Feb 27, 1982 - Sherwood Hospital,. Nottingham NG5 IPD. J R A MITCHELL. Department of Medicine,. University Hospital,. Nottingham NG7 2UH. A H SHORT.
BRITISH MEDICAL JOURNAL
664
Elderly implies that the answers which may be produced by this large multinational, hospital-based trial will be no more clearcut than the ones we reported from our small single-centre trial. If so, we share the hopes of Professor W S Peart and his colleagues (p 1397) that a further "trial of antihypertensive treatment in people aged 65-74 may be mounted attempting to determine the balance between the benefits of pressure reduction and any adverse effects." M E SPRACKLING Department of Health Care of the Elderly, Sherwood Hospital, Nottingham NG5 IPD
J R A MITCHELL Department of Medicine, University Hospital, Nottingham NG7 2UH
A H SHORT Department of Physiology, Medical School, Nottingham NG7 2UH
G C M WATT Glyncorrwyg Health Centre, Glyncorrwyg, W Glam SA13 3DP
SIR,-I must thank Dr Mary R Bliss (30 January, p 347) for her comments on my letter (2 January, p 50) but point out that we are writing of two quite different problems. I wrote of patients referred to me because of severe hypertension with symptoms-left ventricular failure, retinal haemorrhage, subarachnoid bleeds, and transient cerebral episodes were among them. Treatment was essential, but after years of blood pressure control I tried to discover whether the survivors, by then elderly, could be weaned from the drugs. This relates to Dr F J Flint's original question (14 November, p 1336). Dr Bliss, on the other hand, writes of hypertensives whose treatment "may" be continued, and must have in mind the symptomless patients whom we must now aim to discover, and about whose management we have time to debate. I fully agree with her that hypotensive treatment requires most careful follow-up; but withdrawing treatment that was started with clear indications requires even more care. C P PETCH King's Lynn, Norfolk PE31 6HA
Primary health care in residential homes for the elderly SIR,-While Dr M E M Herford (30 January, p 347) does not appear to enjoy an optimal working relationship with the staff of the home for which he is responsible, his point that the care staff act as "arbiters of who should see the doctor" could have far-reaching implications. In a recent study of local authority homes for the elderly, 33 5°h of the total 1154 residents were receiving hypnotic drugs.' Levels of hypnotic usage within the 24 homes varied from 16 6% to 54 5%. From a follow-up study six months later (which showed a similar pattern and prevalence of drug usage) attempts were made to account for these wide betweenhome differences in hypnotic prescribing. We found no relationship between the number of GPs attending a given home and the level or variety of hypnotics prescribed. However, a consistent and somewhat paradoxical relationship did emerge between hypnotic usage and the resident's level of
dependence. As a group, the least dependent residents (that is, those with the least degree of mental or physical impairment) showed the highest probability of receiving hypnotics, and the probability of receiving hypnotics was appreciably less in high-dependency groups. Similar findings have been reported in the USA.2 Various factors, acting singly or in combination, might account for this relationship. Disturbed sleep in a residential home can arise from several non-medical causes (for example, unfamiliar surroundings, sharing a room, departures from an established daily routine, etc). If such problems are, as Dr Herford suggests, selectively referred to the general practitioner, then it is possible that the prescribing of hypnotic drugs is, to some extent, mediated by the staff's perception of need. It is plausible that the needs of the least dependent, perhaps more demonstrative, residents are better communicated to the staff than are the needs of the more dependent individuals. As Dr M J Clarke and others (14 November, p 1307) have pointed out, care staff draw the attention of GPs to the needs of the resident as a relative might if they were at home. This specific example illustrates some of the difficulties in providing primary health care for an increasingly debilitated population in what is, ostensibly, a noiA1-medical institution. Clearly, residential homes for the elderly require special medical support. I would agree with the conclusions of Dr Clarke and others that such support should be flexible, and sensitive to the needs of both the residents and the home. K MORGAN
VOLUME 284
27 FEBRUARY 1982
there been any cases of sinus arrest occurring in patients taking amiodarone alone ? A K JONES Accident and Emergency Department, Warrington District General Hospital, Warrington WA5 IQG
Retroperitoneal fibrosis associated with metoprolol
