics and causes of death were obtained from each deceased patient.

Data from patients who died prior to 1996 were NOT included in the analysis of “current drug use / mortality” nor were they included in the analysis of “cumulative drug use / mortality.”

Resulting Problem (methodological confounder):
Mortality data were excluded for all of the patients who had been admitted between 1973 and 1996, but who died prior to 1996. This suggests that data may have been lost from patients who had been exposed to antipsychotic drugs for as long as 23 years.

Furthermore, as clozapine was the ONLY second-generation antipsychotic on the market prior to 1996, the deletion of this data necessarily DIMINISHED the mortality statistics which would otherwise have been collected for the users of that drug.

Step #4 Calculation of Mortality Rates according to drug use
The numbers of deaths (for each drug vs. “no drug use”) were reported in terms of DEATHS per 1000 PERSON YEARS. [e.g., if a patient consumed Clozaril for 1 year, that person contributed 1 person year]

a) investigators failed to disclose ABSOLUTE #s of DEATHS per drug
e.g., no information was given about the total #s of patients dying in relation o the use of each drug

Resulting Problem (methodological confounder):
The reporting of mortality events in terms of PERSON YEARS of EXPOSURE had the effect of “masking” the actual importance of drug-related toxicity, because the Person Years contributed by patients who are able to survive end up “drowning out” the signal of toxicity provided by those individuals who succumb to drug-lethality sooner.]

b) In the Analysis of Mortality with Current Drug Use, investigators excluded any deaths which occurred AFTER the first 48 hours of a new hospital admission (e.g., if a current Clozaril user died during day #4 of hospitalization, this death was excluded)

Resulting Problem (methodological confounder):
Given the fact that patients on neuroleptics commonly die from pneumonia, heart failure, DKA, Neuroleptic malignant syndrome, or non-pulmonary infections – and given the fact that these deaths often occur AFTER the first 48 hours of hospitalization, this exclusion criterion necessarily confounded the results by DECREASING the detection of drug-related deaths associated with ONGOING antipsychotic use.

c) In the Analysis of Mortality with Cumulative Drug Use, investigators included any and all deaths (including those which occurred AFTER the first 48 hrs of hospitalization)

Resulting Problem (methodological confounder):
Delayed hospital deaths of patients assigned to the “no antipsychotic drugs” category were included in the cumulative drug analysis [yet, delayed deaths were EXCLUDED from the “current drug use” analysis].

The impact of “b” and “c” (confounders) was the inflation of mortality associated with past (but discontinued) drug use; along with the deflation of mortality associated with present (ongoing) drug use.

Step #5 Correlation of Mortality Rates with Respect to Drug Use
The relative risks of death according to cumulative use of specific drugs were pegged (compared) to perphenazine (Trilafon) as the reference drug. The overall conclusion about clozapine as a relative “life saver” was based upon relative mortality, as compared to perphenazine — not relative mortality, as compared to NO antipsychotic treatment EVER.

The investigators then celebrated the shortened longevity of all antipsychotic users in 2006, based upon a claim that the “relative decrease in lifespan” had improved between 1996 and 2006.

In other words, after discounting numerous deaths (as explained above — all of the deaths which occurred prior to 1996; and all of the deaths which occurred after 48 hours of admission for current drug users), chizophrenic patients were reportedly living 1.5 to 2.5 years LONGER in 2006 than they had lived in 1996.

However, for schizophrenics aged 20 (in 1996), the use of antipsychotic drugs was still associated with a LIFE SPAN that was 22.5 years SHORTER than the general population of the same age; for schizophrenics aged 40 (in 1996), the use of antipsychotic drugs was associated with a LIFE SPAN that was 17 years shorter than the general population of the same age.

Overall, the researchers claimed that their results showed that life expectancy among schizophrenics increased by 4.9 years between 1996 and 2006 (while it increased by 2.4 years in the non-schizophrenic population).

Summary

The authors concluded: “We showed that long-term use of any antipsychotic drug was associated with lower mortality than was no use of antipsychotic drug.”

This conclusion was not supported by their study design, based upon numerous confounders (methodological tricks and problems) which mitigated the detection of drug-associated mortality.

1) the claim that “no drug use” resulted in the HIGHEST mortality was misleading

The non-drug users were, in fact, NOT “antipsychotic naïve” patients.
In reality, the patients assigned to the “No Drug Use” group were simply those individuals who did not use antipsychotics as outpatients between 1996 and 2006. An unknown number of these patients (hospitalized between 1973 and 2004) may have been exposed to antipsychotic drugs (including Clozaril) for many years prior to the window of follow-up.

