Relationship between Surgery under General Anesthesia and the Development of Dementia: A Systematic Review and Meta-Analysis

Table 1

Study characteristics.

Study (1st author, year); Region	Study design	Study population	Source of data extraction	Method of exposure assessment	Outcome definition	Lag time; adjusted or matched covariates	Group definition		Statistics; (95% CI)
Cohort study
Kim, 2018 [27]; South Korea	PC	From the South Korean NHIS-NSC database.; free of dementia	Patient records on NHIS-NSC database files between 2002 and 2013	GA operation codes in the NHIS-NSCdatabase	Dementia (clinical diagnosis using ICD-10 codes and history of dementia medication)	Did not included lag time; adjusted for gender, age group, health security system, health care visit frequency, and Charlson comorbidity index			HR
							GA group	44954	1.285 (1.214–1.361)
							Age 60–69	17387	1.216 (1.118–1.322)
								5655	1.162 (1.059–1.276)
							Unexposed group	174469	1.000
							Age 60–69	57867	1.000
							Age> 69	41461	1.000
Teipel, 2018 [31]; German	RC	From the German statutory health insurance database.	Medical records from AOK	History of joint replacement surgery	Dementia (clinical diagnosis using ICD-10 codes)	Included lag time; adjusted for cerebrovascular risk factors, age, sex, the presence of delirium, and regular prescription of sedative or analgesic drugs (SAD)	No surgery	154604	Calculated HR; 3.55 (3.13–4.03)
							Quarter 0	10563
							1-3 quarter		0.95 (0.82–1.11)
							4-6 quarter		0.83 (0.70–0.97)
							≥7 quarter		0.91 (0.83–0.99)
Aiello Bowles, 2016 [30]; Washington, USA	PC	Adult Changes in Thought (ACT) cohort which was randomly selected from members of Group Health (GH).; free of dementia	Self-reported data through interview at baseline and follow-upstudy visits	Self-reported anesthesia data (reviewed by anesthesiologist)	Dementia (DSM-IV); AD (possible or probable AD by NINCDS-ADRDA)	Did not included lag time; adjusted for ACT study cohort, age, age at study entry, sex, education, hypertension, diabetes mellitus, smoking, stroke, coronary heart disease, exercise, self-rated health, body mass index, depression, Parkinson’s disease, Charlson comorbidity index, and difficulty with activities of daily living.	High-risk surgery with GA	248	HR (dementia/AD); 0.86 (0.58–1.28); 0.95 (0.61–1.49)
							Other surgery with GA	3363	0.63 (0.46–0.85); 0.65 (0.46–0.93)
							Other surgery with neuraxial anesthesia	123	0.49 (0.26–0.90); 0.62 (0.32–1.19)
							No anesthesia group	254	1.00; 1.00
Chen, 2014-1 [26]; Taiwan	RC	LHID (a subset of the Taiwan NHIRD).; without history of cancer, dementia, parkinsonism, stroke, and brain surgery	Records from the LHID between 2004 and 2007	Record of anesthesia from the LHID	Dementia (clinical diagnosis using ICD-9-CM)	Did not included lag time; matched for exact age and sex.; Cox regressions adjusted for hypertension, hyperlipidemia, depression, and Charlson index.			HR
							Anesthesia group	24901	1.75 (1.59–1.92)
							General	13715	1.46 (1.28–1.68)
							IV or IM	1686	1.60 (1.11–2.30)
							Regional	8777	1.80 (1.57–2.07)
							Control group (4 or 5 patients selected for each person in anesthesia group)	110972	1.00
Zuo, 2010 [8]; Virginia, USA	RC	From the CDR containing deidentified information of inpatients and outpatients in the University of Virginia Health System.	Medical records from the CDR	Record of spine surgery under GA	AD (clinical diagnosis using ICD-9-CM)	Did not included lag time; none	Spine surgery group (from discectomy to fusion between 1992 and 2004)	2881 ()	Calculated OR using number of patients; 0.67 (0.43–1.06)
							No surgery group	6157; ()	1.00
Lee, 2005 [9] USA	RC	Veterans Affairs (VA) patients undergoing CABG or PTCA between October 1996 and September 1997.; without AD	VA administrative databases of inpatient and outpatient encounters	History of CABG (not mentioned about GA, but necessary)	AD (clinical diagnosis using ICD-9)	Did not included lag time; adjusted for age, number of surgeries, number of diagnoses, and length of stay for index hospitalization.	CABG group (including patients who had both CABG and PTCA)	5216	HR; 1.71 (1.02–2.87)
							PTCA group	3954	1.00
Case-control studies
Strand, 2019 [62]; Sweden	CC	Case: Swedish Dementia Quality Registry, diagnoses of AD, late-onset AD, early-onset AD, and mixed Alzheimer’s and vascular dementia in the county of Östergötland from May 2007 to April 2012; Control: selected from the Statistics Sweden	Medical records	Medical record of prior GA with gas	Dementia (diagnoses of AD, late-onset AD, early-onset AD, and mixed Alzheimer’s and vascular dementia from the dementia registry)	Did not included lag time; adjusted for age category, sex, hypotension under anesthesia, total time anesthesia, and number of exposures of anesthesia.	Cases	457	OR; 2.47 (1.17–5.22)