SIR,-Dr M J Mitchinson (30 January, p 347) makes two comments about the report of an association between retroperitoneal fibrosis and metoprolol. The first and major comment is that "idiopathic retroperitoneal fibrosis is usually such a slowly progressive disease that it seems unlikely to have reached an advanced stage in the 11 months of exposure to the drug." This view is incorrect. In a series of patients with retroperitoneal fibrosis associated with methylsergide therapy and treated surgically four out of 14 had taken the drug for less than one year and all but two for less than two years.' The period of exposure to metoprolol in this case is thus consistent with its being the culpable agent. The second point made is that hypertension itself is a feature of retroperitoneal fibrosis. As Dr Mitchinson has so beautifully demonstrated,' hypertension is a late feature due to severe renal damage and occurs when there is bilateral hydronephrosis or a non-functioning kidney on intravenous pyelography. Hopefully, future reports of retroperitoneal fibrosis occurring in patients taking beta-blockers (and other drugs) will provide urea or creatinine and erythrocyte sedimentation values from near the time beta-blocking agents were started where University Department of possible. Reports of the occurrence of retroPsychiatry, Royal Edinburgh Hospital, peritoneal fibrosis in patients taking betaEdinburgh EH10 5HF blocking agents for angina rather than hypertension are obviously of great value.3 Morgan K, G;illeard CJ. Neuropharmacology 1981; 20:1355-6. The number of case reports of retroperito2 Ingman SR, Lawson IR, Pierpauli PG, Blake P. neal fibrosis associated with beta-blocker usage I Am Geriatr Soc 1975;23:309-16. is steadily growing. Details of all such cases will, I hope, be reported to the Committee on Safety of Medicines. Only in this way will it be determined whether retroperitoneal fibrosis is Sinus arrest during treatment a rare but important side effect of beta-blocker with amiodarone therapy. D W BULLIMORE SIR,-I read with interest the report by Dr Brian McGovern and others (16 January, University Department of Medicine, St James's Hospital, p 160) of two cases of sinus arrest during Leeds LS9 7TF treatment with amiodarone; but I must ask the question "Could these cases be attributed to a lGraham JR, Suby HI, LeCompte PR, Sadowsky NL. N EnglJ Med 1966;274:359-68. direct effect of amiodarone alone ?" Might I 2 Mitchinson MJ. BrJ Surg 1972;59:58-60. direct attention to the fact that both the 3 Pierce JR, Trostle DC, Warner JJ. Ann Intern Med 1981 ;95:244. patients involved were also taking digoxin at the time the reported sinus arrest occurred ? The reported actions of amiodarone include potentiation of digoxin-presumably by a Dobutamine and salbutamol in mechanism of displacement from protein- cardiogenic shock binding sites. One might expect this to be relevant in these instances as digoxin itself is SIR,-The recent article by Dr M B Fowler thought to depress sinoatrial nodal activity and others (9 January, p 73) was a potentially through vagal stimulation and, in higher doses, important study undertaken to compare the through a direct effect on the myocardium- haemodynamic effects of two 5-agonists, with though, admittedly, this latter effect is thought differing specificity for 3,- and r2-adrenoto be due to an atrial conduction failure rather ceptors, in cardiogenic shock. However, the than an effect on sinoatrial nodal rhythmicity. incorrect use of statistical methods in the It follows that sinus arrest and subsequent analysis of the results seriously impairs bradycardia could conceivably be attributed scientific assessment of the haemodynamic to the increased available digoxin in a patient effects of each and comparison between the taking both digoxin and amiodarone. I feel two drugs. sure that Drs McGovern and others have In particular, the use of multiple t tests in a crossconsidered this possibility, but this is by no over and repeated measures design is erroneous as it takes no account of the effect of multiplicity of means clear when one reads the article. To this I might add another question: have such analyses on the estimates of significance.' 2 An
BRITISH MEDICAL JOURNAL
VOLUME 284
appropriate statistical method would have been analysis of variance, followed by either trend-curve fitting or a comparison procedure to partition the variance between patients, time, and that attributable to therapy.3 The statement in the "Results" section that "the haemodynamic response to the two drugs was similar . . ." cannot therefore be uncritically accepted; statistical validation would require analysis of variance on the incremental changes induced by each drug together with a procedure such as Fisher's least-significantdifference test4 to facilitate comparison. Even with such analysis, however, it is difficult to see how adequate allowance for subtle carry-over effects from the first to the second drug period could have been excluded with only 30 minutes between crossover; this appears to take little account of the well-described altered pharmacokinetic profile and impaired drug handling in cardiac decompensatory states.5 A further criticism is that the wisdom of plotting one variable (pulmonary artery end-diastolic pressure-fig 3) against a derivative of itself (stroke-work index) is of questionable value; thus decreasing the former variable will increase the stroke-work index even if other equation variables -heart rate, cardiac output, and mean blood pressure-are completely unchanged. A review for the first three months of 1976
conservatively estimated that 52°h of publications using statistics in the BMJ contained at least one error6; clearly this is still a problem. The publishing of studies where inappropriate statistical methods are employed leaves the reader in a quandary about the meaningful assessment of the results.
