Back to 3.2

3.3 Withdrawal and related problems with SSRIs

One of the key factors in revealing the benzodiazepine dependence problem was the introduction to the UK market of lorazepam (Ativan, Wyeth). Two features of this drug made withdrawal problems more conspicuous. One was that the UK recommended dose for lorazepam was the equivalent of twice the dose of other BDZs (and double the recommended dose in the USA), and this increased the severity of dependence. Also, unlike the well-established brands, (eg Librium, Valium, Mogadon, Dalmane) lorazepam had a relatively short half-life. As the drug cleared the body quite fast, withdrawal effects became evident soon after stopping it and were more acutely felt. By contrast, the leading drugs had much longer half-lives, so withdrawal effects were attenuated and delayed. Clinical experience with lorazepam in effect gave the game away: this is what prompted Tyrer and colleagues to investigate the whole problem of BDZ withdrawal.

This seems relevant today, first because SSRIs are usually prescribed at higher equivalent doses than alternatives and, secondly, because the exemplar, fluoxetine (Prozac), has an exceptionally long half-life. Significant amount of the drug usually persist in the body for weeks, which explains why the manufacturers say that withdrawal problems are rare.

It is true that reported withdrawal problems with fluoxetine are rare, especially in relation to the huge volume of prescribing. However, bearing in mind the low reporting rate for adverse reactions in general - and that patients may abort withdrawal reactions and stay on the drug - one might not expect to encounter more than a handful of reported cases per several million patient months. As the Table shows, paroxetine (with a much shorter half-life) appears to be the greater culprit; yet fluoxetine has attracted over twice the number of Yellow Card reports about suspected withdrawal problems as diazepam, in many fewer years, so it is hardly in the clear. Similar reports from other countries have prompted the WHO Centre on International Drug Monitoring to identify paroxetine and fluoxetine as the major source of concern. (Stahl et al, 1997)

Reports of withdrawal problems with SSRI antidepressant drugs

Drug Market share Half-life Withdrawal reports
(UK/US brand names) (% NHS scripts for SSRIs, England 1995) (drug plus active metabolites) % of all SSRI Yellow Cards relating to withdrawal
Fluvoxamine (Faverin/Luvox)


17-22 hours


Fluoxetine (Prozac)


4-16 days


Sertraline (Lustral/Zoloft)


22-36 hours


Paroxetine (Seroxat/Paxil)


21 hours, variable


Venlafaxine (Effexor/Efexor)


5-11 hours


Citalopram (Cipramil)


1.5 days


Nefazodone (Dutonin/Serzone)


Up to 24 hours


Though paroxetine is associated with more acute and recognisable withdrawal symptoms, it would not necessarily follow that a higher proportion of users stay on the drug. Evidence from general practice suggests that about 30 per cent of patients stay on both paroxetine and fluoxetine for over six months, but it is not known how many can’t or won’t stop, if indeed they want to. The blurred distinction between can’t/won’t is illustrated in the following accounts. The first two were reported by a celebrated US psychiatrist, notwithstanding his belief that: "it is not addictive - patients do not crave Prozac, and there is no known withdrawal syndrome":

"We lowered the dose of medicine and two weeks later Julia called to say that the bottom had fallen out: ‘I’m a witch again’. She felt lousy - pessimistic, angry, demanding ... and then she used the very words Tess had used: ‘I don’t feel myself’ .... Julia resumed taking the higher dose of Prozac. Within two weeks, she felt somewhat better; after five weeks, she was ‘almost there again’, with many more good days than bad. She said work had been torture on the lower dose of medicine" (Kramer, 1993)

"Three weeks after he began the medication, Paul felt back in control. And, as I had hoped, the drug worked on the chronic issue of self-worth. Paul reported he no longer felt globally inadequate and inferior ... The medicine, he said, gave him the will and the means to continue to face himself ... This statement of independence, Paul felt, differed from the others. In the past, he had wanted therapy but denied himself; now he just felt beyond the need for psychotherapy. He did consider Prozac a ‘crutch’ but said, ‘What the hell. Some people need a crutch to walk’. " (ibid.)