2) numerous confounders minimized the detection of drug-associated mortality, including:
     a) the use of Defined Daily Dose as a surrogate for drug exposures
     b) the exclusion of death statistics from all patients who died before 1996
     c) the reporting of mortality data in terms of “deaths per Person Years” rather than in terms of ABSOLUTE #s of DEATH
        per Drug
     d) the mixed (contradictory) reporting of deaths which occurred 48 hours AFTER a hospital admission
     [for current users of drugs, these deaths were excluded; for past users of drugs, these deaths were included]

3) active use of antipsychotic drugs was never shown to “decrease” death, relative to “no use”

In actuality, this study compared death rates among “current drug users” with death rates among “former drug users” whose cumulative exposures prior to 1996) were never considered in the calculations of total drug use.

By way of analogy:
This study was akin to comparing rates of lung cancer among former smokers who had smoked cigarettes for as long as 23 years (1973 through 1996) relative to those who were relatively new smokers (1996 to 2006: average duration of smoking = 8.6 years). Lung Cancer rates were excluded for any individual whose ancer emerged prior to 1996. Meanwhile, in response to the finding that lung cancer rates among “non-smokers’ (people no longer smoking between 1996 and 2006), were higher than current smokers, the researchers erroneously claimed that “current smoking protects against lung cancer.”

4) The study confirmed the fact that former AND current users of antipsychotics — whether use involved OLD drugs, or NEWER drugs — died 17 or 23 years SOONER (in 2006) than age-matched, [presumably] never-neuroleptized members of the Finnish population.

5) Source of Funding:
Although the title page reveals that this study was funded by the Finnish government (government subsidies from the Ministry of Health and Welfare), the authors of the study have served as expert witnesses / lecturers / and or consultants on behalf of the following drug companies:

JT = Lundbeck, Organon, Janssen-Cilag, Eli Lilly, Bristol-Myers Squibb, GSK,
Pfizer, Astra Zeneca
JH = Janssen-Cilag
LN = Eli Lilly, Bristol-Myers Squibb, Janssen-Cilag, Astra Zeneca

Grace E. Jackson, MD

~~~~~~~~~~~

REUTERS
Study may prompt rethink on schizophrenia drugs
Sun Jul 12, 2009

By Ben Hirschler

* Fewer deaths with clozapine than newer drugs
* Call to consider cheap generic as first-line treatment
* Finnish study fuels debate on atypical antipsychotics

LONDON, July 13 (Reuters) – Schizophrenia patients given a cheap older drug are less likely to die prematurely than people on newer treatments, despite the older product’s well-known adverse side effects, Finnish researchers said on Monday.

The finding may lead to wider use of clozapine — sold by Novartis as Clozaril, but also available as a generic — instead of newer drugs like AstraZeneca’s Seroquel, the current market leader.

Clozapine was the first of a new generation of schizophrenia drugs, known as atypical antipsychotics. But its use has been restricted by health authorities because of safety concerns and patients taking it require regular blood tests.

Despite this, an analysis of 10 years’ records for 67,000 patients in Finland found that, compared to treatment with the first-generation drug perphenazine, the risk of early death for patients on clozapine was reduced by 26 percent.

By contrast, mortality risk was 41 percent higher for those on Seroquel, known chemically as quetiapine; 34 percent higher with Johnson & Johnson’s Risperdal, or resperidone; and 13 percent higher with Eli Lilly’s Zyprexa, or olanzapine.

"We know that clozapine has the highest efficacy of all the antipsychotics and it is now clear, after all, that it is not that risky or dangerous a treatment," study leader Jari Tiihonen of the University of Kuopio said in a telephone interview.

"We should consider whether clozapine should be used as a first-line treatment option."

THOUSANDS OF PREMATURE DEATHS
Tiihonen estimates clozapine is given to around one fifth of Finnish schizophrenia patients, but less than 5 percent in the United States.

Clozapine’s side effects include agranulocytosis, a potentially fatal decline in white blood cells, and current rules stipulate the drug can only be used after two unsuccessful trials with other antipsychotics.

Tiihonen and colleagues wrote in the Lancet medical journal that these restrictions should be reassessed in the light of their findings, since not using the drug may have caused thousands of premature deaths worldwide.

Seroquel, Zyprexa and Risperdal are among the world’s top-selling drugs, with a combined sales of $12.5 billion in 2008, although Risperdal now faces generic competition.

Worries about the safety profile of all the atypical antipyschotics have loomed large since 2002, however, following evidence of increased rates of diabetes and cardiovascular disease.

The Finnish study found no pronounced differences in heart deaths between the different atypicals, but patients on clozapine had a substantially lower risk of suicide while those on Seroquel were more likely to kill themselves.

An AstraZeneca spokeswoman said the Anglo-Swedish company was comfortable that Seroquel was safe, effective and an important treatment for mental illness.

Schizophrenia is a severe psychiatric disorder in which patients experience distorted thinking, hallucinations and abnormal emotions. (Editing by Jon Loades-Carter)

© Thomson Reuters 2009. All rights reserved