							Controls	420	1.00
Huang, 2018 [63]; China	CC	Case: residents in Shenyang, China, who were diagnosed with dementia between January 2007 and December 2012; Control: matched for every case via the medical reports archival system	Medical records from Chinese database of inhabitants of Shenyang	Medical record of prior GA	Dementia (DSM-IV); AD (DSM-IV)	Did not included lag time; matched for sex and age (within 1 year).	Cases		OR
							Dementia	577	0.81 (0.71–1.09)
							AD	485	0.89 (0.61–1.01)
							Controls
							Dementia	577	1.00
							AD	485	1.00
Chen, 2014-2 [25]; Taiwan	NCC	LHID (a subset of the NHIRD).; Case: newly diagnosed from 2005 to 2009; Control: 4-fold frequency matched	Medical records from the LHID	Record of endotracheal tube intubation GA	Dementia (clinical diagnosis using ICD-9-CM)	Did not included lag time; matched randomly by age (every 5 years of age), sex, and index year. Adjusted for age, sex, depression, diabetes mellitus, hypertension, stroke, and atherosclerosis.	Dementia group	5345	OR; 1.34 (1.25–1.44)
							Control group	21380	1.00
Sprung, 2013 [59]; Minnesota, USA	NCC	From residents of Olmsted County using Rochester Epidemiology Project (REP).; Case: diagnosed between 1985 and 1994; Control: matched for each case	Medical records from the REP	Medical record of exposure to GA between age 45 and the index date	Dementia (DSM-IV); AD (DSM-IV, NINCDS-ADRDA)	Did not included lag time; matched randomly by sex and age (within 1 year).	Cases		OR (dementia/AD)
							Dementia	877	0.89 (0.73–1.10)
							AD	732	0.88 (0.71–1.11)
							Controls
							Dementia	877	1.00
							AD	732	1.00
Bufill, 2009 [61]; Spain	NCC	From subjects in COGMANLLEU study (belonging to the basic health care area of Manlleu).	Interview with participants and their relatives or caregivers	Self- or surrogate-reported	AD (DSM-IV, NINCDS-ADRDA)	Did not included lag time; matched for age and gender.; Adjusted for age.	Cases	51	OR; 3.22 (1.03–10.09)
							Controls	49	1.00
Yip, 2006 [28]; England and Wales, UK	NCC	From Cognitive Function and Ageing Study (CFAS).; Case/control defined based on two times of interview	Interview with participant	Self-reported exposure to GA	Dementia (AGECAT algorithm)	Did not included lag time; adjusted for age, sex, education, and social class.	Cases: Dementia at wave 2/3	133/142	OR (wave 2/3/both); 0.7 (0.4–1.1); 0.6 (0.3–1.0); 0.6 (0.4–0.9)
							Controls: Wave 2/3	2453/1347	1.0; 1.0; 1.0
Harmanci, 2003 [14]; Turkey	CC	Randomly selected from population registries (records of the Muhtars’ list).; Case: probable AD patients; Control: cognitively normal individuals identified by neurologic examination.	Interview with proxy informants	Surrogate-reported history of GA	AD (probable AD by NINCDS-ADRDA)	Did not included lag time; adjusted for level of education, use of electricity for residential heating, and occupational group.	Cases	57	OR; 1.2 (0.58–2.48)
							Controls	127	1.0
Gasparini, 2002 [29]; Italy	CC	Recruited at the Department of Neurological Sciences of “La Sapienza” University of Rome, who were treated between January 1990 and June 1997.; Each case was matched for 4 controls (2 PD and 2 other disease)	Hospital records	Hospital record of exposure to GA in the 1-year and 5-year periods prior to onset of neurological disease.	AD (probable AD by NINCDS-ADRDA)	Did not included lag time; matched for sex, age (within 3 years), and geographical area of residence.	Cases	115	Calculated OR; 1.03 (0.60–1.75)
							Controls (PD)	230	1.00
							Controls (others)	230
Tyas, 2001 [60]; Canada	CC	Randomly sampled from a list provided by the provincial health insurance plan.	Interview and questionnaire	Self-reported exposure to GA	AD (probable or possible AD by NINCDS-ADRDA)	Did not included lag time; adjusted for age, sex, education.	Cases	36	RR; 0.61 (0.22–1.63)
							Controls	658	1.00
Bohnen, 1994 [10]; Minnesota, USA	CC	Case: selected from patients with AD developed between 1975 and 1984 in Olmsted County by reviewing medical records; Control: matched for each AD case from Olmsted County Mayo Clinic patients	Medical records	Anesthesia records for GA	AD (clinical diagnosis using their own preselected specific criteria)	Did not included lag time; matched for age, sex.	Cases	252	OR; 1.28 (0.82–2.00)
							Controls	252	1.00