27 FEBRUARY 1982 reanalysed our data using an analysis of variance and covariance with repeated measures of the data in tables I and II of our article (computer package BMDP software 1981, program BMDP 2V with Boneferroni theorem for protection against multiple comparison). As expected, the interpretation of our results was not influenced to any important degree. Most p values reflected the same or greater significance and any changes observed did not affect the clinical message in any way. With regard to the comparison between two drugs, we have been advised that trend-curve fitting with small numbers is inappropriate and a futile exercise. But this is not the most important issue. We are primarily clinicians, and so are most of those who will have to read our paper. A glance at figures 2 and 3 shows that there is no distinction of clinical importance between the two drugs save for the greater tendency of dobutamine to lower indirect left atrial pressure. We are surprised that Drs Silke and Nelson should be troubled by possible "subtle carry-over effects from the first to the second drug period...." When making observations on critically ill patients we have a serious obligation to restrict the time without treatment to the shortest which will give us meaningful data. Both drugs have very short biological half-lives, but we recognise the possible complication of a carry-over effect in patients with heart failure, and we therefore varied the order of administration of the drugs using random selection. In any case our data did not suggest any important carry-over effect, and most variables had returned to near-control values after the 30-minute interruption of therapy and at the completion of the study period (in the six patients in whom the drugs were stopped for a second time).
We have sympathy with the objection to B SILKE plotting end-diastolic pressure against strokeG NELSON work index, though it is conventional to do so University Department of and the plot does give one clinical approximaCardiovascular Studies, General Infirmary, tion to a ventricular function curve. We used Leeds LS1 3EX this technique to illustrate that those with poor ventricular performance respond least well to Glantz SA. Circulation 1980;61 :1-7. 'Tukey JW. Science 1977;198:679-84. drugs-though their need is greatest. This is 3 Wallenstein S, Zucker CL, Fleiss JL. Circulation Res readily seen to be the unhappy truth, however 1980;47:1-9. Winer BJ. Statistical principles in experimental design. we present it. 2nd ed. New York: McGraw-Hill, 1971:177-85. M B FOWLER 'Benowitz N, Meister W. Clin Pharmacokinetics 1976; 1:389-405. A TIMMIs 'Gore S, Jones IG, Rytter EC. Br MedJ7 1977;i :85-7. J C CRICK RICHARD VINCENT ***We sent this letter to the authors, who reply D A CHAMBERLAIN below.-ED, BMJ. SIR,-We believe that the comments of Drs Silke and Nelson on statistical methods commonly used in clinical studies justify further debate. Perhaps the most important matter concerns the use of multiple t tests in serial observations at increasing dose levels of a single drug. The more dosage points which are considered the greater the likelihood that an apparently significant result will arise by chance alone. Moreover, a significant result at one point must influence the prospects of a significant result occurring at the next point-so that the assessments are not truly independent. We face inherent difficulties in conducting studies of this design. Some may consider a simple approach in statistical analysis to be justified, especially if results are interpreted conservatively. The simple approach is certainly fashionable. We have reviewed 15 papers of studies using a design similar to ours published in six leading medical journals over the past four years (including papers in Circulation, which carried the editorial on statistical method cited by Drs Silke and Nelson). Only three adopted more complex methods similar to those suggested in the letter above. We do not wish to imply that more appropriate tests of significance should not be employed because they are complex. They would most certainly be indicated when the interpretation of data may be influenced by fine distinctions or the chance appearance of spurious significant values. This was not the case in our study. We have, nevertheless,