"Although Prozac is not addicting or habituating, people often remain on the medication for extended periods. This should not be surprising, since many are suffering from chronic conditions. After being on, off, then back on Prozac, one woman patient told me, ‘If you ever take me off this drug, I’ll break your kneecaps!’." (Manolis, 1995)

Against this background, and for all the problems of interpretation, the numbers of Yellow Card reports do suggest a problem. After 17 years of use, the benzodiazepines had attracted 28 Yellow card reports of suspected withdrawal problems, while the numbers of reports relating to SSRIs (at March 1997) were pushing the 1,000 mark and increasing. Probably all SSRIs present some risk, though the numbers of reports of suspected withdrawal problems with paroxetine must be unprecedented for any drug. This problem was quietly acknowledged in a note from the CSM/MCA in early 1993, when the numbers of Yellow Card reports were one-tenth the level they had reached by March 1997:

"We have received 78 reports of symptoms occurring on withdrawal of paroxetine, including dizziness, sweating, nausea, insomnia, tremor and confusion. Such reactions have been reported more often with paroxetine than with other SSRIs. Reactions tended to start 1 - 4 days after stopping paroxetine and in several patients resolved on re-instating treatment. Paroxetine should not normally be discontinued abruptly." (CSM/MCA, 1993)

Since then, the CSM/MCA have reported the results of a more detailed investigation (Price et al, 1996). This involved a follow-up questionnaire to doctors who had reported suspected withdrawal reactions with paroxetine and included a brief review of reactions with fluoxetine, fluvoxamine and sertraline as well. This was a more searching study than the "systematic review" of the benzodiazepines (CRM, 1980), but reached broadly similar conclusions essentially on the same basis as before. Again, the cardinal error was to assume that the scale of the problem could be assessed by the numbers of Yellow Cards received, and to produce an absurd underestimate as a result:

"It appears that the reports represent genuine withdrawal reactions, but the low frequency of reporting per thousand prescriptions, together with the published comparative studies suggest that, overall, symptoms due to stopping an SSRI are rare. The absolute risk of a withdrawal reaction with any of the SSRIs may be so low that differences are undetectable except through spontaneous reporting where drug exposure is high." (Price et al., 1996)

The flaws in this study are underlined also by some lack of concordance between numbers and words. Doctors in the follow-up survey had reported that 8 out of 10 paroxetine withdrawal reactions were quite severe but, inexplicably, the CSM/MCA concluded the opposite:

Severity of withdrawal reactions with paroxetine?

What 192 doctors reported to the CSM/MCA: "SEVERE" 21%
"MODERATELY SEVERE"  * See comment 58%
"MILD" 21%
How the CSM/MCA interpreted this: "The withdrawal symptoms observed do not appear to be severe ... this study suggests that they are relatively mild ...".


Doctors in the follow-up survey reported that untreated withdrawal symptoms lasted for an average 10 days (range 1-52 days), and about one in five patients needed treatment with another drug (mainly an SSRI or another antidepressant, or a major or minor tranquilliser). About the same proportion had restarted paroxetine and had not been able to withdraw from it within three months. The CSM/MCA’s conclusion was: "There was no evidence of a physical drug dependency syndrome".

Apparently, no evidence of SSRI withdrawal problems was reported in pre-marketing clinical trials, though much has emerged in case reports published since. (Barr et al., 1994; Benazzi, 1996; Berlin, 1966; Black et al., 1993; Bloch et al., 1995; Debattista & Shatzberg, 1995; Dominguez & Goodnick, 1995; Einbinder, 1995; Ellison, 1994; Farah & Lauer, 1996; Fava & Grandi, 1995; Frost & Lal, 1995; Kasantikul, 1995; Koopowitz & Berk, 1995; Lazowick & Levin, 1995; Leiter et al., 1995; Louie et al., 1994, 1996; Mallya et al., 1993; Mareth & Brown, 1996; Phillips, 1995; Pyke, 1995; Rosenblatt, 1994; Rosenstock, 1996; Stoukides & Stoukides, 1991; Szabadi, 1992). Between them, these indicate an intensity and frequency of withdrawal problems greater than reported for other antidepressants, and to compare with those with benzodiazepines.

This literature also includes one or two reports of neonatal withdrawal reactions resulting from maternal SSRI use in pregnancy. (Kent & Laidlaw, 1995; Spencer, 1993) They are especially interesting because they suggest the purely physical nature of the bond between body and drug. The newborn human mind is less developed than that of a mature, higher order, experimental animal, so the question of "psychological dependence" doesn’t arise. Withdrawal symptoms on their own clearly demonstrate the existence of a purely substance-induced phenomenon, of some physical dependence on the drug.


Contents page