CHSA, 1994 [57]; Canada	CC	Recruited from both the community and institutions in Canada.	Risk factor questionnaires completed by proxy respondents	Surrogate-reported exposure to GA	AD (probable AD by NINCDS-ADRDA)	Did not included lag time; frequency matching by study center, residence in community or institution, and age group; Adjusted for age, sex, residence, and education	Cases	258	OR; 1.07 (0.60–1.90)
							Controls	535	1.00
Li, 1992 [15] China	CC	Cases: Clinically diagnosed AD inpatients or outpatients from 1988 to 1989.; Controls: Selected from the neighborhoods of the matched cases.	Direct interview using a structured and standardized questionnaire with surrogate informant.	Surrogate-reported history of GA	AD; (probable AD by NINCDS-ADRDA, ICD-10)	Did not included lag time; Matched by age (within 3 years) and sex.	Cases	70	OR; 0.62 (0.20–1.86)
							Controls	140	1.00
Kokmen, 1991 [11]; Minnesota, USA	CC	Cases: Rochester, Minnesota, residents with onset of AD between 1960 and 1974 using the existing medical records resource.; Controls: matched for each case by searching the registration system at Mayo Clinic.	Entire community medical records.	Medical record of prior GA	AD (clinical diagnosis by reviewing clinical and postmortem data)	Did not included lag time; matched by age (within years), sex, and duration of community medical record.	Cases	415	OR; 0.86 (0.60–1.21)
							Controls	415	1.00
Graves, 1990 [56]; Washington, USA	CC	Cases: patients living in Washington state who were diagnosed with AD between January 1980 and June 1985.; Controls: friend, relative or surrogate of the cases.	Interview with surrogate respondents	Surrogate-reported history of surgery with GA	AD (DSM-III, NINCDS-ADRDA)	Did not included lag time; matched by sex and age (within 10 years).; Adjusted for age in the reference year.	Cases	130	OR; 1.21 (0.54–2.73)
							Controls	130	1.00
Broe, 1990 [58]; Australia	CC	Cases: from consecutive new referrals to dementia clinics in Sydney by general practitioners (GPs).; Controls: a person matched for each case from same GP’s files.	Interview with the informants of the cases and controls	Surrogate-reported exposure to GA	AD (probable or possible AD by NINCDS-ADRDA)	Did not included lag time; matched for sex and age within 2 years.; Matched pairs odds ratio was calculated.	Cases	170	OR; 0.95 (0.50–1.81)
							Controls	170	1.00
Amaducci, 1986 [12] Italy	CC	Cases: Patients admitted to the neurology departments of the seven centers between 1982 and 1983.; Controls: 1 hospital (from same hospital) and 1 population control (neighbor, a friend, or an acquaintance) was identified for each case	Interview with a surrogate respondent.	Surrogate-reported exposure to GA	AD; (clinical diagnosis using their own criteria)	Did not included lag time; Matched for age (within 3 years), sex, and region of residence.; Matched-pair analysis were used.	Cases	116	Calculated OR; 1.00 (0.38–2.66)
							Controls (hospital)	116	1.00
							Controls (population)	97
Heyman, 1984 [13]; USA	CC	Cases: participants in a comprehensive clinical, genetic, and epidemiological study of AD at Duke University Medical Center.; Control: 2 matched subjects selected by the telephone sampling technique of random-digit dialing.	Structured interview with a close family member.	Surrogate-reported history of surgery with GA	AD (clinical diagnosis using their own diagnostic procedure)	Did not included lag time; matched for sex, race, 5-year age interval (50-54, 55-59, etc.), and residential area.	Cases	40	Calculated OR; 0.63 (0.20–1.96)
							Controls	80	1.00

Abbreviations. : number of subjects; CI: confidence interval; HR: hazard ratio; OR: odds ratio; RR: relative risk; AD: Alzheimer’s disease; GA: general anesthesia; PC: prospective cohort study; RC: retrospective cohort study; NCC: nested case-control study; CC: case-control study; ICD-10: International Classification of Diseases, 10th Revision; ICD-9-CM: International Classification of Diseases, 9th Revision, Clinical Modification; DSM-III: Diagnostic and Statistical Manual of Mental Disorders, 3rd Edition; DSM-IV: Diagnostic and Statistical Manual of Mental Disorders, 4th Edition; NINCDS-ADRDA: National Institutes of Neurological and Communicative Disorders and Stroke-Alzheimer’s Disease and Related Disorders; AGECAT: Automated Geriatric Examination for Computer Assisted Taxonomy; NHIS-NSC: National Health Insurance Service-National Sample Cohort; LHID: Longitudinal Health Insurance Database; NHIRD: National Health Insurance Research Database; AOK: Allgemeine Ortskrankenkasse; CDR: Clinical Data Repository; CABG: coronary artery bypass graft; PTCA: percutaneous transluminal coronary angioplasty; PD: Parkinson’s disease; CHSA: The Canadian Study of Health and Aging.