Royal Sussex County Hospital, Brighton BN2 5BE
Hyperosmolar non-ketotic diabetic syndrome precipitated by treatment with diuretics SIR,-Drs Vivian Fonseca and David Phear (2 January, p 36) mention that there is no report of bumetanide precipitating the hyperosmolar non-ketotic diabetic syndrome. Their own case 5 had taken both bumetanide and a beta-blocker. I have recently had a patient with the syndrome whose only previous therapy was bumetanide. A previously fit 69-year-old male caucasian presented in August 1980 with acute left ventricular failure secondary to ischaemic heart disease, which responded to treatment with diuretics alone. He was maintained on bumetanide with a maintenance dose rising to 6 mg daily because of recurrent cardiac failure. His potassium supplement was stopped in November 1980 because of a serum potassium concentration of 5-2 mmol/l(mEq/l). He remained very well, his electrolytes being entirely normal in March 1981. He presented again acutely in June 1981 with confusion and gross dehydration. Investigations at that time showed a blood glucose concentration of 49 mmol/l (883 mg/100 ml), sodium 127 mmol(mEq)/l, potassium 4-6 mmol(mEq)/l, urea >35 mmol/l (211 mg/ 100 ml), and bicarbonate 36 mmol(mEq)/l, with a
665
calculated osmolality of greater than 431 mmol (mosmol)/kg. He responded well to careful rehydration and insulin and has been maintained since on insulin and continues with bumetanide 6 mg daily, still with no potassium supplement. When seen this month he was well and his electrolytes were normal.
This case gives some support to the hypothesis that hyperglycaemia is precipitated by potassium depletion since, although the patient never had a low serum potassium concentration, it is probable that he has depleted total body potassium in view of the prolonged diuretic therapy. STEPHEN HALL Tunbridge Wells, Kent TN1 2DX
SIR,-We wish to endorse the message expounded by Drs V Fonseca and D N Phear (2 January, p 36) that diuretics may precipitate a hyperosmolar non-ketotic diabetic syndrome in the elderly.
Recently we managed a 91-year-old widow who presented as an emergency with a three-week history of polyuria, polydipsia, and mental clouding. Six weeks previously she had been started by her general practitioner on xipamide 20 mg daily because of mild left ventricular failure. Otherwise she was well, was taking no other medication, and was able to look after herself at home. On admission, however, she was dehydrated and conscious but confused, and had atrial fibrillation, with a ventricular rate of 100 beats/min, and bilateral basal crepitations. She had 2 % glycosuria but no ketonuria. Her plasma glucose was 41-3 mmol/l (744 mg/100 ml), urea 25-1 mmol/l (151 mg/ 100 ml), sodium 134 mmol(mEq)/l, potassium 5-0 mmol (mEq)/l, and total carbon dioxide 24 mmol/l (10-8 mg/100 ml). The calculated plasma osmolarity was 344 mmol (mosmol)/kg. With intravenous fluids and soluble insulin her biochemical abnormalities rapidly came under control and her normal mental clarity was restored. It proved possible to stabilise her diabetes on diet and glibenclamide 10 mg daily, and to control her left ventricular failure and ventricular rate with oral digoxin. She was vocal about her desire to return to an independent existence and was able to do so two weeks after admission. Xipamide is a diuretic which has been intro-
duced relatively recently and has found its main use alone or in combination with betaadrenergic-blocking drugs in the treatment of mild or moderate hypertension.1 2 In a dose of 20 mg it provokes a diuresis equivalent to that following 40 mg frusemide but, because this is achieved over a period of 12 or more hours, it is also potentially useful in the management of chronic left ventricular failure. It is a salicylic-acid derivative which bears a superficial structural resemblance to chlorthalidone; but whether it acts, like chlorthalidoneand hydrochlorothiazide,on the distal renal tubule alone or on both the distal tubule
